GaWC Research Bulletin 7

"The first problem with London is to define it. London has never taken kindly to attempts at delimitation ..."

"The city (of London) can still be seen as at the heart of an empire, but now it is an empire of financial information that stretches around the globe"

"London was, as it remains, exceptional. More now even in than in the days when it was the centre of a world empire, the place remains an oddity. The British and English capital is, increasingly, a city which belongs to neither country but to the world"

INTRODUCTION

This chapter reports on an empirical research project that investigates globalization through the study of world cities. This makes the project unusual because most empirical descriptions of world-wide patterns rely on data from states and thus produce country by country map distributions. By focusing on cities, we believe that we can get closer to the trans-state processes that are at the heart of the concept of globalization. This is not to argue that states are no longer important but rather that states are implicated more in 'international' (i.e. inter-state) processes. The latter are premised upon territorial logics that should not be conflated with the network logics which dominate globalization. The world city network is a prime site for trans-state processes and here we focus on one of its leading players, London.

The argument proceeds in four stages. First, we explicate globalization in terms of Castells 1996) 'space of flows'. He defines globalization as a 'network society' with world cities as crucial nodes at one level in the new global space of flows. The world city network is explored as a framework for understanding the space of flows. Using the seminal works of Friedmann (1986) and Sassen (1991), we treat world cities as global service centres providing the necessary services to enable capital to operate globally. Second, we look at London as a world city: it is commonly depicted, with New York and Tokyo, as part of a triad of 'global cities' at the apex of a world city hierarchy. Despite the large literature on the city, we note that there is no measurement of London's relations with other cities. Fourth, we consider how London's connections with other world cities can be measured. Following on from the definition of world cities as service centres, we develop a measurement methodology based upon the office networks of corporate service firms. Fourth, we report on the results of such measurement to show London's position among world cities for both particular services and in aggregate. We claim these to be unique results in showing the inter-city relations of London or any other world city. Finally, in a short conclusion we relate this 'ego-centric' study to a broader agenda in world city network research.

GLOBALIZATION AND WORLD CITIES

Globalization incorporates two fundamental spatial features. First, and most obviously, globalization defines a specific geographical scale. There is the notion that contemporary society is unusual, if not unique, in the quality and quantity of social processes operating at the global scale. Second, there is the more subtle notion that contemporary society is restructuring in such a way that spaces of flows are coming to dominate spaces of places. There is the notion that globalization heralds a new spatial logic. This idea is most associated with Castells' (1996) network society and we follow his argument here.

The new global space of flows, as described by Castells, consists of several layers, two of which are pertinent here. The infrastructure network consists of the hardware and software that electronically connects the world into a global cyberspace. It emerged in the 1970s with the combination of two 'high tech' branches of industry, communications and computing. World-wide telephone communications had been developed via cables and satellites in the 1960s but by combining this communication with the power of computers a qualitative change in the intensity of connections was instituted. With the instantaneous transmission of masses of information, public and corporate 'global knowledges' could be constructed changing a wide range of behaviours. This has culminated in the World Wide Web with its e-commerce, e-marketing and teleports for world-wide conferencing. In particular, this electronic revolution has enabled the dispersion of some activities and has resulted in the concentration of others (Sassen, 1994). The basic point is that this new technology has made possible a new spatial restructuring of society at a global scale.

The world city network is part of a second level of Castells' space of flows. This is the level of new social organizations that the infrastructural level makes possible. These consist of 'hubs' of activities linked together in their workings through electronic linkages. For instance, there are new networks of medical researchers that connect leading universities and hospitals across the world, new global 'invisible colleges' of trans-state collaborations. World cities constitute the most conspicuous example of such hubs and connections.

Despite this incorporation into Castells influential work, the literature on world cities is relatively neglected in most discussions of globalization (e.g key texts such as Waters, 1995; Hirst and Thompson, 1996; Held et al., 1999; Scholte, 2000; and in the only 'Reader' on the subject compiled by Lechner and Boli (2000), cities take up one out of 54 chapters). This is a real pity because in the last decade and a half the 'world city hypothesis', and the related concept of 'global city', have provided an exciting alternative framework for understanding contemporary global change (see for example, Friedmann & Wolff, 1982; Friedmann, 1986, 1995; Knox, 1995; Knox & Taylor, 1995; Feagin & Smith, 1987; Sassen, 1991, 1994a,b; Smith & Timberlake, 1995). Cities as hubs or nodes have been conceptualised in two different ways. Initially Friedmann (1986) termed them 'command and control centres' since they housed the headquaters of many multinational corporations. For Friedmann cities were the nodes of organization for the new international division of labour (i.e. industrialization of many parts of the 'Third World'). However, although many major firms were headquartered in the leading cities, this was not necessarily due to contemporary needs, headquarter locations tend to be dependent on historical factors, notably a firm's geographical origins. In contrast Sassen (1991) focused upon just one type of corporation, the providers of advanced producer services such as advertising, finance, and management consultancy which, she argued, had to locate in the major cities because of the nature of their work. They are knowledge businesses, combining new information with new concepts and ideas to create unique commercial and professional products for their corporate clients. To be successful they have to be part of the knowledge and creative clusters to be found only in world cities. For instance, you can run a car production corporation from offices on the outskirts of Detroit but to create new financial instruments you have to be in New York. It is Sassen's conception of hubs that Castells uses and which we follow here. Contemporary world cities lie at the intersection of vast flows of information that are used in creating valuable commercial and professional knowledge. They have become the service centres of the new global economy

This is, however, a contested intellectual terrain. Cox (1997: 1) refers to 'so-called "world cities"' in his problematizing of the concept. The crucial critique is the weak empirical grounding of globalization arguments: Cox (1997: 3) writes of the need 'to confront the claims of the globalization school with the evidence'. We agree with this assessment; in key respects empirical research has lagged behind theory in the world city literature. However this question is beginning to be addressed by world city researchers (Short et al. 1996; Taylor, 1997, 1999; Beaverstock et al. 2000). Much of this chapter will be about confronting the data problem and trying to overcome it so that we can provide a unique description and analysis of the global reach of a world city. But before we explore the global capacity of London as a major node in the world city network we need to review previous studies of London as a world city, partly to further flesh out the world city concept with a concrete example, and also to illustrate the empirical lacuna on relations between cities even for this most important world city.

LONDON AS WORLD CITY

London is one of the most researched cities in the world. It has had a pivotal role in the development of the concept of world/global city. We will show that despite the important theoretical and empirical researches on London, this work is deficient in one important respect. There is a major empirical inadequacy: relations of London to other world cities is posited but with little or no systematic evidence. This is what we supply in the empirical section below. But to understand and appreciate the importance of the research niche we have claimed, it is necessary to set it within the context of the general literature on London as a world city.

Following Peter Hall's (1966) early, seminal work on World Cities, where he analysed London's internal characteristics as one of the world's major cities, it was Anthony King (1990) who placed London within Friedmann's (1986) world city research agenda by identifying the 'internationalization' processes which accounted for "its global control and co-ordination" in the world economy (page 1). King (1990) focused upon London's development as a world city through an investigation of its role as the major articulator of British colonial expansion, and more recently, as one of the major hinges in the global financial system. Building upon King's (1990) post-imperial stance, three further texts have made a significant contribution to our understanding of London as a world city, all published in the early 1990s. First, London. World City Moving into the 21st Century, co-published by The London Planning and Advisory Committee and Coopers & Lybrand Deloitte (HMSO, 1991), not only discussed London's attributes as a world city (including its wealth creation, jobs, quality of life and infrastructure needs), but, more importantly, also provided a comparative analysis of London's competitiveness, or "overlapping roles", with Paris, New York and Tokyo, and other world cities. Second, Keith Hoggart and David Green's edited volume (1991) London. A New Metropolitan Geography, provided a number of detailed studies of London's world city internal geographies, including: de-industrialization; financial deregulation and office development; the gendered division of labour; and, transport capacity. Third, Susan Fainstein, Ian Gordon and Michael Harloe's (Eds.) (1992) Divided City, examined London's history, economy, labour market restructuring, inequality, housing provision and political structure, as a comparative study with New York City.

An important role of both Hoggart and Green (1991) and Fainstein, Gordon and Harloe (1992) volumes has been to provide a solid impetus for many other detailed economic, social and cultural readings of London as a world city. Important London-specific themes which have received attention, include for example: de industrialization (e.g. Graham and Spence, 1997); producer services (e.g. Daniels, 1991, 1993); labour markets, gender and migration (e.g. Beaverstock and Smith, 1996; Frost and Spence, 1991a-b, 1993; Lyons, 1996); financial and the City of London (e.g. Leyshon and Thrift, 1997; McDowell, 1998; Pryke, 1994); economic development and planning (e.g. Fainstein, 1994); the regulatory environment (e.g. Goodwin, 1996), and social polarization (e.g. Hamnett, 1994, 1996). All of these studies, however, have been overshadowed in their analyses of London as a world city, by the impact of three key contributions: Saskia Sassen's (1991) The Global City; Llewelyn-Davies' (1996) Four World Cities; and the collective works of Nigel Thrift, and his co-authors (e.g. Thrift, 1987; Amin and Thrift, 1992; Leyshon and Thrift, 1997).

Sassen's (1991) reading of London as a global city has had an enormous effect upon our understanding of both globalization and global city formation, and London's comparative geographies with New York City and Tokyo, in particular. Sassen (1991) has suggested that two intertwining processes, the globalization of economic activity in the world economy, especially processes of de-industrialization and the rise of the service sector, and the agglomeration, centrality and concentration of the advanced producer service sector, have both accounted for the position of London as the pre-eminent global city (along with New York and Tokyo). Understanding global city formation by investigating the 'global control capabilities' and 'global reach' of London, New York and Tokyo's advanced producer service sector, has been at the forefront for Sassen' s argument for conceptualizing both global city formation and urban hierarchies. Moreover, throughout Sassen's analysis of the global city (e.g. see 1994a-b, 1995), it has been the intensity of a city's producer service 'complex' which has always been cited as the most important factor in global city reproduction. Of course, if we examine London's producer service complex, not only is it one of the most important international financial centres (e.g. Drennan, 1996; Pryke and Lee, 1995), but also it has a major intensity of other advanced producer service activities (Daniels, 1991, 1993), including for example: accountancy (e.g. Beaverstock, 1996); advertising (e.g. Leslie, 1995; Perry, 1990); law practices (e.g. Beaverstock, Smith and Taylor, 1999a, in press); and other business services (e.g. Daniels et al.,1992). It is this 'complex' which we study below.

In contrast to Sassen's conceptual standpoint, Llewelyn-Davies (1996) provides a wealth of comparative data between London, New York City, Tokyo and Paris. The very clear remit of this study was to quantify London's success, and identify potential weaknesses, in comparison with its three potential heirs as we move toward the next millennium. Data on four major economic groupings are analysed: finance and business services; power and influence; tourism; and; creative and cultural activities. For the context of this paper, we focus on the first grouping. With respect to finance and business, London and New York City were identified as the "two pre-eminent global financial centres" (page 102). London remained very strong in global banking and securities, legal services, accountancy and other "allied" professional services (page 87), in comparison to the other three. Moreover, London's de-regulatory environment was deemed as being very important in its success in attracting foreign institutions and capital, in comparison with New York City, Paris and Tokyo. Llewelyn-Davies (1996), however, does suggest that London's position as a global financial centre does remain precarious. The comparative data analysed by Llewelyn-Davies (1996) reveals that London competes directly with New York in the "higher level financial services" (page 103), but not with Tokyo or Paris. In the European context, they report that London is facing growing competition from Paris and Frankfurt, especially in individual specialised financial services and ancillary services (e.g. in the development of financial software, Paris is moving ahead of London). Moreover, Llewelyn-Davies (1996, 104) comments that "if the U.K. does not join the Euro, Frankfurt or Paris could become the principal European financial centre." The jury, however, still remains out on this thorny issue (other authors suggest that London's position is secure and improving (eg. Sassen, 1999; Thrift, 1997)), which, of course, cannot be adjudicated until well into the early 21st century.

Finally, at a very conceptual, and empirical, level, the collective work of Nigel Thrift, and his co-authors, have extended Sassen's (1991, 1994a-b;1995) work on the producer service complex of London by specifically investigating the economic, cultural and social structures of its international financial centre: The City of London. Like Sassen, Thrift points to the importance of a few places, or world cities (i.e. London, New York and Tokyo), in the "space of flows" (Castells,1989; 1993; 1996) which constitute economic globalization. Unlike Castells who argues that "the global city is not a place, but a process" Thrift argues that the general (economic globalization) produces the specific (world cities), that "greater universalism forces new kinds of particularism" (1996: 231).

This argument is now a familiar one but is relevant here because Thrift uses London to illustrate the key point about places. In 1992 Richard O'Brien proposed, in a succès de scandale, that global financial integration has resulted in "the end of geography", a dramatic statement which has with the emergence of postmodernism been accompanied by an announcement of all sorts of "ends", including the announcement of "the end of the city" (Smith, 1997). Not suprisingly this fashion for hyperbole has generated a considerable response from a number of authors (e.g. Leyshon, 1995; Sassen, 1994b; Thrift, 1994) who, following Harvey (1989: 294), argue that place has become more important, not less important, in a globalizing world because, "as spatial barriers diminish so we become much more sensitized to what the world's spaces contain". However, Thrift (1994; 1996) goes beyond this reassertion of the idiographic to investigate the subtle connection between place and process, to point to the greater reflexivity brought about by globalization to places such as London, world cities as "nodes of reflexivity", where "people work hard to make contacts and to present themselves": "I have shown the way in which the growth of information from new communication technologies has presented fundamental problems of interpretation for workers in the City which have forced greater rather than less face-to-face communication: the City has become, even more than formerly, a key story-telling node for the world as a whole, and, as a result, its importance as a spatially fixed centre has, if anything, been boosted" (Thrift, 1997: 140).

The increasingly cosmopolitanism of London as an international financial system requires actors to negotiate a diverse set of social relationships. Combined with the very nature of economic globalization, with financial instruments which are more rather than less specific, this means that the need for face-to-face contact, for spatial proximity is greater than ever before. This heightened need for social relationships, contacts and networks, to ossify or embed the fragile symbolic disembedded ties formed in electronic space "forces an ever greater reflexivity into the conduct of many meeting places". Thrift illustrates the rise of economic reflexivity through the example of the City of London where he sees a break between the 'traditional' City which existed before the death of Bretton Woods (i.e. the City of London pre-1971) and the 'de-traditionalising' City which exists thereafter. Thrift argues that in the post-Bretton Woods era the number of financial centres which count, "as meeting places and, in general, as significant nodes of reflexivity in electronic discourse networks - has decreased but, in turn, those places that are left in contention have become more important". It is the famous 'triad' of world cities which Thrift argues are becoming "more social, more reflexive, more interpretive" and therefore more powerful as they control the construction, negotiation, and validation of representations. For Thrift, London is viewed as one of, what Law (1994: 104) calls, 'ordering centres' which are 'constituted by gathering, simplifying, representing, making calculations about, and acting upon the flow of information'.

These are all important studies which provide both some detailed empirical insights and innovative theoretical contributions but they have remained comparative in their methodology. Information is gathered on London and other cities and comparisons are made but this is does not provide evidence of relations between world cities. For all the emphasis on cities as nodes in a space of flows, the latter is conspicuous by its empirical absence. This has of course been recognised. For instance, in the Coopers & Lybrand Deloitte (HMSO, 1991) report, it is concluded that to ensure wealth creation a world city must "develop and enhance global links and relations - it must work to further international links and the transfer of knowledge and experience between cities to foster integration in the urban community" (page 196, their emphasis). But they lament that this is where the evidence is lacking because data is generally produced for national needs, not to measure inter-city linkages. They resort to a qualitative market research survey of key decision makers in London and several other key world cities (i.e. New York, Tokyo, Paris, Frankfurt and Berlin) who give opinions on which world cities they see as the most important. The empirical limitations of such research for understanding 'global links and relations' are obvious. In contrast, this chapter provides a solid empirical knowledge of the extent of London's links and relations with other world cities, which provides a substantial framework by which to gauge and measure the global capacity of London as a world city.

London: Measuring Connections

The description of our measurement methodology proceeds in three sections. First, we begin by developing further the argument about the dearth of relational data to study world cities. Second, we describe our global data collection exercise that covers 69 firms in 263 cities. The firms are all from one of four producer services - accountancy, advertising, banking/finance and law - for which we have collected enough information to produce reasonable descriptions of each sector as a whole. Inevitably, the information available is very variable which has necessitated particular concern for different levels of measurement. Third, we discuss how we combine the information to maximize comparable data across cities and sectors. This provides the basic data from which connections between London and other cities can be computed.

The Paucity of Relational Data: "The dirty little secret of world cities research"¹

How exactly is London located within the contemporary space of flows which constitutes the world-economy? Answering this question is not as simple as it may at first seem. The vast majority of international data is about relations between states not relations between cities. Hence, for example, there is much data available on UK-France relations but relatively little on London-Paris links. Furthermore, what is available for inter-city links tends to produce quite general measures of connection. For instance, air traffic flows between cities can be obtained from timetables and a hierarchy produced (Keeling, 1995; O'Connor, 1995), and data on postal flows, telephone calls and internet linkages are also available (Marek, 1992; Michelson and Wheeler, 1994;Warf, 1995; International Telecommunication Union (ITU) World Telecommunication Indicators Database) and can be used to explore city hierarchies. However, there is a basic problem with all such data in that it measures general patterns of flow, but does not differentiate the specific flows within those patterns so that specific processes relating to world-city formation are masked. For instance, in any international airline survey London is easily the leading hub but this clearly over-estimates London's importance relative to New York in inter-city terms: New York's connections to other US cities (e.g. New York-Chicago) are not counted as international whereas London's connections to other EU cities (e.g. London-Paris) are counted.

This data deficiency is responsible for a most curious paradox in the literature on world cities: whereas the essence of world cities is their relations one to another, researchers have generally not focused on this aspect of their being. Hence the internal patterns of world cities are well-researched, but their external patterns, their geographical scopes, are not. We attempt to overcome this deficiency and paradox for our London study through an empirical investigation of the organizational structures of advanced producer services. Sassen (1994b) has argued that the activities of advanced producer services are at the heart of contemporary world city formation: world cities are the necessary places to be to sustain 'state of the art' servicing of global capital. Thus, data on the organizational structure of the firms carrying out the advanced service production is a particularly fertile line of research at this time. This relates to the new economic geography of services which has focused on the recent growth of producer services, and within them, financial services (eg. Corbridge et al., 1994; Daniels, 1991, 1993, 1996; Thrift & Leyshon, 1996). Although information on the geography of offices for producer service across the world does not provide data on the actual transactions of companies, it does reveal their geographical strategies as prime agents in the making of contemporary world cities, or international financial centres (see for example, Coakley, 1992; Budd, 1995; Pryke & Lee, 1995; Thrift & Leyshon, 1994). Typical uses of such data are the pioneering works of Meyer (1986) on how Latin American have stronger ties to the USA than between each other, Daniels (1986) comparative analysis of foreign banks in New York and London, and Thrift (1987) on 'international financial centres' using foreign banks in London and selected London companies' international networks. More recently such data have been used to explore hierarchical systems of cities, for example: ó hUallacháin (1994) uses foreign banks in US cities to define the global scope of six of them, Lynch & Meyer (1992) use law firm branching patterns to suggest a hierarchy among US world cities, and Daniels et al. (1992) uses consultancy services to construct a urban hierarchy in Western Europe. Clearly, here we have a potent type of data to solve the deficiency of relational studies in world city research. 

Data Production: Global Scope and Levels of Measurement

There are three basic questions which arise in the process of producing data for this research: (i) which cities to include? (ii) which firms to include? (iii) what levels of measurement can be employed? We answer these questions in turn in this section. The basis of our decisions in these three areas are very different: for cities we attempt to be as inclusive as possible, for firms we are dependent on the availability of information, for measurement we are dependent on the quality of information. The final outcome of the data production is to cover 263 cities, using information on 69 firms for each of which at least one of four levels of measurement are possible.

Cities

The 263 cities in this study are classified into three groups: 55 world cities (including London), 68 cities exhibiting evidence of world city formation, and 141 other cities (see Appendix A ). Although our focus is on world cities, globalization in practice is most certainly not limited to these important locations which is why we extend our concern beyond world cities.

There has been no generally recognized roster of world cities to guide researchers. Lists of world cities which are available in the literature are unsatisfactory because they are products of individual subjective assessment. For instance, in what has probably been the most commonly used roster, that of Friedmann (1986), It is not at all clear how the variables chosen are combined to produce the hierarchical outcome. Clearly there is a need for a transparent ordering of major cities to identify a category of worlds cities which we have attended to in another paper (Beaverstock, Smith & Taylor, 1999b). Basically this classification is based upon identifying 'significant presences' in cities of major producer service firms from four sectors: advertising, accountancy, banking/finance and law. From the cities thus identified, all those with only a single firm's presence in a sector were removed leaving 123 cities still under consideration. For each of the four sectors 3 levels of service provision were defined and these were combined to provide degrees of 'world city-ness'. The 123 cities scored between 12 and 1, those 55 cities at 4 and above are classified as world cities with the other 68 designated as cities exhibiting evidence of world city formation.

Finally, the category 'other cities' was defined much more subjectively. We decided that there was little utility in recording all service firm presences on three grounds. First, this would involve very large numbers of cities creating extremely sparse data matrices. For instance, KPMG has offices in 828 cities. Second, firms typically have many branches in their home country (or the home countries of other firms which they have taken over). Such local, idiosyncratic, patterns are not relevant to our purposes here. Third, we consider that the additional effort to be locationally comprehensive would have been out of all proportion to any research gains made. Hence we included in our set of other cities, those cities which were recorded but did not qualify in the world city definition exercise, capital cities of most countries plus selected other major cities. This list was modified in an on-going manner as we got to know our information better - cities were added right up to the end of this research. Although, quite subjectively defined, we feel that the 140 cities we have identified in this category in Appendix A do represent an intelligent 'remainder' group which capture almost all the global service provision beyond the first two categories.

Firms

Our universe of interest is made up of London located global firms (LLGFs) who provide advanced producer services. "Global' in this instance is defined arbitrarily as firms with branches in at least 20 other world cities (except for some law firms, see below). Selection of firms from this universe depended solely on data availability. In our final analyses we deal with 69 LLGFs: 5 in accountancy, 11 in advertising, 14 in banking/finance, and 39 in law (see Appendix B). The different numbers of LLGFs among these sectors partly reflects the differing degrees of business concentration with accountancy the most concentrated and law the least. Of course, this process of concentration was on-going during our information gathering but we have decided not to up-date our information with such changes. We are carrying out a cross sectional analysis for 1997-8, we assume any post-merger rationalization of offices will not effect the overall patterns of the geographies we report.

Information on firms comes from two main sources, published directories and web sites. Where firms produce internal directories of their organization we have very good information on their global operations. More often we have had to rely on firm's web pages. Sometimes these will give direct information on global operations but on other occasions individual offices have control of their own pages. In the latter case information can be variable in content and likely not be comprehensive in coverage. For this research it is vital that information is locationally comprehensive for a given firm. A firm with information on offices across the world except for, say, Latin America, is not usable. For instance, we had intended to include insurance as one of our producer services but we were unable to obtain sufficient information for enough firms to cover this sector. The other omission is management consultancy, the fastest growing of all producer services. In fact, key management consultancies are part of the accountancy sector and therefore are recorded in our information sources there. We decided that, for our specific purposes, we did not need to separate management consultancy from accountancy. The final list of firms and their information source is shown in Appendix B.

As we have indicated, it is important that, for each of the four sectors we do deal with, we have enough information from a sufficient variety of LLGFs to constitute a reasonable picture of the global scope of the sector. This we have achieved as follows.

(i) Accountancy: at the time of data collection there were the "Big Six" firms² which totally dominated the sector and we have global information on five of them.

(ii) Advertising: this is the most complicated sector in terms of organization but we have global information on 11 firms who are in the world top 50 or are parts of larger groupings in that listing.

(iii) Banking and Finance: we have global information on 14 firms in the world top 50 including the top 3.

(iv) Legal services: this sector is dominated by UK and USA firms and we have global data on 126 such firms. This is the only case where we go below 20 offices to admit a firm into our data: for London headquartered firms we included all firms in the top thirty list which have overseas offices. This is justified by the smaller scale of law firms- only one London headquartered firm reached the threshold of 20. In contrast 13 US firms in London were over the threshold which we maintained for these LLGFs.

Measurements

From the global information of the 69 firms we can derive 4 levels of measurement.

(i) Nominal: this is a basic presence/absence measure where all we have is information on the city locations of a firm's offices.

(ii) Ordinal: this is a situation where we have some additional information which enables us to order cities in terms of level of service a firm provides (for comparability we rank only 1 to 3).

(iii) Interval-B: this is where we have full information on individual branches so that we have data on the precise number of offices of a firm in a given city.

(iv) Interval-A: this is the most sophisticated data where we have measures of the activity of a firm within a city, for instance the number of practitioners, or the number of fax contact numbers (this is preferred to numbers of telephone contact numbers since it implies business communication rather than, say, domestic banking).

For the 69 firms the levels of measurement are 34 nominal, 5 ordinal, 11 interval-B, and 19 interval-A. (see Appendix B).

It is important in data collection to capture information at the highest level of measurement possible - nothing should be wasted for future possible investigation. However, in this paper we need to combine the various information on different firms into single sets of data, first for the four sectors, and finally as overall aggregate data for London's links with other cities. This can only be achieved by the loss of some information in a combination which respects the properties of the different levels of measurement. With the data we have collected two forms of data combination are possible.

The most basic level of measurement is the simple recording of presence/absence of a LLGF in a city. All measures can be reduced to this level. Obviously, we lose a lot of information in this process but we do produce measures showing the pattern of London's basic service connection with other cities. This is the first measure we use in our description and preliminary analysis below.

The next question is how can we retrieve at least some of the information lost in the reduction to presence? The answer is to convert all measures to ordinal data. Because our nominal data has only two classes (presence/absence), they can be combined with ordinal measurement (obviously 1 ranks above 0). The only problem is that in this combination LLGFs with nominal data become weighted less than other firms because of the upper limit of 1 to the measurement. However this is more than offset by the additional information to be found in ordinal measurement.

In the case of interval data the frequencies of practitioners and offices have to be reduced to just three ordered categories to make them equivalent to the ordinal data we already have. This means defining boundary points between 3 and 2 and between 2 and 1. This has been done separately for each set of interval data by inspecting the distribution of frequencies. (For the specific decisions see the footnote to the table in Appendix B.) The result of these conversions is to create a second more informative set of data for our description and preliminary analysis below.

LONDON IN THE WORLD

Finally we are ready to provide some empirical findings on London's relations with other cities across the world. It is worth repeating again that such inter-city connections have not been reported elsewhere for London, or indeed any other world city. These unique results are presented at two levels of complexity. In the first section, we present a simple description and analysis of the presence in other world cities of firms located in London. This is carried out separately for the four producer services and then combined to give an overall picture of London's 'global presence'. This is complemented in the next section by an attempt to measure the relative strength of the connection of these other cities to London. Using a simple distribution analysis, the key world city connections to London for each service are identified. This culminates in the production of an overall index of London's global connections.

London's Presence

For initial analysis of such a large data set we begin by looking at the basic distribution of measurements. Patterns of city frequencies can suggest the basic processes operating which produce the measures.

The distribution of 263 cities in terms of how many LLGFs they house are shown for each sector in Figure 1. 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.

For the three distributions with zero modes there is a particularly interesting feature within their distributions. In all three cases the distributions begin to tail off smoothly before rising again. Such an 'enhanced tail' is exactly what we would expect to result from world city formation as described by Sassen (1994b). Such a distribution feature is indicative of non-random concentration processes operating within the data. This means that we can identify those cities where concentration processes are acting by focusing on frequencies above where the distribution changes direction. These 'tail cities' are identified in Tables 1 to 3. By converting the frequencies to percentages of the maximum possible presences we have produced a quite interesting result: for each city in these tables the figures indicate the percentage probability that if you enter a London office what are the chances of that firm having a direct link to that city in terms of another of the firm's network of branches. Thus, for banking/finance (Table 2) The probabilites for New York and Tokyo are 100%: you can be certain that the LLGF will have offices in the other two world financial centres. The more interesting results come when you go down the list and when you compare it with the other two sectors. For instance, the dominance of Brussels in law (Table 3) and the prominence of European cities in advertising (Table 1) are noteworthy.

These lists of 'tail cities', not surprisingly, confirm in general terms our previous definition of world cities. In Table 4 these results are brought together by focusing only on our roster of world cities. For accountancy, of course, all 54 of our world cities score the maximum 100, and for the other services the high scores from Tables 1 to 3 are shown again. The interest here, therefore, is on the low scores especially the zeros. These tell the stories of particular service deficiencies in some world cities in terms of their direct office linkages to London. For instance, for advertising seven cities are conspicuous by their low scores (Atlanta, Berlin, Boston, Houston, Munich, Osaka and Washington, DC). Given that these cities are from just three countries, this indicates London links concentrated on New York, Tokyo and Frankfurt are sufficient for this service. Banking/finance firms in London are poorly represented in four world cities in particular (Atlanta, Copenhagen, Minneapolis and Stockholm), all minor American and European world cities. For law there are eight world cities (Atlanta, Copenhagen, Hamburg, Kuala Lumpur, Minneapolis, Montreal, Osaka and Seoul) with no direct office link at all with London. This number is not surprising for this least globalized of the services we are dealing with; in law some firms provide a service through associations and alliances which do not require overseas offices (Beaverstock, Smith, and Taylor, 1999a, in press). Note that the same cities are beginning to reappear in these lists of weak linkages with London; this is measured more precisely below with ordinal data which incorporates levels of service provision beyond simple presence.

Before we leave the question of presence, it is of interest to identify those cities at the other end of the spectrum from the world cities where LLGFs have none or very few offices in aggregate across all four service sectors (Table 5). The cities featuring in this table are either major cities weakly integrated into the global economy for political (e.g. Baghdad) or poverty reasons (e.g. Kinshasa) or both (e.g. Havana) or small European cities (e.g. Bergen). The most interesting case is probably Bonn. When this data was collected Bonn still housed nearly all German government activities but this never operated as an economic magnet in the manner of other capital cities.

The Strength of London's Connections

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 263 cities for each service sector (Figure 2). 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 again 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 6 to 9 for accountancy, advertising, banking/finance and law respectively. Compared to the simple presence analysis, this new analysis shows several noteworthy results especially the recognition of New York as London's main 'partner': in banking/finance New York is now above Tokyo and the latter is joined by Singapore and Hong Kong at the top of the table; in advertising New York's dominance now becomes very clear; in law, New York and Washington are now above Brussels; and finally, for the first time in an analysis of tail distributions, we can see that the high level accountancy linkages includes New York at the top with 4 other cities.

Focusing on just the world cities, the levels of linkage to London are brought together in Table 10. In this table we can observe variations among world cities; this is particularly relevant for analysis of accountancy since no variation was recorded in terms of presences (Table 4). Hence, we now record a range from a high of 93 (Dusseldorf, New York, Paris, Tokyo, Toronto) to a low of 33 (Istanbul, Prague). For the other producer services also, more subtle differentiation is indicated; for instance comparing Table 4 with Table 10, in advertising Amsterdam and Budapest have the same level of presence but the former has a much higher level of connection, in banking/finance Santiago and Seoul have the same presence but the latter better connections, and in law Moscow and Washington, DC have similar presences but differ by a factor of more than two in connections.

The culmination of this preliminary global analysis is to measure the overall level of connection between London and other world cities by averaging across the four different producer services. These are shown in the final column of table 10. This confirms the overall dominance of the New York-London link in the data: it scores 87 compared to Paris-London in second place with 68 followed by Hong Kong-London (64), Tokyo-London (61), Brussels-London (69), Singapore-London (58) and Sydney-London (57). Even listing just this top 7 clearly illustrates the world-wide scope of London's linkages but to show this further requires a regional analysis including all world cities.

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 11. Perhaps the lack of major variation in average scores is the main feature of this table which confirms London's world-wide reach (as above). 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 (1991) terminology, a global city.

CONCLUSION

The prime purpose of this paper has been an empirical one, to report on a relational study of London. Having presented this work, this conclusion can be quite brief. We know of no other study which maps the global reach of London, or any other world city, in as comprehensive a way as we have outlined above. Previous research on world cities has relied on a priori judgements about the 'globalness' of key world cities in terms of their location in the global space of flows. Despite the severe data limitations of extensive global, as opposed to international, analysis, we have shown that real trans-state data can be brought to bear on such important matters. Of course, we have provided only a cross-sectional analysis of just one city. Hence, as well as being informative about London in the world, the paper points towards global empirical research covering more cities and monitoring changes in inter-city relations.

Producing global geographies of offices creates new frameworks for research which identify key nodes in the contemporary global space of flows. This goes a little way towards answering the critics who point to the paucity of evidence on world cities and globalization. In terms of further research, the details of the geographies can be pursued (for law, see Beaverstock, et al., 1999), comparisons between cities can be made, and comparisons over time to produce trends is a particularly necessity as indicated above. Most importantly, we need to move on from an ego-centric network analysis (one node at a time) such as this one to a comprehensive analysis of the world city network as a whole (Taylor, 2001). But ultimately the new geographies can form the indispensable framework for globalization studies that attempt to understand precisely the processes operating at this level in Castells' (1996) network society. New relational studies need to be informed by what goes on in the network nodes so as to combine knowledge on internal city and external inter-city contacts and linkages, as well as the location decision-making which creates the dynamic of the geography we have captured as a cross-section. Our preliminary global analysis provides such a framework for further studies of London, with other such global analyses it can provide the geographical skeleton of contemporary globalization.

ACKNOWLEDGEMENT

We would like to thank the Economic and Social Research Council for funding this research (R000222050). Thank you to the anonymous referees for their comments.