GaWC Research Bulletin 10

By referring to 'so-called "world cities"', Kevin Cox (1997, 1) provides us with a salutary reminder of the dangers of attractive theoretical ideas turning into taken-for-granted conceptions. There is no doubt that John Friedmann's (1986) revival of the notion of world cities by linking them to the recently discovered 'new international division of labour' created a very popular concept for studying 'the urban' in the age of 'the global'. This literature is the subject matter of this editorial comment which I write in the spirit of Cox's scepticism.

Problematising the core concept of a literature suggests fundamental criticism of a research tradition either in terms of poverty of theory or empirical deficiency. In the case of world and/or global cities research both types of critique have been recently made - see, for instance, Storper (1997) on theory and Short et al . (1996) on evidence. In this comment I focus on the latter. Concern for the paucity of evidence for Friedmann's 'world city hypothesis' surfaced in the initial published discussion of paper (Korff, 1987) and it might have been thought that this problem should have been sorted out by now. Below I show that this is most certainly not the case and suggest reasons why empirical deficiency has dogged this literature to the degree that Cox can imply that world cities may not even exist. This requires the construction of what I term an evidential structure of the literature.

By the evidential structure of a literature I mean the nature of the empirical material which is marshalled to describe and test the propositions of a research field. It involves selecting key texts in an area and investigating the data employed with a view to finding particular biases in the resulting patterns. If these patterns are interpreted as more socially embedded than simply the outcome of free research decisions then the literature can be said to be facing structural problems. I have argued previously (Taylor, 1997) that this is indeed the case for the world cities literature in two particular aspects. First, statistics in general have their origin in servicing the information needs of states and this has resulted in our contemporary world being measured through state-centric data. A browse through the publications of the United Nations and its associated institutions will quickly affirm this. Second, statistics in general have developed a critical bias towards measuring attributes at the expense of connections. Although flows - of people, of commodities, of information - are measured they pale in significance when compared to the quantity and quality of attribute data. Both these features of statistics in general have the potential to be quite debilitating for the needs of world cities research. In global research, cities may be subsumed and lost in data on countries which is likely to be particularly disastrous for measuring flows between cities. Inter-city relations has the structural potential to be the Cinderella of evidence of global-scale processes.

For studying the evidential structure of the world cities literature I have chosen seven key texts which will feature in the readings of all who research and teach the subject matter. The two key texts by Saskia Sassen (1991, 1994) need no justification. Similarly there are three collections of essays which are obvious candidates for this exercise (Knox and Taylor 1995, Brotchie et al. 1995, Lo and Yeung, 1998); these have the advantage of adding a variety of authors to the survey. The final two texts focus on the enabling potential of electronic communications which is the key technology connecting world cities: Graham and Marvin (1996) deal explicitly with this technology in relation to the contemporary city, and Castells (1996) in the first volume of his trilogy on the 'information age' charts more generally 'the rise of the network society'. Although cities do feature in the book, the latter text is the only one not specifically on cities but its contents are central to the concerns of world city researchers: Castells' career has been as a key social theorist of cities and here he continues to inform us about the latest trends in 'urban society'. I believe these seven books provide a reasonable cross-section of the world cities literature.

For each book I have concentrated upon the tables and figures used to inform the written text. Eliminating tables and figures which are not based upon specific data (such as a schematic diagram) or which are not place-specific (such as comparing the efficiencies of different technologies) leaves 491 pieces of evidence for analysis. Each table or figure was first classified as to whether the evidence used was state-based or city-based. Where both cities and countries appear in a table or figure a subjective decision was made in terms of which of the entities the table or figure most informs. Second, the evidence was classified in terms of whether the data was of attributes or was relational. For the latter there has to the an origin and a destination. For instance, a table ranking countries by foreign direct investment is deemed attribute, if the locations from where investments came are given it becomes relational. In addition, the subject matter of tables and figures was recorded to aid in interpreting the results.

The resulting evidential patterns are shown in Table 1. The first point to make is that although the origin paper (Friedmann 1986) of this literature was overtly hypothetical in nature with little systemic use of data, the ensuing research has been very empirical in content. 491 pieces of empirical evidence in just seven books does suggest that the authors have not felt themselves to be intimidated by the potential data problems alluded to earlier. However, the resulting patterns are not what you might expect from a 'free' choice of data for studying world cities. Certainly there are more tables and figures on cities (267) than on states (224) but not by that much. In fact, the surplus of city evidence over state evidence is more than accounted for by the 53 tables and maps which feature simple city population figures from state censuses, hardly very sophisticated evidence of world city processes. Of course, there is nothing intrinsically wrong with having information on states in writings on cities, relating processes to different geographical scales is to be applauded. It is the balance of evidence which is at issue here. Obviously it can be argued that states provide the context within which the cities are changing and therefore tables on state trends are necessary. Well, yes again, but it is worthwhile putting the matter the other way round to get a clearer perspective. Cities and their regions constitute countries and national economies are primarily the sum of activities carried out in towhns and cities. But we do not expect, and certainly do not find, social science texts on states and national economies having almost as much evidence on cities as they provide on states. Perhaps they should but that is a different question, the point here is the relative quantity of evidence about states in a literature about cities.

Of course, the cities in this literature are not just any cities, they are world cities with the presumption of incorporating activities which transcend the particular states they are located within. To understand this requires information on connections. Here we see a massive weakness in the evidential pattern of the world city literature with attribute data (436) far exceeding relational data (55). Furthermore, we should not assume that the relational data all relates to cities: this is true of just 31 tables and figures. When you begin to think about it, this number of 31 pieces of evidence on such a key matter - cities and connections - is quite astounding low. In a literature which abounds with concepts such as city networks and global urban hierarchies only 6% of the evidence directly informs such central notions! When we turn to a more detailed look at the figures for individual texts and what they mean, even this meagre grand total gives a rosy picture of the situation.

Looking at the detailed figures for individual texts there are not many surprises given the above grand totals. In terms of city-state ratios, it is to be expected that Castells (1996) will be much lower than the other publications given his wider brief. However, it can be noted that in the discussion of his methodology (p. 26), Castells queries the accuracy of some state statistics but not the efficacy of relying so heavily on state-centric data to understand his 'network society' In fact, this has meant that when we turn to attribute-relational ratios, Castells (1996) is again the lowest: it seems that relying upon state data creates a likelihood of neglect of evidence on connections even if your subject matter is networks. Here the highest ratio is for Graham and Marvin's (1996) work on telecommunications but even here state data still dominates. Furthermore, half of their relational data is about intra-city connections so that we cannot look here for much evidence to inform us on inter-city relations.

In fact much of the data on connections between cities is from analyses of airline flights where information is easily accessible for cities as represented by their airports (Keeling 1995, O'Connor 1995, Rimmer 1998). In Rimmer's work (1998) we can contrast the difference between telecommunications and air passenger networks. For the former he tells us that '(m)any of the propositions on telecommunications and world cities have be taken on trust' (p. 451) and illustrates connections with a diagram of flows between countries. For air passenger networks he is able to produce diagrams which feature cities. However, it should be noted that airline flows are in any case problematic measures of relations between world cities. In a map of European airports presented by Kunzmann (1998, 49) after the big three (London, Paris, Frankfurt) there are 14 second order airports identified such as Munich, Milan, Madrid and ... Parma (Majorca). The latter is, of course, the most popular holiday destination in Europe hence its high ranking among important world cities. Airline passenger flows are constituted by many processes outside world city processes and they will all be reflected in gross figures - for Miami, for instance, vacation, retirement and commercial traffic is all mixed together. Finally it should be noted that airline traffic analyses have all incorporated an insidious form of state-centric bias. All base their analyses on 'international passengers' which results in a down-grading of US world cities in particular. The classic example is Chicago - it only appears on one of Rimmer's maps (Figure 16.10, p. 460) courtesy of 'fourth level' linkage to Toronto while Dublin, for instance, appears on all maps because of its 'first level' link with London. Nobody would want to argue, of course, that Dublin was more important than Chicago as an airport hub in the world-economy, it only appears that way when we rely upon international data.

In conclusion: I think the evidence I have presented on the evidential patterns of the world cities literature does suggest a severe structural problem due to the state-centric biases in data available to social science researchers. The evidence we have for asserting that world cities do in fact constitute nodes in a network of flows within the world-economy is very patchy indeed. Since the concept of world city has no meaning without connections, it seems reasonable to suggest that this literature is suffering from an evidential crisis. Further, since the latter appears to be structural, it is a crisis which has been largely hidden in the literature. For readers who wish to help overcome the crisis please look at our website on World Cities and Globalization (GaWC): http://www.lboro.ac.uk/gawc/