GaWC Research Bulletin 89

GaWC logo
  Gateways into GaWC

This Research Bulletin has been published in Geography, 89(2), (2004),145-151.

Please refer to the published version when quoting the paper.


Urban Hinterworlds Revisited

P.J. Taylor and D.R.F. Walker


This is a short technical note that develops the concept of an urban hinterworld. A hinterworld is the pattern of a city's relations with other cities across the world. Within a framework of a world city network created by global service firms, new data are described and used to define the global network connectivity of 316 cities based upon the office networks of 100 global service firms. 'Absolute' hinterworlds are measured by aggregating the level of service provision that is available in a city for doing business in another city. The geography of this first measure closely mirrors global network connectivity and therefore a second 'relative' measure is devised. This is found in the residuals derived from regressing absolute service provision against global network connectivity. The methodology is illustrated by deriving and mapping the hinterworlds of London and Manchester.


In a previous paper (Taylor, 2001a), the notion of an 'urban hinterworld' was introduced as a means of describing relations between world cities. It was necessary to invent this new concept because the traditional urban geography concept of hinterland was found to be inadequate as a tool for describing the 'urban influence' of world cities under conditions of contemporary globalization. Hinterlands demarcate the service area of cities and towns as local service centres but in recent decades application of IT technologies has enabled some cities to become 'global service centres'. In particular, provision of advanced producer services (financial and business services) by 'global service firms' has meant that clients are being serviced through world cities across all regions of the world. In these circumstances, there are no boundaries to draw around a world city's hinterland, rather provision by its leading service firms can be to all parts of the world. However, the worldwide intensity of service provision will vary from city to city and it is this spatial variation that is captured by the concept of urban hinterworld. Such geographies were constructed in Taylor (2001a) to illustrate the hinterworlds of ten leading world cities.

The purpose of this short supplementary note is to develop our understanding of the concept of hinterworld in two main ways. First, the hinterworlds presented below are based upon a new and much improved set of data. In the conclusion of the earlier contribution (Taylor 2001a, 58) a caveat was inserted with respect of the relatively small number of firms (46) that the results were based upon. It was suggested, therefore, that before any general conclusions could be drawn replication using a larger data set was necessary. This is that replication using much better data: the results presented in this paper are from a large data matrix featuring 316 cities and 100 advanced producer service firms. The first section below describes these new data.

Second, we are much more analytically sophisticated in our measurement of service variation in this paper. Instead of simply mapping service provision from one city to another, we compute the service provision in relation to a city's overall global connectivity. This 'relative' measure of hinterworld, tells us where a city's service links are particularly strong or weak in relation to other cities' positions in the world city network. For instance, every city in our data has its highest two service connections with London and New York for the simple reason that far more service firms are located in these two cities than in other cities. The relative measure takes the latter into account and tells us whether a city has more or less linkage to one of these two cities than expected given London and New York's high level of connectivity. The residual analysis we develop to create these relative measures is described in the second section of this paper. Since one of the features of the new data is that it covers a wide range of cities, we illustrate the techniques by comparing the hinterworld of a 'global city' (London) to that of a much less important world city (Manchester).


Data on the relations between cities are notoriously difficult to find. This is especially true in the case of world cities because of the transnational nature of their connections. In fact the large-scale data we need has to be created. Although this has the obvious disadvantage in being very time-consuming, there is also the advantage that we can customise the data to our specific needs. This requires that we have a precise model of what it is we are measuring thus providing a rigorous rationale for the data creation. Just such a rationale is provided below in the form of a specification of the world city network. This is followed by a practical discussion of how the data creation was implemented. Finally we put the model and the data together to derive measures of global network connectivities for cities.

An Interlocking Network of Cities

World cities are places where financial and business service flows converge in the process of contemporary globalization. They are global service centres. The firms providing these services are the agents of world city network formation. In the last few decades, many advanced producer service firms have 'gone global' meaning that they have set up office networks in many cities across the world. In this way firms are able to provide a 'seamless service' to their global clients. From their tower block offices which typify the contemporary world city, these firms communicate information, knowledge, ideas, plans, instructions, advice, strategies, and risks to their other offices in cities around the world. In this way cities are interlocked in a global web of myriad flows of signs and signals.

The world city network can thus be specified as an interlocking form of network in which there are three levels: the network level in the world economy, the nodal level with the cities as the nodes, and the sub-nodal level with firms as the 'interlockers' (Taylor, 2001b). This tells us that to measure properties of the world city network, such as city hinterworlds, we need information on the office networks of service firms. Ideally we would want data on quantities and types of flows but this is never going to be possible on the scale we need because of the obvious commercial sensitivity of such information. However, it is reasonable to assume that large offices with many and/or important functions generate more flows than small offices with fewer functions. On this basis we can collect data on firms' office networks that indicate the importance of each office and from this infer differential levels of flows between offices. For instance, if two cities both house important offices within a firm's network, we can expect more flows than between another pair of cities that both house only small offices of the firm. Such information for a large number of firms in many cities will enable us to aggregate across firms to quantitatively define the world city network.

To summarise: specification of a world city interlocking network points data collection in the direction of office networks of global service firms and, in particular, we need information on the importance of each city office. For a more detailed discussion of this specification, see Taylor (2001).

Data Construction

We define a global service firm as one that has offices in at least fifteen different cities including at least one in each of northern America, Pacific Asia and western Europe. Using world rankings of firms in different sectors we used firms' websites to find their office networks and thus whether they were 'global' by our definition. In this way we identified 100 global service firms in six different sectors: 18 in accountancy, 15 in advertising, 23 in banking/finance, 11 in insurance, 16 in law and 17 in management consultancy. From the websites we gathered information on the size of a city office (e.g. number of practitioners) and their extra-territorial functions (e.g. regional headquarters) for a total of 316 cities. The latter were chosen to be as inclusive as practically possible: the capital cities of all but the smallest states were included plus many other important cities from larger states.

The key problem with the data obtained from this information gathering is that it varies immensely in form, quantity and quality across firms. To make this information comparable we have devised a simple 6-point coding which we call the 'service value' of a city to a firm. At one end of the scale a city scores zero if the firm has no presence in the city, at the other end the headquarter city scores 5. Other scores are reached as follows. The 'typical office' of a firm is scored a two, this is the presumption of every office so there has to be a reason for allocating another score. For instance, a one is scored for a city office with something lacking (e.g a law office without a partner). On the other hand a 3 is scored for especially large offices, and for those with important extra-territorial functions a four is scored. In this way a 100 x 316 data matrix of service values has been created. Each column is a codified listing of the global locational strategy of a firm; each row is a codified global service mix of a city.

In summary: a large data matrix has been created relating each of 316 cities to each of 100 global service firms in the form of 31,600 service values indicating the importance of a city within a firm's office network. For more details on this data collection exercise, see Taylor et al. (2002).

Global Network Connectivity

The simplest measure we can make of each city is to sum its row of service values and compute a measure of its total service, the 'site service status' of the city (Taylor, 2001b, 186). This simple measure indicates the importance of a city as a node within the world city network but it is not a relational measure, it does not indicate linkages between cities within the network, the 'situational service status' of a city (p. 187).

To develop a relational measure we build upon the 'elemental interlock link' that can be computed between two cities. This is simply the product of a firm's service values for a pair of cities and will range from zero (where either one of the cities has no office of the firm) to 25 (indicating a shared headquarter function). Most non-zero elemental links will have a value of four indicating two cities each with a typical office. The sum of a city's elemental interlock links across all firms with each of the other 315 cities defines the 'total network interlock linkage' (Taylor, 2001b, 187) which we will term a city's global network connectivity. This measure shows how well connected a city is within the world city network. In our data there are three cities that have a global network connectivity of zero (Pyonyang, Lucknow, Alma Ata) indicating the presence of no global service firms and at the other end of the spectrum come London and New York with connectivities of 63399 and 61895 respectively (Taylor et al. 2002, Table 3). In the analyses reported below the global network connectivies of all cities are converted to proportions of the highest connectivity (London's) to ease interpretation. Since we will be using this measure later, these connectivities are shown in Figure 1 on an 'archipelago' cartogram of the 123 cities.

In summary: the data matrix of service values can be used to compute the global network connectivity of each city indicating its relational status - global network connectivity - within the network. For more information on the measure, see Taylor (2001b); for discussion of the connectivity results, see Taylor et al. (2002).


To compute a city's hinterworld we need to specify its external relations with other cities. For this exercise we deal with just the top 123 cities in terms of global network connectivity. These cities have been selected for having at least one fifth of the highest level of connectivity (London's); it is necessary to limit the number of cities in this way because of the sparseness of the data matrix (increasing numbers of zeros). Obviously with less firms represented in computing a measure, a result becomes that much less reliable. Thus the results reported below are derived from a reduced 100 x 123 data matrix.

A city's hinterworld consists of the levels of service it provides for doing business in each of the other 122 cities in the new data matrix. This is initially computed as follows. First, count the number of firms present in each city. For each city multiply this number by 5, the maximum service value. This constitutes the highest possible level of service that a city could expect in another city (i.e. the other city houses the headquarters of every single one of a city's global service firms). Thus in the simple data set shown in Table 1, highest levels of possible service are 20 for London, 15 for Munich, and 10 for Lagos. Now for each city, take other cities in turn and sum their service scores but ONLY for firms present in the original city. For instance, starting with London, the sums for Munich and Lagos are 10 and 4 respectively; starting with Munich the sums for London and Lagos are 12 and 2; and starting with Lagos the sums for London are 10 and 2. The latter sums are expressed as proportions of the highest level of possible service. For instance, the proportions for London are Munich 0.5 (= 10/20) and Lagos 0.2 (= 4/20). All such computations are shown in Table 2.

The interpretation is relatively simple. The columns in this table define the level of service that can be expected in a city when visiting a global service firm in a row city. Thus, going into an office in London to do business in Munich the service level is 0.5, but to do business in Lagos the level falls to 0.2. Notice that from Lagos, doing business in London has a 1.0 service level showing that Lagos's two service firms in Table 1 have their headquarters in London. In contrast, the lowest level of service in this data is a paltry 0.13 for doing business from Munich in Lagos. These columns represent the servicing linkages that form the basis for describing the hinterworld of a city.

We have subsequently recognised that there is a problem when comparing the hinterworlds of cities using results such as those in Table 2. Notice that in this table London appears with very high service levels for the other two cities and Lagos provides low levels. This is obviously reflecting the network position of these cities in this small data set. Thus when computed for the 100 x 123 data set, it is found that every city has its highest external provision in either London or New York. In general, we can note that external service provisions tend to closely follow the level of a city's global network connectivity. Thus mapping column values to a large degree replicates the connectivity map (Figure 1) so that all hinterworlds, although not exactly the same, look very much alike. To overcome this comparative deficiency we have added an extra step to describing hinterworlds.

We will term the external service provisioning results from Table 2 depictions of 'absolute' hinterworlds. Taking out the underlying general influence of global network connectivity from the absolute provisioning values for a city is a relatively simple task. Scatter diagrams of global network connectivity against a city's external provisioning levels shows a strong positive linear relationship in every case. Thus absolute provision can be regressed against connectivity using the simple equation:

y = a + b x

where y is the absolute provision values and x is global network connectivity. Calibrating this equation for any city produces an estimate of the level of service provision given a city's connectivity. The difference between this estimate and the actual level of provision is the residual. These residuals define a 'relative' hinterworld - where a city is strongly serviced and where it is weakly serviced with respect to the servicing city's position in the world city network (i.e. its connectivity).

Results from such an analysis for London and Manchester are shown in Figures 2 through 5. The scatter diagrams (Figures 2 and 3) show that for London, the largest positive residual is Washington, DC and for Manchester it is Birmingham. This suggests a contrast between the London's 'globality' and Manchester's more intense 'local' relations. This idea is augmented by their respective largest negative residuals with London's being in Europe (Athens) and Manchester's being in China (Guangzhou). Looking at the maps of their respective 'relative' hinterworlds (Figures 4 and 5) confirms this interpretation only partly. To be sure London shows itself to be very strongly linked to USA and Pacific Asian cities and certain European cities (particularly German). Manchester, on the other hand, also shows a 'global pattern' of strong linkages but this time it reflects historical British ('old') Commonwealth connections with Canada and Australian cities very well represented. In the case of London, these historical links are missing, overwhelmed by more recent developments in London's global connections, especially in banking. The latter is especially reflected in the Pacific Asian over-linkages, which is exactly where Manchester is under-linked. Both cities have strengths in the USA and Europe but with clear differences: London's greater connections with German cities again reflects its global banking prowess compared to Manchester.


Globalization is an inherently geographical bundle of processes that involve a jump in scale for many human activities. This is most clearly illustrated in the major cities of the world that provide the spatial structure for many new global spaces of flows. In changing scales we will sometimes have to modify our geographical concepts and developing the notion of city hinterworld from city hinterland is a prime example. But altering this basic relational urban concept requires new means of measuring and mapping a city's external relations. This is what we have tried to achieve in this paper.

The paper is a brief technical note whose purpose has not been to explore in detail the substantive findings that hinterworld mappings provide. For examples of the latter, see Taylor (2002a and b) for Low Countries cities and West Asian/Northern African cities respectively. We are only just beginning to empirically research the new geographies consequent upon contemporary globalization; to keep abreast of developments relating to cities in globalization readers may consult


Taylor, P.J (2001a) 'Urban hinterworlds: geographies of corporate service provision under conditions of contemporary globalization', Geography, 86, pp. 51-60.

Taylor, P.J (2001b) 'Specification of the world city network', Geographical Analysis, 33, pp. 181-94.

Taylor, P.J. (2002a) Amsterdam in the World City Network. Amsterdam: AME (University of Amsterdam).

Taylor, P.J. (2002b) 'West Asian/North African cities in the World City network: a global analysis of dependence, integration, and autonomy', The Arab World Geographer, 4, pp. 146-59.

Taylor, P.J., Catalano, G., and Walker, D.R.F. (2002) 'Measurement of the world city network', Urban Studies, (in press)


Table 1: A simple matrix of service values

Firm A

Firm B

Firm C

Firm D

















Table 2: Levels of servicing derived from Table 1

















Figure 1: Global network connectivity

From Taylor et al. (2002)

This cartogram places cities in their approximate relative geographical positions. The codes for cities are:

AB Abu Dubai; AD Adelaide; AK Auckland; AM Amsterdam; AS Athens; AT Atlanta; AN Antwerp; BA Buenos Aires; BB Brisbane; BC Barcelona; BD Budapest; BG Bogota; BJ Beijing; BK Bangkok; BL Berlin; BM Birmingham; BN Bangalore; BR Brussels; BS Boston; BT Beirut; BU Bucharest; BV Bratislava; CA Cairo; CC Calcutta; CG Calgary; CH Chicago; CL Charlotte; CN Chennai; CO Cologne; CP Copenhagen; CR Caracas; CS Casablanca; CT Cape Town; CV Cleveland; DA Dallas; DB Dublin; DS Dusseldorf; DT Detroit; DU Dubai; DV Denver; FR Frankfurt; GN Geneva; GZ Guangzhou; HB Hamburg; HC Ho Chi Minh City; HK Hong Kong; HL Helsinki; HM Hamilton(Bermuda); HS Houston; IN Indianapolis; IS Istanbul; JB Johannesburg; JD Jeddah; JK Jakarta; KC Kansas City; KL Kuala Lumpur; KR Karachi; KU Kuwait; KV Kiev; LA Los Angeles; LB Lisbon; LG Lagos; LM Lima; LN London; LX Luxembourg; LY Lyons; MB Mumbai; MC Manchester; MD Madrid; ME Melbourne; MI Miami; ML Milan; MM Manama; MN Manila; MP Minneapolis; MS Moscow; MT Montreal; MU Munich; MV Montevideo; MX Mexico City; NC Nicosia; ND New Delhi; NR Nairobi; NS Nassau; NY New York; OS Oslo; PA Paris; PB Pittsburg; PD Portland; PE Perth; PH Philadelphia; PL Port Louis; PN Panama City; PR Prague; QU Quito; RJ Rio de Janeiro; RM Rome; RT Rotterdam; RY Riyadh; SA Santiago; SD San Diego; SE Seattle; SF San Francisco; SG Singapore; SH Shanghai; SK Stockholm; SL St Louis; SO Sofia; SP Sao Paulo; ST Stuttgart; SU Seoul; SY Sydney; TA Tel Aviv; TP Taipei; TR Toronto; VI Vienna; VN Vancouver; WC Washington DC; WL Wellington; WS Warsaw; ZG Zagreb; ZU Zurich


Figure 2:

Figure 3:

Figure 4: London's hinterworld

For city codes, see Figure 1.

Figure 5: Manchester's hinterworld

For city codes, see Figure 1.


Edited and posted on the web on 24th June 2002

Note: This Research Bulletin has been published in Geography, 89(2), (2004),145-151