GaWC Project 19

GaWC logo
  Gateways into GaWC

World City Network Formation: Global Connections Audit (2001-04) and Analysis (2000-04)

Funded by ESRC (2000-2006)


Grant holder: P.J. Taylor

Research Associates: N. Gane, R. Aranya


Introduction: from cross-section to time series

The basic premise underlying this proposal is that a key theoretical and practical issue confronting geographical research in the new century is to explicate relations between spaces of places and spaces of flows. Any programme of research that tackles this topic is immediately challenged by a severe case of asymmetry in information availability: attribute data are plentiful, relational data are typically in short supply. In global-level research, for instance there are large amounts of data describing places, notably for countries but also for cities and regions, but there is relatively little on the flows between these places. This project is part of a research programme that is attempting to rectify this serious data limitation by measuring connections between world cities.1

The project builds directly on the current project "World City Network Formation in a Space of Flows" (see Annex). The latter has collected a large amount of information on the office locations of advanced producer firms in cities across the world. Data is in place for 100 firms2 (the "GaWC 100"3) over 320 cities for the summer of 2000. The census date cannot be more precise than this because the data is derived largely from the firms’ web sites which are irregularly updated. But the main limitation of this ‘unofficial’ source of information is that web pages are ephemeral, once revised the past is simply lost. Hence unless copied and stored at the time, there is no way to create a time series of data. If we cannot go back and produce time series we have to plan for collection in the future. And this is what this project is about: using current data for 2000 as the bench mark, four more years worth of information will be collected and stored to provide a five year run of data. Thus, whereas the current project is cross-sectional and has provided the basic framework, the new project will add a time-series dimension to the data and its analysis.

It hardly needs saying that globalization, and the world of corporate service firms in particular, encompasses a very dynamic world where high levels of change are endemic. Even in the cross-sectional study we have to consistently update due to closure and opening of offices, new acquisitions and alliances, and network/association membership revisions. Over a five-year period we expect to achieve a firm empirical handle on the global changes that are occurring in world city network formation.

Specification of world city network formation

Although there is much reference in the literature to concepts such as ‘world city system’ and ‘global urban hierarchy’ the collective relations among world cities has never been precisely specified (Beaverstock, et al., 2000). One of the major outcomes of the cross-sectional study has been to produce a first definitive specification of the world city network (Taylor, 2001)4. This section summarises this argument and extends it to incorporate change.

I specify a world city network as a tripartite structure (Knoke and Kuklinski, 1982) with the world economy defining the network level, with cities as the nodal level, and with global service firms as the sub-nodal level.5 This defines an interlocking network where the sub-nodal level provides the connections which hold the system together. Global service firms each have their own location strategy which has involved setting up an office network across world cities. Connections between any pair of cities can be derived from the quantity and quality of service firm offices the two cities share in common. The basic assumption is that cities with similar office profiles will likely generate more inter-city flows of information and knowledge. However since we cannot measure flows directly, the specification is for connections only. For instance, at the simplest level we can ask the question, if you go into the office of a service firm in city X, what is the probability that the firm will also have an office in city Y? Such information is directly relevant to service companies’ claims to provide ‘seamless service’ with quality assured through remaining ‘in house’. Of course, such measures of connections between cities do not have to be limited to elementary presence data, sizes of offices, for instance, can be taken into account. The point is that by specifying the world city network in this way, inter-city matrices of connections can be created based upon the global location strategies of advanced producer service firms.

In this specification the world city network is the product of decisions by numerous private corporate service firms as they carry out their location strategies of providing world-wide coverage for clients. Empirically this can be described and analysed only through combining the particular strategies of many firms in order to produce general patterns in the aggregate. Thus there has to be an emphasis on quantity of data in any empirical specification which is why analysis in both this project and the cross-sectional one treats a hundred firms.6 Here we define census points and collect office location data over time to track changes in this ‘inter-locking’ of world cities. Matrices describing changes in connections can be constructed to facilitate a time comparison analysis.

Aims and objectives

This project has four general aims within each of which there are several specific objectives.

  1. Data aim: to create inter-city change data matrices for 2000-04.

Objective 1a: to collect information on the offices of GaWC 100 firms in the summers of 2001, 2002, 2003 and 2004.

Objective 1b: to compare information with previous year and list all changes.

Objective 1c: to devise a standardised measure of five-year changes in terms of degree of change.

Objective 1d: to construct two five-year change matrices, first, a ‘city x firm’ standardised measure matrix and, second, a ‘city x city’ connection change matrix

  1. Analysis aim: to investigate changes in the world city network, 2000-04.

Objective 2a: to conduct cross-sectional principal components analyses for different years and interpret changes in findings.

Objective 2b: to conduct principal components analysis on change data over the five year period.

Objective 2c: to experiment with multidimensional scaling techniques to see whether a suitable visualisation of change can be constructed.

Objective 2d: to explore the data using a variety of other multivariate statistics to see whether they throw any additional light on the dynamics of world city network formation.

  1. Theory aim: to provide a definitive specification of the dynamics of world city network formation.

Objective 3a: to incorporate a change dimension into the current tripartite specification.

Objective 3b: to relate the tripartite specification to other processes and actors (e.g. city competition, states) in a more encompassing model of world city network formation.

  1. Practical aim: to foster external contacts within and beyond academia.

Objective 4a: to disseminate ideas and findings through peer reviewed publications and conference presentations.

Objective 4b: to encourage the GaWC network of researchers within and beyond Loughborough to be involved in analysis and interpretation.

Objective 4c: to maintain contact with the GaWC 100 firms in a reciprocal arrangement of providing and interpreting information.

Objective 4d: to invite the GaWC 100 firms to be represented at a one day seminar at the Corporation of London where project results will be presented.

Methodology: measuring inter-city connections

The basic measurement methodology is exactly the same as for the current cross-sectional project. Although it is normally difficult to obtain any sort of relational information from private firms for confidentiality reasons, in the case of corporate service firms and their world city office networks the exact opposite pertains. Their world wide capability is a primary selling point to clients so that their promotion literature proclaims their ‘globalness’. In today’s world this is most explicit, and easily accessible, on the web sites of firms where ‘location’ or ‘contacts’ is always a major gateway. Designed to enable clients and prospective clients to contact offices appropriate for their needs, these web sites are an invaluable mine of information on the global geography of service provision. Thus this is our main source for creating data on the GaWC 100, supplemented through contact with the firms as and when necessary.

As well as simple presence data on firms and cities, we are interested in the importance of each city in a firm’s office network. Here there is a problem: every web site is different and the amount of information varies immensely. Ideally we want to know the number of practitioners working for a firm in each city. This is sometimes available (and may be obtained from internal firm directories if made available). More likely we have to make do with lesser information which, nevertheless gives us insights into the importance of difference cities to firms. We collect two types of information on every firm in every city. First, we extract any information about the size of a firm’s presence in a city (e.g. number of separate offices, number of contact persons given) to create a size matrix. Second, we record all ‘extra-locational’ functions (obviously headquarters but also regional responsibilities, and specialist offices servicing more than their own city) to create an extra-locational matrix. Between them these two types of information provide a good indication of connections between a firm’s office(s) in a city and the rest of the firm’s office network. They constitute the two basic information matrices created at the study’s census points. To make the information comparable between firms these data are combined and standardised. Taking each firm separately, we assess the importance of each city to that firm on a scale of 0 to 5 to create a matrix describing the overall importance of cities to firms (see Annex). This is the basic analytical matrix used for all subsequent analysis.

All three matrices will be available for summer 2000 as output from the cross-sectional project. In this project, data collection for the basic information matrices will be replicated for 2001, 2002, 2003 and 2004 so that all changes can be recorded in annual change matrices. In addition for 2004 a new basic analytical matrix will be constructed. The plan is to compare 2000 with 2004 using the basic analytical matrices and to use the basic information change matrices to aid in interpreting results.

As with the cross-sectional project, analysis will use two different forms of matrix. First, the basic analytical matrix (city x firm) will be subject to a standard principal components analysis to uncover the basic structure in the data. Previous pilot work suggests that this technique is particularly sensitive for picking out hierarchical and regional clusters of cities (Taylor and Walker, 2001; Taylor and Hoyler, 2000; Taylor, Doel et al., 2000). Second, a connectivity matrix (city x city) is derived from the basic analytical matrix and multiple scaling will be employed to investigate patterns at different dimensions. Previous pilot work has shown that good results can be achieved with just two dimensions (Taylor, Hoyler et al., 2001). In addition other techniques will be experimented with to see whether additional findings can be obtained from either matrix (e.g. simple hierarchical analyses, various ecological clustering techniques). Again some pilot work has been done but in this case without, as yet, very interesting results.

Work Schedules

Schedule 1: Audit

Summers of 2001, 2002, 2003, 2004 for ten weeks starting in the first week of July.

This will involve employment of a research assistant7 to carry out web searches through the GaWC 100 sites to record all changes from the year before.


Week 1 - induction to project, equipment, web sites.

Weeks 2-9 - data collection (searching, recording, storing)

Week 10 - debriefing and dealing with loose ends

September – principal investigator reviews all material and creates matrices.

Schedule 2: Analysis

October 1 2004 – September 30 2005

This will involve employment of a post doc. research associate with experience of using multivariate techniques.


October 1 2004 to February 28 2005 – data analyses of basic analytical matrix for 2004 and compare to 2000.

January 1 2005 to May 30 2005 – interpretations of results using inter-year basic information matrices.

March 1 2005 to September 30 2005 – writing up findings for conference presentation, peer-reviewed journals, and professional/trade publication outlets.

Schedule 3: Dissemination

October 1 2005 – April 30 2006

This is primarily a write-up period for the applicant to produce a research monograph entitled "World city network formation, 2000-05". However, expenditure will be involved in personal dissemination. The applicant will present the results at (a) a users seminar in December 2005 hosted by the Corporation of the City of London8, and at (b) key academic conferences (RGS with IBG in January 2006 and at the AAG in March-April 2006).


Beaverstock, J.V., Smith, R.G. and Taylor, P.J. (2000) "World city network: a new metageography?" Annals, Association of American Geographers 90, 123-34

Knoke, D. and Kuklinski, J.H. (1982) Network Analysis. Beverly Hills: Sage

Taylor, P.J. and Hoyler, M. (2000) "The spatial order of European cities under conditions of contemporary globalization" Tijdschrift voor Economische en Sociale Geografie 91, 176-89

Taylor, P. J., Doel, M. A., Hoyler, M., Walker, D. R. F. and Beaverstock, J.V. (2000) "World cities in the Pacific Rim: a new global test of regional coherence" Singapore Journal of Tropical Geography 21, 233-45

Taylor, P.J., Hoyler, M., Walker, D.R.F. and Szegner, M.J. (2001) " A new mapping of the world for the new millennium" The Geographical Journal (in press)

Taylor, P. J. and Walker, D. R. F. (2001) "World cities: a first multivariate analysis of their service complexes" Urban Studies 38, 23-47

Taylor, P. J. (2001) "Specification of the world city network" Geographical Analysis (in press)


1. The GaWC web address is:

2. I use the term ‘firm’ here to describe all organizations that present an integrated network of offices to provide a given corporate service; this includes limited companies (e.g. banking), partnerships (e.g. law), associations (e.g. accountancy), networks of companies (e.g. management consultancy), and agencies (e.g. advertising).

3. The GaWC 100 consists of global corporate service firms in six sectors: 18 accounting, 16 advertising, 22 banking/finance, 11 insurance, 21 law, 16 management consultancy. See Annex.

4. All papers of mine (either published or in press) are available in a pre-publication version in the GaWC web site as GaWC Research Bulletins.

5. This is a very unusual type of network: most social science networks have just two levels: nodes and supra-nodal level. In this analysis there is not only a sub-nodal level but it is this level which is crucial in the construction of the network. Theoretical and practical implications of this unusual structure are described in Taylor (2001).

6. It is important, of course, to prevent particular idiosyncrasies of firms (usually related to a firm’s country of origin where its service will be greater than normal for a ‘global firm’) from distorting description of global strategy. With 100 firms idiosyncrasies are largely lost in the aggregation.

7. I am expecting to hire a new graduate each year, possibly one who is going on to do postgraduate work either at Loughborough or elsewhere.

8. This is a long way ahead but one of GaWC recent research associates (Dr Richard Bostock) has gone to work for the Corporation and we are therefore expecting to consolidate a relationship which will make this sort of activity our prime means for interacting with users.

Annex: Progress Report on Current ESRC Project


Dr Gilda Catalano (Università degli Studi della Calabria, Arcavacata di Rende, Cosenza) is the Research Associate appointed 1-6-00 to 30-7-01.

We are just completing the data collection phase of this research. We have identified 100 advanced producer service firms as ‘global’ using the criterion of having offices in ten or more cities with at least one office in each of northern America, western Europe, and Pacific Asia. These define the "GaWC 100" firms: 18 in accounting, 16 in advertising, 22 in banking/finance, 11 in insurance, 21 in law, 16 in management consultancy. Their offices and related material have been recorded for 320 cities across the world. Two types of information are recorded to ascertain the importance of each city to a firm’s global strategy: (a) a size measure, and (b) the extra-locational functions of offices. From this material a six point scale has been devised: 0 is no presence in a city: 2 is the standard level of presence for that firm; 1 and 3 represent situations where the evidence suggests the city is slightly less or more important than standard; 4 and 5 represent cities of critical importance to a firm’s network, with, for instance, headquarter cities scoring 5.

Theoretically the specification paper has been completed and accepted for publication and we have begun to think through development of ‘global connectivity indices’ for cities.

For results of this project, see GaWC Research Bulletins 77, 192 and 203.