sign in sign up
<<
Home , Global city

Metropolitan Policy Program

U.S. in the ‘World Network’

Peter J. Taylor and Robert E. Lang1

“The rise of Findings An examination of economic connections among cities’ global advanced service firms finds that: transnational ■ While , , and Los non-U.S. cities than U.S. cities in its top Angeles are the U.S. leaders in global ten list of strongest global connections. interactions connectivity, , , Only three non-U.S. cities make Atlanta, and Washington are also Miami’s top 10 list, for example, while important nodes in the world city Pittsburgh’s list contains none. has produced a network. A mixture of regional capi- tals—such as and —and ■ Even the most globally-connected specialist cities—such as , the U.S. cities are more locally oriented new economic world’s energy capital—are found in the than cities in the EU. New York and next tier. Somewhat surprisingly, Miami are the least locally oriented of dynamic and growing Phoenix and San U.S. cities, but even they are far more Jose have relatively low connectivities. U.S.-centric in their connections than most European cities are EU-centric. A ■ U.S. cities overall—and particularly much larger number of EU cities boast in which cities non-coastal cities—are generally less strong connections to Latin America globally connected than their Euro- and Pacific Asia than cities in the U.S. pean Union and Pacific Asian and their counterparts. Sixteen EU cities and 11 ■ Beyond localism, functional linkages Pacific Asian cities rank among the top among firms are more important 50 most globally-connected cities, while determinants of connectedness than regions are the only seven U.S. cities make the cut. geography alone. While has no Chicago is the only high ranking U.S. American Pacific Coast city linkages, city not located in a coastal state. for example, it has many international prime nodes.” financial centers with strong connec- ■ While important service connections tions to the rising banking and exist among certain U.S. cities and hubs of Pacific Asia. Many of these particular global regions, U.S. cities cities, as well as New York, are much are more strongly linked to other U.S. more oriented to the Pacific Rim cities than to cities around the globe. than San Francisco, Seattle, or even New York is the only U.S. city with more .

As cities aim to position themselves better economically, they must remember that they operate 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.

February 2005 • The • Survey Series 1 Introduction but in reality it shares its economic immigrants can have strong global dominance with other global cities, connections due to the ebb and flow ust about anyone with even a notably and possibly of both people and money back to casual interest in the geography and . And other “global” U.S. countries of origin. of the global economy could cities are still far more locally oriented This paper focuses on one type of Jguess that New York, Los Angeles, than their European counterparts. Put economic connection—those between and Chicago are major world cities, simply, the clear supremacy of the advanced producer services firms— with strong ties to other cities and U.S. in international relations and important because of the high value regions worldwide. But what about military power does not translate into they provide to places. This approach other U.S. cities? How do places like a similar position for American world is an extension of the usual under- Houston and Detroit—or even Port- cities within the world city network. standing of world cities and therefore land and Cincinnati—connect to the The findings from this report high- requires some exposition. rest of the world? What is their place light these shifting dynamics, and in the “world city network?” And why suggest multiple implications for Cities, hierarchies, and networks does it matter? elected officials, policymakers, urban World cities studies suffer from a Until recently there were no real planners, and others who are theoretical legacy that assumes all answers to these questions. Most concerned with regional development inter-city relations are necessarily hier- studies on the connections between and competitiveness. How a city plugs archical.3 A product of the national global entities use nations—not into the of advanced urban systems school of research that cities—as their unit of analysis, and/or producer services could impact local flourished in the 1960s and 1970s, tend to focus on political, as opposed growth patterns, for example, and this theory of inter-city hierarchical to economic, relationships. Those that offer opportunity for development of relations views cities as rivals, strug- have focused on world cities have its . It could influence a gling against each other to “reach the failed to provide adequate measures of city’s economic prospects by deter- top.”4 With the advent of globalization, inter-city relations. By contrast, this mining how many high-paying service the competition is now thought to be paper introduces a new approach to jobs are locally created. And it could even more cut-throat, and a large network analysis that allows us to help shape what types of households a “cities competition” urban policy liter- better understand the relative connec- city attracts and retains. In short, ature has emerged as a result.5 These tivity of cities in the world economy. global connectivity has become studies of worldwide inter-city rela- Using this method, we track the global another urban barometer—a new tions offer little or no interrogation of distribution of 100 leading advanced statistic that helps reveal the health how cities actually relate to one services firms to determine one impor- and strength of cities and regions. another in a global space-economy, tant aspect of cities’ economic however. And so we ask: Must cities connectedness to other world cities always form a hierarchy? and the patterns of these linkages Background In a rare exception to the near- across the globe. ubiquitous hierarchical premise, Jane The importance of these relation- his study treats world cities as Jacobs, in her classic Cities and the ships is becoming increasingly the global service centers of Wealth of Nations, developed an apparent. The conditions of contem- the world economy. As such, economic growth theory based upon porary globalization have, in fact, the network of flows between “dynamic cities” whose relations with spurred a renaissance of major cities citiesT described here provides a other cities are primarily mutual across the world, and a new economic skeletal spatial organization of rather than competitive.6 Cities configuration is emerging based upon contemporary globalization. This generate local economic growth by cities.2 This is not to say that states are approach is an extension of the usual replacing imports from other cities in no longer significant in their interna- understanding of world cities and a process that, rather than being a tional relations, but the rise of therefore requires some exposition. zero-sum game, stimulates new growth transnational interactions has However, there are many ways cities in the city losing the exports. A hypo- produced a new economic globaliza- are connected globally that are beyond thetical example might be cities losing tion in which cities and their regions the scope of this paper. Bentonville, a particular manufacturing production are the prime nodes of a nascent AR, headquarters to Wal-Mart and its (e.g. cotton textiles) compensating by network society. In this new context vast supply chain, certainly is very providing the machine tools (e.g. U.S. domination can no longer be globally connected, as may be a large machinery for producing cotton taken for granted. Yes, New York can agribusiness in Iowa. Additionally, textiles) for the new production in the claim to be “ of the world,” cities with large concentrations of new cities. The result is a complex and

February 2005 • The Brookings Institution • Survey Series 2 interdependent network of cities corporations—have become a cutting- understand this new scale of inter-city rather than a simple hierarchical edge industry in the new globalizing relations, confirmed by the fact that structure, with cities cooperating in economy. A key feature of firms none of the above examples are in any economic development, not just providing these services is their sense formally specified. Thus we competing for a share of the spoils. concentration in major cities, espe- cannot be sure that these labels are This project adopts the Jacobs cially their .12 Advanced trying to describe the same phenom- model of synergistic inter-city rela- producer service firms utilize the enon or whether they are suggesting tions. In our interpretation, we adhere knowledge-rich environs of these cities different types of relations. In to the theoretical distinction between while themselves contributing to contrast, we treat the world city hierarchical/competitive processes and downtown development through their network as a formally specified, empir- network/cooperative processes but business practices. ically-based model that measures the understand that in practice the “real For Sassen, major service cities thus interlocking network of global world” is somewhat messier, and become “global cities,” the “strategic advanced service firms.21 “hybrid processes” are the norm.7 But places” of economic globalization. As while we recognize that a world city the locales for production, develop- Data production network can include hierarchical ment, and marketing of advanced The increasing relevance of a city- tendencies, we believe the latter are to services, global cities are contempo- based perspective on worldwide be searched for empirically within a rary manifestations of Jacobs’ dynamic developments has not yet led to an network model.8 cities. This means that their success is increased availability of data to Our theoretical position leads to not a matter of separate city develop- monitor this process. In contrast to two separate but related issues: What ment, rather they function as nodes in comparisons between countries, is the agency that creates the city a worldwide network of cities. worldwide statistics for evaluating network, and how is the network Cities are not themselves the prime cities are not generally available formed by this agency? agents of world city network forma- beyond simple attribute measures tion, however; it is advanced producer such as population sizes. Globalization, cities, and agency service firms that have been largely A study of inter-city relations thus Globalization is constituted by a responsible for creating and main- requires the creation of new data, myriad of transnational processes— taining the network. These firms have customized for describing a new geog- economic, political, and cultural—that offices in important cities across all raphy. Devising ways to meet this need drive much contemporary social world regions, and personnel, informa- has been the principal task of the change. Economic globalization tion, knowledge, intelligence, ideas, Globalization and World Cities converts an “international economy” plans, instructions, and advice flow (GaWC) virtual research program into a “global economy”—with freely among them. As such, these centered at Loughborough University transnational financial markets global service firms “interlock” the and the Metropolitan Institute at commonly cited as evidence for this cities in which they have a presence. Virginia Tech (MI).22 We introduce the change. Cities are central to these Viewed this way, the world city GaWC/MI data approach here as a processes, initially as headquarters network can be measured. necessary base for beginning to under- locales for multinational corporations;9 stand the ways in which U.S. world and later as advanced producer service cities relate to other world cities, centers.10 In this project we follow Methodology within and outside of the U.S. ’s lead by considering In 2000, we collected data on a key financial and business services as here has been a continuous large number of global service firms the hallmarks of contemporary world invention of phrases to across a large number of cities cities.11 describe inter-city relations in covering all regions of the world.23 We One of Sassen’s major contributions the world cities literature, focused on firms from six key producer to economic globalization theory has including:T “world city hierarchy”;13 service sectors—accounting, adver- been her linking of the literature on “world system of cities”;14 “global tising, banking/finance, insurance, the growth of the service sector to network of cities”;15 “world system law, and management consulting— urban change. Service growth relative of ”;16 “transnational choosing only those with offices in to manufacturing has led to an impor- urban system”;17 “new global urban 15 or more cities, including at least tant economic shift in the second half hierarchy”;18 “functional world city one city in each of , of the twentieth century, and some system”;19 and “global urban net- Western Europe, and Pacific Asia. In services—those providing advanced work.”20 The variety of phrases total, we identified 100 such firms— knowledge products to multinational suggests a literature struggling to 18 in accounting, 15 in advertising,

February 2005 • The Brookings Institution • Survey Series 3 23 in banking/finance, 11 in insurance, processes that create an office scored for headquarters) to New York’s 16 in law, and 17 in management network, so that country of origin, overall connectivity. Second, cities consulting—across 315 cities world- initial expansion strategies, take-over with a large number of global service wide. opportunities taken, etc. each produce firm offices will, obviously, tend to be We defined cities broadly as “city- particular results. By combining 100 more connected than cities with less regions” or metropolitan areas, but in such globalization paths, we created a such offices. Thus it is no surprise practice the service offices considered generalized world city network that is that London and New York turn out to were largely concentrated in the not dominated by just a few specific have by far the largest global network central city, especially in downtowns processes. Thus measurements based connectivity measures in our results. (hence the use of the term “city”—as on these data are reasonable reflec- Third and countering the latter opposed to “” or tions of the broad new geography that tendency, cities housing offices of “region”— throughout this report). is economic globalization. firms with the very large office Using information mainly gathered networks (i.e., found in very many from firms’ websites, we were able to Measuring connectivity cities) will be more globally connected determine each firm’s distribution of Using our matrix, we then set out to than cities housing offices of firms offices and code each city in terms of measure the connectivity of the 315 with mainly small networks. Thus its importance, or “service value,” in cities within the world city network. global network connectivity does not the firm’s office network. To compute We based this measure upon the simply reflect the number of global this value, we employed two types of assumption that the more important service firms in a city—one city can information about individual city an office is within a firm’s network, have many more global service offices offices: the size of the office as a the more flows—of information, than another but if the latter’s offices service provided (e.g. the number of knowledge, ideas, plans, etc.—it will are concentrated among firms with the partners based in a law firm’s office), generate. In other words, we expect greatest networks, it may well have the and the extra-locational functions of a greater connection between two cities higher global network connectivity. city office (e.g., a regional headquar- with important offices of a firm than ters).24 between another pair of cities that Defining “hinterworlds” Using this data, we developed an both house more modest offices. Thus In traditional it was interlocking model consisting of a we treat products of service values for common to depict towns and cities as simple matrix arraying world cities pairs of cities as indicating connection service centers, each with its own against global service firms. Each potentials. Summing all such products “service area” (or hinterland) that it matrix cell indicates the importance of for a given city for all firms across all provisioned. In these global urban a given city within the office network other cities defines the city’s global analyses, we also treat cities as service of a given firm. The end result of this network connectivity (GNC). centers, but the area they service exercise was a service values matrix Computing these connectivities through their global service firms is measuring 100 firms times 315 cities creates very large sums, however, so to worldwide. Thus a new term is depicted as an array of service values make comparison easier, we converted required for the new geographical (codes) ranging from 0 to 5 (i.e., from them into relative measures of GNC, scope of advanced producer servicing: no presence to headquarter location). reporting them as proportions of the city hinterworlds. Unlike hinterlands, The interpretation of this matrix is highest GNC score that is computed hinterworlds have no boundaries sepa- quite straightforward. Each column from this data, that for London. rating the areas provisioned between describes a firm’s global office strategy Our measure of global network neighboring cities. However, they are across 315 cities, where it is strong connectivity has three important prop- like hinterlands in that they have a and where it is weak. Each row erties that should be kept in mind geographic basis or a spatial variation describes the particular global service when interpreting results. First, in the intensity of provisioning. Thus mix of a city, in which firms and connectivity is measured irrespective to map a city’s hinterworld will show sectors it is strong and in which firms of international boundaries. For where it is strongly connected and and sectors it is weak. instance, consider a New York law firm where it is weakly connected across In total, the matrix contains 31,500 with offices in, say, Atlanta and the world. pieces of information. This large quan- Caracas that are both given service The computation involved for tity is significant. The measures we values of 2. In computing New York’s mapping the hinterworlds of cities is derive from the data are aggregations global network connectivity the quite complex and is outlined in detail of different office networks of many Atlanta-New York link and Caracas- in the appendix.25 We applied the firms. Each path to becoming a global New York link are identical: they each method one city at a time to define its service firm is distinctive in the mix of contribute 10 (2 x 5, remember 5 is particular hinterworld. This consists of

February 2005 • The Brookings Institution • Survey Series 4 an array of the other cities, for each of which cities are strongly connected to A mixture of regional capitals and which there are values indicating their Pacific Asian cities, for instance. We specialist cities attracting services for degree of connection with the initial derived the regional score for a city by their distinctive businesses is found in city. The values straddle zero: positive simply summing residuals for all cities other tiers, with the size and economic values show a city is strongly in a given region (details are given in importance of region or specialty connected to the initial city, negative the appendix). determining the strata location of a values indicate a weak connection. city. Thus in strata IV, regional capitals The methodology is as follows: first, in New and the Pacific we computed the link between a given Findings Northwest (Boston and Seattle) are city and each other city for all firms joined by the two Texan world cities, they share (i.e., both have offices of A. While New York, Chicago, and one as regional center (Dallas) and the the firm); second, we summed them Los Angeles are the U.S. leaders in other as the world’s energy capital across firms to provide a provisional global connectivity, San Francisco, (Houston). measure of links to the other cities; Miami, Atlanta, and Washington are Similarly, in strata V, leading cities third, we made each link relative to also important nodes in the world of smaller and less important regions the maximum service value that could city network. are combined with Detroit as the have been provided; fourth, the list of Table 1 shows the gross and relative global auto center. Other strata such links for a given city to all other GNC scores for U.S. cities, presented combine sub-regional centers such as cities was compared to the latter’s in order by size of the GNC, and Charlotte and Portland with reviving global network connectivities using a grouped into 10 strata based on their industrial cities like and simple regression model; and finally, connectivity levels. Alone in Strata I, Pittsburgh. There are surprises in both we used residuals from this regression New York is clearly the American types of metropolitan area, with rela- to determine the other cities to which global city par excellence. In fact, with tively low connectivities for the likes of the given city is strongly connected connectivities more than a quarter “old and reviving” Baltimore and “new (positive residual) or weakly connected above all other cities, New York and dynamic” Phoenix.31 (negative residual). In other words, together with London form a sort of The final city of the series, Wilm- hinterworlds are defined relative to global city duplex. ington, forms the bottom strata on its global network connectivity because a New York is not the only important own. This is no splendid isolation like positive residual indicates that the U.S. world city, however: Chicago and New York in strata I: Wilmington was given city has a higher than expected Los Angeles constitute a clear, second included in the data as a relatively connection to another city than the strata in Table 1.28 Globally, these two low-level U.S. city to see how its GNC latter’s global network connectivity cities are ranked in the top ten compared and because, despite its would suggest, with the opposite connected cities with the likes of Paris small size, the city is the center of the implied by negative residuals. and . U.S. credit card industry. Clearly sepa- As explained in the appendix, we To be sure, the dominance of these rate from the other selected cities, carried out this exercise using just 123 three U.S. cities comes as little Wilmington as strata X represents the out of the 315 cities in the original surprise. Our analysis does reveal many other “ordinary” U.S. cities that data.26 This means that hinterworlds some unanticipated, results, however. do not appear in our data, mainly capi- were computed for each of 123 cities; Strata III includes Miami, which tals of smaller states and industrial every hinterworld consists of measures gives the city an elevated position cities that never formed major metro- of strong and weak connections to the compared to how it is perceived within politan areas. other 122 cities. These have been a U.S. domestic urban hierarchy. However, Wilmington’s GNC score mapped and are depicted in the Atlas However, as “capital of Latin does remind us that globalization is of Hinterworlds.27 For this report we America,” it fully warrants its relatively not just the preserve of leading cities, use these measures in two ways. First, high status in the world city network.29 but that all cities participate in global- we constructed tables showing the 10 The other cities in the strata are less ization. In terms of financial and most strongly connected (positive surprising but no less interesting: San business services the provision in residuals) cities and 10 most weakly Francisco, Atlanta, and Washington these less important cities is not large connected (negative residuals) cities to round out this third tier, due by and but neither is it zero. Still, while there show the nature of selected city large to their roles as western are no “unlinked” U.S. cities in global- hinterworlds. Second, we aggregated gateway/financial center, media center ization, only a relatively few can city hinterworlds by major regions to and unchallenged “capital” of the large realistically aspire to global service-led show the general geographical biases and fast growing South, and nation’s economic development. within hinterworlds—this shows capital, respectively.30

February 2005 • The Brookings Institution • Survey Series 5 at least one fifth of London’s GNC. Table 1. Global Network Connectivities of U.S. Cities There are 28 cities in European Union countries that similarly qualify but City GNC score Relative GNC STRATA only 13 Pacific Asian cities. The key New York 61895 0.976 I point this table makes, however, is not Chicago 39025 0.616 II that there are slightly more EU cities Los Angeles 38009 0.600 II and less Pacific Asian cities, but rather San Francisco 32178 0.508 III that the latter two sets of cities are Miami 29341 0.463 III generally ranked higher globally rela- Atlanta 27052 0.427 III tive to equivalent U.S. cities. Washington, D.C. 26522 0.418 III The capitals of the EU and the U.S. Boston 22249 0.351 IV encapsulate this situation: both rank Dallas 21796 0.344 IV 7th in their respective local economic Houston 21424 0.338 IV areas but ranks 15th globally Seattle 19252 0.304 IV compared to Washington’s 37th world 17368 0.274 V ranking. This pattern is consistent Philadelphia 17006 0.268 V throughout the table: and Minneapolis 16914 0.267 V Houston ranked 10th locally, but 32nd St. Louis 16124 0.254 V and 62nd respectively globally; Luxem- Detroit 15818 0.250 V bourg City and Cleveland are both San Diego 14585 0.230 VI 20th locally, but 63rd and 112th glob- Portland 14113 0.223 VI ally. The situation with Pacific Asian Charlotte 13556 0.214 VI cities is similar at the top end of the Cleveland 13442 0.212 VI global ranking—whereas only 7 U.S. Indianapolis 13347 0.211 VI cities are in the world top 50, 11 Kansas City 12772 0.201 VI Pacific Asian cities make the cut. This Pittsburgh 12707 0.200 VI region has just 2 other world cities, Baltimore 11309 0.178 VII however, reflecting a particular Phoenix 11025 0.174 VII concentration of services in capital Cincinnati 10603 0.167 VII cities (excepting several cities in Tampa 10532 0.166 VII China). Columbus 9974 0.157 VII However this tendency of major San Jose 9843 0.155 VII European and Pacific Asian cities to Rochester 9731 0.153 VII rank higher than those in the U.S. is Palo Alto 9078 0.143 VIII not completely regular: there does Hartford 9007 0.142 VIII appear to be a “middle gap” in the Richmond 8845 0.140 VIII U.S. listing (between strata III and IV Buffalo 8798 0.139 VIII in Table 1) that is balanced by a Honolulu 8656 0.137 VIII cluster of lowly ranked cities (strata VI Las Vegas 7911 0.125 IX in Table 1); neither feature is found in New Orleans 7089 0.112 IX the EU listing. This appears to be Sacramento 6870 0.108 IX reflected, at least partially, in a Omaha 6564 0.104 IX geographical dimension. Whereas Wilmington 3740 0.059 X Europe’s heartland is city-rich in + other cities - - X connectivities (notably in Germany), the American heartland has fewer Source: Authors’ calculations cities and with lower connectivities, Chicago being the exception. After Chicago, the next U.S. city from a B. U.S. cities overall—and particu- In Table 2 we compare the connec- non-coastal state is Denver, ranked larly non-coastal cities—are tivity of U.S. cities to those in the EU 73rd globally. generally less globally connected and Pacific Asia.32 The 23 U.S. cities than their European Union and listed are part of the roster of 123 Pacific Asian counterparts. cities from across the world that have

February 2005 • The Brookings Institution • Survey Series 6 Boston, Chicago, and Washington. Table 2. Global Network Connectivity Rankings: London also appears among the top U.S., EU, and Pacific Asian Cities ten well-connected cities to New York and the remaining majority are found U.S. cities EU cities Pacific Asian cities in Pacific Asia. In short, whereas Pitts- New York – 2 London – 1 – 3 burgh’s hinterworld is very “local,” Chicago – 7 Paris – 4 Tokyo – 5 New York’s is global, with specific Los Angeles – 9 – 8 Singapore – 6 strengths in the three leading global- San Francisco – 17 – 11 – 20 ization regions: the U.S. itself, Pacific Miami – 25 – 12 – 22 Asia, and Europe. Atlanta – 33 – 14 – 26 Los Angeles and Miami. The hinter- Washington, D.C. – 37 Brussels – 15 – 28 worlds of these two cities show the Boston – 60 – 27 – 31 expected regional biases beyond the Dallas – 61 – 30 – 36 U.S.: Los Angeles is well-connected to Houston – 62 Barcelona – 32 – 41 Pacific Rim cities in both Asia and Seattle – 68 – 39 – 46 , and Miami is well- Denver – 73 Lisbon – 42 Ho Chi Minh – 88 connected to Latin American cities Philadelphia – 76 Copenhagen – 44 - 109 (Table 3). However, it is also the case Minneapolis –77 Hamburg – 48 that both these cities are, in general, St. Louis – 81 – 49 most strongly linked to other U.S. Detroit – 85 Düsseldorf – 50 cities than those elsewhere around the San Diego – 98 – 51 world. For Los Angeles, nine of the top Portland – 105 – 53 10 well-connected cities are U.S. Charlotte – 108 – 56 cities, the exception being Tokyo. For Cleveland – 112 Luxembourg – 63 Miami, a majority of top 10 well- Indianapolis – 114 Helsinki – 70 connected cities are in the U.S., and Kansas City – 119 – 74 only one is found in Latin America Pittsburgh – 120 Rotterdam – 75 (Rio de Janeiro). Cologne – 92 It is interesting that despite having – 93 a Spanish origin and a large Hispanic Antwerp – 96 population, Los Angeles is not as well- – 101 connected to Latin America as is Birmingham – 106 Miami. We speculate that this may have something to do with the immi- Source: Authors’ calculations gration patterns to both respective places. Many Hispanics in Miami’s business community are political C. While important service connec- important U.S. cities surprisingly refugees from Cuba. When Castro tions exist among certain U.S. cities located in relatively low-level strata in came to power over forty years ago, and particular global regions, U.S. Table 2 (Houston and Seattle). In much of Cuba’s middle class fled to cities are more strongly linked to each case we highlight the top 30 southern Florida. This affluent and other U.S. cities than to cities cities in each of the six hinterworlds educated expatriate community around the globe. (Table 3). In doing so, the “localness” quickly reestablished a business We focused here upon the hinter- of these cities’ hinterworlds becomes network to Latin America. By contrast, worlds of three pairs of U.S. cities, apparent. most Hispanic immigrants to Los selected to show a variety of U.S. city New York and Pittsburgh. As Angeles have been poor. The contacts hinterworlds. First, we contrasted the expected, the hinterworlds of these to Latin America in Los Angeles are country’s definitive global city (New two cities show quite opposite patterns therefore more cultural and are less York) to the U.S. city with the lowest (Table 3). The Pittsburgh hinterworld likely to result in the establishment of connectivity in Table 2 (Pittsburgh). is extremely U.S.-focused, with all of business contacts in the advanced Second, we compared two cities with the top 20 cities in its hinterworld producer services industry. well-known regional proclivities (Los located in the U.S. In contrast, New Houston and Seattle. The hinter- Angeles-Pacific Rim and Miami-Latin York’s hinterworld shows only three worlds of these two important U.S. America). Finally, we considered two U.S. cities among the top ten— cities are strikingly similar despite

February 2005 • The Brookings Institution • Survey Series 7 Table 3. The Top 30 Cities in the Hinterworlds of Six U.S. Cities

Rank New York Pittsburgh Los Angeles Miami Houston Seattle 1 Washington, D.C. Washington, D.C. Washington, D.C. Dallas San Francisco Dallas 2 Tokyo Cleveland Boston Philadelphia Dallas Washington, D.C. 3 Boston Dallas Chicago Washington, D.C. Washington, D.C. Portland 4 Chicago Philadelphia San Francisco Chicago Los Angeles 5 Hong Kong St. Louis Dallas Rio de Janeiro Boston Denver 6 Guangzhou Charlotte Philadelphia Cleveland Philadelphia Boston 7 London Boston San Diego Detroit Los Angeles Minneapolis 8 Singapore Minneapolis Tokyo Frankfurt Minneapolis Philadelphia 9 Ho Chi Min City Indianapolis Minneapolis San Diego St. Louis 10 Beijing Kansas City Denver St. Louis Seattle Montreal 11 Pittsburgh Denver New York Caracas Denver Barcelona 12 San Diego Atlanta Cleveland Bogota Vancouver 13 Cleveland Seattle Pittsburgh New York Pittsburgh Houston 14 Frankfurt Detroit Houston San Francisco Rio de Janeiro Detroit 15 Bangkok New York Calgary Boston Calgary Chicago 16 Minneapolis Chicago Seoul San Diego San Diego 17 Calgary Los Angeles Singapore Kansas City Vancouver Pittsburgh 18 Los Angeles San Francisco Guangzhou Vancouver Melbourne Indianapolis 19 Denver San Diego Panama City Kansas City Miami 20 Dallas Portland Atlanta Adelaide Charlotte San Francisco 21 Bratislava Rotterdam Charlotte Chicago Portland 22 Shanghai Houston Hong Kong Los Angeles Toronto 23 San Francisco Stockholm Vancouver Charlotte Birmingham(UK) Kansas City 24 Nassau Hamilton (Ber.) San Diego Rotterdam Bogota 25 Philadelphia Seattle Copenhagen 26 Cologne Detroit City St. Louis Cleveland 27 Kiev Vancouver Shanghai Sao Paulo Sydney Rio de Janeiro 28 Rio de Janeiro St. Louis Port Louis Sao Paulo Melbourne 29 Rotterdam Miami Manama Cape Town Indianapolis Atlanta 30 Zurich Adelaide Kansas City Barcelona Nassau New Delhi

Non-U.S. cities are shown in italics Source: Authors’ calculations their being from very different parts of Angeles and San Francisco. In general, in Europe. the U.S. (Table 3). Like Los Angeles, neither of Seattle nor Houston has a In Table 4 we report the results both have nine of their top 10 well- hinterworld very different from that of from using the U.S. and EU as “local connected cities in the U.S., and in Pittsburgh. regions” for their respective cities. The both cases the other top 10 city is still findings for U.S. cities are quite stark. North American (Montreal). Thus D. Even the most globally-connected The considerable localism of U.S. these cities are very U.S.-oriented in U.S. cities are more locally oriented cities is confirmed by the fact that their connections and, unlike Los than cities in the EU. three cities, including Pittsburgh, Angeles and Miami, they have no The above examples suggest a degree actually score the maximum in local distinctive extra-U.S. regional pattern of local parochialism in U.S. city link- focus. It is obvious from this list that of well-connected cities. In Seattle, for ages.33 Such localism can be easily Pittsburgh, rather than New York, typi- example, this can be explained by the measured as a particular orientation fies U.S. world cities in terms of their fact that, while it has a big port, it within a hinterworld, allowing us to inter-city relations. All but two U.S. does not have a large producer service both generalize to more U.S. cities and cities—Miami and New York—have a economy relative to such places as Los to make comparisons again with cities local orientation above 0.8 (and there

February 2005 • The Brookings Institution • Survey Series 8 EU cities are relatively well-connected Table 4. Local Orientations of U.S. and EU City Hinterworlds to Latin American cities, including historically linked Lisbon and Madrid. U.S.-ness EU-ness Note that both New York and London U.S. city (local orientation) EU city (local orientation) have appreciably negative orientations to Latin America, as their non-local Denver 1.000 Cologne 0.703 linkage biases are both directed else- Indianapolis 1.000 Munich 0.656 where. Pittsburgh 1.000 Stuttgart 0.640 To further U.S.-EU comparisons, Cleveland 0.981 Antwerp 0.625 we examined cities’ orientation to San Diego 0.981 Hamburg 0.609 Eastern Europe. We defined this St. Louis 0.981 Lyon 0.578 region by examining cities that were in Boston 0.963 Barcelona 0.531 COMECON during the Soviet era. In Minneapolis 0.963 Berlin 0.531 the 1990s they constituted centers for Philadelphia 0.963 Rome 0.500 massive privatizations that fuelled Portland 0.963 Copenhagen 0.375 globalization of advanced producer Kansas City 0.944 Milan 0.360 services, especially in banking/finance Seattle 0.944 Rotterdam 0.360 and law. In short, this was the “global Charlotte 0.926 Lisbon 0.344 opportunity region” of the 1990s and Dallas 0.926 Luxembourg City 0.250 it is clearly EU cities that took advan- Los Angeles 0.926 Vienna 0.250 tage (Table 6). Amongst U.S. cities, Washington, D.C. 0.926 Dusseldorf 0.219 only New York has a positive orienta- Detroit 0.907 Birmingham 0.203 tion, with a level on par with Antwerp, Houston 0.889 Madrid 0.203 which ranks 21st among EU cities. San Francisco 0.851 Amsterdam 0.188 The EU-Eastern European linkages Atlanta 0.833 Paris 0.172 are how we expect a “backyard rela- Chicago 0.815 Dublin 0.156 tionship” to look; it is precisely not Miami 0.741 Frankfurt 0.156 how the U.S.-Latin American linkages New York 0.481 Stockholm 0.141 appear once Miami is allowed for. Manchester 0.109 Athens 0.000 E. Beyond localism, functional Brussels -0.016 linkages among firms are more Helsinki -0.045 important determinants of connect- London -0.078 edness than geography alone. Our examination of hinterwords Source: Authors’ calculations (Table 3), revealed an interesting contrast between New York and the two Pacific cities (Los Angeles and are no values of orientation this high out to be much more American than Seattle). Specifically, while we were found in tables 5, 6, or 7). London is European. able to show a regional bias towards New York’s low level of local orien- European cities generally have the Pacific Rim for Los Angeles, we tation is distinctly un-notable in stronger regional connections across could discern no such pattern for comparison with EU cities, however. the board than cities in the U.S. This Seattle. And yet the Atlantic Coast city In general, EU cities are locally-orien- is even true of inter-city links to Latin of New York is very well connected to tated but to a much lesser degree than America, as shown in Table 5. As we Pacific Asian cities, seemingly much U.S. cities. Some major EU cities have would expect, Miami has the highest more so than even Los Angeles. This low or even negative local orientation Latin American orientation among might be explained by New York’s (e.g., London and Brussels) and the both U.S. and EU cities—but this is premier role in global finance. In same is true of some peripherally- where U.S. preeminence ends. U.S. general, what these contrasts indicate located EU cities (e.g., Helsinki and cities turn out to be weakly connected is that the relative network locations Athens). This comparison highlights to Latin American cities, while EU of Pacific Asian cities (specifically in the extreme nature of the national cities have a balance of strong and banking/finance) are much more parochialism of U.S. cities: Even at weak connections. This is due largely important than their absolute the global city level New York turns to the fact that several medium-sized geographical location (Pacific Rim) in

February 2005 • The Brookings Institution • Survey Series 9 mere geography, driving the formation Table 5. Latin American Orientations of U.S. and EU Cities of city hinterworlds.

Latin American Latin American What do these findings mean for U.S. city orientation EU city orientation U.S. cities and their economies? This paper has been largely a descrip- Miami 0.563 Athens 0.469 tive exercise. We are working in a Cleveland 0.125 Copenhagen 0.469 research area—worldwide inter-city St. Louis 0.094 Helsinki 0.406 relations—where there is little empir- San Diego 0.063 Rome 0.313 ical work, so we need to provide Dublin 0.281 original findings to depict the new Atlanta 0.000 Lisbon 0.188 global geography. At the same time, we Boston 0.000 Hamburg 0.094 must be cognizant of the limits of the Dallas 0.000 Stockholm 0.094 results we have produced. In the Madrid 0.063 myriad connections that constitute Chicago -0.031 Milan 0.063 Manuel Castells’ contemporary “space Denver -0.031 Vienna 0.063 of flows,” the inter-city relations we Kansas City -0.031 have described are but a miniscule Minneapolis -0.031 Amsterdam 0.000 part.34 But they are a very important Philadelphia -0.031 Luxembourg City 0.000 part, as they describe a cutting edge Charlotte -0.063 economic sector in the world Detroit -0.063 Cologne -0.031 economy—advanced services—that San Francisco -0.094 Manchester -0.031 has been crucially instrumental for Houston -0.156 Birmingham -0.063 economic globalization. As such, our Portland -0.156 Dusseldorf -0.125 findings should not be dismissed for Seattle -0.188 Munich -0.125 their narrow economic focus, even if Indianapolis -0.281 Barcelona -0.156 the meaning of the findings still Los Angeles -0.313 Brussels -0.156 remains to be ascertained. We do New York -0.313 Antwerp -0.219 know that, at worst, they cast doubt on Pittsburgh -0.344 Stuttgart -0.219 the continuing viability of U.S. cities Washington, D.C. -0.344 Frankfurt -0.250 in the world economy. In other words, Berlin -0.281 while there are many positive things Paris -0.281 happening in U.S. cities today, we Rotterdam -0.313 can’t neglect the benefits of dense Lyon -0.344 worldwide connections. Intra-city London -0.375 variety and complexity is not an alter- native to inter-city variety and Source: Authors’ calculations complexity, but rather, as Jacobs taught us long ago, they complement each other.35 determining their inter-city links. This might expect: the top six include four To probe the findings further, we can be more definitively illustrated by Pacific Coast cities plus the two domi- need to consider several key questions: considering the Pacific Asian orienta- nant U.S. international financial How relevant is the relative under- tions of U.S. and EU cities. centers, New York and Chicago. Of connectedness and parochialism of Table 7 lists these Pacific Asian course, Europe has no Pacific Coast leading U.S. cities to the future pros- orientations. This is, of course, the cities, but it does have many interna- perity and development of their world region whose rise to economic tional financial centers; thus for metropolitan areas? Why haven’t U.S. prominence in the last quarter of the financial and business services, EU cities made a greater mark on the twentieth century helped popularize cities are generally better connected to world city network? What are the the concept of globalization, and led to Pacific Asian cities than are those in processes or mechanisms operating in the transcendence of the old North the U.S. Clearly, Table 7 confirms the world economy that are producing Atlantic-centered world economy. The what we gleaned from Table 3: It is this inter-city outcome? Does the orientations of cities to this key the financial functions of Pacific Asian degree and nature of connectedness in economic region are ordered as we cities in the world city network, not the world city network translate into

February 2005 • The Brookings Institution • Survey Series 10 but do not wish to invest in service Table 6. East European Orientations of U.S. and EU Cities within the U.S. market itself with its high cost thresholds. Hence, for East European East European global services delivered by non- U.S. city orientation EU city orientation U.S. firms, concentration in New York casts a ‘shadow’ with relatively New York 0.111 Vienna 0.741 few global services provided else- Brussels 0.630 where. Hamburg 0.593 ■ The Comfort Effect: Leading U.S. Detroit -0.148 Helsinki 0.556 financial and business service firms Washington, D.C. -0.185 Athens 0.519 may decide not to develop a global Los Angeles -0.444 Dublin 0.519 strategy and concentrate on San Diego -0.444 Frankfurt 0.482 expanding within the domestic Miami -0.482 Milan 0.444 market. Because this market is so Atlanta -0.519 Amsterdam 0.407 large, such a strategy may be less San Francisco -0.519 Copenhagen 0.407 risky, and it eliminates the need to Cleveland -0.593 Luxembourg City 0.407 cope with expensive “loss leader” Portland -0.593 Stuttgart 0.407 foreign-city offices. Certainly there Seattle -0.593 Dusseldorf 0.304 is much less of an incentive for U.S. Boston -0.630 Berlin 0.297 financial and business service firms Chicago -0.630 Paris 0.297 to embark on a transnational Philadelphia -0.630 Cologne 0.259 strategy than for firms in any other Denver -0.667 Madrid 0.259 country. Alternatively, a firm in Houston -0.667 London 0.222 Switzerland with its limited Dallas -0.704 Munich 0.222 domestic market must operate glob- Kansas City -0.704 Lisbon 0.186 ally to survive. Pittsburgh -0.704 Antwerp 0.111 These two effects must remain St. Louis -0.704 hypotheses for the time being but they Indianapolis -0.741 are, we think, very plausible interpre- Charlotte -0.778 Stockholm -0.037 tations of the situation. If we accept Minneapolis -0.815 Rome -0.111 them as prime reasons for U.S. city Rotterdam -0.222 under-connectedness then it is diffi- Lyon -0.444 cult to see how this global deficiency Barcelona -0.556 of U.S. cities will alter any time in the Birmingham -0.593 near future. But does this matter? Manchester -0.593 Are global connections essential for Source: Authors’ calculations economic success? We can begin to answer this question with a simple comparison. Consider immediate advantages in terms of formation—the global service firms— the world of Atlanta versus economic growth? And what, ulti- use the cities in which they are Phoenix. Both regions are Sunbelt mately, do the findings mean for located. We can, nevertheless, provide boomtowns that expanded rapidly in political and business leaders? some explanation of the processes the past several decades. The two We attempt to address some of behind U.S. cities’ weak connected- metropolitan areas now rank among these issues below. ness: the top 15 regions in the country. But ■ The Shadow Effect: Non-U.S. as “U.S. world cities,” Atlanta—the What might explain the localism of service firms developing a global “New York of the South”— places 6th, U.S. cities? strategy must locate in New York while Phoenix comes in at 25th (Table Understanding the mechanisms (and perhaps Chicago and Los 1). Thus the question arises to through which U.S. city connections Angeles) but often decide on no whether Atlanta has a distinct advan- in the world city network have been further penetration of the U.S. tage over Phoenix because of its created requires additional research domestic market. They need a U.S. superior global connectivity. on how agents of world city network presence for their non-U.S. clients Cities and city networks are

February 2005 • The Brookings Institution • Survey Series 11 lished multiple global offices as part of Table 7. Pacific Asian Orientations of U.S. and EU Cities their overall location strategy. But they remain a small minority of firms in all Pacific Asian Pacific Asian the sectors we cover. Most financial U.S. city orientation EU city orientation and business service firms are still based upon a local client base gener- New York 0.472 Amsterdam 0.611 ated by building a reputation for their Los Angeles 0.306 London 0.528 service provision. Portland 0.306 Dusseldorf 0.472 Overall, in any city, there will be Chicago 0.194 Frankfurt 0.444 access to services operating at Paris 0.444 different levels. The fact that one city Brussels 0.361 has more global connectivity therefore San Francisco -0.056 Luxembourg City 0.333 does not necessarily mean that it is the Seattle -0.111 Athens 0.250 more important service center. In our Houston -0.306 Hamburg 0.083 previous discussion (Table 2), we Philadelphia -0.306 Madrid 0.028 noted that Luxembourg City is much Dallas -0.333 more connected globally than Cleve- Denver -0.333 land (63rd to 112th), however San Diego -0.333 Dublin -0.028 Cleveland is the larger service center Washington, D.C. -0.361 Lisbon -0.083 overall. Luxembourg City has grown Detroit -0.472 Vienna -0.083 an important global niche as an inter- Pittsburgh -0.472 Milan -0.139 national financial center but is weak Charlotte -0.500 Berlin -0.167 in all other sectors and scopes, Atlanta -0.556 Munich -0.167 whereas Cleveland has important Boston -0.556 Helsinki -0.333 local, state, regional, and national Cleveland -0.583 Cologne -0.361 connections in addition to its global Minneapolis -0.583 Rotterdam -0.444 connections. Kansas City -0.611 Barcelona -0.472 Having put our analyses into Miami -0.611 Stuttgart -0.472 economic-geographical perspective, we St. Louis -0.639 Copenhagen -0.611 can begin to answer the question of Indianapolis -0.667 Stockholm -0.694 the importance of our findings to the Birmingham -0.722 growth prospects of U.S. cities. The Rome -0.722 first point to make is that global Lyon -0.750 connections are not necessary for Manchester -0.750 success. Several studies have shown Antwerp -0.778 that medium-sized cities have been particularly prospering in recent years. Source: Authors’ calculations At the global level, Jones Lang LaSalle’s “Winning Cities” project has identified Dublin, , and Las immensely complex phenomenon. taking place at other geographical Vegas as the outstanding growth Certainly it would be naïve in the scales continue to operate. There is centers. Only one of these cities has extreme to assume that more global nothing within globalization processes relatively strong GNC—Dublin’s high connections automatically translate that determine this particular scale to rank (30th) in Table 2 is a “European into more economic growth. It is, of be the most important for cities. City surprise” in our analysis. Dubai has a course, a matter of what these connec- businesses will enter markets at a moderate world rank (54th), but ranks tions contribute to a city’s economic variety of scales and will expand their as the most connected Middle Eastern vibrancy. The only sure thing we can activities at levels where they either city; and Las Vegas’ GNC ranking is say about globalization is that it has make most profit or expect to make only a lowly 197th. Similarly, in a made cities and their networks even most profit. This may or may not be recent study focusing only on U.S. more complex. The key point to note is the global scale depending on a whole metropolitan areas, Jonathan Bowles that these enhanced transnational and array of circumstances. and Joel Kotkin show how many worldwide inter-city relations are an The advanced producer service smaller metropolitan areas are growing addition: Economic transactions firms we have studied have all estab- much faster than New York.37 Taking

February 2005 • The Brookings Institution • Survey Series 12 such results at face value, it would greatest local impact that global Does an Atlanta or Houston seem seem that global connections are not connectivity will have on a city is in its more outwardly focused than Phoenix? just unnecessary for growth; they may central business district. That is hard to say, but it is easy to even be an impediment. First, because much of the advanced imagine that such is the case given However, the global level of producer service industry is located in how isolated Phoenix is from the rest economic activities cannot be central business districts, the growth of of the world. Of course, historically, dismissed so easily. Economic global- global networks could increase the isolation has not been beneficial for ization is a reality, a critical result of demand for downtown Class A office cities: All successful cities have been the massive restructuring of economic space. A larger workforce can subse- cosmopolitan. activities over the last two or three quently impact a host of related decades. What it provides for cities is downtown industries, including retail, new connections, and thus new oppor- entertainment, arts, conventions, and Methods Appendix tunities to reach larger supplier, perhaps even housing. producer, and consumer markets.37 Second, expanding employment in Computing global network Taking opportunities from across a fields like law, advertising, and media connectivity wider range of geographical scales can help attract a highly educated, The world city network is constituted means increased complexity and it is skilled, and “creative” work force to a by m global service firms with offices the latter that is crucial to city world city. In fact, virtually all the jobs distributed across n world cities. The vibrancy and medium-term success. As associated with global producer serv- level of service performed by firm j in business conditions change, it is the ices falls under Richard Florida’s city i is vij which we call the service more complex cities that are best able definition of the “Creative Class.” value. The array of service values to weather the economic storms. John Phoenix and Las Vegas rank low in defines an m x n service value matrix, Short has reported, for example, that a Florida’s scheme despite their rapid V. The world city network is derived large and successful city like Phoenix growth because they are underrepre- from this matrix using the plausible appears to be like much less sented in terms of the creative class. conjecture that the larger a firm’s successful Third World cities in terms In contrast, many of the cities that are service value in a city, the greater the of its population/GNC relation. This is well-connected to the global producer number of the firm’s flows of informa- not to say that Phoenix is on a dead- service economy also maintain a large tion, knowledge, instruction, ideas, end economic trajectory, but its creative class. strategies, plans, etc. will emanate relatively simple, localized economy Third, being a world city can help from that city to connect with offices does make it economically vulnerable. local boosters promote a place. Most in other cities. This assumption allows So while economic globalization is places promote the idea that they are for relations between cities to be not a panacea for cities, it does the world city of something—some- defined. provide new potentials for city times almost anything. Given the The initial relation between each economies that adds to their increasing buzz about the global pair of cities is given by: complexity and therefore to their ulti- economy, not being part of it means a mate sustainability. Thus in the city fails as a real player in business. rab,j = vaj . vbj (1) medium-term U.S. cities cannot afford Cities even bid on such high-profile to be globally under-connected and but often losing ventures as World This defines relations between cities nationally parochial. Because the U.S. Fairs and Olympic Games—think a and b in terms of firm j. This speci- economy is the sum of its city Atlanta in 1996—just to lay claim to fies an elemental interlock link economies, our findings point to a their lofty global status. Atlanta can between two cities. It is multiplicative national economic vulnerability. also now lay claim to out-networking because the potential quantity of flows such Southern rivals as Dallas and between two cities rises geometrically What does this mean for local Houston, and leaving its Sunbelt with the quantity of service provided political and business leaders? bookend Phoenix in the dust. in each city. From equation (1) the The findings presented here have Finally, and perhaps most specula- aggregate city interlock link between several key implications for world city tively, being a well-connected world two cities can be derived as: political and business leaders. Because city may produce a more cosmopolitan the world city network is such a new view. Places with business links rab = ∑rab,j (2) method and concept, it is hard to throughout the world may be less j capture all of its meaning in this parochial in attitude and feel than exploratory paper. But some obvious peer cities that are not well-connected For each city there is n-1 such links, issues do arise. We believe that the to global producer service economy. one to every other city. These can be

February 2005 • The Brookings Institution • Survey Series 13 used to define the overall interlock connectivity of a city so that: Table A. Simple City/Firm Data Set

Ca = ∑rai where a ≠ i (3) (a) Service values i Firm A Firm B Firm C Firm D New York 5 5 5 2 This measure of connectivity picks Hamburg 2 3 0 5 up two features of a city’s service Bogota 2 0 2 0 values. First, and most obviously, cities (b) Levels of servicing derived from (a) where firms locate offices with higher New York Hamburg Bogota service values are more connected. New York - 0.8 1.0 Second, and more subtly, if those high Hamburg 0.5 - 0.2 service offices are for firms with larger Bogota 0.2 0.13 - office networks then the city appears more connected. In other words, a city with several high service offices service firms). Thus, in the simple describing the hinterworld of a city. of firms that has only small office net- data set shown in Table A, highest However, there is a problem when works will not be that well connected levels of possible service are 20 for comparing the hinterworlds of cities as measured by interlock connectivity. New York, 15 for Hamburg, and 10 for using results such as those in Table A. Interlock connectivity indicates the . Now for each city, take other Notice that in this table New York importance of a city in the world city cities in turn and sum their service appears with very high service levels network. Generally we refer to empir- scores but only for firms present in the for the other two cities and Bogotá ical measures of interlock connectivity original city. For instance, starting provides low levels. This is obviously as global network connectivity. with New York, the sums for Hamburg reflecting the network position of and Bogotá are 10 and 4 respectively; these cities in this small data set. Thus Deriving city hinterworlds starting with Hamburg the sums for when this method is used for the 100 To compute a city’s hinterworld we New York and Bogotá are 12 and 2; x 123 data set, we find that every city need to specify its external relations and starting with Bogotá the sums for has its highest external provision in with other cities. For this exercise, we New York and Hamburg are 10 and 2. either London or New York. In deal with just the top 123 cities in The latter sums are expressed as general, we can note that external terms of global network connectivity. proportions of the highest level of service provisions tend to closely This is because the data matrix possible service. For instance, the follow the level of a city’s global becomes sparser (more zero scores) proportions for New York are Hamburg network connectivity. Thus mapping with additional cities creating a partic- 0.5 (= 10/20) and Bogotá 0.2 (= 4/20). direct measures of connections to any ular problem when measuring All such computations are shown in city tends, to a large degree, merely to connections for individual cities. Obvi- Table A. replicate the worldwide map of GNCs. ously with less firms represented in The interpretation is relatively Thus all hinterworlds, although not computing a measure, a result simple. The columns in this table exactly the same, look very much alike. becomes that much less reliable. Thus define the level of service that can be To overcome this comparative defi- the results reported below are derived expected in a city when visiting a ciency an extra step needs to be taken from a reduced 100 x 123 data matrix. global service firm in a row city. Thus, for describing hinterworlds. A city’s hinterworld consists of the going into an office in New York to do We will term the external service levels of service it provides for doing business in Hamburg the service level provisioning results from Table A business in each of the other 122 is 0.5, but to do business in Bogotá depictions of ‘absolute’ hinterworlds. cities in the reduced data matrix. This the level falls to 0.2. Notice that from Taking out the underlying general is initially computed as follows. First, Bogotá, doing business in New York influence of global network connec- count the number of firms present in has a 1.0 service level showing that tivity from the absolute provisioning each city. For each city multiply this Bogotá’s two service firms in Table 1 values for a city is a fairly simple task. number by 5, the maximum service have their headquarters in New York. Scatter diagrams of global network value. This constitutes the highest In contrast, the lowest level of service connectivity against a city’s external possible level of service that a city in this data is a paltry 0.13 for doing provisioning levels shows a strong could expect in another city (i.e., the business from Hamburg in Bogotá. positive linear relationship in every other city houses the headquarters of These columns represent the servicing case. Thus absolute provision in city i every single one of a city’s global linkages that form the basis for for all other cities j can be regressed

February 2005 • The Brookings Institution • Survey Series 14 against their GNCs using the simple might expect less extreme orientation. a special issue of Urban Studies, see W. F. equation: Third, we divide the sum by the Lever and I. Turok, “Competitive Cities: maximum possible sum to measure the Introduction to the Review,” Urban Hj = _ + _ Cj + _ where i ≠ j (4) relative orientation to a region. In the Studies, 36 (1999): 1029–44. case above the orientation of City i to Hj is the absolute provision values Region X is 1.0 which one the limit of 6. , Cities and the Wealth of and C is global network connectivity. the scale, the other is -1.0 when Nations (New York: Vintage, 1984). The regression constants are _ and _, maximum under-linkage is found. and _ is the error term or residual. Producing a scale between plus and 7. A good discussion of this distinction can be Calibrating this equation for any city minus one in this way enables us to found in W. W. Powell, “Neither Market produces an estimate of the level of create comparable measures of the Nor Hierarchy: Network Forms of Organi- service provision given a city’s connec- geographical orientations of cities zation,” Research in Organizational tivity. It is the difference between this across different world regions. Behavior 12 (1990): 295–336. See also estimate and the actual level of provi- G. Thompson, Between Hierarchies and sion that is the residual. These Markets: The Logic and Limits of Network residuals define a ‘relative’ hinter- Endnotes Forms of Organization (Oxford University world—in which other cities that are 1. Peter J. Taylor is associate director of the Press, 2003). ‘over-linked’ to city i are identified (a Metropolitan Institute at Virginia Tech and positive residual) and also where city i co-director of the Globalization and World 8. This position is presented in detail in is ‘under-linked’ (a negative residual) Cities (GaWC) Study Group and Network Taylor, World City Network. with respect to position in the world at Loughborough University, UK. He is city network (i.e., its GNC). author of World City Network: a Global 9. John Friedman, “The World City Hypoth- Such mappings of hinterworlds have Urban Analysis 2003 (London: Routledge). esis.” This approach to defining the “world been analyzed and discussed for North Robert E. Lang is director of the Metropol- city system” has been recently opera- African/West Asian cities,40 for Dutch itan Institute at Virginia Tech and an tionalised in Arthur S. Alderson and Jason cities,41 for British cities,42 and for associate professor of urban affairs and Beckfield, “Power and Position in the Belgium cities.43 The hinterworlds of planning. He is author of Edgeless Cities: World City System,” American Journal of all 123 cities, including the 23 US Exploring the Elusive (Wash- 109 (2004): 811–51. For a cities, are depicted in the GaWC ington: Brookings Institution, 2003). comparison with the approach adopted virtual Atlas of Hinterworlds.44 here, see Peter J. Taylor, “Parallel Paths to 2. See the argument in Susan E. Clarke and Understanding Global Inter-city Rela- Computing regional orientations Gary L. Gaile, The Work of Cities tions,” GaWC Research Bulletin 143 within hinterworlds (Minneapolis: University of Minnesota (www.lboro.ac.uk/gawc). This alternative Geographical orientations within city Press, 1998) approach has previously been applied to hinterworlds, such as a propensity to U.S. inter-city relations by Allan Pred, be over-linked to neighboring cities or 3. The two classic derivations of world city City-Systems in Advanced Economies to a particular world region, can be hierarchies are: J. Friedmann, “The World (London: Hutchinson, 1977) and Christo- measured by converting the hinter- City Hypothesis,” Development and pher Ross, The Urban System and Networks world residual values into ordinal Change 17 (1986): 69–83 and R.P. of Corporate Control (Greenwich: JAI scores in the following way. First, for Camagni, “From City Hierarchy to City Press, 1992). City i, the top ten over-linked cities Network.” In T. R. Lakshmanan and P. are scored 3, the next top 20 are Nijkamp, eds., Structure and Change in the 10. Saskia Sassen, The Global City (Princeton scored 2, and remaining over-linked Space Economy (Berlin: Springer-Verlag, University Press, first edition 1991, second cities score 1. Under-linked cities are 1993). edition 2001). similarly scored as negative values: -3, 2, and -1. Second, identify Region X 4. The leading papers in this research school 11. For Sassen, the production and sale of which includes, say, 10 cities. were brought together in L. S. Bourne and advanced producer services are crucial Summing the values for just these 10 J. W. Simmons, eds., Systems of Cities elements of her global cities. cities provides an aggregate measure (New York: Oxford University Press, 1978). of City i’s geographical orientation to 12. Lang, Edgeless Cities. Region X. If the sum is 30 this means 5. The city competition referred to here is that Region X’s cities actually consti- that between major city regions, not cities 13. Friedmann, “The World City Hierarchy,” tute the top ten over-linked cities. This as in incorporated administrative entities. Development and Change 17 (1986): is a maximum score; in practice we The literature is reviewed and developed in 69–83.

February 2005 • The Brookings Institution • Survey Series 15 14. D. R. Meyer, “The World System of Cities: 25. The technique for measuring and mapping 34. Manual Castells, The Rise of Network Relations Between International Financial hinterworlds is described in detail in Peter Society (Oxford: Blackwell, 1996). Metropolises and South American Cities,” Taylor and David Walker, “Hinterworlds Social Forces 64 (1986): 553–581. Revisited,” Geography 89 (2004): 145–51. 35. Jacobs op cit. Described in detail in Jane The methods appendix provides a summary Jacobs, The Economy of Cities (New York: 15. A. D. King, Global Cities (London: Rout- of this technique. Random House, 1969). See also Jane ledge, 1990). Jacobs, The Nature of Economies (New 26. The United States contains 40 of the 315 York: Vintage, 2000), discussion of Los 16. D. R. Meyer, “Change in the World System cities in the global services/cities data, 23 Angeles page 51. of Metropolises: The Role of Business of which are in the 123 cities used for Intermediaries,” 12 defining hinterworlds. 36. See Jones Lang Lasalle at (1991): 393–416. www.joneslanglasalle.co.uk/ 27. See “Visualization” section at 17. S. Sassen, Cities in a World Economy www.lboro.ac.uk/gawc 37. Jonathan Bowles and Joel Kotkin, Engine (Thousand Oaks, CA: Pine Forge, 1994). Failure (New York: Center for an Urban 28. J. Abu-Lughod, New York, Los Angeles, Future, 2003). 18. W. Wu, “Economic Competition and Chicago, America’s Global Cities Resource Mobilization.” In M. A. Cohen, (Minneapolis: University of Minnesota 38. Short, “Black Holes.” B. A. Ruble, T. S. Tulchin and A. M. Press, 1999). Garland, eds., Preparing for the Urban 39. Richard Florida, The Rise of the Creative Future (Washington: Woodrow Wilson 29. Ed Brown, Gilda Catalano and Peter Class: and How it’s Transforming Work, Center, 1996). Taylor, “Beyond World Cities,” Area 34 Leisure, Community and Everyday Life (2002): 139–148. (New York: Basic Books, 2002). 19. F-c. Lo and Y-m. Yeung, Globalization and the World of Large Cities (Tokyo: University 30. Carl Abbot, Political Terrain: Washington 40. Peter Taylor, “West Asian/North African of United Nations Press, 1998). DC from Tidewater Town to Global Metrop- Cities in the World City Network,” Arab olis, (Minneapolis: University of Minnesota World Geographer 4 (2001): 146–159. 20. J. R. Short and Y. Kim, Globalization and Press, 1999). More generally, he promotes the City (London: Longman, 1999). a historical approach to understanding 41. Peter Taylor, Amsterdam in the World City U.S. world cities: Carl Abbott, “The Inter- Network (Amsterdam: AME, 2002). 21. For full details of the specification, see national City Hypothesis: An Approach to Taylor, “Specification.” The methods the Recent History of U.S. Cities,” Journal 42. Taylor, “Hinterworlds Revisited.” appendix provides a summary of that argu- of Urban History 24: 28–52. ment highlighting aspects necessary for 43. Ben Derudder and Peter Taylor, “Brussels this paper. 31. J. R. Short, “Black Holes and Loose in the World City Network,” Belgeo 4 Connections in the Global Urban (2004): 259–76. 22. See www.lboro.ac.uk/gawc and Network’” GaWC Research Bulletin www.mi.vt.edu No. 76. 44. See “Visualization” section at www.lboro.ac.uk/gawc 23. This data collection effort was funded by 32. The European Union (EU) is chosen for the Economic and Social Research comparison because it is an economic area Council of the and is roughly equivalent to the U.S. in economic described in detail in Peter J. Taylor, Gilda size and wealth. Pacific Asia has no such Catalano and David Walker, “Measurement institutional unity but it does constitute a of the World City Network,” Urban Studies key globalization arena and therefore 39 (2002): 2367–2376. The regional provides a second valuable comparison. breakdown is: Europe 101, Northern America 50, Pacific Asia 37, Latin America 33. Remember that it is only global firms 37, Sub-Saharan Africa 32, Middle included in the study; we do not consider East/North Africa 25, South Asia 13, national or local firms. Thus the ‘national Australasia 12, CIS 8. localism’ recorded shows how global firms are using US cities. 24. Op cit.

February 2005 • The Brookings Institution • Survey Series 16 Acknowledgments The authors would like to thank Jennifer Vey and Jared Lang for their contri- butions to this report. The Brookings Metropolitan Policy Program would like to thank the Surdna and Fannie Mae foundations for their support of our work on competitive cities.

For More Information Robert E. Lang Metropolitan Institute Viriginia Tech phone: 703-706-8101 e-mail: [email protected]

Peter J. Taylor Department of Geography Loughborough University phone: +44 (0)1509 222790 e-mail: [email protected]

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

The Brookings Institution

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

Direct: 202-797-6139 • Fax/direct: 202-797-2965 www.brookings.edu/metro