GaWC Research Bulletin 9

Commercial cities and nation-states are emblematic of the two types of spaces geographers deal with, the spaces of flows and places respectively. Networks of cities and mosaics of states depict two very different ways of imagining the spatial organisation of the modern world, alternative metageographies if you like. Beyond geography, Charles Tilly (1990) has interpreted cities as representing capitalism competing with states which represent coercion. Following his lead, in a previous paper I provided a very brief historical sketch of the changing relations between cities and states in the modern world-system (Taylor, 1995a). The story had three stages. First, antagonistic relations: state formation was a centralisation of power at the expense of other power centres including cities. Second, a mutuality: cities became integral to the creation of 'national economies' and their prosperity depended upon the states they were located in. Third, the return of antagonism: under conditions of contemporary globalization the old mutualities seem to be breaking down as the most important opportunities for cities may not lie within their own state's territory. Thus, in an argument typical of world-systems analysis, we seem to have come full circle, back to 'cities versus states'. Implying as it does a diminution of state power through the 'release' of cities towards a more global destiny, the story fits in well with globalization theories which predict the end of the state. Focusing solely on the third stage of my original story, in this presentation I want to revise the simple 'fall of mutuality thesis' and thereby scotch any notion that I see cities simply replacing states as the key 'spatial' institutional centre of modern life today or in the near future.

I will begin by clarifying what I mean in my title by 'under conditions of globalization'. I interpret globalization as a one among many tendencies which can be identified in contemporary social change. Particularly associated with the new enabling 'distance-shrinking' technologies of the last few decades, at the core of the idea of globalization is the notion of the enhanced importance of trans-state processes. This is important because embedded statism has been a primary force in the development of the modern world-system since its inception. To be sure, trans-state processes have always existed (Taylor, 1995b), but what is different today is the thesis that these processes have grown to such an extent that they not only rival state-based processes but that they are dominating the late capitalist world-economy. Hence, the key problem with contemporary globalization debates is that they have tended to be couched in either/or terms forcing a choice between globalization and state-centric positions. I refuse to make this choice. By viewing globalization as just one tendency in a world-system where the states are integral to social reproduction, we can ensure a more subtle consideration: beyond simple global versus state to consideration of their manifold inter-relationships as illustrated by Brenner (1998) for contemporary Europe.

World/global cities are the most conspicuous geographical manifestation of contemporary trans-state processes. The relationship between such major cities and their contemporary states has always quite complex affairs and the current situation is no exception. Most of the great cities of history were centres of state power and the roster of world/global cities are dominated by capital cities, as we shall see. The empirical core of this paper derives from research with colleagues at Loughborough University on world/global cities to which I have added my own political dimension. The result is what I believe to be a unique political geography which integrates the traditional political mosaic of state territories with the network of world cities in a single analysis. All this is very preliminary, as will be made clear, but combining the world space of territories with the world space of flows is surely a critical way forward for the subdiscipline in the next millennia. Before and after the empirics I deal with some theory and practice for carrying out such research. Doing global research is no straightforward matter.

INTRODUCTORY THEORY AND PRACTICE

World city research is strong on ideas but weak on evidence. Much of the latter is based upon methods drawn from traditional urban and regional development research - using data from national censuses to make comparisons between cities based upon sectoral changes, for instance. What is typically missing from such comparative analysis is any information on the relations between the cities. It is studying the nodes in the space of flows without reference to the flows themselves! But before I address this empirical predicament I want to look again the general theory behind world city research. In his initial formulation, John Friedmann (1986) set world cities within a world-systems framework with cities as the 'basing points' of capital in the world-economy. But world-systems analysis implies much more locating cities in core or semi-periphery. Here I begin by using Braudel to look afresh at capitalism and the state, and cities' relations to them both.

Braudel's Provocation

Fernand Braudel has broken away from the Adam Smith/Karl Marx orthodoxy which equates the rise and spread of capitalism with the successful establishment of competitive markets (Braudel, 1992, 1994; Wallerstein, 1991). This conventional wisdom has, of course, been bolstered in the last decade by the demise of state socialism and the development of economic globalization. Braudel provides us with a lovely bit of unconventional wisdom: capitalism is inherently anti-market. Any sensible capitalist avoids a proper functioning market because that is not where the biggest profits are to be made. Put the other way round, markets are bad for capital accumulation because they do not allow enough of it. If capitalism does not equal market what does it equate with? For Braudel the answer is simple: monopolies, that is where the big profits are to be made. I suppose the orthodox response would be that relying on a monopoly can only be a short-term expedient, without the competition of the market to stimulate efficiency monopoly ultimately leads to stagnation and bankruptcy. Yes, of course, but capital is mobile, sensible capitalists will have long since forsaken a failing enterprise and moved on to new monopolies where returns remain high, temporarily.

Historical capitalism is, for Braudel, a world of multiple monopolies not multiple markets. Capitalism is a mode of production which maximises capital accumulation and this is done through being flexible, being able to take the largest profits available at any one time irrespective of specialisation. In fact long-term successful capitalists are inherently generalist in nature; they are neither financial, nor merchant, nor industrial, nor agrarian but any and all depending on the size of the profits available. So where is the market in all of this? For Braudel this exists within a level of activity below capitalism, in the sphere of economics. It is here that people compete as producers and consumers in markets which set prices. Profits are made to enable reproduction of the institutions and activities but while markets operate fairly there are no large returns to be made to satisfy the historical needs of capital accumulation. Hence the spur for social change is not market competition but how to keep out of such competition. Of course, the boundary between the levels of economics and capitalism is an continuous zone of transition as markets are cornered and monopolies undermined. The whole point of the capitalist's flexibility is to read this frontier so that investment remains outside the market: into Lancashire textiles at one point, return to London finance at the next, for instance.

I have called this argument Braudel's provocation and by now you will have seen why: instead of monopoly being 'a spreading rot in the fabric of competitive capitalism' (Shepherd 1970, 14) as is commonly supposed, it is itself capitalism. Rather than monopoly indicating an imperfect market, the market is a sign of failed monopoly. Turning conventional wisdom upside down is all well and good fun but how credible an argument is it for guiding our researches? I fear positions may be entrenched here but please bear with me as I continue to plough this furrow. The reasons are straightforward. First, the basic mechanism in Braudel's capitalism is power. Power is necessary to get above the market. This means that the state becomes directly implicated as part of the capitalism, not something used for or against capitalism. Bringing the state to the centre of analysis is obviously relevant to political geography. Second, the political manipulation of space has been crucial in the design of monopolies, it is the simplest form among an array of strategies. As Braudel (1982, 447) tells it 'geography accounts for a good deal' because 'overseas trade' relied on distance to evade free market processes, eliminating 'competition by holding a virtual or actual monopoly, and to keep supply and demand so effectively separated that the terms of trade were entirely dictated' (p. 416). This 'primitive monopoly' of distance is potentially of relevance for human geography in general. In fact once we begin to focus on monopolies, not as special exceptions to the market, but as the normal operation of capitalism then much old and new work in geography and the social sciences in general take on a new perspective. Let's begin by describing some general monopoly strategies or what I shall term anti-market nexuses.

Anti-market nexuses are power configurations which enable capitalists to stay above the market. They consist of variable mixtures of geography, knowledge and coercion which provide a holder of capital with access to extra large profits. Table 1 lists eight well known examples of market avoidance in historical capitalism. All eight cases encompass competition, of course, but not market competition. 'Monopoly competitions' are inherent from imperial conflicts through to regional rivalries; the most well-known is Schumpeter's (1942) corporate innovation competition where the promise of superprofits stimulates investment in technology to create monopoly positions.

These eight nexuses give a flavour to the world of multiple monopolies through which capitalism operates. Certainly there will be many other forms and the ones listed are often closely entwined: imperialism, for instance, has worked so well because of its relations to contemporaneous forms of all the other seven nexusus. Anti-markets are easy to spot, although typically opaque in operation, you just have to look for where the really big profits are being made. However, as monopolies they are always under challenge and all of them have a limited life as capitalist bonanzas. To repeat a key point, flexibility means that capitalists avoid being reduced to market competition, by the time the latter arrives, capitalists will have long since gone. One analogy is to think of monopolies as leaking ships. Capitalists are those very big and very intelligent rats who have found themselves a new home long before the leak becomes a serious danger to life.

Obviously this list of nexuses is pregnant with many political geography stories to tell. As already indicated I am interested here in contemporary world cities which are a special case of the regionalism anti-market nexus (Storper, 1997). Although the new distance-shrinking technologies have reduced the scope of 'primitive geography' monopolies, they most certainly do not signify 'the end of geography' as a critical component of capitalist strategy. Geography remains implicated in monopoly formation but in different ways. For Michael Storper globalization has coincided with a new reflexive capitalism where regionally-specific knowledge complexes are sites of continual learning to keep ahead of the market. Production is geared to quality rather than quantity so that instead of the usual market 'economy of scale', monopoly 'economy of variety' is created. In this way regional assets are reproduced and not easily imitated elsewhere which Scott (1997, 325) has called 'the monopoly powers of place'. World cities are key concentrations of such monopoly powers based upon economic reflexivity. In particular, Saskia Sassen (1991, 1994), in her explanation of why world cities have prospered under conditions of potential economic decentralisation through distance-shrinking technologies, identifies advanced producer services as the distinguishing characteristic of contemporary global cities. In services such as finance, accountancy and corporate law, practitioners are not just servicing 'global capital' they are creating new products based upon their unique knowledge collectivities. In other words, they do not aim to operate in a price-setting market but rather develop multiple monopolies of quality knowledge products to reap appropriately large profits. Hence, unlike services in general which operate in markets which keep prices down, advanced producer services are above economics, they concentrate in world/global cities where they enjoy the benefits of capitalism. This is the capitalism - the geographical monopolies - I am concerned with here, but how do you study their locus, the world cities?

Doing Global Research

If you are at the apocryphal party and answer the question "What do you do?" with the words "global research" you will probably not get the usual stunned silence expected for such an 'academic' response. Rather the quiz will continue about what you know of the ozone layer and things getting warmer. Global means climate in the public mind and despite all the globalization hype there seems little notion that there might be global social research. It may be understood that we do international research but that is a rather different thing. The annual reports and statistics put together by the UN and its agencies from information supplied by their members provides the basic data bank for such international research. It describes states and therefore has enabled comparative analysis of states to proceed as a research boom area since the 1950s. But it has provided little or nothing for researching trans-state processes. In the world of state-istics, those social scientists wishing to do global research are largely on their own when it comes to information availability. Hence the penchant in globalization debates for speculation and assertion over facts; quite simply, for trans-state processes the latter are in very short supply.

Global social research as a major empirical project really began in the 1970s with studies of multinational corporations. Building on the finding that some of the new corporations were financially larger than many small countries, a new subject was found which had 'global reach'. In practical research terms, however, corporations and countries are like chalk and cheese. This new subject was a private player in the world-economy: most information which is necessary to understand its behaviours is simply not publicly available. Empirical research proceeded by case study which allows for good comparative analysis but there is no sense in which such global research is as empirically grounded as international research drawing on the UN data base. For the latter there is agreement on the subjects - the list of sovereign states at a given time - and on the information to be collected on each subject. This standardisation allows for relatively easy comparison over time and space for very many important and relevant topics. Now consider the equivalent for global research using corporations as subjects: the nearest counterpart at the level of a simple roster of subjects is a magazine list (the Fortune 500 - see Ikeda (1996) for use of this list to relate corporations to states). We have, of course, learned a vast amount on the strategies of corporations on the back of large numbers of case studies and comparisons allied to international statistical analyses (Dicken, 1998), but we have no overall empirical handle on global changes in the private corporate world.

At Loughborough we have taken a different tack on the problem of doing global social research. Although corporations remain the subjects of our researches, the objects of concern are world cities. Following Sassen's argument presented above, we focus on advanced producer services corporations where their global reach is part of their product. This reach is centred upon their networks of offices and it is this which our preliminary research has targeted. It is these offices which are in world cities and make the latter worthy of their name. We have constructed a data bank on office geographies covering 69 corporations over 263 cities. Corporations are from four sectors, accountancy, advertising, banking/finance and law, and the information for each of the cities ranges from simple presence/absence of a firm through to numbers of practitioners a corporation has located in each city. Notice that we have gone beyond banking in this data collection: we do not focus just upon 'international financial centres', we adopt the broader, original concept of world city in which financial services play only a part, albeit a very important part, in world city formation. We think this is a unique trans-state set of data and which, therefore, enables us to begin some original global empirical research on world city network formation.

CITIES AND STATES IN A SINGLE WORLD-ECONOMY

Because of the state-centric nature of data, the world-economy is commonly equated with the construct 'international economy' which describes economic relations between states. In contrast, I will focus upon inter-city relations as a model of the world-economy. Of course, there is only one world-economy and all models and constructs are partial representations. In this empirical section I begin by exploring basic questions for the neglected cities model but then bring states back in to provide some initial assessments of cities and states in a single world-economy. All of this should be seen as preliminary results to illustrate the global research we are doing. I am going to push the data probably further than is warranted because I am more interested in potentials than findings in this initial foray. I tell a story of tails, rosters, surpluses, regions, and conclude with 'ins and outs'.

Tails: Are There World Cities?

Let us start at the beginning: is there a general process of world city formation operating across the world-economy? Kevin Cox (1997, 1) seems to doubt the existence of such a process by referring to 'the rise of so-called world cities'. Obviously this question has to be carefully specified. Of course there are large cities which dominate flows in the world-economy but this is not necessarily world city formation as previously defined. For instance, a central place-like global process with firms servicing the needs of global capital, with other firms servicing regional capitals, others national capitals, and so on in hierarchical fashion is not what world city formation implies. Such a global to local hierarchy is a market model with central places as the loci for bringing together service providers and users. The result is some form of spatial equilibrium reflecting an integrated urban hierarchy as postulated at the national level many years ago by Berry and Horton (1970). With world city formation no such spatial equilibrium is expected. Rather, a class of cities is defined which is qualitatively different from other cities. World cities are the loci, not just of services in the central place sense, but of unique knowledge complexes as previously discussed. These monopoly conditions are relatively rare, they are not distributed in varying degrees across all cities.

Now that we have global scale data on advanced producer services we can test whether these firms are indeed operating to create world cities or whether they represent merely the latest reshuffling of services in a traditional, albeit extended, hierarchy of cities. In Figure 1 the distributions of presences of our 69 firms across the 263 cities are shown separately for the four service sectors. The distribution for advertising, banking/finance and law are very similar: each display a lumpy and very enhanced tail. In any equilibrium process as postulated by the central place/market model, the distribution would show a smooth and regular decline from the mode; here, in contrast, we have distinct evidence of concentration processes, specific cities being singled out for office location. This is entirely consistent with the world city/monopoly position.

The distribution of accountancy offices is completely different from the other three and requires separate discussion. Accountancy has the most monopolistic attributes of all the advanced producer services and has recently gone through a period of mergers and take-overs reducing the major players from nine to five - there were six during our data collection for five of whom we obtained global data. The result is huge multi-billion dollar corporations with many more offices than other producer services. Hence they are located in far more cities as the distribution in Figure 1 shows. Clearly this illustrates the monopolistic tendencies in this sector as reflected in carving out market shares, but not the world city formation I am concerned with here. We do have more than presence data on the levels of accountancy service in our data bank and this is used in the next section to identify world city processes.

In conclusion: world city formation is in operation - in the three service sectors where the frequency distribution shows a reasonable range of city presences, the common pattern of an enhanced tail implies a process of locational concentration rather than equilibrium.

Rosters: Where are the World Cities?

Having agreed that there are world city formation processes, we can now move on to the second question: where are the world cities? It is indicative of the poverty of global empirical research that there is no generally agreed roster of world cities. For sovereign states we have the generally accepted roster listed for any year in UN publications. How can we begin to compare the network of cities with the mosaic of states if we have no agreement on the members of the former? London and New York yes, Frankfurt and Sao Paulo yes, but what about Manchester and Cape Town? Where do you draw the line? The easy answer is to use a given population size as a threshold but this defines 'mega-cities' as a size category rather than world cities as a functional category. As arbitrary as the Fortune 500 list of corporations, it is like including California and North Rhine-Westphalia in the roster of states at the expense of Equador and Portugal just because they have larger populations and/or economies. That would be wrong, it is not a matter of size but sovereignty which matters in international economic relations which is what the latter two have and the former two do not have. Similarly Amsterdam and Zurich are not mega-cities but are certainly world cities, Calcutta and Lagos are mega-cities but not world cities. Clearly before we can begin looking at the relations between world cities and territorial states we have to produce an empirically sound roster of world cities.

The most widely used list of world cities is that originally produced by John Friedmann (1986; 1995) and which he has subsequently revised. Unfortunately, it is not entirely clear how his world cities are identified and ordered. In his original list we are told 'selection criteria include', which implies an incomplete enumeration, followed by a collection of functions (e.g. corporate headquarters), processes (e.g. growth of business services) and measures (e.g. population size). How these are combined is not made clear. I am reluctant to make the point yet again, but the widespread reporting and use of such a preliminary taxonomy does encapsulate the poverty of contemporary global social research. Here I report on an exercise which uses the data bank of 69 firms over 263 cities to identify a roster of 55 world cities. For each of the four service sectors we go beyond simple office presences in cities by using quantitative measures of the size of offices and the status of an office in the firm's organisation where that is available (for details of the procedures see Beaverstock et al. 1999). The first results of this exercise are to identify three levels of global service - prime, major and minor centres - for each of the four sectors. For instance, we identify 13 cities as prime global accountancy service centres, 37 as major accountancy global service centres and 27 as minor accountancy global service centres. This information is combined across all four sectors to provide measures of the global capacity in terms of these advanced producer services, what we can call 'world city-ness'. Scoring 3/2/1 for prime/major/minor in each sector, when summed all cities are arrayed from 0 to 12. Cities which score at least one point are listed in the 'Inventory of World Cities' in Table 2. The question of where to draw the line to define world cities must be arbitrary to some extent. The score of four was chosen partly because of the distribution of scores but also because to score four a city must be a centre for at least two of the services. The 55 cities scoring 4 or more are designated world cities and these are our main concern in the discussion from this point on (Figure 2). It is important to note that while this methodology orders cities and this has been used to identify three levels of world city, this does not designate a hierarchy. Ranking cities by size, even if it is a global capacity measure, does not define an urban hierachy, that requires evidence of inter-city relations (Taylor, 1997) of which more later.

Given a roster of cities we can begin to contemplate new forms of analysis. For the simplest of illustrations, the cities can be cross-tabulated against the roster of states (Table 3). In fact only 33 of the world's states house world cities and of these 24 have just a single city. These include many medium-sized states but also highly centralised UK and France where historical concentrations on the capital city are today reflected in just one dominant world city in each state where we might expect more. In contrast the USA has 11 world cities and Germany 5 reflecting the combination of decentralised federal polities with large national economies. In between China has three world cities (counting Hong Kong) and another six countries have a brace each: Australia, Canada, Italy, Japan, Spain and Switzerland, all are decentralised polities except Japan whose very large economy can obviously accommodate more than one world city.

For the reasons detailed earlier, even as elementary an empirical exercise as that above has never before been carried out. Of course, we do not have to stay at that level now we have global data on world city formation. In Table 4, levels of world city formation are related to the size of national economies as measured by gross domestic product. There is obviously a positive relationship between the size of national economies and the distribution and levels of world city formation processes. For instance, eight of the ten alpha world cities are located in the big six economies. As usual it is the exceptions which are particularly interesting - Hong Kong and Singapore as 'world city states'. We will look at their regional roles below. All but two (India and Austria) of the next 21 ranked economies have world cities in their territories. The smallest two economies incorporating world cities are Hungary and the Czech Republic and there are only two countries above them with no evidence of world city formation (Pakistan and Nigeria). These examples show that beyond size of economy, what are important are the economic opportunities that exist at in a country. Hence the massive privatisation potentials in the former communist and Soviet republics has attracted advanced producer services so that their cities are generally over represented in world city formation processes in Table 4.

In conclusion: we can identify a roster of 55 world cities which is the necessary first step for studying global social change through the network of cities and for comparing the latter with the mosaic of states.

Surpluses: Where are the Trans-state Processes?

In John Friedmann's 'world city hypothesis' he designates a majority of the cities as articulators of national economies into the world-economy. This process seems to be born out by the analyses of Tables 3 and 4 above. In the theoretical framework being used here, we would say that these world cities have a particular monopoly of the knowledge nexus combining their national with the global. For instance, a law firm will combine professional and commercial knowledge of the local jurisdiction with abilities to work in English and/or New York State commercial law (the two law codes governing inter-jurisdictional business). Hence the 24 world cities which are the only world city in their own country each have constellations of knowledge which are unique to their localities. Certainly this is an important process but where are the trans-state processes which indicate globalization?

To begin with, we can note that there are a majority of countries which do not possess world cities. Hence their access to the specialised knowledges of advanced producer services will have to be trans-state in character. This is illustrated in Table 5 for a firm which gives details of who to contact in countries where they have no offices. This represents an articulation of small countries into the world economy: clearly three world cities dominate, Miami for central America, Paris for francophone Africa, and London for a range of Asian and African countries. However, this is not a major trans-state process: in general we can note that since all the major national economies possess world cities, the non-world city countries actually account for only a small proportion of the world gross domestic product - 12.7% to be precise. There are, of course, trans-state relations among states with world cities so that 12.7% represents the minimum estimate of trans-state processes. But how can you measure trans-state services above and beyond simple national articulations?

In our data bank we have information on firm's offices at different levels of measurement. The only level which gives a full quantitative measure of degrees of differences in provision of services between cities are the interval level measures. This is information on sizes of offices, either numbers of practitioners or numbers of fax connections. Unfortunately such information is relatively scarce: we have it for two accountancy firms, for seven banking/finance firms and for US law firms. With no such information for advertising firms this sector does not figure in the subsequent analysis. In addition, to make the law data comparable with the other two sectors we use only the largest ten firms in this analysis. Obviously numbers of firms are drastically reduced from our original analyses and therefore results are that much less reliable: think of them as suggestive rather than indicative in what follows.

Using our interval data, in Table 6 each world city is measured in terms of its level of service provision for each of the three sectors relative to the top scoring city (London once, New York twice). Summing these scores and rescaling them in relation to the top sum (New York) produces a scale from 0 to 100 as estimates of world city formation for each city. We can compare these measures to our original inventory (Table 2) remembering that this listing does not include advertising and only a proportion of the other sector firms. At the top, the main difference is Hong Kong and Frankfurt marginally overtaking Paris and Tokyo in this analysis. Elsewhere the main anomaly is Madrid which is particularly boosted by a bank headquartered in that city appearing in the data. Conversely, there are no Japanese banks in this data set. However, we will treat this scale as reasonable estimates of localised world city formation.

Given such measures we can compare world city formation within a country to the size of its economy. In Table 7 percentages of global gross domestic product and of global world city formation are listed. From this information, surpluses and deficits of world city formation relative to national demand for the services are computed for each country. It is these numbers which imply trans-state processes. Thus by far the biggest surplus is for the UK implying that London provides services for much more than just its own country. Hong Kong comes next in size of surplus and Singapore also features high in surpluses reflecting the raison d'être of each city as trans-state service centres. In contrast Japan has a massive deficit in this analysis showing that Tokyo and Osaka have advanced producer services which fall far short of what would be expected for the second largest economy in the world. The absolute size of this deficit may be due to the data limitation concerning Japanese banks mentioned above but the relative position of Japan as the highest deficit is consistent with other analyses which have used international data on service exports and which show that Japan's world share is only about half that of the UK (Enderwick, 1989). The USA pretty well breaks even in this comparison, which is surprising given the role of New York as the leading world city. It would seem that New York's global pre-eminence is countered by a relative underdevelopment of other US world cities given the huge size of the US economy.

In conclusion: trans-state processes can be inferred from comparing new measures of world city formations with standard measures of national economy sizes to show service surpluses and deficits.

Regions: Where are the Regional Centres?

Showing surpluses locates trans-state process sources but does not specify the actual processes themselves: where exactly is London's surplus provision employed? To answer such questions fully requires a different methodology than that used to create our data bank. However, we can explore one aspect of trans-state processes through analysis of office geographies by looking at the regional organisations of firms.

Because of the central place heritage of urban studies it seems that when researchers consider systems of cities they think automatically in terms of hierarchies. I have previously suggested that world city formation is distinct and different from central place processes but, nevertheless, hierarchies seem to feature prominently in the world city literature. This is largely due to Friedmann's original specification of world cities which he portrayed as a global hierarchy. For instance, his work shows that the European cities on his diagram (Milan, Vienna and Madrid) all link to Paris while only Madrid links to London. In Pacific Asia, Tokyo is specified as the 'primary city' and Singapore is particularly well linked compared to Hong Kong. However, it is not clear what the evidence is for such specific hierarchical linkages. They are reproduced in part and embellished in Peter Dicken's (1998, 210) classic text Global Shift. Here we have such detail as Zurich and Geneva linking only to Milan which this time links direct to London while Paris fields links from Frankfurt and Dusseldorf, the latter via Brussels. As with Friedmann, Tokyo dominates Pacific Asia followed by Singapore. Again the evidence for these specificities are not clear. Both studies deal with trans-national corporations in general rather than just advanced producer services and this should be kept in mind in what follows. Nevertheless, the imposing of empirical evidence on this hierarchical speculations is surely overdue.

We have good regional organisation data for just 11 firms across all four sectors: caveats of quantity of evidence apply as before. In Table 8 these firms are listed with cities which have more than one regional office. The first point to make is the low number of such cities. In fact, unlike the complicated diagrams reviewed above, there seem to be in reality just five major regional centres: Hong Kong, London, Miami, New York and Singapore. The regional responsibilities of the four other cities in the table are quite limited. Tokyo is only a centre for the region called 'Japan', Paris for francophone Africa (see also Table 5), Johannesburg for Africa, and although Brussels features as a European headquarters in one of its cases most of the personnel for this office are actually located in London. This is indicative of London's dominance amongst European world cities: as well as having regional offices for Europe, London also has more Africa offices than Johannesburg and is also a Middle East office centre. Hong Kong and Singapore share Pacific Asia office headquarters which are often split between them as north east and south east Asia respectively. The relationship between New York and Miami is less equal: where the region is the Americas, New York is the centre but where Latin America is a designated region then Miami is the centre.

This pattern of regional relations are shown in Figure 3 which contrasts sharply with the complex hierarchies in the literature. It is, in fact, in keeping with our dispensing of central place thinking in the theoretical introduction. For instance, two of the main regional centres are not in regions they organise (Miami, London for Africa and Middle East). Furthermore Miami is a major regional centre while not being a particularly important world city in its own right (only a gamma world city in Table 2). Unlike hierarchies in the literature neither Tokyo nor Paris feature as important regional centres here which is compatible with the deficits for Japan and France in Table 7. Given the paucity of regional centres, it may be that in current globalization circumstances world cities focusing upon national articulations do not require an organisation tier above them. If this is the case this would be bad news for Hong Kong, Miami and Singapore in particular. Whatever the regional-national relations, it is clear that there is no integrated hierarchy, rather we have a few specific cities developing monopolies of knowledge for given regions.

In conclusion: although our data is limited in scope, this is an area of world city formation where application of real evidence does seem to suggest overturning current ideas.

Ins and Outs: Who is Where?

In the last section I concentrated upon regional centres rather than firm's actual headquarters because, strangely, the latter is of rather less important to our analysis. Headquarters are usually a product of history, the origin city of a firm. This may be of interest for studying initial conditions leading to a global scale firm but when it reaches the latter status it chooses its regional centres as part of its global strategy. A term used to indicate the relative unimportance of headquarter locations is Wimbledonization. This refers to the situation in the City of London where the fact that only three of the top 50 banks have headquarters there is not detrimental to its overall status as a global financial centre. All top 50 banks (and many more) find it necessary to have offices in London - it is an essential place to be for banking/finance. Hence, like the tennis tournament, London provides a premier site for foreigners to play and win in. But, of course, London wins as well as it is reproduced as a global city.

This question of who is where can be neatly illustrated by a selective regional look at law firms in and out of London. In Table 9 western European countries outside the UK are divided into two groups: large countries (France, Germany, Italy, Spain) and small countries (Denmark, Finland, Netherlands, Norway, Portugal, Sweden, Switzerland - Belgium is not included because of the special case of Brussels as 'capital' of the EU). The number of law firms which have branches in London are roughly the same for these two groups. Although cities within these countries have local monopoly knowledges, it seems they benefit by joining in London's knowledge complex to boost the global side of their national-global articulations. In contrast when it comes to London firms locating in western Europe they focus almost entirely on the big four countries. Paris, Milan, Madrid and Frankfurt are part of the global network of London law firms in a way that Helsinki, Stockholm, Amsterdam and Lisbon are not. The latter's law firms need London but this is not reciprocated. This example could hardly be further from central place expectations but it is explicable in world city/monopoly terms with law firms selectively joining knowledge complexes to further their global strategies.

Finally, there is one very interesting feature of law firms which is particularly relevant to globalization debates: the concept of the 'homeless firm'. In recent years some US law firms have declined to specify a headquarters in their organisation. They argue that as global firms, all their offices are global offices linked together electronically in a single network. This is the case with largest law firm in the world, Baker and MacKenzie, who describe Chicago as their origin city and although it remains their largest office they are emphatic that it is not their headquarters. Rather they operate as a global partnership, with all partners of equal status and with equal say in the running of the firm. Table 10 shows the world regional distribution of partners and it is certainly impressive in its scope. Although the Chicago office has the most partners (70 to London's 40 in second place), there are more partners, and cities with partners, in western Europe than the USA. In Dicken's (1998) critique of the 'global firm' concept he argues that however wide a firm's global links it always maintains the business culture of its origins. Is Baker and MacKenzie a US firm? Table 10 must cast doubt upon this and therefore suggests, in this case at least, a global firm has been developed. This is suggestive of further research.

In conclusion: relations between cities across countries can be quite complex in form, the one thing all the evidence does point to is that there is no global urban hierarchy in the manner predicted by the central place model.

CONTINUING PRACTICE AND THEORY

What I have presented is a glimpse, and only a glimpse, of what a political geography for the twenty first century might look like. The limitations of the analyses are obvious: despite great effort data remains sparse hinting at rather than definitely describing city/state relations. But it is much more than a question of information. We still have problems theorising in a trans-state manner. As I have argued elsewhere, the social sciences have developed as both creations and creatures of the states so that state-centrism became embedded in their concepts and theories (Taylor, 1996). Political geography, by taking the territorial state as its prime subject, found its own particular niche in this state-centric thinking. To break free we do not have to lessen our concern for states, but rather to see them as one important element in a nexus of power which straddles geographical scales. In fact, appreciation of the importance of interlocking scales is an important general mode of dismantling state-centric social sciences.

But we are right at the beginning of trans-state research in both practice and theory. This is an on-going project to keep political geography in particular and social sciences in general relevant to social changes under conditions of globalization. In this final section I consider briefly what I believe to be two critical challenges for practice and theory respectively. First, I return to the data question. I began by arguing that globalization should be interpreted as a tendency within contemporary social change. All I have presented is data for a cross-sectional analysis of aspects of this tendency at one point in time. Studying social change demands more than a single cross-sectional trench of information. Second, we have followed the idea of territorial concentrations of specialised knowledge in world cities but have not confronted the issue of knowledge between cities which has been called 'global learning'. As well as the territorial reflexivity, there would seem to be a need for a network reflexivity under conditions of globalization.

GaWC as Embryo

Whatever the reader's opinion of the empirical analyses reported above, and I have conceded limitations as I have proceeded, it is the best there is for studying the whole global network of cities. This claim is not as immodest as it may seem: these analyses are unique, they are the best of one. All the work reported derives from a research project which is part of the Globalization and World Cities (GaWC) Research Group and Network centred at Loughborough University. GaWC is also a metageographical crusade, an evangelical movement to promote research on the space of flows to counter the state-istics biases in social research. Clearly the data I have presented here can be greatly improved upon in both quantity and quality: I offer an invitation to all to join in our attempts to create trans-state data sets which will enable globalization to be put truly in its place. The web address is: http://www.lboro.ac.uk/gawc/. A visit will provide details about GaWC, here I will just highlight two aspects.

It is not, of course, as if there has been no data collected on trans-state processes since global social research began some quarter of a century ago, it is that the production of such data has been immensely dispersed and unco-ordinated. Without a data collection agency overseeing trans-state data production, it is not just a matter of this information not being brought together, it is not in a form where this is possible. With a myriad of researchers measuring different aspects of different practices and institutions there has been no possibility of comparability over cases or over time. Hence any collection of existing data would likely be of limited utility. At GaWC we have chosen to focus upon world cities to give us a simple empirical hook in an interesting theoretical area. Given this focus, we are going back to basics as will have become obvious. The first step is to have some standardisation in what is measured and how. Only then will an accumulation of data on world cities be possible. Ways of achieving this goal are given in GaWC Briefing Papers on the web site.

Standardisation is, however, a dangerous research procedure for studying social change. Providing common measures over time freezes our concepts in a world of rapid change. GaWC must also, therefore, be a site for on-going debates about what is measured, what the results mean, and how they can be improved. One way in which this may be achieved is through the GaWC Research Bulletins which use the electronic medium in the way it is best suited: to make public research findings much quicker than the slow process of traditional journal publication. These can be accessed on the web site given above.

At present GaWC is only an embryo, please help us bring it to full term.

Network Reflexivity?

The study of networks has traditionally been quite restricted in its scale of analysis. There are numerous empirical analyses and much theorising about small group interactions and linkages but relatively little at larger scales. When we come to the world-economy as a whole we have reached the Cinderella of understanding of networks. And yet the whole notion of such an economy rests upon there being myriad transactions and flows within and across the whole system. Put simply the modern world-system is defined by its networks despite being largely studied through its mosaic of states. For any reasonably advance in our theorising of the world-system we need to consider systemic networks much more seriously than hitherto. I will conclude in a conventional world-systems analysis manner and trace systemic networks before the advent of contemporary globalization: network reflexivity, or something resembling it, has a history. Furthermore, this history concludes with an important change in network structures under contemporary conditions of globalization.

There are two major systemic networks which have operated through the half-millennia history of the modern world-system: the diplomatic network linking governments in states to one another, and the commercial network linking firms in cities to one another. Although one is inter-state and the other, typically, trans-state, structurally both have operated in quite similar ways. Both are based upon investment in collecting information, generically termed 'intelligence', to convert into knowledge to gain advantage in a complex world which is often viewed as anarchic. The latter implies unpredictability and the information, which begins as local, is converted into systemic knowledge to order the 'anarchy'. In the case of diplomacy surveillance, intelligence and opinion are collected at national and regional levels, relayed to the home capital cities where the appropriate part of the state apparatus combines it into a continually updated geopolitical code, the geographical guide to foreign policy. In the case of commercial networks similar processes are set in motion: agents in different cities relay information on prices and prospects to their home firm where the strategic arm plans its investment and growth. In this way the diplomatic network attempts to overcome the problem of operating in an anarchy of states and the commercial network attempts to overcome operating in an anarchy of markets. Recognising that knowledge is power, both networks operate under conditions of secrecy to create what they hope will be information monopolies.

Despite these synergies, there has been one important difference in the two networks. Historically, the commercial network has a tendency to be much more hierarchical than the diplomatic network. In a dispersed market buyers and sellers will soon exceed their marginal costs - paying for large numbers of agents - in seeking information over long distances. A systemic network can only fully develop beyond pure speculative behaviour through a pooling of information in a central information-handling city. This specialised information exchange can facilitate long distance transactions by selling its monopoly information. Smith (1984) describes how Amsterdam provided such a role in the early modern world-system. This city stood at the apex of a hierarchy, functionally above regional cities which all looked to Amsterdam for systemic networking. This is the role of hegemonic cities and later London and New York were to become systemic apexes of knowledge in the nineteenth and twentieth centuries respectively (Lee and Pelizzon, 1991). In contrast, although the United Provinces, Kingdom and States were consecutive hegemonic states their positions in the diplomatic political network was never so dominant. The nominal Westphalian equality of states ensured multiple capital cities as centres of political intrigue even during high hegemony.

Both networks have changed appreciably in recent years. In diplomacy the bi-lateral relations which had dominated the organisation of the network are being undermined by more multi-lateral diplomacy in numerous international institutions and frequent conferences (Berridge, 1995). And this has been reinforced by what Der Derian (1987) calls techno-diplomacy. New communication advances have produced a new speed in what had been traditionally a leisurely pursuit with now 'time replacing space as the significant medium of diplomacy' (p. 208). No longer fearing that distant ambassadors would 'go native' (Berridge, 1995, 10-1), Der Derian suggests this may be the beginning of the end of diplomacy as a means of mediating between states. However, I would argue that whereas time replaces space for information, it does not do so for knowledge; while there are states with different interest and cultures there will be inter-state diplomacy.

It is in the commercial world that the new communication revolution may be genuinely revolutionary. To begin with, it has created a situation where the simple hierarchy of past systemic commercial networks no longer obtains: with instant communication there can be global organisation. Although there is a tendency towards a 'big 2, 3 or 4' 'global cities' as we have seen, there are over 50 other cities we have identified which have world city credentials. What we have is a complex world city network which is forming and interacting with the interstates system as previously illustrated. There can no longer be a single centre at the apex of a knowledge hierarchy as in the past. Rather there is a network world to master of 'interlocal' monopolies of knowledges. For as well as local complexes of knowledge, world cities have the critical relational role of linking into a global network. Advanced producer service firms create global service products. Baker and MacKenzie, the last firm we dealt with above for instance, call themselves 'the global law firm' (Beaverstock, et al.,1999b). But to keep ahead in 'global law' or any other global knowledge-based product requires a 'global reflexivity'. This is much more than the 'global knowledge' any major corporation must now acquire: reflexivity implies a process, in this case global learning practices to develop unique 'service' products. This is a new, or at least an enhanced, form of reflexivity and learning which goes far beyond the practitioners' need to 'negotiating their way through a variety of cultural barriers in different parts of the world', as Scott (1997, 334) describes it.

Operating in a widely separated network requires different means for translating knowledge into power and thus monopoly: we do not know even if it is possible to replicate anything like regional or city territorialism at a global network scale, in whole or in part. We do know that these global relations are not simply electronic, they are literally embodied in the thousands of business trips made between world cities on every working day. Whether the resulting contacts and cultural interactions can produce the required degree of reflexivity to be worth appropriate investment in 'global learning' in a 'deep' sense is a key question for the future of globalization. If the answer is yes, and organisational studies are exploring this issue (Parker, 1996; Weick and Westley, 1996) especially in the new area of 'international human resource management' (Scullion, 1995), then the current tendency towards globalization will be able to flourish in the new century as never before: global network reflexivity will constitute a new anti-market nexus to add to Table 1. On the other hand, if there is an 'inevitable' shallowness to global reflexivity, this may well be the Achilles Heel of globalization: as a firm grows more globally, diminishing returns may set in as the quality of information (and hence knowledge and hence power) declines to create only very leaky monopolies at best. The jury is still out on the future efficacy of reflexive network knowledge monopolies and hence on the future possibilities of globalization.