GaWC Project 81

GaWC logo
  Gateways into GaWC

Benchmarking the World City Network: City Connectivities on the Eve of the Current Financial Crisis

Funded by ESRC (2009-2010)


Grant holders: Peter Taylor, Michael Hoyler and Kathy Pain
Research Associate: Sandra Vinciguerra




This proposal to measure and analyse the world city network in 2008 is part of a research programme that has been on-going for over a decade. There have been four main stages of this research carried out through the Globalization and World Cities (GaWC) group at Loughborough University:

  1. In 1997-98 an ESRC project on measuring London's links to other cities in the global economy led to realization that the methods could be extended to define a world city network.
  2. In 1999-2000 the interlocking network model was devised for studying inter-city relations. This provided a theoretically-informed specification of the world city network that provided an explicit guide to data collection.
  3. In 2000-01 the above model was applied in an ESRC project to measure and analyse the world city network in 2000. This was the first comprehensive description of city-centred globalization.
  4. In 2001-06 an ESRC project used the interlocking network model to monitor changes in the world city network and provide a second comprehensive description for 2004. This enabled 2000-04 changes to be measured and analysed.

Through this work GaWC researchers have become agenda setters for our sub-area of urban studies.

The project plan preceding this application was to extend the above programme of work through another ESRC project proposal for more data collection in 2008. However, like all creative on-going research we were learning on the job and this came to a head when we began comparing the 2000 and 2004 data. In 2000 we collected data on the office networks of 100 global service firms across 315 cities. Firms were chosen in terms of their network (e.g. offices in at least 15 different cities) and the ease with which data could be collected about them (e.g. the quality of their web sites). In 2004 the idea was to find changes through using the same set of firms. However, between 2000 and 2004 the 100 firms were reduced to 80 through mergers and other changes. Clearly our methodology under-estimated the population dynamics among major firms. Thus our initial intention of repeating the exercise in 2008 to provide a further point of change was found to be problematic: how far further would the number of firms fall? This problem was particularly acute because our methodology uses aggregation wherein a large number of firms is necessary to provide credible results. Hence we were forced to make a major reassessment of our methodology.

The key alteration was to define our roster of firms by size alone: for instance, to select for our 2008 data collection the top 25 firms in a sector using the latest (2007) global income figures. The content of top 25 will inevitably change in the future but the ranking will now form the same basis for any subsequent data collection. Thus network change will be measured in terms of not only changes in firms' offices but also in changes to firm membership of the top 25. This was clearly the sensible way to proceed and produce consistent monitoring of change in the world city network.

However, there is a major research labour implication in the new methodology. The initial data collection was designed to be completed by a single researcher over a nine-month period (covering both collection and checking). This was possible because originally, for any firm considered, if data on that firm was hard to come by, the firm was simply dropped from the research. With the new methodology firms are selected by specific criteria and have to be investigated until adequate information on them is found. From our previous data collection experience we estimated that this would increase the time input at least tenfold. Our methodology had moved from being a relatively simple and quick (for global coverage) piece of research work to a much improved exercise but one requiring a very different scale of work. Fortunately GaWC operates as a worldwide network of city researchers and therefore we have been able to put together a global team to accomplish the new data collection needed. The team comprised the Global Urban Competitiveness Project at the Chinese Academy of Social Sciences (CASS, Beijing), researchers in the Department of Geography at Ghent University, and the GaWC group at Loughborough. There have been four stages in this new research:

  1. In 2007 two meetings with all three institutions represented were held first in Loughborough and then in Beijing to plan the collaboration. A division of labour was agreed and the size of the exercise established. It was decided to extend the number of service firms to 175 and to add 25 media companies (the latter relates to some of our earlier work). The cities to be included were reviewed and extended. In addition CASS insisted on further data collection on ‘city market' indicators (headquarters, hotels, events, and science parks).
  2. In November-December 2007 Loughborough GaWC researchers produced a comprehensive 37 page Manual for Data Collectors (with associated Excel matrix frames for cities, networks (firms) and markets).
  3. From January to May 2008 CASS researchers carried out the data collection with files completed for distribution in July. Twenty researchers were assigned to this work under the leadership of Professor Ni Pengfei (CASS) and directed by Huang Jin (Beijing University of Post and Telecommunication), Lu Fengyong (CASS), and Yang Xiaolan (Beijing University of Post and Telecommunication).
  4. In August-September 2008 a thorough check of this data was conducted by the Ghent University team carried out by Anneleen Devos under the supervision of Dr Ben Derudder and Professor Frank Witlox. This work was completed and circulated on September 23rd 2008.

Only with the data checking completed satisfactorily for all parties, have we have begun to create this application. The data are the common property of CASS, the Department of Geography at Ghent University, and the Department of Geography at Loughborough University. We have free access to and use of the data for the purpose of this application.

The main output of this data collection is a 175 service firms x 525 cities matrix showing each firm's use of each city (on a scale of 0 to 5). There are two further stages to bring this research to fruition:

  1. We have a programme of analysis within our current resources and collaborations to carry out the basic analyses on this data equivalent to 2000 and 2004 (mainly connectivity analysis, principal component analyses, and fuzzy set analyses).
  2. This submission is about the Loughborough GaWC researchers extending our repertoire of techniques through bringing in a creative social statistician as a research associate to maximise the knowledge potential of our impressive new data. We see this as very much an exploratory role that will be looking to develop network analyses, visualization techniques and other methods to bring our researches to a wide range of audiences.

Thus this is a rather unusual submission in that the bulk of the research, which is the data collection, has already been done and we are asking for funds to pay for someone who can bring new skills and insights into our researches to fully exploit the research done so far.

One final point of context: the timing of the data collection was fortuitously just right in terms of understanding contemporary globalization dynamics. What we have produced is a set of data that represents the world city network just before the 2008 financial crisis came to a head. We are planning a next data collection for 2010 (informally agreed with our collaborators) so that we will have network change for a before-and-after crisis scenario. In other words with the work planned in this submission complete, we will be in a perfect position, uniquely, to prepare for analysis of the ‘geo-economic transition' we are currently experiencing and thereby answering such questions as to what degree is NYLON (New York-London) losing its globalization reach in the new economic world order.

Aims and Objectives

This research project has one basic aim: To understand the structure of the world city network in 2008 through new insightful analysis of the offices networks of 175 leading advanced producer service firms.

This understanding will be based upon new quantitative analyses interpreted through the theoretical lens of the interlocking network model. This will provide a benchmark for later research on how today's geo-economic transition is changing the global space economy.

There will be six objectives that both improve or extend current methods and experiment with new methods.

A. Analytical improvements:

  1. The measurement objective: to review our current methods of analysis and consider how we might add weightings to firms to improve measurement of the world city network. The current data collection is based upon financial rankings of firms and these data may be incorporated into analyses. These should produce better estimates of spaces of flows and thereby better descriptions of the world city network.
  2. The disaggregating objective: to expand analyses by breaking down the data matrix by sectors and/or world regions to further elucidate the complexity of the world city network. We have done this previously but not very fully because disaggregating soon brought numbers of cities or firms down to unreliably small numbers. With the new larger data set we can sensibly return to this method to produce new breakdowns of analyses with real confidence for their reliability.
  3. The power objective: to measure different forms of power (network, command) operating in and through the world city network. As in objective 2, we have previously experimented with such analyses but only with this larger data set can we fully develop this line of research.

B. New methodological developments:

  1. The technical objective: to experiment with new techniques in network analysis methodologies to enable new insights about the world city network to be realized. We are familiar with other techniques produced in related network research but we do not envisage mechanical replication of such work; techniques will have to be used in ways compatible to our theoretical framework. Nevertheless we expect large strides to be made through this objective with the new data.
  2. The visualization objective: to develop new visualization techniques possibly including animation to show a dynamic world city network on the web. This has long been a neglected theme in our research and the new larger and better data provides the perfect opportunity to make advances for what is potential the key tool for disseminating our research to non-specialist practitioners (in firms and cities).
  3. The continuity objective: to attempt to measure a changing world city network from 2000 through 2004 to 2008. This is highly problematic due to changing data collection methodology but is worth pursuing because of a dearth of change data in this field. We do not want to waste our past data efforts and therefore need to consider creative ways to produce credible change results to 2008.

The overriding purpose in all the objectives is to get the most out of an exceptional set of data.

Research Questions

The specific research questions of this research revolve around the role of cities in globalization interpreted through the interlocking network model. The questions are about the structure of the world city network, descriptions and relations in an extensive research exercise. Here are some key questions.

A. At the network and sub-net levels:

  1. Which city is the most connected city in the world city network and what is the ranking of other cities below the leading city? This is measured by our standard measure of interlocking network connectivity but may be improved and other measures assessed.
  2. How are the rankings in (i) changed when specific sectors are analysed separately? For instance, earlier work has shown a distinctive top 3 for financial services (London, New York, Tokyo) but for other services Tokyo drops down the ranks.
  3. Questions (i) and (ii) can be repeated for sub-nets in the world city network defined as world regions (e.g. Latin America) or large countries (e.g. China).
  4. How do cities cluster together in terms of their inter-city relations? Past researches have shown strong regional patterns plus surprises such as an ‘old' Commonwealth grouping. Do such patterns reappear in 2008 or are they being eroded by continuing globalization?
  5. What are the power configurations in the world city network? This will need to be answered using different concepts of power relations.

B. City-centric analyses:

  1. For a given city, where are its strongest network links and where are its weakest? For instance major cities are likely to have a more widespread pattern of strong links, and lesser cities will likely be more locally connected.
  2. How do specific cities compare within the same region, country or level of connectivity? For instance, compare Sao Paulo with Buenos Aires, Beijing with Shanghai, and London with New York.
  3. How are specific cities changing within the world city network from 2000 through to 2008? This is a key question practitioners (e.g. city policy makers, real estate firms) ask.
  4. In which sectors is a given city particularly strong and where is it weak compared to its peers? This is another question of direct relevance to urban policy makers.

Research Methods  

This is an extensive research exercise in which standard methods will be employed and improved (e.g. principal components analysis); methods previously developed will again be deployed and extended (e.g. interlocking network connectivity); and new methods will be experimented with and developed – customised – for our particular needs (e.g. notably in network analysis and visualization work). Methods are what this research submission is all about – see objectives above.

Uniqueness of the Data

It cannot be emphasized enough that the data for this proposed project are unique in structure, size and content. They constitute a customised empirical input to the theoretical framework defined by the interlocking network model.

Description of the Data

The data consist of a ‘service value matrix' that arrays cities against advanced producer service firms. Each cell is an estimate of how much a firm uses a city in its work (i.e. how important is that city's office in the firm's overall city network?). This is the city's ‘service value' to that firm.

Using publicly available information on firms (usually from their web sites), cities are coded from 0 (no office for a given firm) to 5 (housing the firm's headquarters). Such assessments have been made for 175 firms across 525 cities to provide 91,875 pieces of information as input into the analyses.

Checking the Data

Given the size of the data and the remote nature of the collection process (in Beijing), checking the data is more than usually vital for this research project. The task of checking the data was undertaken by our colleagues at Ghent who did an exceptionally thorough job. Putting together this submission was not started until we were jointly assured with Ghent that the data were of a high quality and that the inevitable errors in such a large exercise had been eliminated as far as reasonably possible.

Framework and Methods of Analysis

The theoretical framework for the research is the interlocking network model and the methods of analysis are the raison d'être of this submission and thus are outlined above in objectives and research methods.


  1. Research papers. The nature of this research means that there will be a substantial number of papers. They will be posted as soon as written as GaWC Research Bulletins. Currently numbering nearly 300, these are publicized though GaWC-i, a newsletter electronically sent to 2,000 plus in our address bank (includes academics and practitioners).
  2. International journal articles. We have an outstanding publication record – our Bulletins are invariably published about a year after posting. We will submit at least four papers to leading international peer-reviewed journals in the fields of Geography and Urban Studies.
  3. Academic conferences. We will cover major relevant international UK and US conferences in presenting our results and reach out to wider global audiences through our international collaborative network.
  4. Practitioner conferences. We are actively involved in two major practitioner conferences: Metropolis 2008 in Sydney and Global City Abu Dhabi in April 2009. We presented our initial early results to the former, and plan to present further results to the latter. We expect links with both conferences to continue into later years for further dissemination of results to practitioners. In addition we expect to make interventions into urban policy debates in China, Belgium, the UK and the USA. Hence dissemination will be ongoing.
  5. Book. It is intended that the results from this submission will combine with a similar exercise in 2010 to create a state-of-the-art book on the 2008 financial crisis and its aftermath.


For results of this project, see GaWC Research Bulletins 335, 350, 368, 369, 377, 378.