"Globalization takes place in cities and cities embody and reflect globalization"
(Short and Kim, 1998, 9)
Short and Kim (1998, 9) have recently complained that much of the literature on globalization is pitched at such a 'stratospheric level' that it misses the actual activities and practices which constitute social change. Hence to understand 'how globalization takes place' requires investigation of real places, in particular large metropolises. These are generally termed 'world cities', so-called because they function to provide services of control and organisation through which the contemporary world-economy operates. It is these world cities which are the subject matter of this paper.
Research on world cities has drawn significantly upon two seminal pieces of work. The foundation article by John Friedmann (1986) set out his 'world city hypothesis' which provided a set of propositions about how the world's major cities have become the 'command centres' of global capital. In the process he identifies 30 cities as world cities (see also his slightly revised list in Friedmann (1995)) which are presented as 'the world city hierarchy': cities are divided into primary and secondary world cities and the distinction between core and semi-peripheral locations is also made. Saskia Sassen's (1991) The Global City focused upon London, New York and Tokyo as the apex of the world city hierarchy and attempted to delineate their convergence as locations for the production of advanced producer services. In the process she took our understanding of these cities beyond their role as the foremost international financial centres to a new comprehension of them as special places, knowledge complexes where advanced producer services - in accountancy, advertising, insurance, law, management consultancy, real estate, etc. as well as banking and finance - are developed and practised (see also Sassen (1994)).
While providing the necessary substructure and groundwork for an exciting research agenda, Friedmann's and Sassen's early contributions complement each other in clearly illustrating the twin limitations of that agenda. On the one hand Friedmann attempts a comprehensive global treatment of world cities but with very little empirical evidence to back up his arguments (Taylor, 1997). In contrast, Sassen marshals lots of evidence for her interpretation of the global city triad but has little to say about all the other world cities and how they fit into the picture (Taylor, 1994). Hence the legacy of their foundation is to be either globally comprehensive and empirically challenged or empirically comprehensive and globally challenged. This dilemma is currently reflected in world city research by the plethora of studies of single city or comparisons of small numbers of cities (e.g. Todd (1995) on Toronto and Abu-Lughod (1995) on New York, Los Angeles and Chicago) in contrast to the few attempts to encompass world-wide patterns which invariably suffer from inadequate data (e.g. Michelson and Wheeler's (1994) use of Federal Express data, Rimmer's (1998) use of international airline data which describe general flows, including, for instance tourism, and omits important domestic links such as New York-Los Angeles). It is the purpose of this paper to overcome this twin limitation by providing an empirically rich and globally comprehensive study of world cities.
Our approach is to unpack standard concepts such as 'world city hierarchy' and 'global urban system' and focus on the fine grain of contemporary urban development under conditions of globalization. What we produce is a complex interweaving of hierarchical tendencies with distinct regional and inter-regional patterns. There is certainly no blanket globalization effect as often portrayed in the non-geographical literature (see also Taylor et al. 2000; Taylor and Hoyler, 2000). Globalization has many geographies and we aim to describe its basic urban structure (Taylor, 2001). In short, while it has been commonplace to identify multiple 'locals' in the global-local nexus, we introduce the idea of there being multiple 'globals' in the sense of different globalization outcomes resulting from variations in corporate location strategies. This paper illustrates this statement through an unusually detailed empirical analysis of world cities.
The title of this paper includes the word 'first'; we use this in both an immodest and modest manner. We do think that by combining the best feature of Friedmann (his global vision) with the best feature of Sassen (her detailed investigation of advanced producer services) we are able to produce the first empirical analysis which seriously goes beyond their legacy. At the same time, we are mindful that in moving into uncharted waters our study is inevitably exploratory producing what can only be preliminary first results. In fact we produce two sets of results, one service-centred and the other city-centred, which constitute the two substantive sections of this paper. These deal with the cross-city profiles of service firms and the corporate service mixes of cities respectively. Before we embark on describing these results it is necessary to outline both our data collection and the quantitative methodology we apply to the data. The first section below describes the data we use, measures of the service supplied by 46 producer service firms across 55 world cities. In the second section we show how this 46 x 55 data matrix is reduced to basic dimensions, Eight cross-city profiles and nine corporate service mixes, by using the standard multivariate statistical technique of principal components analysis. Hence our main general conclusion: how complex globalization turns out to be when studied empirically through producer services in world cities. This is the theme of our concluding section where we draw on our results to suggest an enhanced agenda for world city research set in an empirically rich and globally comprehensive context.
CREATING A SET OF GLOBAL DATA
A major reason why there has been a dearth of global-scale cross-city studies is because there are no easily accessible data available for analysis. Most published data are created by states for states one result of which is a very state-centric social science (Taylor, 1997). Even world city research has succumbed to this debilitating bias (Taylor, 1999). For sure there are census data and related materials collected by state agencies to describe cities: some limited cross-city comparisons are possible once allowance is made for differences in measurement. But there is no comprehensive collection of published data on world cities: to do global research on cities requires the creation of global-scale data. This study is part of a research initiative which does just that (Taylor, Walker and Beaverstock, 2000).
The Globalization and World Cities Research Group at Loughborough University has collected comprehensive information on the offices of 74 advanced producer service firms in 263 cities. These offices are the outcomes of location decision making by firms operating under conditions of globalization. Their cross-city patterns define a world city network (Taylor, 2001). This represents the basic 'skeleton', as it were, of globalization. However because this information is largely obtained firm by firm there is no standardised composition of the data (Beaverstock et al., 2000). The kind of information we have on each firm's offices varies from simple presence in a city through to number of practitioners employed in each city. Turning this into a data set for use in our analysis involved two stages, first selecting cities and second selecting firms.
The creation of a roster of world cities is described in detail in another paper (Beaverstock, Smith and Taylor, 1999). The basic method we have used is to consider four service sectors (accountancy, advertising, banking/finance and law) separately and identify the leading cities in each sector. Cities were scored 3,2,1 depending on their importance in a given sector and combining these scores enabled us to produce an ordering of cities up to a maximum aggregate score of 12. From this list we derived our 'inventory of world cities' using the threshold score of four to qualify. Of the initial 263, 142 cities appeared in the different sector lists but only 55 exceeded the threshold in the aggregate ordering. Cities deemed to have world city status are listed in the appendix where they are divided into three levels in terms of overall importance as service centres. There are no particular surprises in this roster: globalization when viewed through cities is very uneven. For this reason we have found it necessary to illustrate the spatial pattern of the cities in the form of a cartogram depicting the roster as three separate groupings: the Americas, Europe and Pacific Asia (Figure 1).1 This schematic diagram is the key for displaying our results by cities below.
For this analysis we focus upon the major global firms in our data. The threshold we use is that a firm must have offices in at least 15 separate cities. Setting up an office to provide advanced producer services is a very expensive undertaking for a firm since it is the nature of service provision that it takes time to build up clienteles. Furthermore, operating in this many cities inevitably means working in different legal frameworks and cultural settings. In short, to have 15 or more offices shows a firm to be committed to developing a significant cross-city provision for their particular producer service, we will call them global service firms. There are 46 firms in this category in our data and they are listed in the appendix. The variations in information we have on each office is dealt with by converting all the data into a simple ordinal scale (0, 1, 2, 3) as described elsewhere (Taylor, Walker and Beaverstock, 2000) where the higher the score the larger the service provision by a firm in a city for a given sector. For instance, New York and London have more scores of three because most firms have decided they need to locate one of their largest offices in these two very important cities. Note that for eleven firms we have only presence/absence data (scored 0,1).
The end result of these operations is to create a data matrix of 46 global service firms across 55 world cities in which each cell records high (3), medium (2), low (or simple presence) (1), or absence (0) for a given firm in a given city. This constitutes the input to our analyses.
METHODOLOGY: PARSIMONY THROUGH PRINCIPAL COMPONENTS ANALYSES
Faced with a matrix which contains 2,530 (55 x 46) pieces of information we need to reduce the detail into a relatively small number of common patterns for interpretation. Such parsimony is the basic purpose of the factor analytic family of statistical techniques of which the most straightforward (i. e. with least axiomatic baggage attached) is principal components analysis. This is the technique we apply to our cities/firms matrix.
In principal components analysis, a data matrix consisting of x variables is treated as an x-dimensional space to which each variable contributes an axis. Each axis is one unit in length which represents the spread of values (variance) of a variable so that the total variance is x, the number of variables in the matrix. By analysing the co-variance (correlation) amongst the variables, an alternative set of axes of different lengths can be produced ranked by size. These are principal components, in which the first (largest) component describes the biggest cluster of co-variance amongst the original variables, the second component the next biggest of cluster of co-variance, and so on down to a very small final xth component. The idea is that most of the variation in the data is revealed in a relatively small number of large components so that many small components can be discarded as unimportant. The result, therefore, is to transform an original x-dimensional variable space into a much smaller y-dimensional component space defined by selection of only the large and important principal components. This is the parsimony, converting a large number of variables x into a relatively small number of relevant components y which, nevertheless, between them account for a large proportion of the original variance. For example, a very successful parsimonious analysis would transform, say, 80 variables into just 6 principal components which account for 75% of the original variation.
This is a standard statistical procedure, the skill is in the interpretation of the principal components. The chief way of interpreting components is to focus upon the location of these new component axes with respect to the original variable axes. These relations are measured as 'loadings' which are the correlations between each component and each variable. By looking at the high correlations, the cluster of variables which a component represents is revealed. In order to facilitate interpretation, the principal components are usually 'rotated' so as to maximise high loadings. The most common method is called varimax rotation which creates well-defined and orthogonal (independent) patterns of variability as new components. It is these rotated principal components which we use here for parsimonious analysis of the cities/firms data matrix. In effect, these rotated components are new 'super-variables' and as such they can be defined in terms of the original objects over which the input variables were measured. These 'component scores' are like the original variable measures and tell us about the differences between objects for a given component.
There is one crucial decision which has to be made in any principal components analysis: what is y, the number of relevant components to be selected from the x number of components which are actually created. There is no simple statistical answer to this question, ultimately it comes down to the researcher's judgement. In the analyses below, we use the maximum interpretable component method. This involves starting with a small number of components and then adding extra components one at a time and rotating them to see whether the last component is interpretable. To determine the latter we have arbitrarily set a threshold of a component having at least one loading above 0.6. Hence, when we come to a rotated solution where there is a component without a 0.6-or-above loading, We reject that analysis and return to the penultimate analysis as that containing the maximum interpretable components.2
Finally there is the question of variables and objects. Since a matrix has two sides there are two ways of considering its variability, by columns (variable) or by rows (objects). Most principal component analyses use correlations between variables (called-R-mode analysis) as described above. In our study the variables are the 46 firms' distributions of offices across cities and our R-mode analysis uses the correlations between firms to identify 8 interpretable components, that is to say, it reduces the 46 variable-dimensions to 8 component-dimensions. However, it is just as statistically feasible to analyse from the perspective of the objects (called Q-mode analysis), which involves creating components out of the correlations between objects. In our study the objects are the 55 cities' distributions of offices across firms and our Q-mode analysis uses correlations between cities to identify nine interpretable components, that is to say, it reduces the 55 object-dimensions into 9 component-dimensions. In simplest terms, parsimony is achieved by the R-mode analysis seeking out patterns of similarity between firms, and the Q-mode analysis seeking out patterns of similarity between cities. We discuss the results of each analysis in turn.
CROSS-CITY PROFILES OF FIRMS: DIMENSIONS OF GLOBAL SERVICE LOCATION STRATEGIES
Our data describes the location outcomes resulting from 46 different world city office strategies pursued by global service firms. Bi-variate correlations between these outcomes shows the degree of similarity between pairs of firms in terms of their world city locations. For instance, the highest correlation (0.776) is between two London law firms, Allen and Overy and Freshfields, indicating two quite similar distributions of world city offices. Principal components analysis of the 46 x 46 correlation matrix (R-mode) using the extraction and rotation methods described above reduces the 46 different patterns to 8 components which account for 84% of the original variability in the data. We term these principal components cross-city profiles.
The 8 cross-city profiles are independent or separate (orthogonal) patterns of world city office locations. In the first instance these profiles can be interpreted through the firms which load (correlate) highest on each principal component. These describe clusters of firms with similar office location patterns. Table 1 shows a ranked ordering of all loadings above 0.4 although we focus particularly on the top loadings in our interpretation of each case. The scores for the 8 components show the pattern across cities. These are depicted in Figure 2 which uses the cartogram from Figure 1 as its key. These spatial patterns provide a second source for interpreting the components. Hence for each component we have a list of firms ranked by loadings (Table 1) and its geographical pattern across cities measured by its scores (Figure 2).
The first finding of this analysis is that firms from the different service sectors strongly tend to be loaded on the same components. That is to say, there is relatively little cross-sector similarity in corporate service office patterns. Thus we organise the discussion by service sectors in the first instance.
Cross-city Profiles of Law Firms
In this analysis the 16 law firms divide neatly into two components on which 14 of them load above 0.4 (Table 1).
The US law firms' cross-city profile (principal component 3 accounting for 18.2% of the total variance) is statistically the most important component and it is the easiest to interpret: 11 of the 13 US law firms in our data load on this component. Despite the high variance accounted for, the scores (Figure 2(a)) show the simplest cross-city pattern featuring just six cities. What this tells us is that US law firms as a whole have a very simple core global location strategy which can be portrayed as follows. They focus particularly on lobbying in Washington, DC and financial work in New York, plus using London as an international centre, Los Angeles as a West/Pacific centre, and they have also moved into Warsaw and Prague as a result of 1990s privatisation opportunities in eastern Europe.
It is noteworthy that Baker and McKenzie, the world's largest international law firm, is the glaring omission from firms loading on this component. Originating from Chicago, they aspire to be a global law firm and are located in far more cities than any other law firm (Beaverstock, Smith and Taylor, 2000b). Their uniqueness is reflected in the fact that they do not appear at all in Table 1, their location strategy is distinct and separate from all eight general profiles we have created.
London law firm's cross-city profile (principal component 8 (14%)) could be viewed as a mixed sector dimension since there are four banks featured in Table 1. Nevertheless, the three London law firms in our data dominate by the size of their loadings which is how we identify this component. The pattern of scores (Figure 2(b)) indicates a European bias in the profile plus international financial centres (Tokyo and Singapore) with a paucity of offices in north America. It is the European bias which results in the four European-centred banks also loading on this component. The office location strategy represented here can be portrayed as follows. London law firms have taken advantage of their proficiency in English law (as one of two international commercial law codes, the other is New York state law) to expand into European markets (and to a lesser extent Pacific Asia) but have not attempted to compete with US firms in their home market.
Cross-city Profiles of Banking/Finance Firms
Although four of the 14 banking/finance firms in our data load on a law component, this has not prevented two distinctive banking dimensions from being identified.
The major banking/finance cross-city profile (principal component 1(13.2%)) includes half the banks in the data (Table 1). The scores (Figure 2(c)) show a very predictable pattern highlighting the main international banking centres across the world and also, negatively, those cities with a relative paucity of banking/finance functions. Note that there is no regionality to this cross-city profile; as we have come to expect, banking/finance is the most global of the producer services.
The minor banking/finance cross-city profile (principal component 7(6.7%) is statistically the smallest component we interpret and has only two loadings, both banks, in Table 2. It is basically the Citibank profile with a little input from Barclays. The scores (Figure 2(d)) show a global distribution but a very different pattern from the international financial centres of the other banking profile. In this case capital cities are featured, notably Washington, DC and Brussels which can be portrayed as a political location strategy but given this evidence it can only be an idea for possible future exploration.
Cross-city Profiles of Advertising Firms
The advertising firms define two dimensions which are particularly distinctive in their regionalities.
The European advertising cross-city profile (principal component 4 (10.6%)) has five advertising firms loading on it in Table 2. The scores (Figure 2(e) show a very distinctive geography which contrasts Europe (plus Sydney and Toronto) with US and Japanese cities. This cross-city profile can be portrayed as the outcome of locational strategies which have targetted European national markets but have chosen not to compete in the two largest domestic markets, the USA and Japan.
The Latin American advertising cross-city profile (principal component 5 (10.1%)) has four advertising firms contributing to it (Table 2). We have called it Latin American even though only two of the six cities with positive scores in Figure 2(f) are strictly Latin American (in addition, Buenos Aires just misses the 0.4 threshold). However three of the other cities have language and/or political-cultural links to Latin America: the highest scoring city Miami, Los Angeles and Madrid. European and northern USA cities dominate the negative scores. Hence this cross-city profile may be reasonably portrayed as the result of location policies for servicing Latin America.
Mixed Sector Cross-city Profiles
There are two dimensions which feature prominently more than one sector, one is very mixed, the other a dual-sector component.
The general mixed sector cross-city profile (principal component 2 (15.1%)) is statistically the second most important component, firms from all four sectors are featured in the loadings reported in Table1. This is where four of the five accountancy firms appear in this analysis (Arthur Anderson does not appear in any of our components suggesting a unique cross-city profile like Baker and McKenzie). Accountancy is the concentrated of all producer service sectors resulting very large firms having the most widespread pattern of offices of all services. However, we should not treat this component as an incipient 'globalization' dimension as the scores clearly show (Figure 2(g). The positive scores have a North American bias plus Europe's two leading world cities (London and Paris) but it is the negative scores which are particularly interesting in this case. Featuring eastern Europe and Pacific Asia, these are the two 'frontier' globalization arenas of the 1990s, new opportunities but high risks. This cross-city profile may be portrayed, therefore, as bringing together firms whose location strategies have avoided the new globalization arenas.
The Pacific Rim dual-sector cross-city profile (principal component 6 (12.4%) has two services contributing to it, advertising and banking. The key to this component lies in the scores (Figure 2(h)): they show a clear Pacific Rim pattern plus the two leading world cities, London and New York. In many ways this is the most regional of all the cross-city profiles. It can be portrayed as the result of location policies focussing upon the Pacific Rim.
Beyond the intrinsic fascination of the specificities of the cross-city profiles, there are two important empirical findings which inform us about the geography of globalization. First, we have achieved only a relatively modest parsimony: as just noted, there is the great variability which requires 8 components to summarise the office geographies of 46 firms. But this is itself a very interesting finding. The concept of globalization, and with it such notions as world city network and hierarchy, are commonly viewed in the singular as an all-encompassing process. Discussion of the local-global nexus, for instance, implies many locals but only one global. This analysis shows globalization to be an immensely complex process or set of processes. There is not a simple network of world cities out there which global service firms are attaching themselves to in a predictable manner. Even this one narrow element of globalization - the city-directed strategies of just one group of firms - has indicated several layers and patterns of outcomes. Second, and closely related to the variability, there is the high degree of separation in cross-city profiles between the four service sectors we are studying. Different sectors have different office geographies which means that researchers cannot use their relatively extensive knowledge of international financial centres to predict the locational strategies of services for which we have much less knowledge such as law. In short, there are many geographies of globalization and until these are comprehensively and clearly delineated our knowledge of globalization will be partial, dependent on either the geographical myth of a singular pattern or on geographical speculation based upon case studies or even simply anecdotes.
This analysis has done its job to the extent that geographical complexity and sector autonomy have been identified. As a cross-sectional analysis of office geographies producing cross-city profiles at one point in time, there is a limit to how far our interpretation can be extended into explanations. We advocate combining an evolutionary approach with a place-sensitive approach which we might call geohistorical. The different service sectors have different degrees of globalization - among our four sectors accountancy firms are the most globalized, law firms the least - and have different situated information/knowledge needs - banks and financial firms operate in largely trans-state markets whereas advertising is still closely bound to national markets, for instance. These are the sort of factors which geohistorical research would explore to make sense of the sort of results we have produced, but that is for another project. Here we can usefully augment our cross-sectional findings by focusing on the world cities themselves in terms of the types of service mixes to be found within them.
CORPORATE SERVICE MIXES OF CITIES: GROUPING WORLD CITIES BY CORPORATE LOCATIONS
Looking at our data from the perspective of the 55 world cities, each one has a unique combination of service firms located within it. Bi-variate correlations between these city mixtures of firms shows the degree of similarity between pairs of cities in terms of the firms located in each of them. For instance, the highest correlation (0.863) in our data is between Amsterdam and Toronto indicating a large overlap in the service firms to be found in each city. Principal components analysis of the 55 x 55 correlation matrix (Q-mode) using the extraction and rotation methods described previously reduces the 55 different city combinations of firms to 9 components which account for 80.1% of the original variability in the data. We term these principal components corporate service mixes.
The 9 global service mixes are independent or separate (orthogonal) patterns of world city office combinations. In the first instance these mixes can be interpreted through the cities which load highest on each principal component. Table 2 shows a ranked ordering of all loadings above 0.4 although we focus particularly on the top loadings in each case. In this mode of analysis it is the loadings themselves which can be directly mapped to show the geographical pattern of the various components. These are depicted in Figure 3 using the cartogram of cities in Figure 1 as the key again.3
The first finding of this analysis is that despite globalization, corporate service mixes are distinctly regional in nature. Hence the nine principal components can be classified by their geographical scope into inter-regional, regional and specific cities (Figure 3). We use this ordering to structure the discussion below.
Interregional Clusters of Cities
The corporate mix in major transnational and Latin American world cities (principal component 3 accounting for 22.3% of the variance) is in many ways the most complicated partly because it is the most global in its distribution (Table 2; Figure 3(a)). It consists of two distinct groups of cities. First, in terms of loadings it is headed by Zurich, the archetypal transnational centre; Frankfurt and Tokyo are two other such major centres which load high. Second, all five Latin American world cities load high. The reason for these two distinctive sets of cities coming together in a single component is the particular role of Latin America in globalization. Previous related research (Beaverstock, Smith and Taylor, 1999, 2000) has shown that in the late 1990s Latin America has been a relatively minor globalization arena compared to south east Asia, east Asia, western Europe and eastern Europe. However, they have featured in earlier 'internationalization' of services, especially by US firms (see Beaverstock, Smith and Taylor (2000) for law firms. Hence, their corporate mixes have been relatively stable recently and in this way they are similar to the established transnational centres: well-established global service firms will tend to be located in both groups of cities. Note the omission of the two major transnational centres from this component: London and New York have their own distinctive mix which we will come to below.
The corporate mix of minor Pacific Asian world cities (principal component 1 (18.1%)) is much more straightforward with most of the less important Pacific Asian world cities from both south east Asia and east Asia (i.e. from Jakarta and Bangkok to Seoul and Beijing) loading particularly high (Table 2; Figure 3(b)). There are two interesting points here. First, the two regional centres, Hong Kong and Singapore, have relatively low loadings; their corporate mix shares some elements with the transnational mix shown by the modest loadings on the previous component (see Table 2). Second, the two Japanese world cities do not feature in this component; we have already seen that Tokyo belongs to the transnational group and Osaka will appear on a later component. This finding is also consistent with previous related research which showed that although Tokyo is the most important world city in Pacific Asia it does not appear to have developed a leading regional role (Taylor, 2000).
The corporate mix of minor north Atlantic world cities (principal component 2 (16.3%)) is dominated by northern European cities without major transnational linkages which are combined here with some minor north American world cities (Table 2; Figure 3(c)). This is an unexpected cluster of cities and seems to be bringing together those world cities where banking/finance is relatively unimportant (see negative scores in Figure 2(d)).
These are the three most important components and between them account for a little over half the variation in combinations of firms in cities.
Regional Clusters of Cities
The corporate mix of USA world cities (principal component 9 (9.5%) has the political capital, Washington, DC, with the highest loading in a selection of eight US cities, out of a total of 11 US cities in the data, loading above 0.4 (Table 2; Figure 3(d)). There is no internal regional bias in this grouping of cities: the top four loadings cover all major US regions: North (Washington, DC), South (Dallas), Central (Chicago), and West (San Francisco). In addition the three missing US world cities are regionally spread (New York, Minneapolis and Miami).
The first corporate mix of eastern European world cities (principal component 5 (9.2%)) is picking out the organisation centres - Prague, Warsaw, Budapest - of a new globalization arena which has emerged in the post-Cold War world (Table 2; Figure 3(e)). The inclusion of Istanbul in this grouping given its non-communist background but eastern European location is quite interesting. The omission of Moscow is also noteworthy implying different firms being attracted to the ex-USSR than to its former satellites.
The corporate mix of western European cities (principal component 6 (8.5%)) only has partial coverage of the region because of previous selections of European cities in other components (Table 2; Figure 3(f)). Hence there is a southern bias but, like the USA component, this selection is headed by the political capital, Brussels, with the highest loading.
The second corporate mix of eastern European world cities (principal component 8 (5.1%)) confirms the separate nature of Moscow's attraction to service firms compared to other eastern European cities (Table 2; Figure 3(g)) which has figured previously (Figure 3(e)). In some ways this city's role is similar to that of Tokyo, the leading city of its region but without regional leadership.
The addition of these four principal components means that all major globalization arenas are covered in the clusters of cities dealt with so far. However, there are still some important, indeed very important, cities to be accounted for.
Specific City Dimensions
The final two components are relatively unimportant in statistical terms but describe crucial elements of the world city network.
The global city corporate mix (principal component 3 (5.1%)) highlights New York and London (Table 2; Figure 3(h)). These two cities appear at the top of the rankings in almost all comparative studies of world cities (Beaverstock, Smith and Taylor, 1999b) and are clearly worthy of the title global city as bequeathed by Saskia Sassen (1991). However, it is noteworthy that the third of Sassen's triad, Tokyo, does not load on this component, we found it earlier as a transnational centre equivalent to Zurich and Frankfurt rather than New York and London. The appearance of Los Angeles with a higher loading on this component rather than the USA component is interesting; this is consistent with other related work which shows Los Angeles to be more globally linked than the USA's other major world city below New York, Chicago (Beaverstock, Smith and Taylor, 2000b).
The Miami corporate mix (principal component 7 (4.3%)) is the smallest statistically but one of the most interesting of the components uncovered (Table 2; Figure 3(i)). With only a single loading over 0.4, it shows this city to have a unique corporate mix. This is consistent with previous research which identifies Miami as a very unusual regional corporate centre (Taylor, 2000). Unlike other key regional centres (London, New York, Singapore, Hong Kong), Miami is not of itself a high ranking world city but nevertheless serves as headquarters to many firms' Latin American business. It does not load on the first inter-regional component with the Latin American cities because it has the special and distinctive gateway role.
From these results on corporate mixes of groups of cities we can begin to infer something of the roles and functions of world cities within the overall context of a globalising world-economy. Certainly the regional nature of our city groupings is instructive, globalization as evidenced by world city formation seems to be developing through a small number globalization arenas: it comprises largely of a collection of distinct regional patterns rather than the general global zones (such as core and semi-periphery). However, most comparative studies of world cities emphasise the hierarchical features of the system. Principal components analysis is not designed to identify hierarchies but nevertheless our results do show features which are consistent with such ordering. It is particularly noteworthy that the specifically identified regional components are generally missing their leading cities or else they have relatively small loadings, for instance: London and Frankfurt are missing from 'northern Europe'; Paris and Milan have small loadings for 'western Europe'; New York is missing from both 'northern America' and 'USA', Los Angeles is missing from the former and has a small loading on the latter; and Tokyo is missing from Pacific Asia, Hong Kong and Singapore have relatively small loadings on this component. These cities are all what we have termed 'alpha' world cities (Beaverstock, et al., 1999a and see appendix A).
Looking to where these alpha cities do feature in our analysis, the specific global city component does imply New York and London at the very top of the world city hierarchy. Obviously no surprise there, and similarly the same can be said of the five alpha cities which occur in the largest component as transnational cities (Frankfurt, Tokyo, Milan, Singapore and Hong Kong). Of the other three alpha world cities, Los Angeles loads moderately on the global city component and that leaves Chicago and Paris as the only alpha cities which appear solely in their respective regional groupings. Paris is perhaps the most problematic with its weak loading on 'western Europe' perhaps preventing it having its own unique component like Miami. Chicago is different, despite its high world city status it appears as a US world city no different from lower status American cities.
What is clear is that our analysis and interpretation, by combining regional and hierarchical elements, has found an order within the corporate mixes of world cities but it is an order which is much more complex than commonly supposed. We should be very suspicious of presentations of simple world city hierarchies with neat national and regional articulations nested under global cities such as John Friedmann's (1986, 1995) pioneering and oft-quoted formulation tells it.
These standard multivariate statistical analyses of office geographies have produced two sets of intriguing results. As we emphasised in the introduction, this is a first quantitative analysis of contemporary world cities at a global scale and therefore the findings are specifically interesting as a comprehensive empirical glimpse of the comparative roles and functions of world cities. We invite replications of this research using other firms (and possibly other cities) to show which of our results are reasonably robust and which may be specific artefacts of our particular data set. Here we will proceed on the assumption that our results are tolerably robust thus allowing us to develop some general conclusions. These come in two parts: first, we consider the complexity of globalization as evidenced in our analyses; and, second, we suggest research agendas which derive from this work.
The Geographical Complexity of Globalization
This research challenges the idea of globalization as a single all-pervasive spatial process. Neither of our analyses produce a large 'first factor' which we could label 'globalization' or 'global hierarchy'. Rather, although we have used a technique which searches for parsimony, by producing results requiring first eight and then nine interpretable components, we have been left with a representation of the real complexities of globalization. Clearly there is not a simple hierarchy of cities which office network geographies fit into.
Principal components analysis provides groupings of like items but these are not discrete categories: there is always overlap between components. In Figure 4 the nine city components are depicted in venn-like diagrams with sets defined by city loadings above 0.4. Cities with loadings above 0.7 define the 'cores' of the component-sets. This diagram divides the cities into three types: 26 are component-core cities, 19 are overlap cities which lie between components, and10 are found in a single component-set but outside its core. The overlap category is the interesting one, it is neglected in our previous component by component description. Here we see that although the component-core cities define the structure of the data (i.e. provide the basic identifications of components), the overlap cities re-inforce the meanings of the components in terms of which components are linked together: there is an order to the complexity of Figure 4. For instance, the international/Latin America component is linked to other regional components through key important regional cities: note Madrid and Milan for western Europe, Singapore and Hong Kong for Pacific Asia, Toronto for northern Atlantic, and Houston for USA. Similarly, the nothern Atlantic component links to western Europe through Amsterdam, to USA through Minneapolis and Atlanta, and to the two eastern Europe components through Stockholm and Berlin. Two other features are noteworthy: first, the relative isolation of the global city component linking only to USA through Los Angeles and Boston, and second, the absolute isolation of Miami. In summary, there is a meaningful pattern to the components beyond the cores but this does not lessen our overall argument for geographical complexity.
Further Research Agendas
Finding an order in the geographical complexity provides research leads to new questions for investigation. As mentioned previously, there is a need for a geohistorical approach to supplement this cross-sectional analysis. For instance, looking at the two central sets in Figure 4, they both represent established world cities but of different statuses (more alpha and beta cities in the international/Latin America component, see Appendix A) whereas the other large set, Pacific Asia is defined by relatively new world cities. Hence the latter does not include the two Japanese world cities in our data. This Asian component, along with the two eastern European components, represent the 'frontiers' of expanding globalization in the last decades or so. Such ideas can only be explored with data which allows analysis over time. However, perhaps the main benefit of such dynamic analysis would be to see whether the complexity we have found at the very end of the 1990s represents a dimunition of complexity compared with previous recent times, say 1990 and 1985. Is the complexity we have found an incomplete globalization as corporate service firms move towards similar city hierarchies in their office networks? We will only know the answer when the extra research is carried out.
These system-wide considerations can be complemented by a more city-specific research agenda. We will conclude with some very concrete suggestions to illustrate how our general analysis does inform our knowledges of specific cities. Here is a list of ten research leads which can be easily identified from our analyses. Some will be relatively familiar where there has been research on the topic but other themes would seem to be new. The suggestions fall into two groups, five concerning single cities and five concerning comparisons between cities.
Particularly interesting single cities that emerge from this work are as follows.
Paris as a distinctive alpha world city because it hardly registers in our analyses.
Istanbul as a link between east and west (resurrecting its traditional role) given that it is surprisingly grouped with post-communist eastern European cities.
Johannesburg as the most isolated world city in that it has a continent to itself and no clear similarities with other world cities.
Miami as the most peculiar world city being distinctive because it combines low world city status with a major regional centre role.
Osaka as the most difficult city to interpret because it seems to be unlike other Pacific Asia cities and also unlike the other Japanese world city, Tokyo, with its transnational role.
Some particularly interesting city comparisons to emerge from this work are as follows.
London and New York as the global city duo given their unique similarities which produces a distinctive and separate city grouping.
Washington, DC and Brussels as political centres which lead regional groupings separate from the global and transnational cities within their respective world regions.
Hong Kong and Singapore as Asian gateway cities and both cities exhibiting remarkably similar affiliations as both Pacific Asian and transnational cities.
Moscow and Tokyo as leading regional centres which do not lead their regions, an unlikely pairing but their common failures to have local regional dominations is clearly intriguing.
Los Angeles and Chicago as US alpha world cities but in the shadow of New York, it seems that the former is the more successful in expanding its role beyond the national scale.
No doubt, there are many more 'loose ends' which can be identified from our analyses but the point is that a new integrated world city research agenda is in the process of being developed out of the geographical complexity of globalization.
This research builds upon an ESRC funded project "The geographical scope of London as a world city" (R000222050). We thank Jon Beaverstock (co-applicant) and Richard Smith (research associate) for their contributions to the original project and Michael Hoyler for his contributions to the analysis.
Abu-Lughod, J L (1995) Comparing Chicago, New York and Los Angeles: testing some world city hypotheses, in P N Knox and P J Taylor (eds) World Cities in a World-System. Cambridge, UK: Cambridge University Press
Beaverstock, Lorimer, H, J V, Smith, R G, Taylor, P J and Walker, D R F (2000) Globalization and world cities: some measurement methodologies, Applied Geography, 20, 43-63
Beaverstock, J V, Smith, R G and Taylor, P J (1999a) A roster of world cities, Cities 16, 445-58
Beaverstock, J V, Smith, R G and Taylor, P J (1999b) The long arm of the law: London's law firms in a globalising world-economy, Environment and Planning A 31, 1857-76
Beaverstock, J V, Smith, R G and Taylor, P J (1999c) The global capacity of a world city: a relational study of London, GaWC Research Bulletin 7
Beaverstock, J V, Smith, R G and Taylor, P J (2000a) World city network: a new meta-geography?, Annals, Association of American Geographers , 90, 123-34
Beaverstock, J V, Smith, R G and Taylor, P J (2000b) Geographies of globalization: US law firms in world cities, Urban Geography 21. 95-120
Friedmann, J (1986) The world city hypothesis, Development and Change 17, 69-83
Friedmann, J (1995) Where we stand: a decade of world city research, in P N Knox and P J Taylor (eds) World Cities in a World-System. Cambridge, UK: Cambridge University Press
Michelson, R L and Wheeler, J O (1994) The flow of information in a global economy: the role of the American urban system in 1990, Annals, Association of American Geographers 84, 87-107
Rimmer, P J (1998) Transport and telecommunications among world cities, in F-C Lo and Y-M Yeung (eds) Globalization and the World of Large Cities. Tokyo: United Nations University Press
Sassen, S (1991) The Global City. Princeton, NJ: Princeton University Press
Sassen, S (1994) Cities in a World Economy. Thousand Oaks, CA: Pine Forge
Short, J R and Kim, Y-H (1998) Globalization and the City. London: Longman
Taylor, P J (1997a) Hierarchical tendencies amongst world cities: a global research proposal, Cities 14, 323-332
Taylor, P J (1997b) Embedded statism and the social sciences: opening up to new spaces, Environment and Planning A 28, 1917-28
Taylor (1999) So-called "world cities": the evidential structure within a literature, Environment and Planning A, 1901-04
Taylor, P J (2000) World cities and territorial states under conditions of contemporary globalization, Political Geography 19, 5-32
Taylor, P J (2001) A specification of the world city network, Geographical Analysis, 23,
Taylor, P J, Doel, M A, Hoyler, M, Walker, DRF, and Beaverstock, J V (2000) World cities in the Pacific Rim: a new global test of regional coherence, Singapore Journal of Tropical Geography
Taylor, P J and Hoyler, M (2000) The spatial order of European cities under conditions of contemporary globalization, Tidjschrift voor Economische en Sociale Geografie
Taylor, P J, Walker, D F R and Beaverstock, J V (2000) Introducing GaWC: researching world city network formation, in S Sassen (ed) Cities and their Cross-Border Networks. Oxford: Blackwell Publishers
Todd, G. (1995) 'Going global' in the semi-periphery: world cities as political projects. The case of Toronto, in P N Knox and P J Taylor (eds) World Cities in a World-System. Cambridge, UK: Cambridge University Press
1. It should be understood that this form of depicting the cities in no way impinges on our analysis. Like all mappings we have made decisions on portrayal to best display our information. For instance, for cartographic convenience we have located Johannesburg in the 'Europe box' but it can be noted that this has no influence on the subsequent input and output concerning this city in our analysis.
2. This is the final decision of many which have to be decided upon before a principal components analysis can be carried out. It is our judgement that this set of decisions provides the best analysis of the data for the purposes we have set ourselves here. Other decisions will produce different specifics in the findings but we think the overall general pattern of results is reasonably robust. For readers who want to explore this matter further, the original data matrix is available from the GaWC web site.
3. In this particular analysis the matrix has more inter-correlated 'variables' (the 55 cities) than 'objects' upon which they are measured (the 46 firms) which means that scores cannot be generated in any case.
Table 1: Cross-city profiles (R-mode loadings)
(including all loadings above 0.4)
Sectors are indicated by text style: accountancy, advertising, banking/finance, law. For names of firms, see Appendix B.
Table 2: Corporate service mixes (Q-mode loadings)
(including all cities with loadings above 0.4, for city codes see Appendix A2)
A. INTER-REGIONAL DIMENSIONS
B. REGIONAL DIMENSIONS
C. CITY DIMENSIONS
Figure 1: World cities: three major zones
Figure 2: Cross-city profiles of service sectors
(a) US law (b) London law (c) Major banking/finance (d) Minor banking/finance
(e) European advertising (f) Latin American advertising (g) General mixed sector (h) Pacific Rim dual-sector
Figure 3: Cities with similar corporate service mixes (for city identification see Figure 1)
(a) Major transnational and Latin American world cities (b) Minor Pacific Asian world cities (c) Minor north Atlantic world cities (d) USA world cities
(e) First eastern European world cities (f) Western European world cities (g) Second eastern European world cities (h) Global city
Figure 4: The complexity of globalization
A. WORLD CITIES
1 World cities by strata: cities are ordered in terms of 'world city-ness' with a maximum value of 12
ALPHA WORLD CITIES
12: London, Paris, New York, Tokyo
10: Chicago, Frankfurt, Hong Kong, Los Angeles, Milan, Singapore
BETA WORLD CITIES
9: San Francisco, Sydney, Toronto, Zurich
8: Brussels, Madrid, Mexico City, Sao Paulo
7: Moscow, Seoul
GAMMA WORLD CITIES
6: Amsterdam, Boston, Caracas, Dallas, Dusseldorf, Geneva, Houston, Jakarta, Johannesburg, Melbourne, Osaka, Prague, Santiago, Taipei, Washington
5: Bangkok, Beijing, Montreal, Rome, Stockholm, Warsaw
4: Atlanta, Barcelona, Berlin, Buenos Aires, Budapest, Copenhagen, Hamburg, Istanbul, Kuala Lumpur, Manila, Miami, Minneapolis, Munich, Shanghai
2. World cities listed alphabetically, with abbreviations.
B. ADVANCED PRODUCER SERVICE FIRMS
This is a list of 46 firms with offices in 15 or more cities. Firms for which we have only presence/absence data are marked *.
SECTOR, TABLE CODE AND FIRM
AA Arthur Anderson
AM Abbott Mead Vickers (BBDO)
Banking and Finance
AO Allen & Overy*
Edited and posted on the web on 25th October 1999; last update 5th July 2000