U.S. Cities in the ‘World City Network’Feb 22, 2005 · Angeles are the U.S. leaders in global...

February 2005 • The Brookings Institution • Survey Series 1

M e t r o p o l i t a n P o l i c y P r o g r a m

“The rise of

transnational

interactions

has produced a

new economic

globalization

in which cities

and their

regions are the

prime nodes.”

■ While New York, Chicago, and LosAngeles are the U.S. leaders in globalconnectivity, San Francisco, Miami,Atlanta, and Washington are alsoimportant nodes in the world citynetwork. A mixture of regional capi-tals—such as Boston and Seattle—andspecialist cities—such as Houston, theworld’s energy capital—are found in thenext tier. Somewhat surprisingly,dynamic and growing Phoenix and SanJose have relatively low connectivities.

■ U.S. cities overall—and particularlynon-coastal cities—are generally lessglobally connected than their Euro-pean Union and Pacific Asiancounterparts. Sixteen EU cities and 11Pacific Asian cities rank among the top50 most globally-connected cities, whileonly seven U.S. cities make the cut.Chicago is the only high ranking U.S.city not located in a coastal state.

■ While important service connectionsexist among certain U.S. cities andparticular global regions, U.S. citiesare more strongly linked to other U.S.cities than to cities around the globe.New York is the only U.S. city with more

non-U.S. cities than U.S. cities in its topten list of strongest global connections.Only three non-U.S. cities makeMiami’s top 10 list, for example, whilePittsburgh’s list contains none.

■ Even the most globally-connectedU.S. cities are more locally orientedthan cities in the EU. New York andMiami are the least locally oriented ofU.S. cities, but even they are far moreU.S.-centric in their connections thanmost European cities are EU-centric. Amuch larger number of EU cities boaststrong connections to Latin Americaand Pacific Asia than cities in the U.S.

■ Beyond localism, functional linkagesamong firms are more importantdeterminants of connectedness thangeography alone. While Europe has noAmerican Pacific Coast city linkages,for example, it has many internationalfinancial centers with strong connec-tions to the rising banking and financehubs of Pacific Asia. Many of thesecities, as well as New York, are muchmore oriented to the Pacific Rim than San Francisco, Seattle, or evenLos Angeles.

FindingsAn examination of economic connections among cities’ global advanced service firmsfinds that:

U.S. Cities in the ‘WorldCity Network’Peter J. Taylor and Robert E. Lang1

As cities aim to position themselves better economically, they must remember that theyoperate in a global marketplace. Cities able to grow and attract globally-connected, high-value service firms can access, and benefit from, a worldwide array of customers, workers,and contracted services, ultimately boosting quality growth at home.

Introduction

Just about anyone with even acasual interest in the geographyof the global economy couldguess that New York, Los Angeles,

and Chicago are major world cities,with strong ties to other cities andregions worldwide. But what aboutother U.S. cities? How do places likeHouston and Detroit—or even Port-land and Cincinnati—connect to therest of the world? What is their placein the “world city network?” And whydoes it matter?

Until recently there were no realanswers to these questions. Moststudies on the connections betweenglobal entities use nations—notcities—as their unit of analysis, and/ortend to focus on political, as opposedto economic, relationships. Those thathave focused on world cities havefailed to provide adequate measures ofinter-city relations. By contrast, thispaper introduces a new approach tonetwork analysis that allows us tobetter understand the relative connec-tivity of cities in the world economy.Using this method, we track the globaldistribution of 100 leading advancedservices firms to determine one impor-tant aspect of cities’ economicconnectedness to other world citiesand the patterns of these linkagesacross the globe.

The importance of these relation-ships is becoming increasinglyapparent. The conditions of contem-porary globalization have, in fact,spurred a renaissance of major citiesacross the world, and a new economicconfiguration is emerging based uponcities.2 This is not to say that states areno longer significant in their interna-tional relations, but the rise oftransnational interactions hasproduced a new economic globaliza-tion in which cities and their regionsare the prime nodes of a nascentnetwork society. In this new contextU.S. domination can no longer betaken for granted. Yes, New York canclaim to be “capital city of the world,”

but in reality it shares its economicdominance with other global cities,notably London and possibly Tokyoand Paris. And other “global” U.S.cities are still far more locally orientedthan their European counterparts. Putsimply, the clear supremacy of theU.S. in international relations andmilitary power does not translate intoa similar position for American worldcities within the world city network.

The findings from this report high-light these shifting dynamics, andsuggest multiple implications forelected officials, policymakers, urbanplanners, and others who areconcerned with regional developmentand competitiveness. How a city plugsinto the world economy of advancedproducer services could impact localgrowth patterns, for example, andoffer opportunity for development ofits downtown. It could influence acity’s economic prospects by deter-mining how many high-paying servicejobs are locally created. And it couldhelp shape what types of households acity attracts and retains. In short,global connectivity has becomeanother urban barometer—a newstatistic that helps reveal the healthand strength of cities and regions.

Background

This study treats world cities asthe global service centers ofthe world economy. As such,the network of flows between

cities described here provides askeletal spatial organization ofcontemporary globalization. Thisapproach is an extension of the usualunderstanding of world cities andtherefore requires some exposition.

However, there are many ways citiesare connected globally that are beyondthe scope of this paper. Bentonville,AR, headquarters to Wal-Mart and itsvast supply chain, certainly is veryglobally connected, as may be a largeagribusiness in Iowa. Additionally,cities with large concentrations of

immigrants can have strong globalconnections due to the ebb and flowof both people and money back tocountries of origin.

This paper focuses on one type ofeconomic connection—those betweenadvanced producer services firms—important because of the high valuethey provide to places. This approachis an extension of the usual under-standing of world cities and thereforerequires some exposition.

Cities, hierarchies, and networksWorld cities studies suffer from atheoretical legacy that assumes allinter-city relations are necessarily hier-archical.3 A product of the nationalurban systems school of research thatflourished in the 1960s and 1970s,this theory of inter-city hierarchicalrelations views cities as rivals, strug-gling against each other to “reach thetop.”4 With the advent of globalization,the competition is now thought to beeven more cut-throat, and a large“cities competition” urban policy liter-ature has emerged as a result.5 Thesestudies of worldwide inter-city rela-tions offer little or no interrogation ofhow cities actually relate to oneanother in a global space-economy,however. And so we ask: Must citiesalways form a hierarchy?

In a rare exception to the near-ubiquitous hierarchical premise, JaneJacobs, in her classic Cities and theWealth of Nations, developed aneconomic growth theory based upon“dynamic cities” whose relations withother cities are primarily mutualrather than competitive.6 Citiesgenerate local economic growth byreplacing imports from other cities ina process that, rather than being azero-sum game, stimulates new growthin the city losing the exports. A hypo-thetical example might be cities losinga particular manufacturing production(e.g. cotton textiles) compensating byproviding the machine tools (e.g.machinery for producing cottontextiles) for the new production in thenew cities. The result is a complex and


interdependent network of citiesrather than a simple hierarchicalstructure, with cities cooperating ineconomic development, not justcompeting for a share of the spoils.

This project adopts the Jacobsmodel of synergistic inter-city rela-tions. In our interpretation, we adhereto the theoretical distinction betweenhierarchical/competitive processes andnetwork/cooperative processes butunderstand that in practice the “realworld” is somewhat messier, and“hybrid processes” are the norm.7 Butwhile we recognize that a world citynetwork can include hierarchicaltendencies, we believe the latter are tobe searched for empirically within anetwork model.8

Our theoretical position leads totwo separate but related issues: Whatis the agency that creates the citynetwork, and how is the networkformed by this agency?

Globalization, cities, and agencyGlobalization is constituted by amyriad of transnational processes—economic, political, and cultural—thatdrive much contemporary socialchange. Economic globalizationconverts an “international economy”into a “global economy”—withtransnational financial marketscommonly cited as evidence for thischange. Cities are central to theseprocesses, initially as headquarterslocales for multinational corporations;9

and later as advanced producer servicecenters.10 In this project we followSaskia Sassen’s lead by consideringkey financial and business services asthe hallmarks of contemporary worldcities.11

One of Sassen’s major contributionsto economic globalization theory hasbeen her linking of the literature onthe growth of the service sector tourban change. Service growth relativeto manufacturing has led to an impor-tant economic shift in the second halfof the twentieth century, and someservices—those providing advancedknowledge products to multinational

corporations—have become a cutting-edge industry in the new globalizingeconomy. A key feature of firmsproviding these services is theirconcentration in major cities, espe-cially their downtowns.12 Advancedproducer service firms utilize theknowledge-rich environs of these citieswhile themselves contributing todowntown development through theirbusiness practices.

For Sassen, major service cities thusbecome “global cities,” the “strategicplaces” of economic globalization. Asthe locales for production, develop-ment, and marketing of advancedservices, global cities are contempo-rary manifestations of Jacobs’ dynamiccities. This means that their success isnot a matter of separate city develop-ment, rather they function as nodes ina worldwide network of cities.

Cities are not themselves the primeagents of world city network forma-tion, however; it is advanced producerservice firms that have been largelyresponsible for creating and main-taining the network. These firms haveoffices in important cities across allworld regions, and personnel, informa-tion, knowledge, intelligence, ideas,plans, instructions, and advice flowfreely among them. As such, theseglobal service firms “interlock” thecities in which they have a presence.Viewed this way, the world citynetwork can be measured.

Methodology

There has been a continuousinvention of phrases todescribe inter-city relations inthe world cities literature,

including: “world city hierarchy”;13

“world system of cities”;14 “globalnetwork of cities”;15 “world system of metropolises”;16 “transnationalurban system”;17 “new global urbanhierarchy”;18 “functional world citysystem”;19 and “global urban net-work.”20 The variety of phrasessuggests a literature struggling to

understand this new scale of inter-cityrelations, confirmed by the fact thatnone of the above examples are in anysense formally specified. Thus wecannot be sure that these labels aretrying to describe the same phenom-enon or whether they are suggestingdifferent types of relations. Incontrast, we treat the world citynetwork as a formally specified, empir-ically-based model that measures theinterlocking network of globaladvanced service firms.21

Data productionThe increasing relevance of a city-based perspective on worldwidedevelopments has not yet led to anincreased availability of data tomonitor this process. In contrast tocomparisons between countries,worldwide statistics for evaluatingcities are not generally availablebeyond simple attribute measuressuch as population sizes.

A study of inter-city relations thusrequires the creation of new data,customized for describing a new geog-raphy. Devising ways to meet this needhas been the principal task of theGlobalization and World Cities(GaWC) virtual research programcentered at Loughborough Universityand the Metropolitan Institute atVirginia Tech (MI).22 We introduce theGaWC/MI data approach here as anecessary base for beginning to under-stand the ways in which U.S. worldcities relate to other world cities,within and outside of the U.S.

In 2000, we collected data on alarge number of global service firmsacross a large number of citiescovering all regions of the world.23 Wefocused on firms from six key producerservice sectors—accounting, adver-tising, banking/finance, insurance, law, and management consulting—choosing only those with offices in 15 or more cities, including at leastone city in each of North America,Western Europe, and Pacific Asia. Intotal, we identified 100 such firms—18 in accounting, 15 in advertising,


23 in banking/finance, 11 in insurance,16 in law, and 17 in managementconsulting—across 315 cities world-wide.

We defined cities broadly as “city-regions” or metropolitan areas, but inpractice the service offices consideredwere largely concentrated in thecentral city, especially in downtowns(hence the use of the term “city”—asopposed to “metropolitan area” or“region”— throughout this report).Using information mainly gatheredfrom firms’ websites, we were able todetermine each firm’s distribution ofoffices and code each city in terms ofits importance, or “service value,” inthe firm’s office network. To computethis value, we employed two types ofinformation about individual cityoffices: the size of the office as aservice provided (e.g. the number ofpartners based in a law firm’s office),and the extra-locational functions of acity office (e.g., a regional headquar-ters).24

Using this data, we developed aninterlocking model consisting of asimple matrix arraying world citiesagainst global service firms. Eachmatrix cell indicates the importance ofa given city within the office networkof a given firm. The end result of thisexercise was a service values matrixmeasuring 100 firms times 315 citiesdepicted as an array of service values(codes) ranging from 0 to 5 (i.e., fromno presence to headquarter location).The interpretation of this matrix isquite straightforward. Each columndescribes a firm’s global office strategyacross 315 cities, where it is strongand where it is weak. Each rowdescribes the particular global servicemix of a city, in which firms andsectors it is strong and in which firmsand sectors it is weak.

In total, the matrix contains 31,500pieces of information. This large quan-tity is significant. The measures wederive from the data are aggregationsof different office networks of manyfirms. Each path to becoming a globalservice firm is distinctive in the mix of

processes that create an officenetwork, so that country of origin,initial expansion strategies, take-overopportunities taken, etc. each produceparticular results. By combining 100such globalization paths, we created ageneralized world city network that isnot dominated by just a few specificprocesses. Thus measurements basedon these data are reasonable reflec-tions of the broad new geography thatis economic globalization.

Measuring connectivityUsing our matrix, we then set out tomeasure the connectivity of the 315cities within the world city network.We based this measure upon theassumption that the more importantan office is within a firm’s network,the more flows—of information,knowledge, ideas, plans, etc.—it willgenerate. In other words, we expectgreater connection between two citieswith important offices of a firm thanbetween another pair of cities thatboth house more modest offices. Thuswe treat products of service values forpairs of cities as indicating connectionpotentials. Summing all such productsfor a given city for all firms across allother cities defines the city’s globalnetwork connectivity (GNC).Computing these connectivitiescreates very large sums, however, so tomake comparison easier, we convertedthem into relative measures of GNC,reporting them as proportions of thehighest GNC score that is computedfrom this data, that for London.

Our measure of global networkconnectivity has three important prop-erties that should be kept in mindwhen interpreting results. First,connectivity is measured irrespectiveof international boundaries. Forinstance, consider a New York law firmwith offices in, say, Atlanta andCaracas that are both given servicevalues of 2. In computing New York’sglobal network connectivity theAtlanta-New York link and Caracas-New York link are identical: they eachcontribute 10 (2 x 5, remember 5 is

scored for headquarters) to New York’soverall connectivity. Second, citieswith a large number of global servicefirm offices will, obviously, tend to bemore connected than cities with lesssuch offices. Thus it is no surprisethat London and New York turn out tohave by far the largest global networkconnectivity measures in our results.Third and countering the lattertendency, cities housing offices offirms with the very large officenetworks (i.e., found in very manycities) will be more globally connectedthan cities housing offices of firmswith mainly small networks. Thusglobal network connectivity does notsimply reflect the number of globalservice firms in a city—one city canhave many more global service officesthan another but if the latter’s officesare concentrated among firms with thegreatest networks, it may well have thehigher global network connectivity.

Defining “hinterworlds”In traditional urban studies it wascommon to depict towns and cities asservice centers, each with its own“service area” (or hinterland) that itprovisioned. In these global urbananalyses, we also treat cities as servicecenters, but the area they servicethrough their global service firms isworldwide. Thus a new term isrequired for the new geographicalscope of advanced producer servicing:city hinterworlds. Unlike hinterlands,hinterworlds have no boundaries sepa-rating the areas provisioned betweenneighboring cities. However, they arelike hinterlands in that they have ageographic basis or a spatial variationin the intensity of provisioning. Thusto map a city’s hinterworld will showwhere it is strongly connected andwhere it is weakly connected acrossthe world.

The computation involved formapping the hinterworlds of cities isquite complex and is outlined in detailin the appendix.25 We applied themethod one city at a time to define itsparticular hinterworld. This consists of


an array of the other cities, for each ofwhich there are values indicating theirdegree of connection with the initialcity. The values straddle zero: positivevalues show a city is stronglyconnected to the initial city, negativevalues indicate a weak connection.The methodology is as follows: first,we computed the link between a givencity and each other city for all firmsthey share (i.e., both have offices ofthe firm); second, we summed themacross firms to provide a provisionalmeasure of links to the other cities;third, we made each link relative tothe maximum service value that couldhave been provided; fourth, the list ofsuch links for a given city to all othercities was compared to the latter’sglobal network connectivities using asimple regression model; and finally,we used residuals from this regressionto determine the other cities to whichthe given city is strongly connected(positive residual) or weakly connected(negative residual). In other words,hinterworlds are defined relative toglobal network connectivity because apositive residual indicates that thegiven city has a higher than expectedconnection to another city than thelatter’s global network connectivitywould suggest, with the oppositeimplied by negative residuals.

As explained in the appendix, wecarried out this exercise using just 123out of the 315 cities in the originaldata.26 This means that hinterworldswere computed for each of 123 cities;every hinterworld consists of measuresof strong and weak connections to theother 122 cities. These have beenmapped and are depicted in the Atlasof Hinterworlds.27 For this report weuse these measures in two ways. First,we constructed tables showing the 10most strongly connected (positiveresiduals) cities and 10 most weaklyconnected (negative residuals) cities toshow the nature of selected cityhinterworlds. Second, we aggregatedcity hinterworlds by major regions toshow the general geographical biaseswithin hinterworlds—this shows

which cities are strongly connected toPacific Asian cities, for instance. Wederived the regional score for a city bysimply summing residuals for all citiesin a given region (details are given inthe appendix).

Findings

A. While New York, Chicago, andLos Angeles are the U.S. leaders inglobal connectivity, San Francisco,Miami, Atlanta, and Washington arealso important nodes in the worldcity network.Table 1 shows the gross and relativeGNC scores for U.S. cities, presentedin order by size of the GNC, andgrouped into 10 strata based on theirconnectivity levels. Alone in Strata I,New York is clearly the Americanglobal city par excellence. In fact, withconnectivities more than a quarterabove all other cities, New Yorktogether with London form a sort ofglobal city duplex.

New York is not the only importantU.S. world city, however: Chicago andLos Angeles constitute a clear, secondstrata in Table 1.28 Globally, these twocities are ranked in the top tenconnected cities with the likes of Parisand Singapore.

To be sure, the dominance of thesethree U.S. cities comes as littlesurprise. Our analysis does revealsome unanticipated, results, however.

Strata III includes Miami, whichgives the city an elevated positioncompared to how it is perceived withina U.S. domestic urban hierarchy.However, as “capital of LatinAmerica,” it fully warrants its relativelyhigh status in the world city network.29

The other cities in the strata are lesssurprising but no less interesting: SanFrancisco, Atlanta, and Washingtonround out this third tier, due by andlarge to their roles as westerngateway/financial center, media centerand unchallenged “capital” of the largeand fast growing South, and nation’scapital, respectively.30

A mixture of regional capitals andspecialist cities attracting services fortheir distinctive businesses is found inother tiers, with the size and economicimportance of region or specialtydetermining the strata location of acity. Thus in strata IV, regional capitalsin New England and the PacificNorthwest (Boston and Seattle) arejoined by the two Texan world cities,one as regional center (Dallas) and theother as the world’s energy capital(Houston).

Similarly, in strata V, leading citiesof smaller and less important regionsare combined with Detroit as theglobal auto center. Other stratacombine sub-regional centers such asCharlotte and Portland with revivingindustrial cities like Cleveland andPittsburgh. There are surprises in bothtypes of metropolitan area, with rela-tively low connectivities for the likes of“old and reviving” Baltimore and “newand dynamic” Phoenix.31

The final city of the series, Wilm-ington, forms the bottom strata on itsown. This is no splendid isolation likeNew York in strata I: Wilmington wasincluded in the data as a relativelylow-level U.S. city to see how its GNCcompared and because, despite itssmall size, the city is the center of theU.S. credit card industry. Clearly sepa-rate from the other selected cities,Wilmington as strata X represents themany other “ordinary” U.S. cities thatdo not appear in our data, mainly capi-tals of smaller states and industrialcities that never formed major metro-politan areas.

However, Wilmington’s GNC scoredoes remind us that globalization isnot just the preserve of leading cities,but that all cities participate in global-ization. In terms of financial andbusiness services the provision inthese less important cities is not largebut neither is it zero. Still, while thereare no “unlinked” U.S. cities in global-ization, only a relatively few canrealistically aspire to global service-ledeconomic development.


B. U.S. cities overall—and particu-larly non-coastal cities—aregenerally less globally connectedthan their European Union andPacific Asian counterparts.

In Table 2 we compare the connec-tivity of U.S. cities to those in the EUand Pacific Asia.32 The 23 U.S. citieslisted are part of the roster of 123cities from across the world that have

at least one fifth of London’s GNC.There are 28 cities in European Unioncountries that similarly qualify butonly 13 Pacific Asian cities. The keypoint this table makes, however, is notthat there are slightly more EU citiesand less Pacific Asian cities, but ratherthat the latter two sets of cities aregenerally ranked higher globally rela-tive to equivalent U.S. cities.

The capitals of the EU and the U.S.encapsulate this situation: both rank7th in their respective local economicareas but Brussels ranks 15th globallycompared to Washington’s 37th worldranking. This pattern is consistentthroughout the table: Barcelona andHouston ranked 10th locally, but 32ndand 62nd respectively globally; Luxem-bourg City and Cleveland are both20th locally, but 63rd and 112th glob-ally. The situation with Pacific Asiancities is similar at the top end of theglobal ranking—whereas only 7 U.S.cities are in the world top 50, 11Pacific Asian cities make the cut. Thisregion has just 2 other world cities,however, reflecting a particularconcentration of services in capitalcities (excepting several cities inChina).

However this tendency of majorEuropean and Pacific Asian cities torank higher than those in the U.S. isnot completely regular: there doesappear to be a “middle gap” in theU.S. listing (between strata III and IVin Table 1) that is balanced by acluster of lowly ranked cities (strata VIin Table 1); neither feature is found inthe EU listing. This appears to bereflected, at least partially, in ageographical dimension. WhereasEurope’s heartland is city-rich inconnectivities (notably in Germany),the American heartland has fewercities and with lower connectivities,Chicago being the exception. AfterChicago, the next U.S. city from anon-coastal state is Denver, ranked73rd globally.


Table 1. Global Network Connectivities of U.S. Cities

City GNC score Relative GNC STRATANew York 61895 0.976 IChicago 39025 0.616 IILos Angeles 38009 0.600 IISan Francisco 32178 0.508 IIIMiami 29341 0.463 IIIAtlanta 27052 0.427 IIIWashington, D.C. 26522 0.418 IIIBoston 22249 0.351 IVDallas 21796 0.344 IVHouston 21424 0.338 IVSeattle 19252 0.304 IVDenver 17368 0.274 VPhiladelphia 17006 0.268 VMinneapolis 16914 0.267 VSt. Louis 16124 0.254 VDetroit 15818 0.250 VSan Diego 14585 0.230 VIPortland 14113 0.223 VICharlotte 13556 0.214 VICleveland 13442 0.212 VIIndianapolis 13347 0.211 VIKansas City 12772 0.201 VIPittsburgh 12707 0.200 VIBaltimore 11309 0.178 VIIPhoenix 11025 0.174 VIICincinnati 10603 0.167 VIITampa 10532 0.166 VIIColumbus 9974 0.157 VIISan Jose 9843 0.155 VIIRochester 9731 0.153 VIIPalo Alto 9078 0.143 VIIIHartford 9007 0.142 VIIIRichmond 8845 0.140 VIIIBuffalo 8798 0.139 VIIIHonolulu 8656 0.137 VIIILas Vegas 7911 0.125 IXNew Orleans 7089 0.112 IXSacramento 6870 0.108 IXOmaha 6564 0.104 IXWilmington 3740 0.059 X+ other cities - - X

Source: Authors’ calculations

C. While important service connec-tions exist among certain U.S. citiesand particular global regions, U.S.cities are more strongly linked toother U.S. cities than to citiesaround the globe.We focused here upon the hinter-worlds of three pairs of U.S. cities,selected to show a variety of U.S. cityhinterworlds. First, we contrasted thecountry’s definitive global city (NewYork) to the U.S. city with the lowestconnectivity in Table 2 (Pittsburgh).Second, we compared two cities withwell-known regional proclivities (LosAngeles-Pacific Rim and Miami-LatinAmerica). Finally, we considered two

important U.S. cities surprisinglylocated in relatively low-level strata inTable 2 (Houston and Seattle). Ineach case we highlight the top 30cities in each of the six hinterworlds(Table 3). In doing so, the “localness”of these cities’ hinterworlds becomesapparent.

New York and Pittsburgh. Asexpected, the hinterworlds of thesetwo cities show quite opposite patterns(Table 3). The Pittsburgh hinterworldis extremely U.S.-focused, with all ofthe top 20 cities in its hinterworldlocated in the U.S. In contrast, NewYork’s hinterworld shows only threeU.S. cities among the top ten—

Boston, Chicago, and Washington.London also appears among the topten well-connected cities to New Yorkand the remaining majority are foundin Pacific Asia. In short, whereas Pitts-burgh’s hinterworld is very “local,”New York’s is global, with specificstrengths in the three leading global-ization regions: the U.S. itself, PacificAsia, and Europe.

Los Angeles and Miami. The hinter-worlds of these two cities show theexpected regional biases beyond theU.S.: Los Angeles is well-connected toPacific Rim cities in both Asia andAustralia, and Miami is well-connected to Latin American cities(Table 3). However, it is also the casethat both these cities are, in general,most strongly linked to other U.S.cities than those elsewhere around theworld. For Los Angeles, nine of the top10 well-connected cities are U.S.cities, the exception being Tokyo. ForMiami, a majority of top 10 well-connected cities are in the U.S., andonly one is found in Latin America(Rio de Janeiro).

It is interesting that despite havinga Spanish origin and a large Hispanicpopulation, Los Angeles is not as well-connected to Latin America as isMiami. We speculate that this mayhave something to do with the immi-gration patterns to both respectiveplaces. Many Hispanics in Miami’sbusiness community are politicalrefugees from Cuba. When Castrocame to power over forty years ago,much of Cuba’s middle class fled tosouthern Florida. This affluent andeducated expatriate communityquickly reestablished a businessnetwork to Latin America. By contrast,most Hispanic immigrants to LosAngeles have been poor. The contactsto Latin America in Los Angeles aretherefore more cultural and are lesslikely to result in the establishment ofbusiness contacts in the advancedproducer services industry.

Houston and Seattle. The hinter-worlds of these two important U.S.cities are strikingly similar despite


Table 2. Global Network Connectivity Rankings: U.S., EU, and Pacific Asian Cities

U.S. cities EU cities Pacific Asian citiesNew York – 2 London – 1 Hong Kong – 3Chicago – 7 Paris – 4 Tokyo – 5Los Angeles – 9 Milan – 8 Singapore – 6San Francisco – 17 Madrid – 11 Taipei – 20Miami – 25 Amsterdam – 12 Jakarta – 22Atlanta – 33 Frankfurt – 14 Kuala Lumpur – 26Washington, D.C. – 37 Brussels – 15 Bangkok – 28Boston – 60 Stockholm – 27 Shanghai – 31Dallas – 61 Dublin – 30 Beijing – 36Houston – 62 Barcelona – 32 Seoul – 41Seattle – 68 Vienna – 39 Manila – 46Denver – 73 Lisbon – 42 Ho Chi Minh – 88Philadelphia – 76 Copenhagen – 44 Guangzhou - 109Minneapolis –77 Hamburg – 48St. Louis – 81 Munich – 49Detroit – 85 Düsseldorf – 50San Diego – 98 Berlin – 51Portland – 105 Rome – 53Charlotte – 108 Athens – 56Cleveland – 112 Luxembourg – 63Indianapolis – 114 Helsinki – 70Kansas City – 119 Stuttgart – 74Pittsburgh – 120 Rotterdam – 75

Cologne – 92Lyon – 93Antwerp – 96Manchester – 101Birmingham – 106


their being from very different parts ofthe U.S. (Table 3). Like Los Angeles,both have nine of their top 10 well-connected cities in the U.S., and inboth cases the other top 10 city is stillNorth American (Montreal). Thusthese cities are very U.S.-oriented intheir connections and, unlike LosAngeles and Miami, they have nodistinctive extra-U.S. regional patternof well-connected cities. In Seattle, forexample, this can be explained by thefact that, while it has a big port, itdoes not have a large producer serviceeconomy relative to such places as Los

Angeles and San Francisco. In general,neither of Seattle nor Houston has ahinterworld very different from that ofPittsburgh.

D. Even the most globally-connectedU.S. cities are more locally orientedthan cities in the EU.The above examples suggest a degreeof local parochialism in U.S. city link-ages.33 Such localism can be easilymeasured as a particular orientationwithin a hinterworld, allowing us toboth generalize to more U.S. cities andto make comparisons again with cities

in Europe. In Table 4 we report the results

from using the U.S. and EU as “localregions” for their respective cities. Thefindings for U.S. cities are quite stark.

The considerable localism of U.S.cities is confirmed by the fact thatthree cities, including Pittsburgh,actually score the maximum in localfocus. It is obvious from this list thatPittsburgh, rather than New York, typi-fies U.S. world cities in terms of theirinter-city relations. All but two U.S.cities—Miami and New York—have alocal orientation above 0.8 (and there


Table 3. The Top 30 Cities in the Hinterworlds of Six U.S. Cities

Rank New York Pittsburgh Los Angeles Miami Houston Seattle1 Washington, D.C. Washington, D.C. Washington, D.C. Dallas San Francisco Dallas2 Tokyo Cleveland Boston Philadelphia Dallas Washington, D.C.3 Boston Dallas Chicago Washington, D.C. Washington, D.C. Portland4 Chicago Philadelphia San Francisco Melbourne Chicago Los Angeles5 Hong Kong St. Louis Dallas Rio de Janeiro Boston Denver6 Guangzhou Charlotte Philadelphia Cleveland Philadelphia Boston7 London Boston San Diego Detroit Los Angeles Minneapolis8 Singapore Minneapolis Tokyo Frankfurt Minneapolis Philadelphia9 Ho Chi Min City Indianapolis Minneapolis San Diego Montreal St. Louis

10 Beijing Kansas City Denver St. Louis Seattle Montreal11 Pittsburgh Denver New York Caracas Denver Barcelona12 San Diego Atlanta Cleveland Bogota Toronto Vancouver13 Cleveland Seattle Pittsburgh New York Pittsburgh Houston14 Frankfurt Detroit Houston San Francisco Rio de Janeiro Detroit15 Bangkok New York Calgary Boston Calgary Chicago16 Minneapolis Chicago Seoul Buenos Aires San Diego San Diego17 Calgary Los Angeles Singapore Kansas City Vancouver Pittsburgh18 Los Angeles San Francisco Guangzhou Vancouver Melbourne Indianapolis19 Denver San Diego Sydney Panama City Kansas City Miami20 Dallas Portland Atlanta Adelaide Charlotte San Francisco21 Bratislava Rotterdam Charlotte Chicago Portland Chennai22 Shanghai Houston Hong Kong Los Angeles Zurich Toronto23 San Francisco Stockholm Vancouver Charlotte Birmingham(UK) Kansas City24 Nassau Brisbane Hamilton (Ber.) San Diego Rotterdam Bogota25 Philadelphia Riyadh Seattle Copenhagen Manama Karachi26 Cologne Perth Detroit Mexico City St. Louis Cleveland27 Kiev Vancouver Shanghai Sao Paulo Sydney Rio de Janeiro28 Sofia Rio de Janeiro St. Louis Port Louis Sao Paulo Melbourne29 Rotterdam Miami Manama Cape Town Indianapolis Atlanta30 Zurich Adelaide Kansas City Barcelona Nassau New Delhi

Non-U.S. cities are shown in italics


are no values of orientation this highfound in tables 5, 6, or 7).

New York’s low level of local orien-tation is distinctly un-notable incomparison with EU cities, however.In general, EU cities are locally-orien-tated but to a much lesser degree thanU.S. cities. Some major EU cities havelow or even negative local orientation(e.g., London and Brussels) and thesame is true of some peripherally-located EU cities (e.g., Helsinki andAthens). This comparison highlightsthe extreme nature of the nationalparochialism of U.S. cities: Even atthe global city level New York turns

out to be much more American thanLondon is European.

European cities generally havestronger regional connections acrossthe board than cities in the U.S. Thisis even true of inter-city links to LatinAmerica, as shown in Table 5. As wewould expect, Miami has the highestLatin American orientation amongboth U.S. and EU cities—but this iswhere U.S. preeminence ends. U.S.cities turn out to be weakly connectedto Latin American cities, while EUcities have a balance of strong andweak connections. This is due largelyto the fact that several medium-sized

EU cities are relatively well-connectedto Latin American cities, includinghistorically linked Lisbon and Madrid.Note that both New York and Londonhave appreciably negative orientationsto Latin America, as their non-locallinkage biases are both directed else-where.

To further U.S.-EU comparisons,we examined cities’ orientation toEastern Europe. We defined thisregion by examining cities that were inCOMECON during the Soviet era. Inthe 1990s they constituted centers formassive privatizations that fuelledglobalization of advanced producerservices, especially in banking/financeand law. In short, this was the “globalopportunity region” of the 1990s andit is clearly EU cities that took advan-tage (Table 6). Amongst U.S. cities,only New York has a positive orienta-tion, with a level on par with Antwerp,which ranks 21st among EU cities.The EU-Eastern European linkagesare how we expect a “backyard rela-tionship” to look; it is precisely nothow the U.S.-Latin American linkagesappear once Miami is allowed for.

E. Beyond localism, functional linkages among firms are moreimportant determinants of connect-edness than geography alone.

Our examination of hinterwords(Table 3), revealed an interestingcontrast between New York and thetwo Pacific cities (Los Angeles andSeattle). Specifically, while we wereable to show a regional bias towardsthe Pacific Rim for Los Angeles, wecould discern no such pattern forSeattle. And yet the Atlantic Coast cityof New York is very well connected toPacific Asian cities, seemingly muchmore so than even Los Angeles. Thismight be explained by New York’spremier role in global finance. Ingeneral, what these contrasts indicateis that the relative network locationsof Pacific Asian cities (specifically inbanking/finance) are much moreimportant than their absolutegeographical location (Pacific Rim) in


Table 4. Local Orientations of U.S. and EU City Hinterworlds

U.S.-ness EU-ness U.S. city (local orientation) EU city (local orientation)

Denver 1.000 Cologne 0.703Indianapolis 1.000 Munich 0.656Pittsburgh 1.000 Stuttgart 0.640Cleveland 0.981 Antwerp 0.625San Diego 0.981 Hamburg 0.609St. Louis 0.981 Lyon 0.578Boston 0.963 Barcelona 0.531Minneapolis 0.963 Berlin 0.531Philadelphia 0.963 Rome 0.500Portland 0.963 Copenhagen 0.375Kansas City 0.944 Milan 0.360Seattle 0.944 Rotterdam 0.360Charlotte 0.926 Lisbon 0.344Dallas 0.926 Luxembourg City 0.250Los Angeles 0.926 Vienna 0.250Washington, D.C. 0.926 Dusseldorf 0.219Detroit 0.907 Birmingham 0.203Houston 0.889 Madrid 0.203San Francisco 0.851 Amsterdam 0.188Atlanta 0.833 Paris 0.172Chicago 0.815 Dublin 0.156Miami 0.741 Frankfurt 0.156New York 0.481 Stockholm 0.141

Manchester 0.109Athens 0.000Brussels -0.016Helsinki -0.045London -0.078


determining their inter-city links. Thiscan be more definitively illustrated byconsidering the Pacific Asian orienta-tions of U.S. and EU cities.

Table 7 lists these Pacific Asianorientations. This is, of course, theworld region whose rise to economicprominence in the last quarter of thetwentieth century helped popularizethe concept of globalization, and led tothe transcendence of the old NorthAtlantic-centered world economy. Theorientations of cities to this keyeconomic region are ordered as we

might expect: the top six include fourPacific Coast cities plus the two domi-nant U.S. international financialcenters, New York and Chicago. Ofcourse, Europe has no Pacific Coastcities, but it does have many interna-tional financial centers; thus forfinancial and business services, EUcities are generally better connected toPacific Asian cities than are those inthe U.S. Clearly, Table 7 confirmswhat we gleaned from Table 3: It isthe financial functions of Pacific Asiancities in the world city network, not

mere geography, driving the formationof city hinterworlds.

What do these findings mean forU.S. cities and their economies? This paper has been largely a descrip-tive exercise. We are working in aresearch area—worldwide inter-cityrelations—where there is little empir-ical work, so we need to provideoriginal findings to depict the newglobal geography. At the same time, wemust be cognizant of the limits of theresults we have produced. In themyriad connections that constituteManuel Castells’ contemporary “spaceof flows,” the inter-city relations wehave described are but a minisculepart.34 But they are a very importantpart, as they describe a cutting edgeeconomic sector in the worldeconomy—advanced services—thathas been crucially instrumental foreconomic globalization. As such, ourfindings should not be dismissed fortheir narrow economic focus, even ifthe meaning of the findings stillremains to be ascertained. We doknow that, at worst, they cast doubt onthe continuing viability of U.S. citiesin the world economy. In other words,while there are many positive thingshappening in U.S. cities today, wecan’t neglect the benefits of denseworldwide connections. Intra-cityvariety and complexity is not an alter-native to inter-city variety andcomplexity, but rather, as Jacobstaught us long ago, they complementeach other.35

To probe the findings further, weneed to consider several key questions:How relevant is the relative under-connectedness and parochialism ofleading U.S. cities to the future pros-perity and development of theirmetropolitan areas? Why haven’t U.S.cities made a greater mark on theworld city network? What are theprocesses or mechanisms operating inthe world economy that are producingthis inter-city outcome? Does thedegree and nature of connectedness inthe world city network translate into


Table 5. Latin American Orientations of U.S. and EU Cities

Latin American Latin American U.S. city orientation EU city orientation

Miami 0.563 Athens 0.469Cleveland 0.125 Copenhagen 0.469St. Louis 0.094 Helsinki 0.406San Diego 0.063 Rome 0.313

Dublin 0.281Atlanta 0.000 Lisbon 0.188Boston 0.000 Hamburg 0.094Dallas 0.000 Stockholm 0.094

Madrid 0.063Chicago -0.031 Milan 0.063Denver -0.031 Vienna 0.063Kansas City -0.031Minneapolis -0.031 Amsterdam 0.000Philadelphia -0.031 Luxembourg City 0.000Charlotte -0.063Detroit -0.063 Cologne -0.031San Francisco -0.094 Manchester -0.031Houston -0.156 Birmingham -0.063Portland -0.156 Dusseldorf -0.125Seattle -0.188 Munich -0.125Indianapolis -0.281 Barcelona -0.156Los Angeles -0.313 Brussels -0.156New York -0.313 Antwerp -0.219Pittsburgh -0.344 Stuttgart -0.219Washington, D.C. -0.344 Frankfurt -0.250

Berlin -0.281Paris -0.281Rotterdam -0.313Lyon -0.344London -0.375


immediate advantages in terms ofeconomic growth? And what, ulti-mately, do the findings mean forpolitical and business leaders?

We attempt to address some ofthese issues below.

What might explain the localism ofU.S. cities?Understanding the mechanismsthrough which U.S. city connectionsin the world city network have beencreated requires additional researchon how agents of world city network

formation—the global service firms—use the cities in which they arelocated. We can, nevertheless, providesome explanation of the processesbehind U.S. cities’ weak connected-ness: ■ The Shadow Effect: Non-U.S.

service firms developing a globalstrategy must locate in New York(and perhaps Chicago and LosAngeles) but often decide on nofurther penetration of the U.S.domestic market. They need a U.S.presence for their non-U.S. clients

but do not wish to invest in servicewithin the U.S. market itself with itshigh cost thresholds. Hence, forglobal services delivered by non-U.S. firms, concentration in NewYork casts a ‘shadow’ with relativelyfew global services provided else-where.

■ The Comfort Effect: Leading U.S.financial and business service firmsmay decide not to develop a globalstrategy and concentrate onexpanding within the domesticmarket. Because this market is solarge, such a strategy may be lessrisky, and it eliminates the need tocope with expensive “loss leader”foreign-city offices. Certainly thereis much less of an incentive for U.S.financial and business service firmsto embark on a transnationalstrategy than for firms in any othercountry. Alternatively, a firm inSwitzerland with its limiteddomestic market must operate glob-ally to survive.These two effects must remain

hypotheses for the time being but theyare, we think, very plausible interpre-tations of the situation. If we acceptthem as prime reasons for U.S. cityunder-connectedness then it is diffi-cult to see how this global deficiencyof U.S. cities will alter any time in thenear future. But does this matter?

Are global connections essential foreconomic success?We can begin to answer this questionwith a simple comparison. Considerthe world city status of Atlanta versusPhoenix. Both regions are Sunbeltboomtowns that expanded rapidly inthe past several decades. The twometropolitan areas now rank amongthe top 15 regions in the country. Butas “U.S. world cities,” Atlanta—the“New York of the South”— places 6th,while Phoenix comes in at 25th (Table1). Thus the question arises towhether Atlanta has a distinct advan-tage over Phoenix because of itssuperior global connectivity.

Cities and city networks are

February 2005 • The Brookings Institution • Survey Series 1 1

Table 6. East European Orientations of U.S. and EU Cities

East European East European U.S. city orientation EU city orientation

New York 0.111 Vienna 0.741Brussels 0.630Hamburg 0.593

Detroit -0.148 Helsinki 0.556Washington, D.C. -0.185 Athens 0.519Los Angeles -0.444 Dublin 0.519San Diego -0.444 Frankfurt 0.482Miami -0.482 Milan 0.444Atlanta -0.519 Amsterdam 0.407San Francisco -0.519 Copenhagen 0.407Cleveland -0.593 Luxembourg City 0.407Portland -0.593 Stuttgart 0.407Seattle -0.593 Dusseldorf 0.304Boston -0.630 Berlin 0.297Chicago -0.630 Paris 0.297Philadelphia -0.630 Cologne 0.259Denver -0.667 Madrid 0.259Houston -0.667 London 0.222Dallas -0.704 Munich 0.222Kansas City -0.704 Lisbon 0.186Pittsburgh -0.704 Antwerp 0.111St. Louis -0.704Indianapolis -0.741Charlotte -0.778 Stockholm -0.037Minneapolis -0.815 Rome -0.111

Rotterdam -0.222Lyon -0.444Barcelona -0.556Birmingham -0.593Manchester -0.593


immensely complex phenomenon.Certainly it would be naïve in theextreme to assume that more globalconnections automatically translateinto more economic growth. It is, ofcourse, a matter of what these connec-tions contribute to a city’s economicvibrancy. The only sure thing we cansay about globalization is that it hasmade cities and their networks evenmore complex. The key point to note isthat these enhanced transnational andworldwide inter-city relations are anaddition: Economic transactions

taking place at other geographicalscales continue to operate. There isnothing within globalization processesthat determine this particular scale tobe the most important for cities. Citybusinesses will enter markets at avariety of scales and will expand theiractivities at levels where they eithermake most profit or expect to makemost profit. This may or may not bethe global scale depending on a wholearray of circumstances.

The advanced producer servicefirms we have studied have all estab-

lished multiple global offices as part oftheir overall location strategy. But theyremain a small minority of firms in allthe sectors we cover. Most financialand business service firms are stillbased upon a local client base gener-ated by building a reputation for theirservice provision.

Overall, in any city, there will beaccess to services operating atdifferent levels. The fact that one cityhas more global connectivity thereforedoes not necessarily mean that it is themore important service center. In ourprevious discussion (Table 2), wenoted that Luxembourg City is muchmore connected globally than Cleve-land (63rd to 112th), howeverCleveland is the larger service centeroverall. Luxembourg City has grownan important global niche as an inter-national financial center but is weakin all other sectors and scopes,whereas Cleveland has importantlocal, state, regional, and nationalconnections in addition to its globalconnections.

Having put our analyses intoeconomic-geographical perspective, wecan begin to answer the question ofthe importance of our findings to thegrowth prospects of U.S. cities. Thefirst point to make is that globalconnections are not necessary forsuccess. Several studies have shownthat medium-sized cities have beenparticularly prospering in recent years.At the global level, Jones LangLaSalle’s “Winning Cities” project hasidentified Dublin, Dubai, and LasVegas as the outstanding growthcenters. Only one of these cities hasrelatively strong GNC—Dublin’s highrank (30th) in Table 2 is a “Europeansurprise” in our analysis. Dubai has amoderate world rank (54th), but ranksas the most connected Middle Easterncity; and Las Vegas’ GNC ranking isonly a lowly 197th. Similarly, in arecent study focusing only on U.S.metropolitan areas, Jonathan Bowlesand Joel Kotkin show how manysmaller metropolitan areas are growingmuch faster than New York.37 Taking


Table 7. Pacific Asian Orientations of U.S. and EU Cities

Pacific Asian Pacific Asian U.S. city orientation EU city orientation

New York 0.472 Amsterdam 0.611Los Angeles 0.306 London 0.528Portland 0.306 Dusseldorf 0.472Chicago 0.194 Frankfurt 0.444

Paris 0.444Brussels 0.361

San Francisco -0.056 Luxembourg City 0.333Seattle -0.111 Athens 0.250Houston -0.306 Hamburg 0.083Philadelphia -0.306 Madrid 0.028Dallas -0.333Denver -0.333San Diego -0.333 Dublin -0.028Washington, D.C. -0.361 Lisbon -0.083Detroit -0.472 Vienna -0.083Pittsburgh -0.472 Milan -0.139Charlotte -0.500 Berlin -0.167Atlanta -0.556 Munich -0.167Boston -0.556 Helsinki -0.333Cleveland -0.583 Cologne -0.361Minneapolis -0.583 Rotterdam -0.444Kansas City -0.611 Barcelona -0.472Miami -0.611 Stuttgart -0.472St. Louis -0.639 Copenhagen -0.611Indianapolis -0.667 Stockholm -0.694

Birmingham -0.722Rome -0.722Lyon -0.750Manchester -0.750Antwerp -0.778


such results at face value, it wouldseem that global connections are notjust unnecessary for growth; they mayeven be an impediment.

However, the global level ofeconomic activities cannot bedismissed so easily. Economic global-ization is a reality, a critical result ofthe massive restructuring of economicactivities over the last two or threedecades. What it provides for cities isnew connections, and thus new oppor-tunities to reach larger supplier,producer, and consumer markets.37

Taking opportunities from across awider range of geographical scalesmeans increased complexity and it isthe latter that is crucial to cityvibrancy and medium-term success. Asbusiness conditions change, it is themore complex cities that are best ableto weather the economic storms. JohnShort has reported, for example, that alarge and successful city like Phoenixappears to be like much lesssuccessful Third World cities in termsof its population/GNC relation. This isnot to say that Phoenix is on a dead-end economic trajectory, but itsrelatively simple, localized economydoes make it economically vulnerable.

So while economic globalization isnot a panacea for cities, it doesprovide new potentials for cityeconomies that adds to theircomplexity and therefore to their ulti-mate sustainability. Thus in themedium-term U.S. cities cannot affordto be globally under-connected andnationally parochial. Because the U.S.economy is the sum of its cityeconomies, our findings point to anational economic vulnerability.

What does this mean for local political and business leaders?The findings presented here haveseveral key implications for world citypolitical and business leaders. Becausethe world city network is such a newmethod and concept, it is hard tocapture all of its meaning in thisexploratory paper. But some obviousissues do arise. We believe that the

greatest local impact that globalconnectivity will have on a city is in itscentral business district.

First, because much of the advancedproducer service industry is located incentral business districts, the growth ofglobal networks could increase thedemand for downtown Class A officespace. A larger workforce can subse-quently impact a host of relateddowntown industries, including retail,entertainment, arts, conventions, andperhaps even housing.

Second, expanding employment infields like law, advertising, and mediacan help attract a highly educated,skilled, and “creative” work force to aworld city. In fact, virtually all the jobsassociated with global producer serv-ices falls under Richard Florida’sdefinition of the “Creative Class.”Phoenix and Las Vegas rank low inFlorida’s scheme despite their rapidgrowth because they are underrepre-sented in terms of the creative class.In contrast, many of the cities that arewell-connected to the global producerservice economy also maintain a largecreative class.

Third, being a world city can helplocal boosters promote a place. Mostplaces promote the idea that they arethe world city of something—some-times almost anything. Given theincreasing buzz about the globaleconomy, not being part of it means acity fails as a real player in business.Cities even bid on such high-profilebut often losing ventures as WorldFairs and Olympic Games—thinkAtlanta in 1996—just to lay claim totheir lofty global status. Atlanta canalso now lay claim to out-networkingsuch Southern rivals as Dallas andHouston, and leaving its Sunbeltbookend Phoenix in the dust.

Finally, and perhaps most specula-tively, being a well-connected worldcity may produce a more cosmopolitanview. Places with business linksthroughout the world may be lessparochial in attitude and feel thanpeer cities that are not well-connectedto global producer service economy.

Does an Atlanta or Houston seemmore outwardly focused than Phoenix?That is hard to say, but it is easy toimagine that such is the case givenhow isolated Phoenix is from the restof the world. Of course, historically,isolation has not been beneficial forcities: All successful cities have beencosmopolitan.

Methods Appendix

Computing global network connectivityThe world city network is constitutedby m global service firms with officesdistributed across n world cities. Thelevel of service performed by firm j incity i is vij which we call the servicevalue. The array of service valuesdefines an m x n service value matrix,V. The world city network is derivedfrom this matrix using the plausibleconjecture that the larger a firm’sservice value in a city, the greater thenumber of the firm’s flows of informa-tion, knowledge, instruction, ideas,strategies, plans, etc. will emanatefrom that city to connect with officesin other cities. This assumption allowsfor relations between cities to bedefined.

The initial relation between eachpair of cities is given by:

rab,j = vaj . vbj (1)

This defines relations between citiesa and b in terms of firm j. This speci-fies an elemental interlock linkbetween two cities. It is multiplicativebecause the potential quantity of flowsbetween two cities rises geometricallywith the quantity of service providedin each city. From equation (1) theaggregate city interlock link betweentwo cities can be derived as:

rab = ∑rab,j (2)j

For each city there is n-1 such links,one to every other city. These can be


used to define the overall interlockconnectivity of a city so that:

Ca = ∑rai where a ≠ i (3)i

This measure of connectivity picksup two features of a city’s servicevalues. First, and most obviously, citieswhere firms locate offices with higherservice values are more connected.Second, and more subtly, if those highservice offices are for firms with largeroffice networks then the city appearsmore connected. In other words, a city with several high service offices of firms that has only small office net-works will not be that well connectedas measured by interlock connectivity.

Interlock connectivity indicates theimportance of a city in the world citynetwork. Generally we refer to empir-ical measures of interlock connectivityas global network connectivity.

Deriving city hinterworldsTo compute a city’s hinterworld weneed to specify its external relationswith other cities. For this exercise, wedeal with just the top 123 cities interms of global network connectivity.This is because the data matrixbecomes sparser (more zero scores)with additional cities creating a partic-ular problem when measuringconnections for individual cities. Obvi-ously with less firms represented incomputing a measure, a resultbecomes that much less reliable. Thusthe results reported below are derivedfrom a reduced 100 x 123 data matrix.

A city’s hinterworld consists of thelevels of service it provides for doingbusiness in each of the other 122cities in the reduced data matrix. Thisis initially computed as follows. First,count the number of firms present ineach city. For each city multiply thisnumber by 5, the maximum servicevalue. This constitutes the highestpossible level of service that a citycould expect in another city (i.e., theother city houses the headquarters ofevery single one of a city’s global

service firms). Thus, in the simpledata set shown in Table A, highestlevels of possible service are 20 forNew York, 15 for Hamburg, and 10 forBogotá. Now for each city, take othercities in turn and sum their servicescores but only for firms present in theoriginal city. For instance, startingwith New York, the sums for Hamburgand Bogotá are 10 and 4 respectively;starting with Hamburg the sums forNew York and Bogotá are 12 and 2;and starting with Bogotá the sums forNew York and Hamburg are 10 and 2.The latter sums are expressed asproportions of the highest level ofpossible service. For instance, theproportions for New York are Hamburg0.5 (= 10/20) and Bogotá 0.2 (= 4/20).All such computations are shown inTable A.

The interpretation is relativelysimple. The columns in this tabledefine the level of service that can beexpected in a city when visiting aglobal service firm in a row city. Thus,going into an office in New York to dobusiness in Hamburg the service levelis 0.5, but to do business in Bogotáthe level falls to 0.2. Notice that fromBogotá, doing business in New Yorkhas a 1.0 service level showing thatBogotá’s two service firms in Table 1have their headquarters in New York.In contrast, the lowest level of servicein this data is a paltry 0.13 for doingbusiness from Hamburg in Bogotá.These columns represent the servicinglinkages that form the basis for

describing the hinterworld of a city. However, there is a problem when

comparing the hinterworlds of citiesusing results such as those in Table A.Notice that in this table New Yorkappears with very high service levelsfor the other two cities and Bogotáprovides low levels. This is obviouslyreflecting the network position ofthese cities in this small data set. Thuswhen this method is used for the 100x 123 data set, we find that every cityhas its highest external provision ineither London or New York. Ingeneral, we can note that externalservice provisions tend to closelyfollow the level of a city’s globalnetwork connectivity. Thus mappingdirect measures of connections to anycity tends, to a large degree, merely toreplicate the worldwide map of GNCs.Thus all hinterworlds, although notexactly the same, look very much alike.To overcome this comparative defi-ciency an extra step needs to be takenfor describing hinterworlds.

We will term the external serviceprovisioning results from Table Adepictions of ‘absolute’ hinterworlds.Taking out the underlying generalinfluence of global network connec-tivity from the absolute provisioningvalues for a city is a fairly simple task.Scatter diagrams of global networkconnectivity against a city’s externalprovisioning levels shows a strongpositive linear relationship in everycase. Thus absolute provision in city ifor all other cities j can be regressed


Table A. Simple City/Firm Data Set

(a) Service valuesFirm A Firm B Firm C Firm D

New York 5 5 5 2Hamburg 2 3 0 5Bogota 2 0 2 0

(b) Levels of servicing derived from (a)New York Hamburg Bogota

New York - 0.8 1.0Hamburg 0.5 - 0.2Bogota 0.2 0.13 -

against their GNCs using the simpleequation:

Hj = _ + _ Cj + _ where i ≠ j (4)

Hj is the absolute provision valuesand C is global network connectivity.The regression constants are _ and _,and _ is the error term or residual.Calibrating this equation for any cityproduces an estimate of the level ofservice provision given a city’s connec-tivity. It is the difference between thisestimate and the actual level of provi-sion that is the residual. Theseresiduals define a ‘relative’ hinter-world—in which other cities that are‘over-linked’ to city i are identified (apositive residual) and also where city iis ‘under-linked’ (a negative residual)with respect to position in the worldcity network (i.e., its GNC).

Such mappings of hinterworlds havebeen analyzed and discussed for NorthAfrican/West Asian cities,40 for Dutchcities,41 for British cities,42 and forBelgium cities.43 The hinterworlds ofall 123 cities, including the 23 UScities, are depicted in the GaWCvirtual Atlas of Hinterworlds.44

Computing regional orientationswithin hinterworlds Geographical orientations within cityhinterworlds, such as a propensity tobe over-linked to neighboring cities orto a particular world region, can bemeasured by converting the hinter-world residual values into ordinalscores in the following way. First, forCity i, the top ten over-linked citiesare scored 3, the next top 20 arescored 2, and remaining over-linkedcities score 1. Under-linked cities aresimilarly scored as negative values: -3,2, and -1. Second, identify Region Xwhich includes, say, 10 cities.Summing the values for just these 10cities provides an aggregate measureof City i’s geographical orientation toRegion X. If the sum is 30 this meansthat Region X’s cities actually consti-tute the top ten over-linked cities. Thisis a maximum score; in practice we

might expect less extreme orientation.Third, we divide the sum by themaximum possible sum to measure therelative orientation to a region. In thecase above the orientation of City i toRegion X is 1.0 which one the limit ofthe scale, the other is -1.0 whenmaximum under-linkage is found.Producing a scale between plus andminus one in this way enables us tocreate comparable measures of thegeographical orientations of citiesacross different world regions.

Endnotes1. Peter J. Taylor is associate director of the

Metropolitan Institute at Virginia Tech and

co-director of the Globalization and World

Cities (GaWC) Study Group and Network

at Loughborough University, UK. He is

author of World City Network: a Global

Urban Analysis 2003 (London: Routledge).

Robert E. Lang is director of the Metropol-

itan Institute at Virginia Tech and an

associate professor of urban affairs and

planning. He is author of Edgeless Cities:

Exploring the Elusive Metropolis (Wash-

ington: Brookings Institution, 2003).

2. See the argument in Susan E. Clarke and

Gary L. Gaile, The Work of Cities

(Minneapolis: University of Minnesota

Press, 1998)

3. The two classic derivations of world city

hierarchies are: J. Friedmann, “The World

City Hypothesis,” Development and

Change 17 (1986): 69–83 and R.P.

Camagni, “From City Hierarchy to City

Network.” In T. R. Lakshmanan and P.

Nijkamp, eds., Structure and Change in the

Space Economy (Berlin: Springer-Verlag,

1993).

4. The leading papers in this research school

were brought together in L. S. Bourne and

J. W. Simmons, eds., Systems of Cities

(New York: Oxford University Press, 1978).

5. The city competition referred to here is

that between major city regions, not cities

as in incorporated administrative entities.

The literature is reviewed and developed in

a special issue of Urban Studies, see W. F.

Lever and I. Turok, “Competitive Cities:

Introduction to the Review,” Urban

Studies, 36 (1999): 1029–44.

6. Jane Jacobs, Cities and the Wealth of

Nations (New York: Vintage, 1984).

7. A good discussion of this distinction can be

found in W. W. Powell, “Neither Market

Nor Hierarchy: Network Forms of Organi-

zation,” Research in Organizational

Behavior 12 (1990): 295–336. See also

G. Thompson, Between Hierarchies and

Markets: The Logic and Limits of Network

Forms of Organization (Oxford University

Press, 2003).

8. This position is presented in detail in

Taylor, World City Network.

9. John Friedman, “The World City Hypoth-

esis.” This approach to defining the “world

city system” has been recently opera-

tionalised in Arthur S. Alderson and Jason

Beckfield, “Power and Position in the

World City System,” American Journal of

Sociology 109 (2004): 811–51. For a

comparison with the approach adopted

here, see Peter J. Taylor, “Parallel Paths to

Understanding Global Inter-city Rela-

tions,” GaWC Research Bulletin 143

(www.lboro.ac.uk/gawc). This alternative

approach has previously been applied to

U.S. inter-city relations by Allan Pred,

City-Systems in Advanced Economies

(London: Hutchinson, 1977) and Christo-

pher Ross, The Urban System and Networks

of Corporate Control (Greenwich: JAI

Press, 1992).

10. Saskia Sassen, The Global City (Princeton

University Press, first edition 1991, second

edition 2001).

11. For Sassen, the production and sale of

advanced producer services are crucial

elements of her global cities.

12. Lang, Edgeless Cities.

13. Friedmann, “The World City Hierarchy,”

Development and Change 17 (1986):

69–83.


14. D. R. Meyer, “The World System of Cities:

Relations Between International Financial

Metropolises and South American Cities,”

Social Forces 64 (1986): 553–581.

15. A. D. King, Global Cities (London: Rout-

ledge, 1990).

16. D. R. Meyer, “Change in the World System

of Metropolises: The Role of Business

Intermediaries,” Urban Geography 12

(1991): 393–416.

17. S. Sassen, Cities in a World Economy

(Thousand Oaks, CA: Pine Forge, 1994).

18. W. Wu, “Economic Competition and

Resource Mobilization.” In M. A. Cohen,

B. A. Ruble, T. S. Tulchin and A. M.

Garland, eds., Preparing for the Urban

Future (Washington: Woodrow Wilson

Center, 1996).

19. F-c. Lo and Y-m. Yeung, Globalization and

the World of Large Cities (Tokyo: University

of United Nations Press, 1998).

20. J. R. Short and Y. Kim, Globalization and

the City (London: Longman, 1999).

21. For full details of the specification, see

Taylor, “Specification.” The methods

appendix provides a summary of that argu-

ment highlighting aspects necessary for

this paper.

22. See www.lboro.ac.uk/gawc and

www.mi.vt.edu

23. This data collection effort was funded by

the Economic and Social Research

Council of the United Kingdom and is

described in detail in Peter J. Taylor, Gilda

Catalano and David Walker, “Measurement

of the World City Network,” Urban Studies

39 (2002): 2367–2376. The regional

breakdown is: Europe 101, Northern

America 50, Pacific Asia 37, Latin America

37, Sub-Saharan Africa 32, Middle

East/North Africa 25, South Asia 13,

Australasia 12, CIS 8.

24. Op cit.

25. The technique for measuring and mapping

hinterworlds is described in detail in Peter

Taylor and David Walker, “Hinterworlds

Revisited,” Geography 89 (2004): 145–51.

The methods appendix provides a summary

of this technique.

26. The United States contains 40 of the 315

cities in the global services/cities data, 23

of which are in the 123 cities used for

defining hinterworlds.

27. See “Visualization” section at

www.lboro.ac.uk/gawc

28. J. Abu-Lughod, New York, Los Angeles,

Chicago, America’s Global Cities

(Minneapolis: University of Minnesota

Press, 1999).

29. Ed Brown, Gilda Catalano and Peter

Taylor, “Beyond World Cities,” Area 34

(2002): 139–148.

30. Carl Abbot, Political Terrain: Washington

DC from Tidewater Town to Global Metrop-

olis, (Minneapolis: University of Minnesota

Press, 1999). More generally, he promotes

a historical approach to understanding

U.S. world cities: Carl Abbott, “The Inter-

national City Hypothesis: An Approach to

the Recent History of U.S. Cities,” Journal

of Urban History 24: 28–52.

31. J. R. Short, “Black Holes and Loose

Connections in the Global Urban

Network’” GaWC Research Bulletin

No. 76.

32. The European Union (EU) is chosen for

comparison because it is an economic area

roughly equivalent to the U.S. in economic

size and wealth. Pacific Asia has no such

institutional unity but it does constitute a

key globalization arena and therefore

provides a second valuable comparison.

33. Remember that it is only global firms

included in the study; we do not consider

national or local firms. Thus the ‘national

localism’ recorded shows how global firms

are using US cities.

34. Manual Castells, The Rise of Network

Society (Oxford: Blackwell, 1996).

35. Jacobs op cit. Described in detail in Jane

Jacobs, The Economy of Cities (New York:

Random House, 1969). See also Jane

Jacobs, The Nature of Economies (New

York: Vintage, 2000), discussion of Los

Angeles page 51.

36. See Jones Lang Lasalle at

www.joneslanglasalle.co.uk/

37. Jonathan Bowles and Joel Kotkin, Engine

Failure (New York: Center for an Urban

Future, 2003).

38. Short, “Black Holes.”

39. Richard Florida, The Rise of the Creative

Class: and How it’s Transforming Work,

Leisure, Community and Everyday Life

(New York: Basic Books, 2002).

40. Peter Taylor, “West Asian/North African

Cities in the World City Network,” Arab

World Geographer 4 (2001): 146–159.

41. Peter Taylor, Amsterdam in the World City

Network (Amsterdam: AME, 2002).

42. Taylor, “Hinterworlds Revisited.”

43. Ben Derudder and Peter Taylor, “Brussels

in the World City Network,” Belgeo 4

(2004): 259–76.

44. See “Visualization” section at

www.lboro.ac.uk/gawc


AcknowledgmentsThe authors would like to thank Jennifer Vey and Jared Lang for their contri-butions to this report. The Brookings Metropolitan Policy Program wouldlike to thank the Surdna and Fannie Mae foundations for their support ofour work on competitive cities.

For More InformationRobert E. LangMetropolitan InstituteViriginia Techphone: 703-706-8101e-mail: [email protected]

Peter J. TaylorDepartment of GeographyLoughborough Universityphone: +44 (0)1509 222790e-mail: [email protected]

For General InformationThe Brookings Institution Metropolitan Policy Program(202) 797-6139www.brookings.edu/metro

The Brookings Institution

1775 Massachusetts Avenue, NW • Washington D.C. 20036-2188Tel: 202-797-6000 • Fax: 202-797-6004

www.brookings.edu

Direct: 202-797-6139 • Fax/direct: 202-797-2965

www.brookings.edu/metro

U.S. Cities in the ‘World City Network’Feb 22, 2005 · Angeles are the U.S. leaders in global...

Documents

Transcript of U.S. Cities in the ‘World City Network’Feb 22, 2005 · Angeles are the U.S. leaders in global...