GaWC Research Bulletin 88

This Research Bulletin has been published in Regional Studies, 37 (9), (2003), 875-886.

doi:10.1080/0034340032000143887

Please refer to the published version when quoting the paper.

(Z)

Hierarchical Tendencies and Regional Patterns in the World City Network: A Global Urban Analysis of 234 Cities

B. Derudder^*, P.J. Taylor^** F. Witlox^* and G. Catalano^***

Abstract

Previous explorations of the world city network have tended to focus on analyses of its upper ranks. Using a data base on the global strategies of 100 financial and business firms across 234 cities, fuzzy classification analysis is used to define 22 urban arenas. With their 'core' and 'hybrid' memberships, these arenas are constituted by interweaving hierarchical and regional processes. The results include evaluations of the previously unassessed outer reaches of the world city network.

Keywords: Globalization, urban hierarchy, world cities, world city network, world regions.

INTRODUCTION

'One of the most salient propositions in John Friedmann's seminal World City Hypothesis (1986, p. 71) refers to the fact that "key cities throughout the world are used by global capital as 'basing points' in the spatial organization and articulation of production and markets. The resulting linkages make it possible to arrange world cities into a complex spatial hierarchy." However the lack of theoretical agreement on the defining characteristics of world cities has resulted in ad hoc taxonomies (e.g. Friedmann, 1986; Knox, 1995), often limited to the highest ranks of the hierarchy (e.g. Sassen, 1991; Abu-Lughod, 1995). Apart from the lack of an undisputed definition of world cities in and by itself, the main reason for these somewhat eclectic approaches has been a lack of data (Smith & Timberlake, 1995; Short et al., 1996), a problem that is, of course, related to the absence of undisputed defining characteristics of world cities. One of the major consequences of the problems pertaining to a description of Friedmann's "complex spatial hierarchy" is that the lower rungs of this transnational urban hierarchy have remained unassessed. The prime purpose of this paper is to rectify this limitation: we investigate a very large number of cities, many of which have never figured at all in previous discussions of world cities. The rationale behind this work is to be found in researches of the Globalization and World Cities (GaWC) Study Group and Network's suite of studies at Loughborough University.

Empirical Rationale

Drawing on Sassen's (1991, 1994, 1995) work on the role of advanced producer services in world city formation, a theoretically grounded endeavour of data acquisition has been undertaken by GaWC (http://www.lboro.ac.uk/gawc). Treating world cities as global service centres, GaWC has developed a methodology for studying world city network formation. Rather than assuming world cities form an urban hierarchy, this approach specifies a network in which 'hierarchical tendencies' may be revealed. The empirical researches have been based upon first, a rectification of the data deficiency problem, and second, incorporation into the analysis of many more cities than heretofore in world cities research. The methodology is described in some detail in the next section but to put this paper into context it is necessary to briefly describe previous related GaWC researches.

In initial work, 125 cities were assessed for 'world city-ness' in devising a 'roster' of 55 world cities (Beaverstock et al., 1999). These 55 cities were subsequently used for creating some experimental data on 46 service firms for a first global multivariate analysis of world cities (Taylor and Walker, 2001). The success of the latter stimulated a new data collection exercise covering 316 cities and 100 global service firms (Taylor et al., 2002a) and new analyses of the 123 most connected cities (Taylor et al., 2002b). These cities were selected for having network connectivities at least one fifth of the highest city connectivity. Analysis was limited to 123 because as the size of the data matrix increases (i.e. inclusion of more cities) it becomes relatively 'sparse' (lots of zero entries) which make analysis less reliable. Nevertheless, this is the largest published analysis of cities across the world that we are aware of but it still leaves several regions relatively unrepresented. For example, in inter-tropical Africa there are only two cities (Lagos and Nairobi) that qualify in the top 123 connected cities.

In this paper we go beyond these previous studies in two ways. First, using the same data based upon 316 cities we analyse 234 of them: we include all cities in which at least 20 of the 100 service firms have a presence. Second, for the 234 cities set we employ fuzzy set analysis in order to cope with the resulting sparser data matrix better: this is a technique that deals with this problem reasonably well. Thus we are able to include more than 100 additional cities. The result is that we have a fuller representation of regions across the world. For example, there are now 12 African cities included outside South Africa and the Arab north.

The end-product is an analysis of the world city network at a level of geographical detail never before attempted. We create a global ordering of cities into 22 urban arenas. These exhibit both hierarchical tendencies - the arenas are arranged into a centre surrounded by five bands of different levels of connectivity - and regional patterns with arenas showing specific geographical concentrations of cities. But before we describe these results in detail this empirical rationale requires conceptual problems to be spelt out.

The contemporary study of world cities started with Friedmann and Wolff's (1982) identification of 'command centers' to control and articulate the 'new international division of labor' being created by multinational corporations. This model reflected the later recognition by Amin and Thrift (1992) of a shift from an international to a global economy, characterized by increasingly integrated global networks of production and services. World cities are then the basing points in these networks, and their specification therefore relates to the identification of 'cities in global matrices', as Smith and Timberlake (1995) reminded us. Related discourses focus on a 'global network of cities' (King, 1990, p. 12), a 'transnational urban system' (Sassen, 1994, p. 47), 'functional world city system' (Lo and Yeung, 1998, p. 10), or a 'global urban network' (Short and Kim, 1999, p. 38). The implications of these various different conceptualisations have never been fully discussed in the world city literature.

Formal specifications of the cities acting as 'command centers' are fraught with difficulties. While it is obvious that cities like London and New York are world cities, there has hardly been a consensus as to the status of less significant cities in this context (Beaverstock et al., 1999). In other words, the highest ranks of the world city network may very well stand out, but the absence of more detailed taxonomies of cities below the leading cities in the world economy is a problem. As a consequence, somewhat vague discourses on patterns of global competence in the outer reaches of the world city network have dominated the debates. For instance, some authors have reified the lower ranks of the hierarchy as 'sub-global cities' (e.g. Daly and Stimson, 1992; O'Connor and Stimson, 1995), while this is particularly problematic given the pervasive nature of globalization. In a special issue of Urban Geography (Knox, 1996), it has been indicated that these 'medium cities' have just as much need to respond to globalization trends as their larger neighbours. In other words, it would be wrong to designate the outer reaches of the world city network as 'sub-global'. Rather, their capacity to operate in a global economy is merely less when compared to the upper ranks of world cities. All cities operate at a myriad of scales, and although the importance of their respective functions may vary, it would be wrong to pin down a clear-cut border between world cities, thoroughly influenced by and influencing globalization, and other cities deemed 'sub-global.'

Conceptual Problems

The outlined problems pertaining to the formal identification of world cities can hence be traced back to a lack of characteristics that have been agreed upon to qualify for world city status. As a result, in the original outset of the World City hypothesis, Friedmann (1986) restricted the associated taxonomy to 30 world cities (with some hierarchical notes on these cities), in addition to an accompanying suggestion on regional patterns in the world city network (i.e. three subsystems: North America, Europe, Pacific Asia). World cities outside these three subsystems primarily served as an outlet of other economies into these subsystems, providing a basic spatial structure for the world-system, as outlined by Wallerstein (1979). Friedmann (1986, p.71) asserted, however, that the lower rungs were highly uncertain, for the importance of some semi-peripheral World Cities was derived from the articulation of significant economies, whereas secondary structures in Europe where difficult to specify "because of their relatively small size and often specialized function". As a result, "the list of secondary cities (.) is meant only to be suggestive".

The GaWC approach to overcoming these discourse and classification problems has been to build upon aspects of Sassen's (1991, 1994, 1995) work on place and production in a global economy, which pertains to the wider debate over the extraordinary changes that have occurred in global economic systems in the last few decades. One of the most fundamental changes, according to DANIELS (1995), has been the self-accelerative transformation of the economic bases of cities from manufacturing to services, a conversion that can be traced back to the observation that a growing number of manufacturing and service industries, unable to cope with the accelerated pace of structural change and the increasing pressure for product innovation on their own, are becoming more and more dependent on specialized business services. For Sassen (1991, 1994, 1995), the salient point is that these business services are in and by themselves an indispensable production factor that has a growth potential of its own, as opposed to other domains of service sector growth that is the strict result of derived demand of other sectors. The reason for this is that such corporate service firms have benefited immensely from the technological advances in computing and communications that have allowed them to broaden the geographical distribution of their service provision. For instance, law firms have been traditionally associated with a particular city and its local client base - a 'New York law firm', a 'Boston law firm' and so on - but under conditions of contemporary globalization a few firms have chosen to pursue a strategy of providing legal services across the world (Taylor et al., 2002a , see also Coffey and Bailly, 1992 and Wood, 1991). Based on these observations, Sassen (1995, p. 63) has ascertained that "(.) a focus on the production process in service industries allows us (.) to examine the proposition that there is a producer services complex which, while catering to corporations, has distinct locational and production characteristics. It is this producer services complex more so than headquarters of firms generally that benefits and often needs a city location". Following this lead, GaWC conceives world cities as the production sites for the leading service industries of our time, i.e. places where knowledge-based (expert/profession/creative) services to other corporations are concentrated. It has indeed been argued elsewhere that new technologies, heavily used by advanced producer services firms, may neutralize distance as an impediment in locational decision making (e.g. O'Brien, 1992), but the reality of the locational strategies of these service industries seems to imply a new form of concentration in the face of economic globalization (Swyngedouw, 1997; for an early account of the relation between enabling technologies and location strategies, see Goddard and Pye, 1977). The fact that these new forms explicitly offer their global networks to possible clients, underpins that locational decisions are of the utmost importance to students of world cities. In other words, the concentration of these services in specific places should be at the basis of research on the world city network.

WORLD CITIES AS GLOBAL SERVICE CENTRES

Conceptualization

In GaWC research world cities are treated as global service centres, locales where advanced producer services are concentrated for servicing their global corporate clients. From this starting point, the world city network is formally specified as an inter-locking network. Full details of the specification are given in Taylor (2001), here we summarise the argument to provide the rationale for data collection and measurement of connectivity.

An inter-locking network has three levels: a network level, in this case cities connected in a world economy, a nodal level, the cities, and a sub-nodal level, which are the firms providing the advanced producer services. It is at the latter level that world city network formation takes place. Through their attempts to provide a seamless service to their clients across the world, financial and business service firms have created global networks of offices in cities around the world. Each office network represents a firm's global strategy for dispensing its services, it is an outcome of location decision making at the scale of the world-economy. The world city network is therefore defined as the aggregate of the many service firms pursuing a global location strategy. In this way it is global service firms that 'inter-lock' world cities into a network of global service centres: they are the prime producers of the contemporary world city network.

This model can be formally represented by a matrix V_ij defined by n cities x m firms where v_ij is the 'service value' of city i to firm j. Service value is the importance of a city to a firm's office network which depends upon the size and functions of an office or offices in a city. Thus every column denotes a firm's global strategy and every row describes each city's mix of services. This allows for two types of study: a focus upon the columns will inform our knowledge of firms; a focus on the rows will inform our knowledge of cities. Here our analysis concentrates on the latter, understanding the configuration of cities within the data.

Connectivity

The advantage of a precise specification of the world city network is that techniques of network analysis can be used. Using elementary network analysis, the most basic measure of a city is its connectivity in relation to all other cities in the matrix. Drawing on Taylor (2001), for an interlocking network this is defined as follows:

In matrix V_ij, the basic relational element for each pair of cities is r_ab,j = v_aj . v_bj

where r_ab,j is the elemental interlock link between two cities a and b in terms of firm j defined in terms of v, a service value of a firm in a city. These links can be aggregated to produce an inter-city interlock link

r_ab = ∑ r_ab,j

Each city has such an interlock link with every other city. Aggregating all the inter-lock links of a city produces the global network connectivity (G) of the city

G_a = ∑ r_ai

for city a across all i cities in matrix V (a ≠ i).

The limiting case is a city that shares no firms with any other city so that all of its elemental links are 0 and it has zero connectivity. In practice with large data sets the global network connectivities can be quite large numbers. To make them manageable in our use below, we express city connectivities as proportions of the largest computed connectivity in the data thus creating a scale from 0 to 1. These scores will be used below to indicate hierarchical tendencies within our analysis.

METHODOLOGY

Data Collection

Precise specification guides our data collection: data are required on the city office networks of global service firms. This exercise in data collection is described in detail in Taylor et al. (2002a) and will be summarised here as it is the input to our subsequent analyses.

Global service firms were defined as firms with offices in 15 or more different cities, including at least one in each of the prime globalization regions: northern America, western Europe and Pacific Asia. Firms meeting this criterion were selected from rankings of leading firms in different service sectors. The other key criterion was purely practical - whether adequate information could be found on the firm's website. A supplementary criterion was to only consider sectors in which at least ten firms could be included to facilitate inter-sector comparisons. In the event 100 firms were identified in six sectors: 18 in accountancy, 15 in advertising, 23 in banking/finance, 11 in insurance, 16 in law, and 17 in management consultancy.

Selecting cities was much more arbitrary and was based upon previous GaWC experience in researching global office networks. Capital cities of all but the smallest states were included plus many other important cities in larger states. A total of 316 cities were selected.

Data collection focussed on two features of a firm's office(s) in a city: first, the size of office (e.g. number of practitioners), and second, their extra-locational functions (e.g. regional headquarters). The main problem with this type of data collection exercise is that the exact nature of the information collected for each firm differed to that for every other firm. The solution was to standardize the information. Information for every firm was simplified into 'service values' ranging from 0 to 5 as follows. The city housing a firm's headquarters was scored 5, a city with no office of that firm was scored 0. An 'ordinary ' or 'typical' office of the firm resulted in a city scoring 2. With something missing (e.g. no partners in a law office), the score reduced to one. Particularly large offices were scored 3 and those with important extra-territorial functions (e.g. regional offices) scored 4. The end-result was a 316 x 100 matrix V_ij where v_ij ranges from 0 to 5.

Fuzzy Classification

One of the most popular data analysis techniques for studying large data matrices is cluster analysis. Applied to the matrix produced here, a traditional clustering algorithm would compute mutually exclusive clusters of cities, based on the various service values for the world cities. However, this classical approach towards cluster analysis is fraught with various sources of problems (Derudder and Witlox, 2002). First, the world city network does not exhibit a simple hierarchical urban structure (Taylor et al., 2002a). Plotting either sizes of cities (Ettlinger and Archer, 1987) or their connectivities (Taylor et al., 2002a) against their rankings produces neither a ranked size nor a primate distribution as conventionally found in studies of 'national urban systems' (e.g. Berry and Horton, 1970, chapter 3 for a discussion, and Bourne, 1975, chapter II for examples). Neither form of hierarchy is found at the global scale, instead the curve takes the form of the inverse of the primate city pattern. This is empirical support for the argument made in Taylor (2001, p. 192) that world cities constitute a complex network rather than a simple hierarchical structure. Although the two first cities ranks may stand out (London and New York), the rest of the curve shows that this is not a 'double-primate' city pattern (Taylor et al., 2002a). As a consequence, clear-cut patterns, as provided by a traditional clustering algorithm, are unlikely to provide a sensitive specification of patterns in this complex network.

Second, mixing hierarchical and functional patterns augment this inherent complexity. There may or may not be hierarchical patterns within the spatial organisation of individual firms at the global scale (depending on their particular strategies), but when aggregated, the result is a complex network exhibiting multiple hierarchical and regional patterns. An exploration of the world city network should therefore include an assessment of intertwining hierarchical and regional patterns of global competence. Again, it is unlikely that a crisp cluster allocation will provide a sensitive specification of these scrambling patterns.

Third, the main reason for the limitation of previous assessments to the upper and medium rungs of the hierarchy lays in the fact that the outer reaches of the world city network are to be classified based on sparse data, yielding vagueness in any classification based on this data. As a result, minor shifts in the sparse data may yield completely different outcomes, and mutually exclusive clusters are therefore unlikely to be unbiased.

In order to represent the data structures more comprehensively, therefore, we propose to replace the crisp separation of clusters, defined by u_ic ∈ {0,1} for all i = 1,...n and c = 1,...C, by a fuzzy notion, defined by μ_ic ∈ [0,1], where

u_ic is the crisp membership of a city (taking only the values of 0 or 1)

n is the number of cities that need to be classified;

C is the number of clusters;

μ_ic is the membership value of city i in cluster c (ranging from 0 to 1).

A fuzzy classification scheme computes grades of membership in different clusters rather than providing information on mere membership (Sato et al., 1997; Höppner et al., 1999). This approach can reflect the expected complexity of multiple and intertwined profiles in any classification, since it is reflected by hybrid membership in different clusters. Furthermore, previous results have indicated that the sensitivity of a fuzzy clustering algorithm can assess both functional and hierarchical tendencies (Derudder and Witlox, 2002). And finally, minor shifts in the data will be reflected by a minor shift in the resulting classification, reducing the bias problem. However, since classifications based on the presence of a very small number of firms will suffer from this bias, such a very sparse section of the matrix is avoided by restricting the analysis to cities scoring over 20 'points' in the presence of advanced producer firms. This yields a roster of 234 cities for analysis.

Different number of clusters can yield different salient results, and hence there is no firm theoretical basis for selecting the number of clusters from the classification analyses. Here, we will focus on the results for C=22. This is a pragmatic choice after assessing several solutions of different classes. With 22 clusters we find a broad diversity in hierarchical and regional patterns in the world city network, that provides for a particularly insightful interpretation. For the clarity of the argument, the results below are reported in a simplified form that identifies 'core' and 'hybrid' members of clusters.

THE GLOBAL SUPER-RICH AND THE GLOBAL ELITE

The 22 clusters in the fuzzy classification of 234 cities are grouped together in three different ways. First, there is a strong hierarchical dimension to the clusters: cities with similar levels of global network connectivity tend to be classified together. Second, there is a strong regional dimension to the clusters: cities from the same part of the world tend to be classified together. Third, there is a tendency for interaction between these two dimensions: clusters with low average connectivity tend to be more regionally restricted in membership. These geographical features mean that our results show more than clusters in an abstract 'service space', they represent urban arenas in a geographical space that is the world city network.

This is an important interpretation because it indicates that cities are not creating and reacting to a simple process of globalization leading to an overarching world city hierarchy. There is a multifaceted geography of arenas through which cities operate as service centres for global capital. Hence, as well as the commonplace notion that individual world cities represent critical local-global nexuses, there are also urban arenas that represent regional-global nexuses within contemporary globalization.

This new complex global urban geography is shown in Table 1 and Figure 1. The table highlights the hierarchical tendency in the results with arenas listed in terms of average global network connectivity for cluster members. These cluster connectivities are in turn used to denote five bands of arenas to represent the hierarchical tendency around cluster A, which is by far the most important arena in terms of connectivity. The latter is called the 'centre' of the bands for reasons that will become clear when we look at Figure 1. The largest gap in connectivity is between cluster A and band I but all the bands are identified using gaps in the levels of connectivity. To get a feel for the structure and geography of the fuzzy classification, Table 1 shows also the size of each cluster, including overlapping cities, and also the most typical city in each cluster.

The regional tendency in the results is added to the hierarchical tendency in Figure 1: arenas are depicted in their respective bands around the centre and in addition they are located in roughly their geographical position. The latter are articulated about a trans-Atlantic centre of London and New York. Two member cities are shown for all arenas to aid in initial reading of the cartogram. In addition to the centre arena, there are only three other arenas that have strong trans-regional membership, two in band I, and one, perhaps surprisingly, in band IV. This means that 18 of the arenas have relatively clear-cut regional identities thus showing the strength of the regional tendency in these results.

Further interrogation of the results requires a detailed look at the content of the arenas. For each cluster/arena we have searched for four sets of cities.

• The cluster nucleus is made up of those cities with affiliations above 0.7.
• Other singular members are cities with affiliations between 0.3 and 0.7 and with no membership of another cluster.
• Hybrid members are the other members of the cluster who share membership with another cluster. Clusters without hybrid members we refer to as distinctive.
• Near isolates are cities that are members of no clusters (they have no affiliation as high as 0.3) but have their highest affiliation to a given cluster.

As above, we begin by ordering the argument through bands before focussing on the regional patterns.

The Centre and Band I: Leading World Cities

Table 2 shows the four sets of cities for the Centre arena and Band I arenas. The former is simple to the extreme. It consists of a two-city nucleus and nothing else. The Centre arena is 'main street, world-economy'; by far the most important link in the world city network in terms of connectivities (Table 1), it is also a wholly distinctive dyad. This is the justification for terming the arena the centre and so locating it as the pivot of our cartogram (Figure 1).

The Band I arenas are also relatively small and simple. This is especially the case with Arena C which includes just the three US cities that rank below New York. It is a distinctive arena with no hybrids. The other two Band I arenas are cross-regional and link western European cities with cities in other parts of the world. Arena B links Paris and Frankfurt with the leading Pacific Asian cities. Although Sassen (1991) has suggested that Tokyo should be ranked next to New York and London, our analysis does not show this. This is because there is a wide range of business services in our data and Tokyo's global prowess is largely restricted to banking/finance firms. Arena D links other leading European cities with leading world cities outside the other two main globalization arenas (USA and Pacific Asia) in European-settler regions, notably Latin America. These two arenas share Brussels and Barcelona as members. The distinction between the two arenas is the particular dominance of banking/finance services in the arena including Pacific Asia.

The Centre and Band I arenas define the 21 most important cities within this analysis of the world city network. They also suggest specific relations amongst these leading world cities. Beyond the Centre, it is the European cities that appear pivotal in linking to other regions, the leading US cities arena appears as relatively isolated in global service provision.

Bands II and III: Major Regional World Cities

The Band II and III arenas (Tables 3 and 4) are regional clusters of important world cities. There are three classic examples in Band II: Arena E is a distinctive cluster that includes all the important Pacific Asian cities not in Band I arenas; Arena G is a distinctive cluster that includes all important German cities not in Band I arenas; and Arena H is a distinctive cluster that includes all the important eastern European world cities. The latter has a near isolate, relatively unimportant city but appropriately located geographically for this arena, Kiev. Not quite distinctive but otherwise similar to the above arenas, Arena F includes the important US cities not in the Centre or Band I. In this case there is one hybrid member, Miami, which is linked into Arena J in Band IV. The latter is the arena of leading Latin American cities that are not in Band I. Clearly this analysis is picking up the regional articulation role of Miami between the USA and Latin America (Brown et al., 2002).

Arena J also links down the bands to less important Latin American arenas. This non-distinctive structure, sharing cities with other clusters, is typical of the other arenas in Bands II and III. Arena I brings together important Asian cities outside the Pacific Rim but also has links to a lower band arena of Asian cities and to Arena K which is a cluster of important 'outer' European cities. This odd combination combining northern, south-eastern and south-western European cities replicates previous findings based on just European cities (Taylor and Hoyler 2000). Arena M is more distinctive than the others and is clearly a south Asian arena with just two hybrids. Finally Arena L is a cross-regional cluster that covers the old British Commonwealth. This 'cultural' historical throwback arena again replicates previous findings (Taylor et al. 2002b). Membership covers Australian, Canadian, New Zealand and South African cities not found in Band I arenas. Note the dearth of British cities, only three appear as hybrids. They link to a particular British arena in Band V.

With the exception of the latter unusual cluster, these arenas show that below the top echelons of the world city network, important cities tend to be very regional in the focus of their global service provision.

Bands IV and V: Major Cities on the Edge

In Bands IV and V we come to cities that are rarely if ever mentioned as world cities (Tables 5 and 6). This does not mean, of course, that they are not involved in the same globalization processes as the cities we have dealt with above, but they are less intensively connected to the world city network. Given our conclusion concerning Band II and III arenas, we would expect the arenas in these two lower bands to be even more regional in their memberships. And this is indeed the case: there are three European arenas, two each from Asia and Latin America and one from Africa and the USA. The latter, Arena O, is distinctive and large and incorporates all remaining continental US cities if the near isolates are included. The cluster of less important German cities, Arena S, is similarly distinctive but much smaller. It also includes some neighbouring European cities as singular members and near isolates. Both the UK (Arena R) and France (Arena T) have their own urban arenas of less important cities albeit less distinctive in nature than arena S. The UK arena includes the Commonwealth arena hybrids, France includes other neighbouring European singular members, hybrids and near isolates.

The Latin American and Asian clusters are much less clear-cut as geographical arenas. In particular, Arena N and Arena P both include less important Latin American cities from across the region with no obvious geographical division of the region. For instance, central American and Caribbean cities are found in both clusters. In contrast, the Asian clusters, Arena Q and Arena V, have geographical concentrations in west and east Asia respectively. The former thus includes Middle Eastern cities not previously appearing in a cluster and the latter includes almost all the Pacific Asian cities not included in earlier clusters. Perhaps appropriately they share central Asian cities as hybrid members. Arena Q also shares less important outer European cities with arena K. Finally, there is an African cluster: Arena U. All inter-tropical African cities belong to this arena except for Nairobi. Furthermore all these African cities are part of the nucleus or are singular members, the hybrids of this arena are non-African. This reflects the lowly and relatively isolated position of inter-tropical African cities in the world city network.

Arena 'Gaps': Regional and National Geographies

The findings are from a global urban analysis and this is how we have interpreted them above. However, the identification of urban arenas does point to some interesting 'sub-global' conclusions. Specifically we can comment upon how contemporary globalization seems to be impinging on long-established national and regional 'urban systems', the traditional concern of urban geographers studying relations between cities. Figure 1 is particularly informative in this respect.

First we can contrast the location of the arenas featuring cities in Europe's leading three economies. Whereas German cities are featured in three bands (I, II and V), British and French cities are concentrated in two bands (Centre and V for the UK, I and V for France). This clearly shows the different national patterns of cities with Germany's distinctive 'horizontal' city relations compared with the UK and France's 'vertical' city relations. Although, this is by no means a surprising result it is relevant to the workings of contemporary globalization: whereas London and Paris cast an inhibiting 'shadow' over their compatriot cities, Frankfurt has no such effect on other German cities. Clearly this is of vital importance to how each of these national economies relates to globalization processes.

US urban arenas have a similar pattern to German arenas in being represented in three bands (Centre, I, II, and IV). This is perhaps to be expected given the large number of US cities within the data. However this fact makes the lack of a US arena in Band III interesting. There appears to be a gap created in the globalization of US cities between the likes of relatively important cities such as Boston and less important cities such as Baltimore. This certainly implies policy incentives for cities in Arena O to try and 'move up' and create a new US arena in a higher band. More generally, The US arenas are typified by their high levels of distinctiveness. New York, as half of the Centre, and Miami as an important hybrid city linking to Latin America, are the only continental US cities to share arenas outside their own country. This relates to the sheer scale of the US economy and its long-developed, massive market in financial and business services that provides less of an incentive for firms 'to go global' to the same degree as global service forms from other world regions. This highlighting of the ambiguous role of US cities in contemporary globalization is an important result of this research.

Finally the most important result of this research is the light it shines on erstwhile 'third world' cities in the world city network. The main point is that both Asian and Latin American urban arenas are quite well represented across all bands beyond the centre. In the case of Africa, its Arab cities and South African cities feature in relatively important Asian and Commonwealth arenas but the inter-tropical cities all cluster in one Band V arena. The exception is Nairobi which is a hybrid in two Asian clusters in Bands II and IV. Here we have the only signs of an authentic African world city emerging.

CONCLUSION

Conceptualising world cities as global service centres, GaWC has developed a methodology for studying world city network formation. In this paper, we have complemented previous exploratory analyses of the world city network with as main objectives (i) the unravelling of the fuzzy spatial dimensions behind world city formation and (ii) the description of the network at a level of geographical detail never before attempted.

Although we specify world cities as an interlocking network to which we apply a global level analysis, our results clearly reveal both hierarchical and regional tendencies within the data. First, the hierarchical tendencies show, contra Sassen's (1991) famous global trilogy, a dual city arrangement heading the network: London and New York form a distinct arena and Tokyo appears in an arena at the next level of connectivity. Overall, hierarchical tendencies interact with regionality: clusters with low average connectivity tend to be more regionally restricted in membership. These geographical features suggest that our results show more than clusters in an abstract 'service space', they represent urban arenas in a geographical space that is the world city network.

Second, the regional tendencies show the three world-economy core regions - northern America, western Europe and Pacific Asia - to be quite different in their allocations to arenas. In northern America, US cities seem to be typified by (i) high levels of distinctiveness (except for Miami) and (ii) with a connectivity gap created in their globalization. In contrast, European cities are both linked into other continents and cover all levels of arena. Pacific Asia shows a globalization gap larger than the USA's. Arguably the most important result is the light our analysis shines on the non-core ('third world') cities in the world city network. While both Asian and Latin American urban arenas are quite well represented across all bands beyond the centre, and Arab cities and South African cities feature in relatively important Asian and Commonwealth arenas, inter-tropical African cities (except for Nairobi) all cluster in a single arena with a marginal position in the network.

In conclusion, we have tried to enhance insight into globalization through the depiction of a new and detailed geography of the world city network. We have not limited 'globalization forces' to just 'global cities' but have incorporated a very large number of cities into a single global urban analysis. Contemporary globalization is not an end-product in itself but an on-going bundle of processes. This means that the gaps in the pattern that we have identified may be filled in the coming years as connectivity within the world city network intensifies. On the other hand the gaps may widen as global services become more concentrated in fewer cities. We cannot know which of these future scenarios will come to pass but we do know that we will not be able to assess such changes unless we have a good empirical understanding of the contemporary world city network.

ACKNOWLEDGEMENT

This research derives from the project "World City Network Formation in a Space of Flows" for which we acknowledge the support of the Economic and Social Research Council (UK).

REFERENCES

ABU-LUGHOD J.L. (1995) Comparing Chicago, New York and Los Angeles: testing some world city hypotheses, in KNOX P.L. and Taylor P.J. (Eds) World Cities in a World-System, pp.171-191. Cambridge University Press, Cambridge (UK).

AMIN A. and THRIFT N. (1992) Neo-Marshallian nodes in global networks, International Journal of Urban and Regional Research, Vol. 16, 571-87.

BEAVERSTOCK J.V., SMITH R.G. and Taylor, P.J. (1999) A roster of world cities, Cities, Vol. 16, 445-58.

BOURNE, L.S. (1975) Urban Systems. Clarendon, Oxford (UK).

BERRY B.J. and HORTON F.E. (1970) Geographic Perspectives on Urban Systems. Prentice Hall, Englewood Cliffs (NJ).

BROWN E., CATALANO G. and TAYLOR P.J. (2002) Beyond World Cities. Central America in a space of flows, Area, Vol. 34, pp. 139-48.

COFFEY W.J. and BAILLY A.S. (1992) Producer Services and Systems of Flexible Production. Urban Studies, Vol. 29, 857-868.

DALY M. T. and STIMSON R.J. (1992) Sydney: Australia's gateway and financial capital, in BLAKELY E.J. and STIMSON R.J. (Eds) New Cities of the Pacific Rim. Institute for Urban and Regional Development (University of California), Berkeley (CA).

DANIELS P. (1995) The EU internal market programme and the spatial development of producer services in Great Britain. European Urban and Regional Studies, Vol. 2, 299-316.

DERUDDER B. and WITLOX F. (2002) Classification Techniques in Complex spatial Databases: Assessment of Vagueness and Sparsity in a Network of World Cities. Solstice, Vol. XIII, 1-32.

ETTLINGER N. and ARCHER J.C. (1987) City-size distributions and the world urban system in the twentieth century, Environment and Planning A, Vol. 19, 1161-74.

FRIEDMANN J. and WOLFF G. (1982) World city formation: an agenda for research and action, International Journal of Urban and Regional Research, Vol. 6, 309-44.

FRIEDMANN J. (1986) The world city hypothesis, Development and Change, Vol. 17, 69-83.

GODDARD J.B. and PYE R. (1977) Telecommunications and office locations. Regional Studies, Vol. 11, 19-30.

HÖPPNER F., KLAWONN F., KRUSE R. and RUNKLER T. (1999) Fuzzy Cluster Analysis. John Wiley & Sons, Chichester.

KING A.D. (1990) Global Cities: Post-Imperialism and the Internationalism of London. Routledge, London.

KNOX P.L. (1995) World Cities in a World-System. In KNOX P.L. and TAYLOR P.J. (Eds) World Cities in a World-System, pp. 3-20. Cambridge University Press, Cambridge (UK).

KNOX P.L. (1996) Globalization and Urban Change, Urban Geography, Vol. 17, 115-17.

KNOX P.L. and TAYLOR P.J. (1995) World Cities in a World-System (Eds). Cambridge University Press, Cambridge (UK).

LO F. and YEUNG Y. (1998) Globalization and the world of large cities (Eds). UNU Press, Tokyo.

O'BRIEN R. (1992) Global Financial Integration. The End of Geography. Pinter, London.

O'CONNOR K. and STIMSON R. J. (1995) The Economic Role of Cities: Economic Change and City Development, Australia 1971-1991. AGPS, Canberra.

SASSEN S. (1991) The Global City. Princeton University Press, Princeton (NJ).

SASSEN S. (1994) Cities in a World Economy. Pine Forge Press, Thousand Oaks (CA).

SASSEN S. (1995) On concentration and centrality in the global city. In KNOX P.L. and Taylor P.J. (Eds) World Cities in a World-System, pp. 63-78. Cambridge University Press, Cambridge (UK).

SATO M., SATO Y. and JAIN L.C. (1997) Fuzzy Clustering Models and Applications. Physica-Verlag, Heidelberg.

SHORT J.R., KIM Y., KUUS M. and WELLS H. (1996) The dirty little secret of world cities research - data problems in comparative analysis,. International Journal of Urban and Regional Research, Vol. 20, 697-719.

SHORT J.R. and KIM Y-H (1999) Globalization and the City. Longman, New York.

SMITH D.A. and TIMBERLAKE M. (1995) Cities in Global Matrices: Toward Mapping the World-System's City System. In KNOX P.L. and Taylor P.J. (Eds) World Cities in a World-System, pp. 79-97. Cambridge University Press, Cambridge (UK).

SWYNGEDOUW E. (1997) Neither Global nor Local: "Glocalization" and the Politics of Scale. In COX K.R. (Ed) Spaces of Globalization - Reasserting the Power of the Local, pp. 137-166. Guilford, New York.

TAYLOR P.J. (2001) Specification of the World City network, Geographical Analysis, Vol. 33, 181-94.

TAYLOR P.J. and HOYLER M. (2000) The spatial order of European Cities under conditions of contemporary globalization, Tijdschrift voor Economische en Sociale Geografie, Vol. 91, 176-189.

TAYLOR P. J. and WALKER D.R.F. (2001) World cities: a first multivariate analysis of their service complexes, Urban Studies, Vol. 38, 23-47.

TAYLOR P.J., CATALANO G. and WALKER D.R.F. (2002a) Measurement of the world city network, Urban Studies, Vol. 39, 2367-2376.

TAYLOR P.J., CATALANO G. and WALKER D.R.F. (2002b) Exploratory analysis of the world city network, Urban Studies, Vol. 39, 2377-2394.

WALLERSTEIN I. (1979) The Capitalist World-Economy. Cambridge University Press, Cambridge (MA).

WOOD P. (1991) Flexible accumulation and the rise of business services. Transactions of the Institute of British Geographers, Vol. 16, 160-172.

NOTES

^* Department of Geography, Ghent University, Belgium.

^** Department of Geography, Loughborough University, UK.

^*** Dipartimento di Sociologia e Scienza Politica, Università degli Studi della Calabria, Arcavacata di Rende, Cosenza, Italy.

Table 1: Bands of Arenas in the World City Network

CLUSTER/ ARENA	AVERAGE CONNECTIVITY	BAND	NO. OF MEMBERS*	TYPICAL CITY**
A	0.988	Centre	2(0)	London
B	0.613	I	7(2)	Frankfurt
C	0.574	I	3(0)	Chicago
D	0.539	I	11(2)	Amsterdam
E	0.438	II	8(0)	Bangkok
F	0.401	II	5(1)	Atlanta
G	0.384	II	4(0)	Berlin
H	0.379	II	6(0)	Warsaw
I	0.371	II	9(5)	Istanbul
J	0.297	III	7(4)	Caracas
K	0.297	III	12(8)	Copenhagen
L	0.231	III	23(6)	Adelaide
M	0.225	III	12(2)	Calcutta
N	0.201	IV	9(4)	Montevideo
O	0.193	IV	23(0)	Baltimore
P	0.180	IV	16(10)	La Paz
Q	0.179	IV	19(12)	Kuwait
R	0.158	V	14(5)	Leeds
S	0.157	V	8(0)	Dresden
T	0.148	V	13(3)	Lille
U	0.141	V	22(8)	Accra
V	0.121	V	13(3)	Osaka

*membership is defined as affiliation of 0.3 and above, figures in brackets refer to hybrid cities with membership of other clusters.

** member with the highest affiliation

Table 2: Centre and Band I Arenas

Table 3: Band II Arenas

Table 4: Band III Arenas

Table 5: Band IV Arenas

Table 6: Band V Arenas

Figure 1: Urban Arenas of the World City Network

Edited and posted on the web on 19^th June 2002; last update 6^th May 2003

Note: This Research Bulletin has been published in Regional Studies, 37 (9), (2003), 875-886