GaWC Research Bulletin 21

There has already been far too much speculation about the twenty-first century being the Pacific Century (Segal,1990, 385)

Pacific Rim was a discourse in search of the research funds to prove the thing itself existed ... (It) is neither a self-contained region nor a community but just a rim (Cummings, 1994, 402, 405)

As associated geohistorical constructs, the Pacific's Century and Rim are unique: they generated a consensus that they were over before they began (cf. Segal, 1990; Dirlik, 1992, 1993; Krugman, 1994; Cummings, 1994; Current History, 1994; Cook et al., 1996; Berger and Borer, 1997). There were good reasons for the widespread critique, indeed cynicism, towards these constructs which emanates from their shared providence. Triggered by the fact that the volume of trade across the Pacific overtook trans-Atlantic trade in 1983 (Daly and Logan, 1989, 2), a particular geopolitical discourse emerged in the context of concern for the relative decline of the USA. The rise of, first, Japan and then the 'four Asian tigers' seemed to presage a geoeconomic shift in the world economy towards east Asia (Berger and Hasiao, 1988; Wade, 1990; Haggard, 1990; Appelbaum and Henderson, 1992; Leipziger and Thomas, 1993; World Bank, 1993a, 1993b; Naisbitt, 1996; Walton, 1997; Fischer, 1997). Extrapolation of growth rates pointed clearly to the world economy in the next century no longer being dominated by either Europe or the USA. For the latter, there was a simple rhetorical means of keeping up with the Asians: by designating the next century 'Pacific' rather than 'Asian' the USA would remain part of the centre of the world economy courtesy of its west coast (Daly and Logan, 1989; Dirlik, 1992, 1993; Borthwick, 1992). Hence the regional construct Pacific Rim was the necessary geographical complement to the Pacific Century.

Future time constructs are less vulnerable to criticism than spatial constructs because while we cannot know the future, geographical investigation can interrogate the coherence of regional propositions. Thus whereas the economic travails in east Asia in the 1990s may have made a dent in the idea of a dawning Pacific Century, it can remain a possibility. In contrast, the Pacific Rim has been empirically evaluated and found wanting.¹ For instance, Segal (1990, 377) has concluded that no one should believe in 'a coherent Pacific region' because '(t)here is no important cultural, ideological, political, or even military sense in which it is particularly useful to talk of "the Pacific"'. No wonder such 'talk' has been denigrated as 'Rimspeak' (Cummings, 1994, 402). While this reaction is understandable, it is, nevertheless, not quite as straightforward as such dismissals suggest. All regions are constituted through their own times - they are always historical regions (Taylor, 1991) - and the 'Pacific Century-Rim' is an exemplary case of this 'time-space' theory of regions. Hence, the question is not whether the Pacific Rim is a coherent region now but whether it is becoming such a region for its presumed time. Certaining, the dawning of the new century makes it particularly apt to reopen this matter to fresh empirical scrutiny.

Our reason for looking again at the Pacific Rim is not, of course, simply time dependent. Rather, we are illustrating a new way of considering such regional questions under conditions of contemporary globalization. Both the origins of the Pacific Rim construct and the testing of its veracity have been state-centric in nature. The Pacific Rim is envisaged as existing or not existing in a world mosaic of states where regions are defined as combinations of states. One of the effects of this state predilection is that the boundaries of the Pacific Rim as a region are very unsatisfactory since it is made up of countries which encompass large 'non-Pacific' areas (e.g. China, Canada, USA, Mexico, Australia). This way of geographical thinking is to work ideas through a 'space of places' (i.e. the international mosaic of states). An alternative approach has been provided by Castells (1996) where his emerging network society is constituted as a 'space of flows'. Among the myriad of types of flows in this new space, world cities as 'nodal centres of the new global economy' (Castells, 1993, 250) act as a critical organising network. We follow Castell's lead here: the state-mosaic thinking is replaced by a city-nodal approach. However, while Castells and others have provided plausible descriptions and concepts for this emerging space of flows, we have developed an empirical and quantitative specification of the world city network (Taylor and Walker, 1999; Taylor and Hoyler, 2000; Taylor, 2000). Given the initial stimulus for 'Pacific thinking' was the 1983 overtaking of Atlantic trade volume, it would seem to be entirely appropriate to evaluate the Pacific Rim within a space of flows framework.

A WORLD-CITY PERSPECTIVE

The most obvious advantage of beginning with cities rather than states is in allowing a tighter geographical specification of the Pacific Rim (Los Angeles but not Chicago, Shanghai but not Chonqing, etc.). However, the key gain in our approach is the new theoretical framework for examining relations between cities which our perspective provides. In this we differ from previous research on Pacific Rim cities where the empirical analysis has been largely about ranking cities by population size and change (Douglass, 1989; Yeung and Lo, 1996; 1998). At about the same time that the concept of the Pacific Rim was being invented, Friedmann (1986; Friedmann and Wolff, 1982) was developing his 'world city hypothesis' which linked world-wide urban development to the new international division of labour (Fröbel, et al. 1980). World cities were the 'command and control centres' of the new world economy being developed by multinational corporations. Friedmann's world city hierarchy encompassed cities from around the Pacific Ocean, viewed as a part of a global pattern: at the top of Friedmann's (1986) original hierarchy were Tokyo and Los Angeles among the 'core/primary' cities, followed by Singapore as a 'semi-periphery/primary' city, and five other Pacific cities designated 'semi-periphery/secondary'. This approach was consolidated by Sassen's (1991, 1994) identification of the 'global city' in which Tokyo joined London and New York as Pacific Asia's representative at the apex of the world urban hierarchy. It is with her work that world cities have been intrinsically linked to the form and content of contemporary globalization.

Sassen has provided a more precise definition of world/global cities than was offered in Friedmann's original identification. Whereas the latter treated multinational corporations in general, Sassen focuses upon those providing corporate services. She argues that the production of advanced producer services (such as financial instruments and inter-jurisdictional law) constitute a cutting edge of the contemporary global economy. Furthermore, this production can only be successful in 'special places' where concentrations of specialised knowledges intersect with flows of the latest informations. World cities are these special places. They are global centres servicing capital to enable it to operate 'seamlessly' across the world.

In our work at GaWC (the Globalization and World Cities Study Group and Network) Sassen's approach has been extended to define world city network formation (Taylor 2000). It follows from Sassen's argument that it is corporate service firms which specify the world city network by their office location strategies. Any world city is the creation of the service firms located there, and its position in the world city network depends upon those firms' particular patterns of offices across other world cities. In this specification, the world city network is, therefore, the amalgam of global service firms' office networks. This is the first formal specification of the world city network; its very tight definition is a necessary first step for the empirical investigation of this network.

WORLD CITY DATA AND ANALYSIS

We originally collected data for 69 producer service firms (in accountancy, advertising, banking/finance and law) for 263 cities across the world. From this material we identified 55 world cities and another 73 cities showing evidence of world city formation (Beaverstock et al. 1999a). Of the 55 world cities, 17 are Pacific Rim cities and there are a further 11 Pacific Rim cities which were found to have evidence of world city formation (Table 1). These are the 28 cities in our analysis below. In a related study we have identified 46 'global service firms' defined as having offices in at least 15 different cities (Taylor and Walker, 1999); they are listed in the appendix. These are the firms in our analysis below. For each of the 46 firms we have a measure, from 0 to 3, of the level of service it provides in each of the 28 cities.² The end result is a 28 (cities) x 46 (firms) data matrix made up of 1288 pieces of information.

To make sense of such a large set of data requires application of techniques which draw out recurring patterns in the data. This is what the factor analytic family of multivariate statistics does and here we employ the most basic of this group: principal components analysis. This technique is based upon correlations. The data matrix is converted into a 28 x 28 correlation matrix where each city is correlated with every other city. These correlations measure the degree of similarity between two cities' mix of corporate services: two cities sharing many of the same firms will have a high, positive correlation, pairs of cities with very little overlap in firms will have a high, negative correlation. What the principal components analysis does is to use these correlations to identify principal components which abstract distinct patterns of corporate service mixes that recur across cities. (For specific details of this methodology, see Taylor and Walker (1999) and Taylor and Hoyler, 2000.)³ The particular corporate mix of each city can be compared to these components as new correlations (instead of the original city-city comparisons, these are city-component comparisons), which, in this context, are called 'loadings'. In this way cities are found to 'load' high on some components and low on others. The high loadings are used both to allocate cities to components and then to interpret and label the components. The result is a typology of cities based upon similar corporate mixes as identified in terms of components.

Such an analysis produces five important outputs which we will use below: (i) the percentage of original total variance in the data accounted for by the components collectively; (ii) the percentage of 'explained' variance accounted for by each component individually; (iii) the heart of the analysis which is the table of loadings relating cities to components; and (iv) the table of component scores that relates the firms to the components, which enables us to specify the distinguishing features of the service mix associated with a particular group of cities.

As previously described we use this method of analysis to identify groups of cities that provide similar mixes of corporate services. Returning to our specification of the world city network, these groups represent clusters of nodes in the network, cities which are located on many of the same office networks of firms. Hence the groupings of cities can be interpreted functionally. If we make the plausible assumption that offices belonging to the same firm will communicate more than offices belonging to different firms, it follows that evidence of two cities with many offices in the same firms' office networks will be a surrogate indicator of high levels of inter-city communications between the cities. In this way the similarities we have analysed may be interpreted in terms of relations between cities: a pair of cities sharing few firms indicates less inter-city flows of business information and knowledge compared to a pair of cities with a large overlap in the firms located within them. Thus groups of cities derived from a principal components can be viewed, tentatively, as sub-networks within the larger world city network. Accordingly, the methodology used here is particularly suited to determining the coherence or otherwise of the Pacific Rim as viewed through its leading cities operating in a world-wide space of flows.

TESTING THE REGIONAL COHERENCE OF THE PACIFIC RIM

We use principal components analysis to test for the existence of a coherent Pacific Rim macro-region in terms of inter-city relations. We refer to this as a 'global test' because our data is based upon firms with world-wide patterns of offices. Hence we are exploring regional coherence within the larger global economy.

In this principal components analysis, regional coherence can be represented in one of two ways. The first and simplest depiction of coherence would be if there was one dominant component which accounted for nearly all the original variance. This was the original way this type of technique was used - looking for the 'general factor' in a set of data (Taylor 1981). Were such a component to be found, it would indicate very similar mixes of corporate services across the whole set of cities thus suggesting a coherent Pacific Rim sub-network of cities. We know from our previous analysis at the world scale (Taylor and Walker 1999) that such a result is highly unlikely. The second outcome that may suggest a coherent region would be for the cities to be grouped by size or importance but for each group to straddle continents and be represented in both east and west rims. This would imply some sort of common hierarchy of service provision across the Pacific thus suggesting a systemic regional coherence. Neither outcome was to be found with the present data matrix. Rather, the principal components analysis generated five components which between them accounted for 74.2% of the original variance in the data. This is a good parsimonious statistical result and the substantive findings are equally impressive: city loadings are displayed in Table 2 and component scores of firms in Table 3. The latter show the relative surplus or dearth of a service by firms for a given component mix. In addition, the five components are easily interpretable in terms of the geographies of the city groupings (Figure 1).

Component I brings together most of the Asian cities in the data, along with Sydney and Melbourne. We have labelled this the Western Rim Cities. With 30.6% of the explained variance within this component, twelve cities are recorded in Table 2 and on Figure 1. Table 3 shows which firms score high or low. The main surplus here is the presence of Baker and McKenzie, the world's largest law firm, which is reciprocated by the dearth of selected other US law firms. There is also a noticeable surplus among some banks. The presence of the two Australian world cities here creates a component which is the closest we get to producing a coherent 'rim' component. However, this largest component is by no means a comprehensive grouping of cities of the western rim. The remaining Australian cities group with New Zealand and Canadian cities to constitute Component II which we call Old Commonwealth Cities. Accounting for 25.7% of the explained variance, all five accountancy firms in our data exhibit surpluses in the scores along with selected advertising companies. This component obviously shows a British imperial legacy. Sydney and Melbourne do load on the component in Table 2 but lower than for component I.

Component III, accounting for 23.2% of the explained variance, is the most interesting component. Its top four cities are from China and Vietnam and so we can label the component Market Communism Cities. Presumably, these cities involve very special circumstances in terms of relations with the party-states which creates a specific mix of service firms. Banks are well represented in the scores for surpluses along with the Japanese (Osaka) advertising firm Dentsu; Baker and McKenzie is conspicuously in deficit with this group of cities.

Asia's leading world city, Tokyo, although to be found loading low on Component I, is picked out by this analysis as singular enough to have what is largely its own component. We call Component V,⁴ which accounts for 10.8% of the explained variance, Global City following Sassen (1991). This finding is entirely in keeping with our previous global analysis which linked Tokyo to other international financial centres, notably Zurich and Frankfurt, rather than other Asian cities (Taylor and Walker 1999: Table 2). Predictably the scores show a surplus for firms in accountancy and banking, and a relative dearth in advertising.

The smallest component, accounting for 9.7% of the explained variance is Component IV. This component defines the most distinctive grouping: it includes all the US cities in our data and no others. We label this Pacific US Cities. The scores show that this service mix is marked by a surplus of US law firms plus accountancy firms, and a dearth of some advertising firms and banks. Although Los Angeles developed in the 1980s as the 'gateway' between the USA and the Pacific Rim (Ong and Blumenberg, 1998, 315), especially reflected in immigration patterns (Soja, 1998, 442), this is not to be found in the city's corporate service mix. From the latter perspective, there is no evidence in our analysis that Los Angeles is becoming a leading Pacific world city. Rather it appears as specifically US in nature.

In addition to these five principal components, notice that the Latin American cities do not feature prominently in this analysis. The fact that they load only weakly across all components (except for the US one where they do not feature at all) is a particular indictment of the geographically comprehensive implications of the concept of Pacific Rim. Quite simply, the consistently low loadings show that Latin American cities really do not really fit anywhere in this analysis. Of course, despite their geographical location, the Pacific Rim concept was never really about their region in any case.

Finally, we can note that this analysis provides clear support for the scepticism of the critics of the Pacific Rim concept who we cited in the introduction. Our contribution has been to present a detailed, rigorous empirical evaluation from a different perspective, both globally and city oriented. There are two particular ironies in our results. First, although the debate on the Pacific Rim dates from the 1980s, the only example of a strong trans-Pacific grouping of cities (courtesy of Vancouver) has a much longer pedigree: Britain's imperial legacy of the old Commonwealth. Second, although the USA is in many ways the 'author' of the Pacific Rim (Borthwick, 1992), its cities are the least related to other Pacific Rim cities, or, in other words, they are the most nationally introverted, in terms of corporate service provision. Overall, our results show a series of largely separate groupings of cities around the Pacific, each with its own specific corporate service mix.

CONCLUSION: THE PACIFIC RIM AS A GEOGRAPHICAL CHAOTIC CONCEPTION

At its simplest level this study has confirmed earlier scepticism regarding the concept of a Pacific Rim. Certainly, we can say that as we enter the twenty-first century there is little empirical indication to suggest the emergence of a Pacific Rim region to complement a possible Pacific Century. However, we have taken the evaluation a stage further by breaking away from traditional state-mosaic description and analysis by introducing a city-nodal approach. This has provided a necessary theoretical framework in terms of world city and world city network formation. This use of the office networks of firms to specify relations between cities allows us to provide a generic geographical interpretion of the Pacific Rim.

A chaotic conception is an arbitrary abstraction from a wider range of relations which define a system (Sayer, 1992). It contrasts with rational conceptions which are constituted by meaningful relations within the larger whole. Thus, a chaotic conception combines together relatively unrelated parts while also dividing highly connected parts. In terms of spatial concepts, ideally geographical regions should be rational conceptions which define part of a whole in terms of its inter-relations. Much of the effort in defining regions focuses upon rigorous depiction of spatial boundaries. In contrast, a geographical chaotic conception is produced when relatively unrelated areas are arbitrarily combined and/or when similar or related areas are arbitrarily divided. It is important to appreciate the existence of geographical chaotic conceptions because they define not simply 'poorly specified regions', but 'anti-regions', spatial categories which confuse rather than elucidate. The Pacific Rim is just such a space.

ACKNOWLEDGEMENTS

We acknowledge the support of the ESRC for this research: the data derives from project R000222050 and the ideas for the application from project R000222693.

REFERENCES