This Research Bulletin has been published in E. Sheppard and R.B. McMaster (eds) (2004) Scale and Geographic Inquiry: Nature, Society, and Method Oxford: Blackwell, 213-235.
At the end of this paper my answer to the title question is going to be no. This may at first seem somewhat disingenuous. Certainly in EU terms there is much talk of a Europe of states, a Europe of nations and a Europe of regions, so why not a Europe of cities? Basically my answer is in the nature of the spaces implied by these various terms: like Europe itself, states, nations and regions define spaces of places, in contrast cities constitute a space of flows. Obviously this is very simplistic when baldly stated - the rest of this paper attempts to explicate this position further - but recognising this difference is critical for understanding the limitations of geographical scale analyses. Scale is usually associated with boundaries so that focusing on relations, connections and networks problematises geographical scale as traditionally conceived.
There are four geographical scales present in the concept of a 'Europe of cities', the two which are explicit plus two which are implicit: global and state. The concept raises implicitly the question of the future of European states in a globalising world. First, Europe is a world region in which the EU is an institution of globalization par excellance with its transnational structures providing a platform for Europeans to confront the global economy. Second, cities provide an alternative focus to the state in a globalising world; the world city network may be becoming a rival metageography to the familiar mosaic of nation-states (Beaverstock, Smith and Taylor, 2000). These four different 'scales' - city, state, world region, global - are familiar to many social science contexts and will appear throughout this discussion. The argument proceeds in four stages. First, I will consider types of scale problems as identified in geography and relate them to wider social science concerns. Second, I identify an 'embedded statism' as the crucial scale implicated in the very nature of social science knowledge. The result is an analytic 'disembedding of cities'. I provide nine propositions which link this state of affairs to the neglect of cities as the crossroads of society. This leads, thirdly, to the contemporary literature on world/global cities in which I argue cities are only partially 're-embedded' due to a particular empirical deficit. I provide a new conceptual and empirical framework for world city network studies. Finally, these themes are illustrated by new analyses of European cities in a world of cities. The question that arises out of all this is: how relevant is geographical scale when studying networks in a space of flows?
AGGREGATIONS, LEVELS, RANGES, SPANS, SCOPES ... AND GEOGRAPHICAL SCALES
All geographical studies are imbued with issues of scale; choosing scales of analysis, comparing outputs at different scales, and describing constructions of scale are all common practices by geographers. Often these issues were left implicit in a study; in recent years there has been an explosion of studies which treat scale as the object of study in its own right. The latter work can be interpreted in two contrasting ways. First, it can be seen to represent, to some degree, a 'geographicalising' of problems common to all social science research. In other words issues of 'size' and 'level' are treated as spatial categories, such as nested regions, so that general problems of analysis or interpretation are presented in specifically areal terms. Second, studies of geographical scale can be seen as making a fundamental contribution to social sciences as part of its so-called spatial turn. In other words scale is treated as intrinsic to all social process as, for instance, when economic restructuring is interpreted as spatial 're-scaling'.
My starting point is to set geographical scale in its wider social science context. To do this I delve back into the literature before the latest burst of geographical scale expositions. I identify five basic scale concepts in social science - aggregation, level, range, span and scope - in which the two interpretations above can be found: both viewpoints make some sense although the balance between them varies across type. In brief descriptions of each concept I begin outside geography before proceeding to geographical scale per se.
Statistical Aggregations: Scale in Spatial Analysis
Scale implies measurement as in 'scaling' which converts a process into a variable. If the objects for which the measures are made are aggregated, the total amount of variance in the variable declines until the aggregation is completed leaving one composite object and thus zero variance. Since the pattern of decline in variance will differ between different variables over the same set of objects, it follows that the co-variance between any pair of variables will alter as the aggregation proceeds. Hence correlations, and all the methods building upon correlations, are potentially unstable being dependent on the level of aggregation at which the analysis takes place (Openshaw and Taylor, 1979).
This well-known statistical phenomenon need not be a research problem where the social scientist has a basic theoretical rationale for choosing a level of analysis, for instance individuals as decision makers in survey data. Geography is not so fortunate; what exactly is a 'geographical individual' has been long debated (Chapman, 1977), today we might express doubt as to whether there is such a thing. Often geographers let the data source define the 'individual' such as census tracts or counties. Such empiricism is particularly susceptable to the aggregation problem since there are no grounds for choice of scale. This is the 'geographical scale problem' as identified by geographers in their quantitative revolution. As particular statistical aggregations, changing the scale of analysis produced alternative results meaning that all findings were scale-dependent. There were two reactions to this. First, most spatial analysts ignored it, witness the tradition of analysing urban patterns using census tracts. Second, some analysts attempted to use the scale dependence theoretically to suggest that different co-variation at different scales represented different processes operating at different scales (Haggett, 1965). Here the lack of a geographical individual was used positively to provide a flexibility in multiscalar analysis to imply a model of alternative explanations at different levels of analysis. However for the development of this model we have to look outside geography to International Relations and its behavioural-quantitative revolution where the 'level of analysis problem' was a central issue.
Levels of Analysis: Systems in Political Geography
In International Relations the stimulous to identifying levels of analysis was theoretical. Three levels were identified: the international system, the state, and the individual, although in practice most treatments focused on the first two levels. In recent years the original conception of levels of analysis has been criticised for conflating two different meanings (Buzan, 1995). On the one hand the levels depict distinct spatial units of analysis or objects of study while, on the other hand, they represent the more abstract notion of sources of explanation. This conflation has been endemic in geography. However, without a meaningful notion of geographical individual this conflation was not explored in quantitative geography but it had catastrophic consequences for political geography.
In the attempts to make political geography a respectable social science the idea of political system was borrowed from political science to provide an ordering frame. However, when combined with the idea of three distinct scales - international, national and sub-national - the result was to create a political geography consisting of three autonomous systems (e.g. Johnston, 1979; Short, 1982). This was saying much more than explanation is scale dependent, geographical scales have been elevated to the status of separate ontological entities. This is IR's conflation of object and explanation taken to its geographical extreme. My own intervention was to reconstitute a single system, capitalist world-economy as an historical system, so that levels of analysis are theoretically integrated: my materialist political geography posits a scale of reality where capital is ultimately realized in the world market, a scale of ideology where the state distorts the market, and a scale of experience where market outcomes are felt in localities (Taylor, 1982). This is to use scale to spatialise ideology separating experience from reality. Such functionalism, albeit from a very different perspective, has been most developed in spatial economics.
Ranges of Goods: Central Place Hierarchies
In the tertiary economic sector, the distribution of service provision has been the subject of normative theory which posits a functional hierarchy of spatial scales. With consumer behaviour governed by the range of goods (how far they are willing to travel) and suppliers governed by the threshold of goods (number of consumers making selling a good viable), clusters of like service goods in central places surrounded by their hinterland markets are produced. In this way, ranges and thresholds define spatial scales in the economic landscape.
Central place theory represents a most sophisticated treatment of geographical scales although it has been out of fashion in geography for some time as urban geography has taken its 'internalist turn'. However, the service sector continues to grow in relative size in the world economy and services are an intregal part of globalization. Hence I will return to cities and hinterlands below.
Spans of Control: Administrative Geographies
Within central place theory there is an 'administrative principle' which produces nested hinterlands across scales. Clearly this is a system for bureaucratic neatness and is more generally related to the idea of span of control in classical organizational theory. For an efficient hierarchical organisation, a superior can only supervise 5, or at most 6, subordinates whose work interconnects. Such ideas of administrative efficiency were at the heart of the English reorganization of local government in 1974 (Dearlove, 1979).
Redrawing the local government map as an exercise in organization science meant defining units in terms of optimal population sizes for a given service. In general there was a policy to equate size with efficiency so that larger units were proposed in order to obtain 'economies of scale'. This was also presumed to produce higher calibre representatives and officials. And all this theory had to be imposed geographically on very uneven and complicated population patterns. Thus the 'boundary problem' as 'solved' by organization theory created arbitrary spaces of government often identifiable by their artificial names. Thus organizational 'efficiency' far outweighed the idea that administrative geography should reflect the daily lives of the citizenry (Honey, 1981). According to Dearlove (1979), this was an attempt to control public power with abstract neatness replacing messy local conflict.
Scopes of Conflict: Production of Geographical Scale
Bringing power into consideration, local government reform can be viewed as an exercise in defining the scope of conflict. Power accrues to those who define the scope of a conflict according to the seminal political analysis of Schattschneider (1960). By controlling who is in a conflict and who is outside, the relative balance of strength between combatants is defined and the outcome thus determined. American federal politics has a long tradition of limiting the scope of conflicts to contain radical challenges to established power. A basic means to contain in this sense is geographical and this means defining the scale at which disputes are resolved and decisions are made (Taylor, 1984).
This traditional political science approach relates to the main recent theorising of scale in geography: the production of geographical scale. This social constructionist approach argues that there are no 'natural scales', the assignment of activities to specific scales is historically contingent upon power forces which create scales amenable to their ends. Thus contemporary economic restructuring can be viewed as a conscious 're-scaling' of markets and regulations both 'upwards' and 'downwards' away from the nation-state (e.g. Brenner, 1999). This 'glocalization' has brought forth a new dyad of scales and lessened the traditional dominance of the state as the 'natural' scale of political activities and much else besides.
* * * * * * * *
Aggregations, levels, ranges, spans and scopes are all social science concepts of relative size whose arguments have been mimicked in terms of geographical scale. In what follows I am going to put some of these different ways of thinking about scale together in a relational argument which moves us from an areal focus to a network one. In other words, using another social science perspective (Castells, 1996), what happens to these relative size concepts and geographical scale when we think in terms of a space of flows instead of a space of places?
EMBEDDED STATISM AND DISEMBEDDING CITIES
This section summarises a set of arguments which I have made elsewhere in various papers. Here they are brought together as a sequence of propositions which begin with states and finishes with cities. There is just a minimal justication for each statement given other available discussions. My purpose here is merely to provide the intellectual context for the empirical contribution in the next section where I return to Europe and its cities.
Proposition 1. The nation-state defines the pivotal scale in geographical scale analyses. This is reflected often in the language used to express other scales such as 'international' and sub-national'. Where the language is not explicit, as in global-local nexus or glocalization, the importance of the concept relies on its omission of the state: even in its silence the state remains pivotal. Our thinking about geographical scale has been state-dependent (Taylor, 1982).
Proposition 2. 'Society' is normally conceived as a people geographically bounded by the sovereign limits of a nation-state. This is part of a remarkable spatial congruence assumption where society, economy and polity are deemed to match state territory. Since these concepts lie at the heart of our orthodox social theories, it follows that there has been a critical lacuna at the heart of the social sciences. The unexamined geographical discourse has meant that core social science disciplines - economics, political science and sociology - have been unconsciously state-dependent. Social analyses have been fundamentally state-centric in nature privileging one geographical scale over all others (Taylor, 1996, 1997a).
Proposition 3. There is a 'states-metageography' which dominates the way the world is routinely perceived at the global scale. The world political map of states - it appears on every geography school rooom - is by far the most familiar of all human geography maps. It is sometimes said the lines on the map which depict state boundaries appear to be as natural as rivers and coasts. In this case the state-dependence is expressed as a mosaic structure of space, a space of places (Taylor, 1994, 1995).
Proposition 4. States are the great producers of publically-accessible evidence on all social activities which is why the results are called stat-istics without the hyphen. Our social world is described by states for state purposes. Whether UN statistics at the international scale or local statistics at the sub-national (regional or urban) scale, state agendas are intrinsic to the information gathered (Taylor, 1996, 1997b).
Proposition 5. As territorial political entities, states are largely interested in 'taking stock', counting their populations and resources in state-defined spaces. This means that the vast majority of publically-accessible data provides attribute measures of areas to the relative neglect of relations, connections and flows. We are provided with a space of places, but not with a space of flows (Taylor, 1999).
Proposition 6. Despite the raison d'être of cities being their connectivity, there is very little data on relations between cities. State census volumes focus upon the cities as places in which to take stock, not as the crossroads of society. Hence although there are numerous references to relational concepts such as city networks, urban hierarchies and city systems in the literature, the evidential basis of these ideas has been limited. Cities are 'de-networked' in stat-istics (Taylor, 1996, 1997b).
Proposition 7. At times cities have been lost in social analyses which focus upon 'urbanization', a stock-taking areal concept eminently suited to stat-istical description. To view the growth of cities in a simple areal manner is the ultimate degradation of the city as a crossroads in a wider world. Instead it is represented as an essentually local problem, the 'invader' of rural idylls. We need to think, in Murray Bookchin's (1995) terms, of 'cities against urbanization': stat-istics have promoted anti-city ideologies (Taylor, 2000a).
Proposition 8. Where cities are conceptualised in relational terms they have, until recently, been geographically truncated as 'national urban systems'. This nationalisation of cities has removed connections beyond the state-space from the city-system analysis. Theorised as 'primate-city' distributions developing into 'rank-size city systems' on a state-by-state basis totally downgraded, if not wilfully ignored, relations beyond the state boundary. Cities and their relations can only be so truncated under absolutely autarkic state conditions (Taylor, 1996, 2000a).
Proposition 9. For studies of a world city network as an alternative metageography it follows that there will be a veritable evidential desert. This is because data on relations between world cities suffers from a 'double whammy', a state-centric data bias complemented by an attributional measurement proclivity (Figure 1). It has been shown that the evidential structure of the world cities literature includes very little on inter-city relations. Despite much reference to networks, hierarchies and systems of cities at a global scale, in fact we do not know much about how world cities connect with one another (Taylor, 1999, 2000a)
* * * * * * * *
One final point should be made: social knowledges are more fluid today than for several generations. Hence many of my propositions are becoming more historical statements than contemporary positions as changes kick in. Praise for this should, in part, be laid at an unlikely source. Although we can criticise the extreme globalization gurus who have proclaimed the end of the state, they have provided one important service: they have produced the necessary jolt to social science to end its unexamined statism. Now there is a vibrant debate on the importance of states so that studies at the state scale have to justify that choice. I choose to experiment at the global scale using the propositions above as justification. In particular, my finishing point, proposition 9, as Figure 1 shows, still holds: the remainder of the paper is about cultivating this evidential desert.
WORLD CITIES AS NETWORKS IN A NETWORK OF WORLD CITIES
The title of this section is unashamedly taken from Brian Berry's (1964) seminal paper "Cities as systems within systems of cities". Obviously I have 'globalised' the original and as we proceed you will see that I will offer a much less all-encompassing model of cities and their systems than Berry attempted. Nevertheless Berry provides an ideal starting point for considering contemporary world cities. At first thought, it might seem seem odd to hark back to the 1960s in this context but, compared to most current research on world cities, in the urban geography of the quantitative revolution cities were less analytically disembedded. Certainly they were nationally truncated but they were not spatially de-networked. The 'internalist turn' in urban geography (Taylor, 2001a) has meant that relations within cities has far out-weighed concern for relations between cities over the last few decades - Berry and Horton's (1970) famous urban geography text is the last to seriously attempt to provide a balance between the internal and external relations of cities. Hence my derivation of a title from Berry is an explicit signal of my intention to return to his balance.
Bringing back the external relations of cities into focus allows me to build upon foundations laid a generation ago by two of the University of Minnesota's most famous geographers, John R Borchert and Fred Lukermann. Borchert (1967) bemoaned the relative neglect of research on the development of systems of cities compared to individual cities and set out to right this omission in his seminal study of US city systems over four epochs from 1820 to the 1960s. Lukermann (1966) made the critical point that you cannot understand nodality and hierarchy simply by enumerating populations and functions, there needs to be measures of flows, exchanges, connections or relations. In what follows I will derive lessons from both these Minneapolis geographers for the study of contemporary world cities starting with the latter.
A Lesson from Lukermann: Beyond the World City Hierarchy
It seems to be accepted that it is in the nature of cities and towns that they form urban hierarchies. That's what central place theory says and studies of national urban systems seemed to have confirmed this truism. For instance, diffusion of innovations were modelled as a hierarchical process from national metropolis through other metropolises to regional and local centres (e.g. Bourne, 1975). With the identification of new world cities in the 1980s, their inter-relations were therefore treated as a given: there must be a world city hierarchy (Friedmann, 1986).
However, if we have learned nothing else from geographical scale analysis, we should be suspicious of research which simply moves processes from one scale to another. Changing scales means constructing a different set of social relations so that moving up from national hierarchy to world hierarchy model must at least be problematised. This has not been the case. From the beginning when world cities were placed in a hierarchy by John Friedmann, his ordering of cities, despite being based on a paucity of evidence, has dominated world city thinking. This has been subsequently augmented by Roberto Camagni's (1993) famous diagram of 'the hierarchy of city networks' with world cities at the top although again no strong evidence is provided. The problem with these structural conceptions is illustrated in Peter Dicken's ( 1999) Global Shift, a key textbook whose value lies to a large extent in its bringing together reams of evidence to describe contemporary transformations in the world economy. Not so with world cities, in his diagram which is intended 'to give an impression of a connected network of cities' (p. 209) we are told that 'the links shown are diagrammatic only'. Looking at just the European part of the diagram (Figure 2), it is not at all clear why the link between Dusseldorf and London goes through first Brussels and then Paris. Why such a three-step connection in this electronic communication age? This example shows both the poverty of data given the need to resort to 'diagrammatic links', and the assumption that there must be some hierarchical structure in a distribution of cities.
According to Lukermann (1966), to define a hierarchy there needs to be more than counting, of producing attribute measures. For a hierarchy to exist there has to be some notion of control up and down different levels: 'each hierarchical level has autonomy over orders below itself, while being dependent of those above' (p. 17). In other words there has to be evidence of 'a line of command' (p. 18, Lukermann's emphasis). World city studies are full of rankings of cities which are deemed to show a hierarchy. Lukermann's lesson is that such information is totally inadequate for this task: to rank is merely to order by a size measure, it need have no relation to hierarchical structure (Taylor, 1997b). Friedmann in his initial work which produced the world city hierarchy did indeed allocate 'command and control' functions to his world cities in the form of locations of corporate headquarters. However in this situation it is not the cities which are doing the commanding - London telling Paris what to do - but rather the hierachies are in the firms. This is much more complicated than a simple world city hierarchy structure (Taylor, 2001b).
A Specification of Inter-city Relations at the Global Scale
There are many ways we can begin to specify relations between world cities. The most common is to map infrastructure patterns notably airline routes. These may provide good general pictures of world city links but they are best seen as enabling mechanisms rather than the particular social relations which define a world city network. The latter have been identified as deriving from the emergence of advanced producer services as a cutting edge of the new globalising economy (Sassen, 2000). Cities have always been service centres, of course, but in the new circumstances of economic globalization particular services have had to be produced to satisfy complex and original needs emanating from intense multi-jurisdictional activities. Whether it is new financial products or inter-jurisdictional commercial law, service firms have growth to provide necessary inputs to make operating in a global market feasible and profitable. To carry out this task, service corporations need to be located in places where there are large stocks of knowledge continually augmented by flows of new information. These are world cities where face-to-face communications facilitate economic reflexivities in dense local networks of service practitioners (Sassen, 2000).
This is the world city not as the home to all kinds of multinational corporations but rather as the locus of one particular type: the producer service providers - accountants, realtors, advertisers, bankers and financiers, lawyers, management consultants, insurers and so on. Located together as knowledge clusters, this is the world city as itself a network (Storper, 1999). But what of the relations between cities? The firms providing these services cannot do their job from just one city. Tying together deals across several jurisdictions requires being located in many cities each with its own distinctive special knowledge mix. Hence there are global service firms which have networks of offices across world cities. The idea is that they can provide a seamless service for a client across all, or most, parts of the world economy. World cities are therefore nodes in information flows within global service firms which link together knowledge clusters to create unique products for global clients (Taylor, 2001b). In Figure 3 a miniscule section of this network is illustrated for ten cities and three firms1.
At GaWC2 we have focused upon the office networks of service firms as the most promising window into inter-city relations. As well as its theoretical rationale, this is also an area where data can be collected reasonably easily to fill the empirical void in inter-city studies, and for world cities in particular. We have put together a data set for 46 global service firms defined as having offices in at least 15 different cities. This is a unique data set which covers over 250 cities across the world although we will focus below on those we define as world cities, 55 in all, plus other European cities for which we have found some evidence of world city formation (see Beaverstock, Smith and Taylor, 1999).
A Lesson from Borchert: From Hinterlands to Hinterworlds
An urban hierarchy assumes cities with different sized hinterlands at different levels of service. In his tracing of the development of the US urban system Borchert (1967) noted that rates of city growth were dependent upon the size and resource base of a city's hinterland and the changing technology especially in transport. These two factors were, of course, seen as being related: Borchert tells us that transport technology directly effects the size of hinterland. Thus it is hardly surprising that his four epochs are all distinguished by a major change in transport technology. He illustrates this as a changing urban hierarchy from 'sail and wagon' through to the internal combustion engine. Clearly since Borchert's work we have entered another technological epoch in his terms: electronic communication is revolutionising the delivery of services in many different ways. As he states: 'unforeseen major changes affect old cities and new alike' (p. 331) so that we need to think afresh for every epoch. This is particularly true for the current epoch with its instantaneous global communication capability. Although this facility has existed for some time now, it is not clear that we have got to the stage of revising our concepts to accommodate this remarkable new circumstance. I will suggest one such change illustrated by using the GaWC data.
One simple measure which can be derived from our data is to estimate the quality of global service available in a given city (Taylor, 2001a). If a client were to walk into a global service firms office in city i to do business in city x, what sort of service can be offered by the latter city? City x may have no office for that particular service firm so that no seamless service can be offered. On the other hand city x could have a particularly large office for that service firm assuring a very high service for this particular business. This is how the outside world views city x as a global service centre. If this exercise is carried out for all 46 firms in the GaWC data and summed then we have an estimate of city x's ability to service business from city i. The same exercise can be carried for cities other than i to provide a global picture of city x's pattern of global service. In Figure 4 such a measure is portayed over the set of 55 world cities for Minneapolis. All measures are converted to proportions of the maximum score possible to give a range from 0 to 1. Minneapolis's scores are relatively low but none falls below 0.1: this minor world city definitely has a global reach.
I call these distributions of service capability hinterworlds (Taylor, 2001a). With the advent of electronic communications and global service firms there are no longer boundaries around service areas as distinct hinterlands for advanced producer services. Rather there are world-wide relative intensities of service capabilities which are city hinterworlds. In this case it is not the resources in the physical sense within hinterlands which matter but rather the information/knowledge resources within its hinterworld which is vital to a city. We can say that behind every successful world city there is an intensive hinterworld. This is illustrated by London and New York in Figure 5 which both have, predictably, very intensive global hinterworlds.
Although these analyses make the obvious point that London and New York rank above Minneapolis as global service providers, the fact the hinterworlds exist at Minneapolis's level suggests that to search for a single, nested, world urban hierarchy is hardly appropriate for world cities defined as advanced producer service centres.3 World cities are constituted as complex networks of service provision. Although hierarchies can be interpreted as a particular type of network, it is common to view them as a very different form of organization: hierarchies, as we have seen, are about directing and control whereas networks are based upon co-operation between nominally equal units created through some mutuality of interest. Viewed in this way hierarchy and network, although commonly conflated in the urban literature, are incompatible concepts. In reality, of course, most systems are hybrids, hence my reference to the possibility of a world city network with hierarchical tendencies (Taylor, 1997b). Returning to Figure 3, it is immediate clear how complicated this network is and how futile it would seem to be to search for simple hierarchies. The next section explores such complexity multiplied manifold.
EUROPEAN CITIES IN AN INTERLOCKING WORLD CITY NETWORK
And so I return to the idea of a Europe of cities. There are 53 cities in Europe which have been identified as world cities or as having evidence of world city formation processes (Beaverstock, Smith and Taylor, 1999). Each of these cities has their own particular mix of global service firm offices. It is these global service complexes which are my subject matter here. They are the result of locational decision making by global service firms when they decide to make the large investment of setting up an office in a world city. The resulting office, of course, has to be staffed by professional practitioners who add to the knowledge concentration in the chosen world city.
The assumption I will be basing analyses on is that in their striving for a seamless service, global service firms office networks represent fundamental distributions of flows of information and knowledge between cities. Thus it follows that two cities with similar service complexes will likely have more commercial relations than two cities with very dissimilar service complexes. This assumption can be empirically tested but this has not been done. However, it is plausible enough to warrant going forward on this basis in this particular research area - inter-city relations - with such an evidential deficit as discussed previously.
Parsimony and Components
The assumption about similar and dissimilar service complexes suggests the need for a study of co-variances between cities. Every one of the 53 European cities can be correlated with the others in terms of service firm provisions creating 1378 measures of relations between pairs of cities. These define the total co-variation in the data. The resulting correlation matrix can be simplified using a standard principal components analysis. This is an exercise in parsimony which is very necessary given the complexity of the situation being faced here.
The basic output of a principal components analysis is as follows:
1. a reduction in the number of dimensions of variability - the cities are replaced by a relatively small number of composite dimensions of variability, or 'components';
2. each original city correlates, or 'loads', with each of the components - normally a city has a 'loading' (correlation) on one component which is much higher than all its other loadings;
3. from the pattern of loadings, components can be interpreted in terms of which cities load highest;
4. this produces a clustering or classification of cities by their service complexes;
5. because cities load on all components it is not a strict classification which is produced as some cities will load relatively high on two components hence creating overlapping groups of cities;
6. for each city their component loadings produce a measure of 'communality' which shows how much of a city's variability is described by the components analysis; and,
7. from city communalities the total variation in the data accounted for by the components analysis can be computed.
For details of this methodology and its application to the GaWC data, see Taylor and Walker (2001).
A European-scale Analysis
The 53 European city service complexes are reduced to just 5 components which account for 69% of the original variance. Although there is no logical reason for these results to define a spatial order this is, in fact, what happens here (Taylor and Hoyler, 2000). Globalization does not simply brush aside history, it builds upon past processes in creating new geographies. In this case the oft-identified 'city-studded' north-south 'spine of Europe' with its origins, according to Rokkan (1970), going back to the earliest modern developments is clearly reproduced as the spatial order in this analysis of contemporary decision-making by global firms. Figure 6 shows this order as five types of service complex which fall into two spatial groups. First, there are the cities of the the central spine of Europe: (a) minor spinal cities such as Arhus, Hamburg and Genoa, and (b) major spinal cities such as London, Paris and Milan. Second, there are the more peripherally located cities which fall into three types: (c) cities of the outer triangle such as Oslo, Lisbon and Budapest, (d) far west (British Isles) cities such as Dublin, Edinburgh and Leeds, and (e) far east (ex-Soviet bloc) cities such as Moscow, Warsaw and Kiev. Each of these types has a common mix of global service offices. Note the hierarchical tendency where the core spine of European cities is divided between two levels of city. One notable feature of these results is that London is the only British Isles city not to load on the British factor. London is not, it would seem, very British in its service complex mix.
However this is not the whole story. Table 1 shows the cities with the lowest communalities. I will call these 'un-European cities' on account that they have complex mixes with less similarities to other European cities. The eight cities listed fall into two groups, eastern European, and more iinterestingly, cities with important transnational functions. London is not only un-British, it is also un-European. The implication is clear and hardly surprising: we do not capture the essence of London's corporate mix in an analysis at the European scale. London is a global city and can only be understood fully at that scale. In other words, even though the analysis is at a world regional scale, London is being truncated just like the national analyses reported previously. The same is true to varying degrees of all the major transnational European cities with their important extra-European linkages.
A Global Scale Analysis
What happens to these European cities when cities from the rest of the world are brought into the analysis? This analysis includes 55 world cities which includes 22 European world cities from the previous analysis (Taylor and Walker, 2001). Using the same methodology as above, nine components were identified as service complex types which account for 80% of the initial variation in the data.
The pattern of cities in this global-level typology is much more complex than at the European scale as we might expect (Figure 7). Although the results are quite regional in nature indicating a spatial patterning to service complex types, European cities feature in a very fragmented way. There are three minor components which are clearly European, one for western Europe and two for eastern Europe but they involve only a few European cities. Northern European cities, particularly German ones combine with smaller north American cities to create a 'North Atlantic' cluster. However the many major transnational centres such as Frankfurt, Milan and Madrid are part of another major cluster with Tokyo, Sydney and Latin American cities. London is separate again, in this analysis it is able to quite appropriately join with New York as a global city cluster whose links are with other US cities more than with Europe.
In other words, European cities soon become dispersed once the rest of the world is brought into the analysis. Focusing on the 'major spine cities' from the earlier analysis, Table 2 shows their dispersion to a range of service complex types at the global scale. Of the ten cities half are to be found in the transnational cluster with key links far beyond Europe. This is the geographical reality behind the complexity of globalization: a world city network with both regional and hierarchical tendencies.
There is no Europe of Cities
This seems to be the clear answer to my title question after these analyses. Quite simply, you cannot contain cities within bounded spaces. Cities by their nature are networked and their analysis should not be truncated. Even as large an area as Europe cannot contain its indigenous cities.
CONCLUSION: SCALE AND NETWORK
So there is no Europe of cities but what does the specification and analysis of the world city network mean for geographical scale? I identify four implications of this work for studying geographical scale.
First, in terms of globalization, two general approaches can be found in the literature: one which emphasizes globalization as a special scale, and another which emphasizes the new intensities of transnational flows. Clearly globalization is both of these but there remains the question of balance between them. Territorialist traditions in the core social sciences and human geography may lead to a privileging of scale, the largest space of place, my argument is for taking this opportunity of post-statism to reassert the importance of the space of flows (Taylor, 2000b). The social sciences are ripe for a new balance between attribute and relation, between places and flows.
Second, re-balancing approaches need not lead to a neglect of the previous dominant mode of thinking. The territorialist approach creates boundaries to be sure, but there is still the matter of how the boundaries are interpreted. Strict regional demarcations, as in administrative geographies, lead to impervious boundaries, whereas in a space of flows all boundaries are porous. Analyses of world city complexes show a very complex geography of globalization which is highly regional in nature. But the patterning is no simple mosaic: there is interpenetration of spaces related to status levels of cities. The resulting spatial orders which are delineated represent sub-networks of cities, groups of cities, often from the same world region, which share similar mixes of service firms. As part of a larger whole, a sub-network is, of course, fundamentally porous in its boundaries.
Third, in terms of the relations of geographical scale to different social science conceptions of size, it is clearly levels of analysis which is the concept implicated in world city networks. The specification of the world city network defines three levels of analysis in a systemic framework. Although the quantitative analysis treats these levels essentually as 'objects', the ultimate interest here will be in defining different sources of explanation through these levels. The specification privileges the subnodal firms but there are many other processes operating which require understanding such as inter-city competition and the insertion of other levels such as state and service sector.
Fourth, this leaves the most imaginiative geographical scale approach, the production of scale, implicit rather than explicitly treated in this analysis. The latter is essentually political as I indicated earlier by linking it to Schattschneider's (1960) work. Clearly it is necessary to bring political geography back into consideration. What is the political geography of a network? What is the political geography linking spaces of places to spaces of flows (Taylor, 2000a)? These questions can begin to be addressed through looking at the world city network as an alternative metageography to the states mosaic and in relation to the latter (Beaverstock, Smith and Taylor, 2000). Moving from territorialist alliances to network-based alliances produces a different politics and a new political geography we have hardly begun to acknowledge let alone study.
Beaverstock, J V, Smith, R G and Taylor, P J (1999) 'A roster of world cities', Cities, 16, 445-58
Beaverstock, J V, Smith, R G and Taylor, P J (2000) 'World city network: a new meta-geography?' Annals, Association of American Geographers
Berry, B J L (1964) 'Cities as systems within systems of cities', Papers of the Regional Science Association, 13,147-65
Berry, B J L and Horton, F E (1970) Geographic Perspectives on Urban Systems. Englewood Cliffs, NJ: Prentice-Hall
Bookchin, M (1995) From Urbanization to Cities. New York: Cassell
Borchert, J R (1967) 'American metropolitan evolution', Geographical Review, 58, 301-23
Bourne, L S (1975) Urban Systems. Oxford: Clarendon Press
Brenner, N (1999) Globalization as reterritorialization: the re-scaling of urban governance in the European Union', Urban Studies, 36, 431-52
Buzan, B (1995) 'The level of analysis problem in international relations reconsidered' in K Booth and S Smith (eds) International Relations Theory Today. Cambridge: Polity
Camagni, R.P. (1993) 'From city hierarchy to city network: reflections about an emerging paradigm', in T.R. Lakshmanan and P. Nijkamp (eds.) Structure and Change in the Space Economy. Berlin: Springer-Verlag
Chapman, G P (1977) Human and Environmental Systems. London: Academic Press
Dearlove, J (1979) The reorganisation of British local government Cambridge: Cambridge University Press
Dicken, P (1999) Global Shift. London: Paul Chapman
Friedmann, J (1986) 'The world city hypothesis', Development and Change, 17, 69-84
Haggett, P (1965) Locational Analysis in Human Geography. London: Arnold
Honey, R (1981) 'Alternative approaches to local government change', in A D Burnett and P J Taylor (eds.) Political Studies from Spatial Perspectives. New York: Wiley
Lukermann, F (1966) 'Empirical expressions of nodality and hierarchy in a circulation manifold', East Lakes Geographer, 2, 17-44
Nystuen, J D and Dacey, M F (1961) 'A graph theory interpretation of nodal regions', Papers and Proceedings of the Regional Science Association, 7, 29-42
Openshaw, S and Taylor, P J (1979) 'A million or so correlation coefficients: three experiments on the modifiable areal unit problem' in N Wrigley (ed.) Statistical Applications in the Spatial Sciences. London: Pion
Sassen, S (2000) Cities in a World Economy. Thousand Oaks, CA: Pine Forge Press
Schattsschneider, E E (1960) The Semi-Sovereign People. Chicago: Dryden
Storper, M (1997) The Regional World. New York: Guilford Press
Taylor, P J (1982) 'A materialist interpretation for political geography', Transactions, Institute of British Geographers, ns7, 15-34
Taylor, P J (1984) 'Introduction: political scale and political geography' in P J Taylor and J W House (eds) Political Geography: Recent Advances and Future Directions. London: Croom Helm
Taylor, P J (1994) 'The state as container: territoriality in the modern world-system', Progress in Human Geography, 18, 151-62
Taylor, P J (1995) 'Beyond containers: inter-nationality, inter-stateness, inter-territoriality' Progress in Human Geography, 19, 1-15
Taylor, P J (1996) 'Embedded statism and the social sciences: opening up to new spaces', Environment and Planning A, 28, 1917-28
Taylor, P J (1997a) 'The crisis of boundaries: towards a new heterodoxy in the social sciences', Journal of Area Studies,11, 11-31
Taylor, P J (1997b) 'Hierarchical tendencies amongst world cities: a global research proposal', Cities, 14 , 323-32
Taylor, P J (1999) '''So-called world cities": the evidential structure within a literature', Environment and Planning A, 31, 1901-04
Taylor, P J (2000a) 'World cities and territorial states under conditions of contemporary globalization', Political Geography, 19, 5-32
Taylor, P J (2000b) 'Embedded statism and the social sciences 2: geographies (and metageographies) in globalization', Environment and Planning A, 32,
Taylor, P J (2001a) 'The study of "urban influence" under conditions of contemporary globalization: intensities of service provision by "alpha" world cities', Geography
Taylor, P J (2001b) 'Specification of the world city network', Geographical Analysis,
Taylor, P J and Hoyler, M (2000) 'The spatial order of European cities under conditions of contemporary globalization', Tijdschrift voor Economische en Sociale Geografie,
Taylor, P J and Walker, D R F (2001) 'World cities: a first multivariate analysis of their service complexes', Urban Studies,
Taylor, P J, Walker, D R F and Beaverstock, J V (2000) 'Introducing GaWC: researching world city network formation' in S Sassen (ed.) Telematics and Global Cities. Oxford: Blackwell
1. This specification is spelt out in detail in Taylor (2001b). World cities constitute an interlocking type of network which is relatively unusual in networks of social relations. The implication is that the key 'actors' are not the city nodes but rather the 'sub-nodal' firms.
2. GaWC is the Globalization and World Cities Study Group and Network organized from Loughborough University and available at http://www.lboro.ac.uk/gawc/. The key features of its research section are data sets and research bulletins. See Taylor, Walker and Beaverstock (2000) for an introduction.
3. A Nystuen and Dacey (1961) graph theory analysis can be carried out on this data but the results are uninteresting. Using 55 cities just two levels emerge: London and New York, and the rest. In fact this technique automatically 'finds' a hierarchy whatever the structure of the data.
Table 1: The most 'un-European' cities
Table 2: Major European 'spine cities' in the global-level analysis
Figure 1: The evidential desert in world city research (derived from data in Taylor, 1999)
Figure 2: 'Diagrammatic links' among European cities (part of Figure 7.4 in Dicken, 1999)
Figure 3: A miniscule section of the world city network (from Taylor, 2001b)
Figure 4: The hinterworld of Minneapolis (with location code for cities)
Figure 5: The hinterworlds of London and New York (for location code, see Figure 4) (from Taylor, 2001a)
Figure 6: The spatial order of European cities (from Taylor and Hoyler, 2000)
Figure 7: Hierarchical and regional tendencies in world city service complexes (from Taylor and Walker, 2001)
Edited and posted on the web on 5th July 2000
Note: This Research Bulletin has been published in E. Sheppard and R.B. McMaster (eds) (2004) Scale and Geographic Inquiry: Nature, Society, and Method Oxford: Blackwell, 213-235