GaWC Research Bulletin 6

Although the lineage of world city studies can be traced back to well before 1980, it is only in the last two decades that a concerted research effort has emerged. This development has been intimately related to new ways in which social scientists have conceptualized the world-economy, first as 'the new international division of labor' and subsequently as 'globalization'. Specifically, the emergence of advanced computer-communication technologies has been emphasized for creating powerful control capabilities in a new global space of flows. The most explicit example of these developments has been in financial markets. This has led to important work on cities as 'international financial centers' (Cohen 1981; Thrift 1987). A second strand has taken a broader view of what is happening in the world's major cities. John Friedmann (1986; Friedmann and Wolff 1982) has identified a process of 'world city formation' of which financial markets are only one, albeit important, aspect. We follow Friedmann's lead in adopting a more holistic approach to the study of today's major cities.

A key question to be answered in contemporary urban studies is why under conditions of instantaneous communication, which would seem to favor decentralization or even homogenization, there is still a functional need for large world cities. Answering this question has produced two distinctive schools of thought, one specifically 'urban', the other keeping the 'network' to the fore. These are best represented by the work of Michael Storper (1997) and Saskia Sassen (2000; 2001) respectively. For Storper the new enabling technologies-he calls them new 'metacapacities'-have not resulted in dispersion, rather particular concentrations of economic activities have developed. He argues that the contemporary world-economy is characterized by a reflexive economics in which to be successful entrepreneurs have to be part of a knowledge-rich, continual learning process. Such processes can only be collective and very specific to different places: they result in local assets, which are difficult to duplicate elsewhere. He uses Hollywood as an obvious example of such processes. Cities enter his argument as privileged sites of such reflexivity, dense networks of learning practices, which steer the contemporary world-economy. This urban theory provides a very good explanation of why cities are prospering as places, but has little or nothing to say about networks of cities. For this we have to turn to Sassen's identification of global cities. In an argument similar to the concept of economic reflexivity she identifies the production of advanced producer services as the distinguishing characteristic of contemporary world cities. Narrower in conception to Storper's definition-reflexivity is to be found in industrial production as well as services-Sassen's global cities are, nevertheless, much more than places where large numbers of financial transactions take place. Advanced producer services provide world-wide assistance to global capital but they are much more than that; they are creative in making new products in finance, law, accountancy and so on, which require the rich knowledge base in cities which Storper describes. The key difference is that for Sassen these cities are part of a network of strategic locations, hence the learning has to transcend particular places. The network of world cities, and not just single cities, is intrinsic to the process. We follow Sassen's lead here by focusing upon world city network formation.

The Globalization and World Cities (GaWC) Research Group and Network is a real and virtual organization¹ dedicated to the study of inter-city relations under conditions of contemporary globalization. We introduce GaWC in this paper in two stages. In the first section we describe the need, origin, rationale and operation of GaWC. In the second section we present a selection of results from GaWC research projects culminating in a preliminary quantitative presentation of a world city network.

OVERCOMING EMBEDDED STATISM

The most familiar of all world maps is that which depicts the sovereign states of the world. This mosaic of territories defines a world of boundaries, emphasizing national differences while totally ignoring spatial connections. This omission of the space of flows from a basic popular cartography is much more than a geographer's problem. It reflects a taken-for-granted-world image, which prioritizes nation-states as the basic units of humanity. As such its influence far transcends the map which hangs on all geography classroom walls. We describe this as embedded statism, a geopolitical feature of the modern world, which entails viewing social relations through the distorting lens of state-centrism. GaWC is part of a research program, which is dedicated to overcoming embedded statism.

This section is divided into three parts, which develop an argument from quite abstract considerations through to specific measurement questions. Embedded statism can be described as the dominant metageography of contemporary society: in the first part we set this in the context of alternative metageographies. However today, state-centrism is so pervasive that even the information we have to understand our world is indelibly linked to a metageography of states. Part two introduces a research initiative that attempts to go 'beyond stat-istics'. Loughborough University's Globalization and World Cities (GaWC) Study Group was set up to develop a web site that provides data alternatives to counter the inherent bias in the usual data sources. In the final part we focus on the data problems to be found in world city research as a prelude to the second empirical section of the paper.

Alternative Metageographies

During one of the early Apollo flights one of the astronauts experienced a fundamental revelation: looking back to Earth he noticed there were no political boundaries. In fact it is often argued that the Great Wall of China is the only human feature on the Earth's surface which can be seen from space. This is not strictly correct; once day turns to night the archipelago of modern human settlements is visible from the artificial light, which is created. And this shows a pattern dominated by the great cities of the world. The apocryphal visitor from another planet would certainly get the impression of a city-dominated civilization on Earth. For the people living on Earth, however, it is the political boundaries, which loom large in their geographical imaginations.

Metageographies are the basic large-scale spatial frames by which people order their world (Lewis and Wigen 1997). These exist at different scales and may be vague in conception but are no less important for that. Hence broad dualities such as 'North-South' and 'East-West' have been crucial to recent geopolitical positions but have not fully replaced the traditional division into 'continents'-the geographical entity 'Africa' currently exists as a social construct with largely negative images for instance. However, underpinning all such representations there is the map of sovereign states, the multi-colored wall map which, despite recent political upheavals, seems to continue to exude an almost 'natural' aura (Jackson 1990): this is the metageography behind embedded statism. This is because territorial sovereignty is the basic building block of the politics of the modern era (Taylor 1999a). It is a metageography rooted in bounded spaces. Globalization is a direct challenge to the dominance of such territorial thinking.

Geographical boundaries are never completely sealed. There are links and connections between areas, which will vary by circumstance. These define a space of flows to exist alongside the space of territories. Such a space of transnational transactions has always been crucial to the development of the modern world-system (Arrighi 1994) but it has not figured prominently in modern peoples' geographical imaginations. Globalization is changing this as it provides an alternative metageography. This is best represented by Manuel Castells (1996) with his work on the 'network society'. He describes a space of flows existing at several levels starting with the basic electronic infrastructure; the world city network represents one of the higher levels of spatial organization. It is the latter that has the potential to create an alternative metageography and which is our concern here.

In summary, we juxtapose our alternative metageography of a network of world cities-a space of flows- against the dominant, conventional, metageography of nation-states-a space of territories (Taylor 2002).

Beyond State-istics: GaWC

Despite its critical credentials, social science has not escaped embedded statism (Taylor 1996). In fact it can be argued that the reason for the success of the three core social sciences-economics, political science and sociology-in the twentieth century is that they have each met the policy reform agendas of states. The social sciences are both the creations and, thus have become, the creatures of states. This has become particular evident when their state-centric theories have been confronted by trans-state processes of globalization (Taylor 1996; 1997). Quite simply, theory begets data; data begets theory. This elementary synergy makes breaking out of state-centric thinking a difficult undertaking. We have chosen world cities with their associated flows as the route 'beyond stat-istics', hence the creation of GaWC.

When Riccardo Petrella (1995), the 'official futurist of the European Union', was asked recently to speculate about the near future he chose to contrast 'two mental maps of the world system' in which major city regions (our world cities) would have either a positive or negative impact by the year 2025. The details of his predictions need not concern us here but what is of interest is the fact that neither prediction is state-centric, both are city-centric. It is entirely appropriate that this metageography should be coming to the fore in this continent where states first subsumed cities in the construction of modern capitalism (Tilly 1990; Taylor 1995). The irony is that the end-result of a European Union of states is likely to be a new 'Europe of Cities'. Hence although the launch of the euro was seen as a pooling of state financial sovereignties, perhaps its most important long-term effect will be the setting up of the new European Bank in Frankfurt, not London. 'London versus Frankfurt' as Europe's leader within global finance and banking is just the sort of issue, the type of world, which GaWC has been set up to study.²

GaWC is organized around three major related activities. First, we are a research group based around a collection of projects about world cities. Currently projects have been Loughborough-based but we are actively working on new projects, which will involve research partners from other universities. Ultimately this type of 'global research' will require multiple site studies so as to create the necessary broad knowledge base. Currently our projects have been focused on London (to meet British ESRC 'national efficiency' concerns) with comparisons reaching New York and Singapore. Plans are in development for further work on Amsterdam, Caracas, Chicago, Frankfurt, Hong Kong, Istanbul, Miami and Shanghai. Second, beyond our research partners, we aspire to be the world clearing house for world city research. In this we take advantage of the speed and flexibility inherent in GaWC's electronic existence. In the slow world of conventional article publication, with two years between initial writing and publication not uncommon, we offer an electronic outlet of immediate issuing as 'GaWC Research Bulletins'. In this electronic publication series papers are made available in early form (as submitted to journals, for instance) and then as regularly revised through to the final version, which may still have to wait more than six months for hard-copy publication. The latter now represents only 'the final solid record' of the paper, as it were.

The third major GaWC activity is also its initial raison d'être: the production and development of inter-city data. This derives from our previous arguments; for instance we have much information on relations between France and the UK but relatively little on relations between Paris and London. Data comes from three sources. First, the data from all our projects is automatically posted in our data bank. Second, we invite researchers with inter-city data to post it on our site or else allow us to point to it from GaWC. Third, we are proactive in trying to stimulate the generation of inter-city data. This has involved creating standardized data collection procedures for use in graduate research projects so that generated data is comparable across case studies (Beaverstock et al 2000). The end result is a data bank accessible to anyone across the world with Internet technology. It is therefore public data but not state-centric data. As with all other information of this type, we promote its use in research and teaching and ask only that appropriate acknowledgement be made.

In summary, GaWC aspires to be the virtual center for the study of relations between world cities under contemporary conditions of globalization. Further details of activities listed above can be found at http://www.lboro.ac.uk/gawc/.

The Paradox of World City Research

Before we look at some of the early results of GaWC projects, it is useful to say a little more about the research niche we have occupied. Our starting point is the identification of a curious paradox in the literature on world cities. This can stated simply: whereas the essence of world cities is their relations one to another, researchers have generally not focused on this aspect of their being.³ Studies of world cities belong largely to two types: case studies and comparative studies. Whether looking at the one or the few, such studies leave relations between cities as either assumed or asserted. This is a specific case of the data problem described above: researchers can use urban data from national census returns to describe and analyze patterns within cities but there is little information on trans-state inter-city relations. Ipso facto, inter-city relations are neglected hence the paradox we began with.

Smith and Timberlake (1995) have constructed a typology of inter-city linkages based upon the form (human, material, information) and function (economic, political, cultural and social) of flows. Twelve types are thus identified but the authors have to admit that they constitute little more than a 'wish list' for world city research. There is some inter-city flow data available, for instance on airline traffic and telephone calls, but these tend to be very general in nature so that their relevance to world city formation is compromised. For instance, scheduled flights into Miami includes the holiday and the seasonal retirement market as well as the economic growth associated with the city's role as an important center in the world city network. For the most part, until GaWC is fully developed, researching inter-city relations means collecting and creating new and original sets of data.

The research we have been funded for thus far has been within an area of study known as the new economic geography of services. This attempts to understand the growth of services in the contemporary economy through studies of service firms. We combine this with Sassen's insight that it is advanced producer services, which particularly distinguish world cities to turn the focus from firm to city: firms are our subjects, cities are our objects of study. In practice this means collecting information on the office network of producer service firms. Although this does not directly measure flows between cities, a firm's global strategy -in which cities it chooses to set up branch offices-must imply both control and information connections. Aggregating large numbers of such strategies across several services will, therefore, indicate inter-city relations. This is what we report upon in the next section.

In summary, world city research is a classic example of state-centric data distorting research output: the dearth of inter-city data causes a neglect of inter-city analysis. This is GaWC's niche.

GLOBAL STRATEGIES, WORLD CITIES

This section is a review of early findings of some GaWC research. It is intended to show the empirical potential of GaWC; given the neglect of inter-city studies, some relatively elementary measurement exercises can provide new and original insights into world city network formation. Hence the emphasis below is upon empirical results to the detriment of both details of production of the data⁴ and full reflection on the theoretical implications of the research.⁵

The data we employ consists of the distribution of offices of producer service firms across cities. Hence we are looking at the location strategies of firms under conditions of globalization. However, although firms are our subjects, it is cities, which are the objects of our analyses. We report on six separate analyses, with a core data set of 69 firms (5 accountancy, 14 advertising, 11 banking/finance and 39 law) across 263 cities for 1997/8. The choice of firms is that they are global producer service firms, otherwise the choice is purely arbitrary in terms of data availability (i.e. we have been able to find information on their complete global network of offices). The numbers in each sector reflects degrees of corporate concentration from the 'Big Six' in accountancy to law as the least 'globalized' service. Data collected varied by firm from simple presence/absence in a city to information on the relative importance of offices, notably in terms of how many practitioners in a firm are based there. Not all analyses use the same information from the data which will become clear as we proceed.

The GaWC Inventory of World Cities

There is no agreement as to the roster of world cities. There is obvious agreement for the leading cities (London, New York, etc.) but no clear cut off point as to when a city should not be considered a world city. The most commonly referenced rosters are those of Friedmann (1986; 1995) but these are not based upon any substantial or systematic analysis of the evidence. Is Lisbon a world city? Is Kuala Lumpur or Mumbai? It depends who you read (Beaverstock; Smith and Taylor 1999, Table 2). Hence the first task of GaWC has been to create a preliminary roster of world cities, empirically-based on our initial data collection. This is carried out in two stages. First, we estimate the global capacity of cities separately for each of the four services we have data on. We use information on the presence, size and roles of offices in cities across the world. Second, we aggregate these sector results to define levels of 'world city-ness'. We designate 55 cities to be world cities and find another 67 cities showing evidence of world city formation.⁶

The first stage consisted of finding global service centers at various levels for the four sectors. In order to avoid any idiosyncratic feature of a single firm (i.e. additional offices in their 'home' country), for each sector only cities in which at least two firms are represented were considered to be global service centers. This resulted in 78, 67, 68 and 72 cities for accountancy, advertising, banking/finance and law respectively. With much overlap between sectors this produced 122 cities as candidates for world city status. For each service sector three levels of presence were identified-prime, major and minor-on the basis of size and importance of offices. Thus each of the 122 cities can be scored 0 (not qualified) to 3 (prime center) for each sector. Sums of these scores produces a figure which indicates a city's 'world city-ness' up to a maximum of 12. For instance, London is a prime center for all four sectors and thus scores 12; Wellington, New Zealand scores just 1, featuring only as a minor center in the advertising data. Table 1 shows all 122 cities from London to Wellington allocated to the twelve levels of world city-ness.

The division of these cities into different classes of world city formation has been carried out using simple logical criteria. Any city scoring 10 or above must feature as a service center in all four sectors. Furthermore, if it is a minor center in one sector, this would have to be compensated for by it being a prime center in the other three sectors. In addition it must be prime in at least two sectors, and the other two would have to be major designations. Ten cities qualify in this top class which are designated alpha world cities. As we might expect, there are no surprises at this end of the distribution with 4 cities from western Europe, 3 from the USA, and 3 from Pacific Asia.

For the second class of world cities, we identify any city scoring 7 to 9. Such a city must be a global service center for at least 3 of the four service sectors and must be a prime or major center in at least two sectors. Ten cities qualify as beta world cities. The same three world regions are represented as for prime world cities but with 'outer' cities appearing such as Sydney, Toronto, Mexico City and Moscow. In addition, a third world region appears: Sao Paulo of South America.

As mentioned previously, it is at the bottom end of the scale where uncertainty reigns for designating world cities. We have decided to define cities scoring 4 to 6 as gamma world cities. With these scores, all cities must be a global service centers for at least two service sectors. This definition catches 35 further cities again distributed largely across the three main regions but with another 3 representatives of South America. Africa has its first city in our list, Johannesburg, but there are still no world cities found in South Asia or the Middle East (if we count Istanbul as European). The remaining 68 cities are designated as having evidence of world city formation processes but the evidence is not strong enough to really call them world cities. The interesting cities are those 12 scoring 3 and here we find one Middle East city, Tel Aviv, and two south Asian cities, Mumbai and New Delhi, which perhaps, signify the early stages of filling in the voids on the global world city map (see Figure 1).

In summary, this is the first time at roster of world cities has been identified and classified on the basis of a systematic analysis of empirical data.

The Question of Regions and Hierarchies

As originally suggested by Friedmann (1986), many world cities seem to operate by articulating their national economy into the world-economy. But there are others which have a wider regional role. Perhaps the most obvious example of this is Singapore, which is a city-state and the most important world city in south east Asia. Hence between the global headquarters of producer service firms and the national economy articulators there are regional world cities. Data for 11 of our firms (2 in accountancy, 2 in advertising, 5 in banking/finance and 2 in law) provide some preliminary evidence for this aspect of world city network formation.⁷

As before, we avoid idiosyncratic results by including only cities which feature in at least two firms' regional organization. Despite being based upon relatively little data, the results are quite stark: we find only 9 cities with regional responsibilities in our data. Of these there are 5 that dominate: London has a total of 9 regional responsibilities, Hong Kong, Miami and Singapore have 6 each, and New York has 5. Discussion will be based around them.

London. At the regional level London operates as the main center for both Europe (with only Brussels as minor rival), sometimes including the Middle East, and as extra-mural center for Africa (with Paris [for francophone Africa] and Johannesburg as minor rivals).

New York/Miami. There seems to be a simple division of labor here: where there is a single center for the Americas it is New York, where there is a division New York is the North America center and Miami appears as the center for Latin America. Hence Miami operates as an 'extra-mural center' with respect to Latin America in the way London does for Africa.

Hong Kong/Singapore. This is the example with explicit rivalry. Although these two cities are centers for their own regions, north east and south east Asia respectively, when no such division is made (for instance as in Asia or Asia-Pacific office designations) Hong Kong edges out Singapore only by 3 to 2 for the larger regional responsibility. It should be noted that where Tokyo appears as a 'regional headquarters' its region is limited to just Japan; this city seems not to have developed responsibilities beyond its own state.

Our preliminary results are clear here: the world city network is regionalized into three 'pan-regions' with few smaller world cities emerging as regional centers (see Figure 2). The great exception is Miami.

In summary, this is the first time the world hierarchy of world cities has been as defined and described by world regional organization (rather than the simple ranking of cities to be found in the literature).

Detecting Concentration Processes: The 'Enhanced Tail' Feature

Our full data set records 2,925 presences of firms across the 263 cities. Such large sets of data lend themselves to elementary distribution analyses. This involves creating a frequency distribution showing the numbers of cities housing different totals of firms. Such distributions can be modeled as probability functions to provide insights into the empirical pattern.

The frequency distribution for this data is shown in Figure 3. One very noticeable feature is the long right-hand tail indicating many cities housing relatively large numbers of firms. This can be compared to a random allocation of firms to cities that would generate a Poisson probability distribution. In such a theoretical distribution no cities would be expected to have more than 15 firms which means our empirical distribution has a very enhanced tail: 65 cities have more than 15 firms. This indicates that there are exceptional non-random concentration processes operating to produce such distribution. This is not surprising, of course, in an urban analysis where some form of hierarchy is the norm, but the length and size of the tail in this case will reflect the particular concentration processes which Sassen (2000) argues are operating under contemporary conditions of globalization.

All the global firms we are dealing with are 'London-located global firms' (LLGFs). Looking at the data from the perspective of London does produce interesting results. For instance, if the particular numbers of firms are converted into percentages of the maximum (69), then we can interpret the results as the chances of a firm in a given city having a direct (intra-firm) office link to London. For instance, Hong Kong has the highest total of LLGFs (52) in our data which translates into a 75% probability that if a client enters an advanced producer service office in Hong Kong, there will be a direct organizational link to London. The top ten cities in terms of these probabilities are shown in Table 2. No surprises here, all the alpha world cities appear except Chicago and they are joined by Brussels and Moscow, two cities with obvious but different appeals to London based firms (EU headquarters, and center for post-Soviet transition respectively). This points to further London-centric analysis but before we present this, it is instructive to look at the service-specific distributions of offices.⁸

The frequency distributions of 262 cities (i.e. not including London) in terms of how many LLGFs they house are shown for each sector in Figure 4. In these diagrams accountancy differs from the other three services in having its mode at its maximum frequency: all five accountancy firms are to be found in 136 of our cities. This distribution reflects the fact that accountancy is the most globalized of producer services. Although this is not a surprising result, the stark contrast with the other services is instructive. The three other services have their mode at zero presence although the dominance of this mode decreases in the order law-banking/finance-advertising which suggests an ordering in their respective levels of globalization with law the least developed. Again not a surprise but this has never been illustrated so directly before.

Most of our data contains much more information than merely presence as analyzed above. In order to retain as much information as possible we have created an ordinal-level set of measures by combining all the different types of information we have.⁹ Since the data combined to ordinal level maintains more information, analysis of this data will be more instructive. We begin the analysis, as before, with frequency distributions for all 262 cities for each service sector (Figure 5). Again the accountancy distribution is distinctive but, compared to simple presences, it is much more suggestive with its central mode and two-tails. The other three distributions continue with their zero modes and enhanced tails but the cities are more widely dispersed compared to simple presences.

The tails of all four distributions have been abstracted from the distributions in this analysis. For advertising, banking/finance and law turning points for the enhanced tail are used; for accountancy the abrupt change in declining levels is used as break point. All scores are converted to percentage of the maximum possible to facilitate comparison. The lists of cities with high levels of connectivity to London are shown in Tables 3 to 6 for accountancy, advertising, banking/finance and law respectively. This analysis shows New York as London's main 'partner' in all sectors as might be expected but there are noteworthy contrasts in degree of dominance and which other cities are at the top of the rankings. For instance, accountancy produces the only table where New York shares top spot with other cities. In banking/finance New York is dominant and is followed by three Pacific Asian cities (Singapore, Hong Kong and Tokyo). In advertising and law New York's dominance is similarly clear; in law, the important political capitals of Washington, DC and Brussels feature prominently.

In summary, this is the first time a comprehensive analysis of cities has detected concentration processes and the cities where they are manifest.

The GaWC Index of London's External Relations

Focusing on just the other 54 world cities, the levels of linkage to London are brought together in Figure 6 where average scores across the four producer services are computed to define the GaWC Index of London's External Relations (Beaverstock, Smith and Taylor 2000). The 54 world cities (i.e. other than London) are arrayed in five groups. The overall dominance of the London-New York link, with score out on its own at 87, is designated as London's 'prime link'. The next two highest links (Paris-London at 68 and Hong Kong-London at 64) each connect London to the other two major world regions and are designated 'major links'. Other connections scoring over 50 then become 'important links' with two lower levels of links also identified. The clear message of this diagram is worldwide scope of London's linkages and this is neatly confirmed by a regional analysis.

The 54 world cities fall neatly into 5 world regions plus an additional British Commonwealth non-contiguous class of cities: 16 are in Western Europe, 12 are from Pacific Asia, 11 are from the USA, and there are five each from Eastern Europe, Latin America, and the old Commonwealth (Australia, Canada, South Africa). Average levels of connection for these groups of cities are shown in Table 7. Perhaps the lack of major variation in average scores is the main feature of this table, which confirms London's worldwide reach. The highest score for the old Commonwealth indicates longer-term linkages than contemporary globalization but the three main 'globalization arenas' (Western Europe, Pacific Asia and the USA) have very similar scores. This shows that London is much more than just 'Europe's representative', articulating Europe into the world-economy, in a 24-hour globalizing financial market. The geographical reach of London confirms its location in the contemporary space of flows as truly, using Sassen's (2001) terminology, a global city (Beaverstock, Smith and Taylor 2001).

In summary, this is not only the first time London as a major world city has been analyzed comprehensively in terms of its linkages with other world cities, it is the first time this type of analysis has been done for any world city.

Local Mixes of Firms across World Cities

In the final two sections we report on on-going work that uses a slightly different data set, one which focuses upon the larger firms. We have identified 46 firms which have offices in 15 or more different cities. The main difference from the previous list of firms is the removal of many small London-based law firms and the addition of three US law firms without London offices. In addition this data is limited to just the 55 world cities previously identified. The result is a 55 x 46 data matrix with each cell scoring a firm's position in a city from 0 (no presence) to 3 (major presence).

The simplest analysis of such data is to correlate the cities in terms of how similar they are in terms of the mix of firms they house. The higher the correlation between two cities, the more firms they share and with similar levels of presence. To illustrate this form of analysis, Table 8 shows correlations between the alpha world cities; the higher correlations are displayed in Figure 7. The two highest correlations are between London and New York and between Hong Kong and Singapore which is what we might expect although no such analysis has previously been shown this. However, when we look at the overall pattern of the higher correlations, some interesting features appear (Figure 7). Certainly the former pairing seems more isolated than the latter. This implies the global cities of London and New York have between them quite distinctive mixes of firms.

A full analysis of the data matrix involving all 55 cities has been carried out using a principal components analysis. This uses the correlations to define the main dimensions in the data, in this case grouping cities in terms of their mixes of firms. This work is in progress¹⁰ but preliminary results show a complex pattern of inter-regional, regional and city dimensions. For instance, London and New York appear as a separate 'global city component' replicating Figure 7, and two other cities define their own individual components: Moscow and Miami ('gateways' to post-Soviet economies and Latin America (see Figure 2) respectively). Other components highlight major regions such as Latin America, Pacific Asia, western Europe, eastern Europe and the USA. In fact the locational clusters in these results show that even among large firms, location strategies under conditions of globalization seem to be more regional than world-wide in nature.

In summary, this analysis is the first to group world cities into types in terms of the firms they house. Classification is a necessary beginning to understand world cities beyond hierarchies.

A Preliminary Description of a Network of World Cities

Assuming intra-firm knowledge and information flow in servicing clients is greater than inter-firm flows, the previous analysis can be used to infer a network pattern. This is no doubt the reason for the regional findings and also some elements of hierarchy which are found in the structures (Taylor and Walker 2001). But there is a way of going beyond such inferences to a more direct measure of inter-city relations. This relates to some of our previous London analysis, which we extend to all cities.

From each city, we can compute the total level of intra-firm service to each other city. Currently this has been carried out for just alpha cities and using only presences (Beaverstock et al 2000). Table 9 is an asymmetric matrix showing probabilities of connections between cities. Each cell contains the percentage probability that a firm in city X will have an office in city Y. Thus, Table 9 shows that if you do business with a Chicago-based firm, then there is a 91% probability that that firm will also have an office in Frankfurt. On the other hand, go to a Frankfurt-based firm and the probability of it having an office in Chicago is only 66%. These probabilities of connection define vectors which are shown for two different levels: prime vectors highlighting the dominance of London and New York, and secondary vectors in which the Pacific Asian cities feature prominently (Figure 8). Notice the marginal positions of Los Angeles and especially Chicago. We describe this as the New York shadow effect: it seems that many firms who do business in the USA focus upon just one single office in New York (Beaverstock et al 2000). This is unlike other globalization arenas which are more politically fragmented necessitating firms to locate in cities in different countries. This may have profound effects on US cities in world city competition.

The end result of this research will be structural analyses of directed graphs for all 55 cities; work is in progress.

In summary, we have presented a preliminary glimpse at a world city network defined by office relations between cities. This will be the culmination of this strand of research, until work such as this is completed we do not even have the basis for understanding world city network formation.

CONCLUSION: FIRST STEPS ONLY

The fact that every summary in the empirical second section of the paper has been able to claim 'firsts' in what are very basic and ultimately elementary empirical analyses provides very strong support for the arguments in the first section of the paper on the dearth of trans-state data and analysis. All that is being reported is the barest skeleton of a global structure, but one not previously revealed despite the two decades of world city research. It is an indicator of the difficulty of 'doing global research' in the social sciences that these first steps have been so delayed. Hence the need for an institution like GaWC.