GaWC Research Bulletin 152

GaWC logo
  
 
  Gateways into GaWC

This Research Bulletin has been published in Urban Studies, 42 (13), (2005), 2371-2388.

doi:10.1080/00420980500379503

Please refer to the published version when quoting the paper.


(Z)

An Appraisal of the Use of Airline Data in Assessing the World City Network: A Research Note on Data

B. Derudder and F. Witlox*

Abstract

Information on air passenger flows is potentially a prime data source for assessing spatial patterns in the world city network, but previous analyses of this issue have been hampered by inadequate and/or partial data. The ensuing analytical deficiencies have reduced the overall value of these analyses. Therefore, this paper examines how some of these deficiencies might be rectified. First, we review the rationale for using airline data to analyse the world city network. Second, we assess the problems encountered by previous research. The third section elaborates on the construction of a global intercity matrix based on the so-called Marketing Information Data Transfer database and explains how this matrix can circumvent some previously identified problems.

Keywords: world city network, airline data, globalization.


I. INTRODUCTION: AIRLINE DATA AND THE FORMATION OF THE WORLD CITY NETWORK

The world city system has frequently been described as a hallmark of a global network. This is because the most seminal contributions to this body of research (Friedmann, 1986; Castells, 1996; Sassen, 1991, 2001) consider worldwide intercity relations crucial in linking urban development to the restructuring of the world economy. Widespread theoretical discussions based on the notion of a ‘world city network’ (WCN) have rarely been accompanied by genuine empirical network analyses. (Notable exceptions include Smith and Timberlake, 2002; Alderson and Beckfield, 2004; and Derudder and Taylor 2005.) There are few fully fledged network analyses of the world city system mainly because of the paucity of data detailing intercity flows at the global level (Smith and Timberlake, 1995a, 1995b; Short et al., 1996; Taylor, 1997, 1999). While there have been recent attempts to rectify this situation, the major exemption to the empirical conundrum at the heart of research on the WCN has been the compilation of information on communications networks in general and air transport in particular.

The potential usefulness of airline data for analysing flows between major world cities is clear: data are comparatively easy to obtain; air transport is traditionally organized through cities rather than through states; and transportation is mainly about connections and flows. However, previous assessments of the WCN based on airline data have faced problems (see also Taylor, 1999; Beaverstock et al., 2000). For example, in spite of its apparent association with global flows, some analyses have incorporated a subtle form of state-centrism. Rimmer (1998, p. 460), for instance, has presented a world city analysis based on data on ‘international passengers’. Important connections, such as flights between Los Angeles and New York, are missed by this approach, which tends to underestimate the "global cityness" of some major US urban centers. Another problem in applying airline data arises because a number of passenger flows reflect processes outside the formation of the WCN. An illustration of this is Kunzmann’s (1998, p. 49) mapping of the European urban hierarchy based on air passenger flows. Kunzmann lists 14 airports that are secondary to the big three (London, Paris, and Frankfurt), including Munich, Milan, Madrid, and Palma de Mallorca. The high ranking of the latter obviously reflects its role as one of the most popular holiday destinations in Europe; nobody would seriously argue that it is a major world city. While there are other problems, the main point is that these two examples show that previous spatial analyses of the WCN based on airline statistics were hampered by inadequate data. The main purpose of this paper is to explain how this problem may be, at least partially, solved.

This paper is organized as follows. First, we address the possibilities and drawbacks of airline data in a general discussion of the chief theoretical assumptions and existing data sources in contemporary WCN research. The WCN is made up of multiple intercity networks, and a full identification of its structure requires multiple data sources; hence, there is the need to discuss the relevance of airline data beyond the more specific shortcomings outlined in the remainder of the paper. In the second section, we present an overview of previous explorations of the WCN based on airline data and show that most of these analyses have not fulfilled their potential. Rather than providing an extensive overview of all existing airline analyses, we explain how the inadequacies of previous databases and frameworks for analysis have prevented a clear-cut translation of air transport databases into spatial analyses of the WCN. The third section outlines the construction of an intercity matrix based on the so-called Marketing Information Data Transfer (MIDT) database, which contains detailed flight information of bookings made through Global Distribution Systems (GDS). The discussion in this section primarily focuses on how the information contained in this database might circumvent some of the difficulties described in the second section.

II. THE RELEVANCE OF GLOBAL INTER-CITY MATRICES

II.A. Rationale

The initial identification of the formation of world cities by Friedmann and Wolff (1982) was based on the identification of a major restructuring of the world economy from the late 1970s onwards. This restructuring was based on the global reach that subsequently characterized the production strategies adopted by major corporations. The organization of the global commodity chains in this so-called ‘new international division of labour’ required a number of critical command and control points in order to function. World cities were considered such entities. Since Friedmann and Wolff’s (1982) seminal paper, research on world cities has evolved enormously in both depth and scope. That is, while the literature developed through cross-fertilization with, for instance, research on global migration (Samers, 2002) and gentrification (Smith, 2002), the conceptualization of world cities has, in itself, led to advances. For example, Sassen’s The Global City (1991) marked a shift of attention to global flows resulting from the critical servicing of worldwide production rather than to its formal command through corporate headquarters of multinational enterprises (MNEs). Sassen’s approach focuses on the attraction of advanced producer-service firms (providing professional, financial, and creative services for businesses) to major cities with their knowledge-rich environments and specialist markets. In the 1980s and 1990s, many such service firms followed their global clients to become important MNEs in their own right. Consequently, these advanced producer-service firms created worldwide office networks covering major cities in most or all world regions. According to Sassen, the myriad of interconnected service complexes led to the formation of the WCN. In the mid 1990s, Castells (1996) described contemporary globalization as a ‘network society’, in which a network of world cities was deemed an important example of the so-called ‘space of flows’. Hence, Castells (1996, p. 380) argued that world cities accumulate and maintain their wealth and power because of the process “that connects advanced services, producer centres, and markets in a global network”. More recently, Smith (2003a, 2003b) has developed ideas from post-structuralism, actor-network theory, non-representational theory, and complexity theory to begin to produce a topological consideration of cities in global networks.

There are many other important and interesting world city conceptualizations beyond the seminal contributions briefly discussed here. However, in the context of this paper, the main issue is that, regardless of the broad diversity of world city conceptualizations, authors have increasingly stressed the importance of a relational stance in world city analyses. That is, it is generally acknowledged that world cities are based on what flows through them (information, knowledge, money, and cultural activities) rather than what is contained in them. Moreover, what lies at the root of world city formation is the existence of a large number of global relations (Allen, 1999; Taylor, 1999).

From an empirical point of view, the consequences of this clear-cut relational standpoint are self-evident. As all measurement and data should be products of theory, empirical analyses of the WCN should reflect the relational perspective that lies at the root of its conceptualization. To develop the view of world cities as a process based on global networking and connectivity, relational data are required. This is most clearly explained by Smith and Timberlake (1995a, 1995b), who stress the importance of constructing and analysing a variety of databases that take the form of ‘cities in global matrices’. However, the twelve types of linkage identified by Smith and Timberlake (1995a, 1995b) have largely remained a ‘wish list’ for world city research. This confirms the low quality of existing data sources in this research field (Short et al., 1996; Taylor, 1997, 1999). Generally, data on flows between cities are conspicuous by their absence, and therefore, with few exceptions, empirical world city research has neglected the relations and linkages between world cities. Taylor (1997, 1999) suggests that there is a lack of suitable data because most data collection agencies focus on attribute data because it is generally easier to collect, and most users prefer information in this format. Furthermore, the relational data that are available (such as trade data) are inadequate for world city research because they mainly cover states—the prime generators of data—rather than cities. Thus, although there are data on flows between countries, there are little on flows between cities located in different countries.

II.B. Existing Relational Data

Although empirical analyses of the WCN have long suffered from this “iron grip of the nation-state on the social imagination” (Taylor, 1996, p. 1923), there have been some exceptions to this empirical conundrum. Some of the most well-known examples are the airline databases dealt with in this paper. These include Beaverstock’s (2004) analysis of the intercity career trajectories of managerial elites and Taylor et al.’s (2002a) creation of intercity matrices from the location strategy of producer service firms. As these and other databases are based on different foundations, we discuss their rationale and relevance against the background of the theoretical underpinnings of world city research. A second reason to present an overarching discussion of existing relational data sources is that any empirical analysis of the WCN is by definition partial and one sided. The WCN is made up of multiple intercity networks, and a full identification of its structure must address this inherent multiplicity (see Smith and Timberlake, 1995a, 1995b). Thus, a discussion of the inherent conceptual relevance of airline analyses must go beyond the endemic data deficiencies outlined in the remainder of the paper.

To consider this conceptual relevance, we consider the theoretical foundations of WCN research, which reveals that world cities have been rooted in different frameworks. Some of the chief concepts include the “new international division of labour” and “global commodity chains” (Friedmann, 1986), “globalization” (Sassen, 1991, 2001), and the “network society” (Castells, 1996). It is self-evident that the merits of each data source will vary with the conceptualization. We therefore confine ourselves to an assessment of the existing data sources against the background of Castells’s (1996) influential theoretical work entitled ‘The Rise of the Network Society’. Castells’s assumptions are of particular interest for such a comparative assessment, as his prime contribution to this literature is to locate WCN within a rich and comprehensive theoretical context. That is, the WCN is an important network within one particular layer of the so-called ‘space of flows’, which represents the dominant spatial logic of a ‘network society’.

Castells (1996, p. 412) argues that the network society is organized around “a space of flows”, which is “the material organization of time-sharing practices that work through flows”. The space of flows has three layers: (i) the material basis for those networks; (ii) the places that constitute the nodes and hubs of networks; and (iii) the spatial organization of cosmopolitan elites in terms of work, play, and movement. The middle layer is of crucial importance, as Castells (1996, p. 415) identifies world cities as “the most direct illustration” of hubs and nodes that make up the space of flows1. Hence, the various relational data sources lead to the measurement of several types of contingent flows that are thought to provide information on the spatiality of an aspect of the middle layer in this three-layer structure. As airline data, the subject matter of this paper, is firmly rooted in the first layer, we work our way back from the third to the first layer.

Table 1: Inter-city data sources in Castells’s three-layered space of flows.

Layer 1

material basis for space of flows

Airline network

Internet backbone

Smith and Timberlake (2002)

Moss and Townsend (2000)

Layer 2

nodes and hubs of the space of flows

MNE-generated city system

Connections between post-industrial production complexes

Alderson and Beckfield (2004)

Taylor et al. (2002a)

Layer 3

spatial organization of dominant, managerial elites

Exposing the geographies of the super-rich

Expatriation of managerial elites

Beaverstock et al. (2004)

Beaverstock (2004)

 

The spatial organization of dominant, managerial elites

The main thrust of the empirical approach taken in Beaverstock (2004) is that expatriation of managerial elites embodies a major globalization strategy of service firms. As a deliberate organizational policy expatriation helps to develop, manage, and diffuse idiosyncratic knowledge between all units in a global network in order to service the client and increase profitability and market share. For example, Perkins’ (1997) review of expatriate organizational strategy shows that the rationale for employing expatriates is that a firm requires expatriation to fulfil its organizational and operational strategies. As a consequence, “[e]very successful major business will invest heavily in the development of a distinctive International Cadre of executives, capable of transferring the enterprise’s commercial and operational philosophies and systems into every location in which they wish to do business.… This group—capable of thinking global, acting local, and vice versa—will be among the premium capital an organization will wish to have access to” (Perkins, 1997, pp. 62–63, quoted in Beaverstock, 2004, p. 167). Following this lead, Beaverstock unpacks the spatiality of expatriation within corporate networks of some major London-based law firms. The major implication for this overview of relational intercity data is that Beaverstock presents, in essence, a three-layer analysis of the space of flows. This analysis reveals some basic tendencies in the overall structure of the WCN, such as London’s strong links with Hong Kong, New York, Paris, Singapore, and Frankfurt. (For a similar but more encompassing approach, see Beaverstock et al., 2004.)

A corporate organization approach

Databases detailing intercity relations in the second layer of the space of flows are more direct in their treatment of world city connections because they focus explicitly on the agents that constitute world city networks. The most important network makers in this respect are firms that pursue global strategies and thereby use individual world cities as bases for their worldwide location schemes. Hence, efforts to construct data adopt a ‘corporate organization approach’. The two most elaborate examples of this approach are: (i) the research pursued by the Globalization and World Cities Group and Network (GaWC, http://www.lboro.ac.uk/gawc), which is epitomized by Taylor et al. (2002a), and (ii) the recent paper by Alderson and Beckfield (2004).

The GaWC has developed a methodology for studying the formation of the WCN, which has largely been based on Sassen’s (2000, 2001, 2002) work on place and production in the global economy. The GaWC’s approach starts from the observation that advanced producer-service firms ‘interlock’ world cities through their intrafirm communications of information, knowledge, plans, directions, advice, etc. to create a network of global service centres. Building on this specification of the WCN, information was gathered from global service firms about the size of their presence in a city and about any ‘extraterritorial’ functions of their offices. This information was converted into data that summarize the location strategies of 100 firms across 315 cities. Applying a formal social network methodology, this information was then converted into a 315 x 315 matrix, which enabled a formal quantitative social network analysis of the WCN. (For more details, see Taylor, 2001, 2004; Taylor et al., 2002a; Derudder and Taylor, 2005). By contrast, Alderson and Beckfield (2004) draw on Hymer (1972), Friedmann (1986), and Godfrey and Zhou (1999) to interpret world cities as command centres that host the headquarters of major MNEs. On this basis, they analyse links between 3,692 cities based on the organizational locations of 446 of the largest MNEs and their subsidiaries in 2000.

Although there are considerable differences between these two analyses, both base their city-centred spatial analysis on an assessment of the location strategies of firms that have transnational fields of activity (Taylor, 2005). It is suggested that world city relations can be derived from intrafirm communications between their constituent parts. Alderson and Beckfield (2004, pp. 813–814) consider this a “key relation” in “an MNE-generated city system”. Taylor (2004, p. 59) argues that it is “firms through their office networks that have created the overall structure of the (world city) network.” Thus, collecting data on the organizational geography of global corporate activity enables the estimation of intercity relations. Conceptually, this approach comes closest to measuring the actual formation of the WCN. However, there are problems with these intercity matrices. This is because their construction requires far-reaching assumptions on the organizational structure of firms (see, for instance, Derudder and Taylor, 2005, pp. 72–73).

The material basis for the space of flows

The empirical research carried out on the second and third layers of Castells’s space of flows has evolved greatly in recent years. However, research focusing on the first layer has spawned most empirical WCN analyses. This first layer provides the material basis for the network society; for instance, by enabling electronic networks (e.g., the Internet) and transport networks (e.g., air transport). As in Beaverstock (2004), it is assumed that that the contingent spatiality of these facilitating networks allows one to make informed inferences on the geographical tendencies within the WCN. This overall approach seems reasonable, as advanced telecommunications and transportation infrastructures are unquestionably tied to key cities in the world economy. The most important cities in the world economy also harbour the most important airports. In addition, extensive fibre-optic-based telecommunications networks that support the Internet have developed across the globe. The latter networks have predominantly been deployed within and between major cities. They have created a planetary infrastructure web on which the global economy has come to depend almost as much as it depends on physical transport networks (Rutherford et al., 2004). These enabling (tele)communications and transportation networks are the foundation on which the connectivity within the WCN is built. Therefore, it is not surprising that the geographical structures of these networks have been used to invoke a spatial imagery of the world city system. Examples include the global urban hierarchy based on air traffic flows between cities (e.g., Keeling, 1995; O’Connor, 1995; Kunzmann, 1998; Rimmer, 1998; Smith and Timberlake, 2001, 2002; Matsumoto, 2004; Witlox et al., 2004), and those based on postal flows, telephone calls, and internet linkages (Marek, 1992; Warf, 1995; Graham and Marvin, 1996).

Apart from the fact that airline data are (comparatively) easy to obtain, the usefulness of this data source is summarized by the fact that “because of its relatively rapid capacity to reply in terms of supply and demand, air traffic provides a pertinent indicator in the quest to evaluate the international character of … cities” (Cattan, 1995, p. 303). More detailed appraisals of the usefulness of airline data are elaborated in Smith and Timberlake (2001, 2002) and Keeling (1995), who present five interrelated arguments for why airline linkages are a suitable empirical source for assessing the WCN:

  1. global airline flows are one of the few indices available of transnational flows of interurban connectivity;
  2. air networks and their associated infrastructure are the most visible manifestation of world city interaction;
  3. great demand still exists for face-to-face relationships, despite the global telecommunications revolution;
  4. air transport is the preferred mode of inter-city movement for the transnational capitalist class, migrants, tourists, and high-value goods; and
  5. airline links are an important component of a city’s aspirations to world city status.

Although the potential usefulness of air transport data in spatial analyses of the world city network is obvious, it can equally be noted that some of the data sources and frameworks for analysis have not been able to live up to their inherent potential. Before we can begin to (partially) rectify this situation, we need a more elaborate overview of the various problems in previous research efforts, and this is the subject matter of the next section.

III. PREVIOUS RESEARCH BASED ON AIR PASSENGER FLOWS

The previous section discussed the potential usefulness of air transport data in a broad framework. From this overview, it is clear that airline data are by no means an exclusive starting point for measuring worldwide intercity relations. However, this data source has some obvious advantages over other data. (i) A relatively limited set of assumptions is required for its application. (ii) Airline data are relatively easy to obtain. However, some of the data sources and frameworks for analysis have not fulfilled their potential. Before we can (partially) rectify this situation, we need a more elaborate overview of the problems encountered by previous research. This is the subject of this section.

The problems associated with the use of airline statistics are sometimes obvious, but sometimes complex and subtle. In addition, many different databases and measures can be used, and they have been applied in a variety of different contexts. Hence, it does not seem worth discussing previous research by undertaking an exhaustive overview of all existing airline analyses. Rather, we outline the main obstacles preventing a clear-cut ‘translation’ of air transport databases into spatial analyses of the WCN. The major impediments in this context have been the following:

  • the lack of origin-destination data;
  • implicit state-centrism;
  • the incorporation of non-world city processes; and
  • airline data are not always provided and/or analysed in an appropriate framework.

III.A. The Lack of Origin/Destination Data

A major obstacle to the use of airline data has been the lack of origin/destination information in the databases. Most airline databases record the individual legs of trips rather than the trip as a whole. Thus, in the case of a stopover, a significant number of ‘real’ intercity links are replaced by two or more links that reflect corporate strategy rather than world city relations. Another problem is that the lack of origin/destination information makes geographically detailed assessments of the WCN difficult, as direct connections become less likely as one goes down the urban hierarchy. According to the airline database described in the next section, 28% of passengers make one or more stopovers, which reveals an important bias in previous datasets.

A case in point is Keeling’s (1995) world city map, which is based on an analysis of the dominant linkages in the global airline network. This map was created from a matrix of scheduled air services between 266 cities of more than a million people. However, only non-stop and direct flights between two cities were taken into account. This means that the measures used by Keeling are not necessarily a reflection of actual intercity relations. That is, such an analysis is likely to overstate the relational importance of cities that function as airline hubs, such as Amsterdam (KLM) and Frankfurt (Lufthansa) at the expense of cities such as Brussels and Berlin. Furthermore, direct links between, say, Brussels and Rio de Janeiro cannot be measured, as passengers are likely to go through São Paulo to make this trip. Thus, relations between second-tier world cities are difficult to measure by using a data source that only has information on individual and direct trips. Data on the origins and destinations of passengers would measure the connectivity of airline hubs and first-tier world cities more appropriately, and at the same time, it would enable the assessment of relational patterns in lower rungs of the WCN2.

III.B. Implicit State-centrism

A second obstacle to translating air transport databases into global intercity analyses arose because some of these data sources have incorporated a subtle form of state-centrism. That is, despite their global aspirations, most analyses are based on databases that contain information on international flows. This delicate bias towards interstate rather than transstate flows tends to undervalue relations between world cities that are situated in large and/or significant nation-states.

For example, Rimmer (1998, p. 460) has based his world city analysis on data on ‘international passengers’. This results in a downgrading of US world cities in particular because important connections such as Los Angeles–New York and Chicago–New York are not incorporated in this approach. Hence, Chicago only appears on one of Rimmer’s maps as a ‘fourth-level’ link to Toronto, while Dublin appears on all maps because of its ‘first-level’ link with London. Nobody would argue that Dublin is more important than Chicago as a world city; it only appears to be when one relies on international rather than global data. Another clear example of this problem is explained by Smith and Timberlake (2002, p. 123), who lacked information on the volume of air passenger traffic between Hong Kong and London. This admittedly important global link did not feature in pre-1997 databases of the International Civil Aviation Organization (ICAO) because flights between London and Hong Kong were considered national.

While the classification of the London–Hong Kong route and the downgrading of US cities are extreme examples, they clearly reveal how data on international passenger flows may hamper a global urban analysis. Smith and Timberlake (2002) were able to overcome the London–Hong Kong problem by estimating the importance of this link. The relegation of US cities was dealt with by the use of an additional data source that contained information on major routes in the US (namely, data provided by the Air Transport Association in Washington DC). While circumventing the most obvious gaps in the initial database has produced one of the most refined databases used to date, in general, this problem continues to affect the major Canadian, Chinese, and Brazilian cities (among others). A database detailing global rather than international air passenger flows would overcome the problems associated with the introduction of this subtle form of state-centrism.

III.C. Non World-city Processes

A third obstacle to the straightforward application of air transport statistics arises from the observation that such data measure general flow patterns. Since airline statistics are unable to differentiate between specific flows within the various linkages, it is doubtful that the flows that define the WCN can straightforwardly be deduced from such data. A key example is the inclusion of major tourist destinations in previous analyses. In his mapping of the European urban hierarchy based on air passenger flows, Kunzmann (1998, p. 49) lists 14 airports that are secondary to the big three (London, Paris, and Frankfurt), including Munich, Milan, Madrid, and Palma de Mallorca. However, the high ranking of the latter merely reflects its role as one of the most popular holiday destinations in Europe; nobody would argue that it is a major world city. While it is likely that most researchers would agree that destinations such as Palma de Mallorca should be omitted from the analysis, such data manipulation becomes increasingly difficult when non-world city processes intersect with world city-formation. The rising importance of Miami, for instance, can be traced back to its control functions vis-à-vis the Carribean (Grosfoguel, 1995; Nijman, 1996; Brown et al., 2002) and its function as a retirement centre and major holiday destination. Whether the latter intersection distorts the analysis depends, of course, on the exact – and therefore possibly debatable – specification of the characteristics of a world city.

In his initial formulation of ‘The World City Hypothesis’, John Friedmann (1986, p. 74) maintained that the major driving forces behind world city growth were found in a limited number of rapidly expanding sectors. Although Friedmann identified world cities as major tourist destinations, it seems that tourism is merely an ancillary function, since

“[m]ajor importance attaches to corporate headquarters, international finance, global transport and communications; and high level business services, such as advertising, accounting, insurance, and legal servcices. An ancillary function of world cities is ideological penetration and control. New York and Los Angeles, London and Paris, and to a lesser degree Tokyo are centres for the production and dissemination of information, news, entertainment and other cultural artefacts.”

Hence, although it is clear that cities such as New York, London, Los Angeles, and Tokyo have become major tourist attractions in their own right, this it is a secondary function at best, so that questions can be raised on the alleged importance of tourist flows in the various databases. We agree that trying to single out the tourist functions may be perceived by some as a questionable move, but it seems nonetheless clear that existing clear-cut tourist destinations should be omitted from the data. Irrespective of the potential controversy over this point, we maintain that airline linkages reflect myriad processes of which world city formation is only one element, so that educing the WCN from airline databases is not a straightforward matter. Admittedly, this problem may be the hardest to overcome, since we have no clear procedures for estimating the amount of world city traffic in overall air travel, while such a procedure would at the same time be open for debate3.

III.D. Airline Data are not Always Provided and/or Analysed in an Appropriate Framework

A final impediment to the use of airline data in WCN analyses is that the statistics that can be derived from them are not necessarily available and/or analysed within an appropriate framework. This general problem can be divided into three interconnecting problems. (i) Airline data are not, by definition, relational. (ii) Analyses do not always fulfil the potential offered by a genuine network-analytical framework. (iii) Assessments of area subsets require unreasonable assumptions.

Airline data are not necessarily provided in the relational form presupposed by WCN research. Cattan (1995), for instance, has determined the global importance of 90 major European cities in terms of their international exchanges. Cattan presents an assessment of the European urban hierarchy based on the computation of various attribute measures, such as the number of international flights, the rate of international travel per head of population, the percentage of international traffic in overall traffic, and the number of direct international links. While these measures provide good proxies for ranking the connectivity of cities, they do not provide information on how overall connectivity can be disentangled into spatial patterns. As a consequence, while such an analysis may convey the hierarchical trends between the major European cities in the context of the WCN, the broader geography of this part of the network remains obscure. Only data in the form of intercity matrices can unravel the spatiality behind overall connectivity. However, this requires a network-analytical framework that is able to fulfil the potential of these datasets. For example, Smith and Timberlake (2002) base their world city assessment on a time-series analysis of the network prominence of world cities. As the centrality measure used in their analysis is a fully fledged network measure, their approach is superior to that of Cattan (1995). Nevertheless, their framework for analysis does not exploit the full potential of the data.

The analysis of Shin and Timberlake (2001) represents a welcome exception to the use of attribute or general centrality measures. Using ICAO-data and three other sources, Shin and Timberlake (2001) analyse how the relative position of Asian world cities has changed over time. Their analysis focuses on 18 Asian cities. For these cities, they investigate the flight connections within and outside the region. Perhaps the most interesting feature of this paper is the structure of their quantitative analysis. As well as using overall centrality measures, they investigated the spatiality of the overall connectivity by using clique analysis4. Consequently, Shin and Timberlake (2001) were able to disentangle hierarchical, regional, and functional tendencies in the Asia-Pacific segment of the WCN.

There is, however, a subtle problem with the analyses of area subsets by Boberg and Collisson (1989), Ivy (1995), Cattan (1995), Burghouwt and Hakfoort (2001), and Shin and Timberlake (2001). That is, analyses that focus on area subsets of the WCN, such as Asia-Pacific world cities (Shin and Timberlake, 2001) or major European cities (Cattan, 1995) are problematic in the context of the WCN. The notion that there is a European or an Asian ‘system of cities’ or ‘urban hierarchy’ may initially seem an attractive idea because it appears to provide a coherent subset of cities to study within a regional context. However, one cannot contain city networks within bounded spaces. They are, by their nature, networked, and hence, analysis of them should not be truncated (Jacobs, 1984; Taylor, 2004). Thus, especially within the framework of a new global scale of production and business servicing, one can neither assert nor justify the existence of area subsets such as the ‘European city system’ within the context of the WCN. Rather, it has to be shown that traditional historical world regions still define a valid pattern of intercity relations (Taylor and Derudder, 2004). For example, while every GaWC analysis of global office networks provides evidence of strong regional patterns (Taylor et al., 2002b; Derudder et al., 2003), these are more complicated than those based on simple delineation by continent. For instance, Derudder et al. (2003, p. 883) emphasize that the connectivity of lower-ranked world cities is largely constituted by more ‘regional’ connections, but at the same time, this overall tendency is densely interwoven with other, non-regional features, such as the relatively strong connections between lower-ranked cities of the British Commonwealth. The main point is that it is somewhat problematic that, in studies of the functioning of the WCN, we can think of Asia-Pacific or European cities as representing something more than a set of cities sharing a common geographical location. Therefore, depicting the patterns of intercity relations within the Asia-Pacific and Europe is only a first step in understanding how these cities operate as world cities. Simply invoking the concept of the world city means that we must extend our vision beyond these area subsets.

IV. AN INTER-CITY MATRIX BUILT ON THE MIDT DATABASE

In the previous section, we showed how early airline-based analyses of the WCN have often been hampered by biased statistics and/or a limited framework for analysis. In this section, we describe the construction of a geographically detailed intercity matrix that is able to circumvent and/or overcome some of the problems outlined. After a brief introduction on the content and the manipulation of the initial dataset, we explain in what respect this new dataset is able to overcome the limitations highlighted in the previous section. We do not engage in a thorough analysis of this database, but mainly introduce it is as an alternative to other, more commonly used air traffic databases.

IV.A. The MIDT Database

The MIDT database contains information on bookings made through so-called Global Distribution Systems (GDS) such as Galileo, Sabre, Worldspan, Amadeus, Topas, Infini, and Abaccus. GDS are electronic platforms used by travel agencies to manage airline bookings (i.e., the selling of seats on flights offered by different airlines), hotel reservations, and car rentals. Using a GDS-based database therefore implies that bookings made directly with an airline are excluded from the system. Airlines choose this direct booking option to avoid commissions charged by travel agencies. Direct bookings via the Internet are estimated to cost an airline 1$, while bookings at travel agents cost an estimated 10$ (Goetzl, 2000). Southwest, EasyJet, Virgin, and Ryanair, which are particularly low-cost carriers, have many direct sales, and consequently, their flights do not feature prominently in GDS-based databases. However, in 1999, 80% of all reservations continued to be made through GDS (Miller, 1999). This suggests that our data source might provide a biased picture of airline transport. However, there is no reason to assume that the spatiality of the reservations made by direct bookings differs fundamentally from that for reservations made through a GDS.

With the cooperation of an airline, we were able to obtain a partial5 MIDT database that covers the period from January to August 2001 and contains information on a total of 3.7 million trips. Each MIDT record is made up of an entire airline trip, and comprises information on the IATA-airport codes of origin/destination, the air carrier, the connecting airports (if any), and the number of passengers6.

Airlines purchase the MIDT database for a variety of reasons, the most important of which is its ability to forecast demand. It is also a helpful tool for assessing the market share and the competitive position of an airline in a specific geographical area. In the context of our research, however, the database is used to construct a global intercity matrix. The first step in the creation of this matrix is to transform the information because we are mainly interested in the total volume of passenger flows between cities. To achieve this, we relabelled the airport codes as city codes. These city codes are needed to compute meaningful intercity measures because a number of cities have more than one major airport7. The particular airport used by a passenger is not important in this context because, for recording the London–New York relation, it is irrelevant whether a flight goes from Heathrow to JFK or from Gatwick to Newark. Having summed the directional information into a single measurement detailing the total volume of passengers, we created a global intercity matrix that focuses on the most important cities in the world economy. That is, we omitted key holiday destinations and less important cities. For this, we used the tentative world city list compiled by GaWC, which contains 315 cities and includes the capital cities of all but the smallest states and numerous other cities of economic importance (Taylor et al., 2002a). Nine of these 315 cities were excluded either because they had no airport (e.g., Bonn and Kawasaki) or because the airport was not serviced in the period under consideration because of political instability (e.g., Kabul). This reconfiguration produced a 306 x 306 matrix that quantifies the relations between the most important cities in the world economy. Tables 2–3 and Figure 1 present an overview of the chief connections. Table 2 presents a 10 x 10 sub-matrix from the 306 x 306 matrix and Table 3 presents the 20 most important intercity relations in the dataset. Figure 1 depicts the connections between the 30 most important cities in terms of total passenger flows. The size of the nodes varies with the total number of incoming or outgoing passengers; the size of the edges varies with the number of passengers flying between two cities. For reasons of clarity, only the most important links are shown.

Table 2: Excerpt of intercity matrix based on the MIDT database.

 CITY

Abidjan

Abu Dhabi

Accra

Addis Abeba

Adelaide

Ahmedabad

Algiers

Almaty

Amman

Amsterdam

Ankara

Antwerp

...

Abidjan

 

 

 

 

 

 

 

 

 

 

   

...

Abu Dhabi

45

 

 

 

 

 

 

 

 

   

...

Accra

15517

5

 

 

 

 

 

 

 

 

   

...

Addis Abeba

2573

173

4227

 

 

 

 

 

 

 

   

...

Adelaide

1

14

1

3

 

 

 

 

 

 

   

...

Ahmedabad

0

35

1

2

2

 

 

 

 

 

   

...

Algiers

860

103

16

18

0

0

 

 

 

 

   

...

Almaty

7

1

2

5

17

0

8

 

 

 

   

...

Amman

31

39562

58

126

28

2

1103

96

 

 

   

...

Amsterdam

4784

9096

26289

2324

1952

414

664

6520

18059

 

   

...

Ankara

0

30

43

12

18

0

283

1647

4152

8562

 

...

Antwerp

79

233

921

18

0

2

0

0

57

8307

2

 

...

...

...

...

...

...

...

...

...

...

...

...

...

...

 

Table 3: Most important inter-city links in the MIDT-database.

 

Rank

Connection

Number of passengers

1.

Hong Kong

Taipei

2138484

2.

Los Angeles

New York

1697593

3.

London

New York

1609337

4.

Melbourne

Sydney

1563106

5.

Milan

Rome

1518767

6.

Cape Town

Johannesburg

1406897

7.

Los Angeles

San Francisco

1375660

8.

Amsterdam

London

1242550

9.

Chicago

New York

1182326

10.

Bangkok

Hong Kong

1141062

11.

London

Paris

1060999

12.

Dublin

London

1050940

13.

Marseilles

Paris

1044128

14.

Bangkok

Singapore

1024818

15.

Rio De Janeiro

Sao Paulo

992775

16.

Boston

New York

988976

17.

Miami

New York

955838

18.

Atlanta

New York

935265

19.

Las Vegas

Los Angeles

924732

20.

New York

San Francisco

909514

Figure 1: Most important cities and links in the world city network.

IV.B. Appraisal of the MIDT Database

With the problems discussed in the previous section used as a checklist, the MIDT-based intercity matrix may be able to overcome a number of obstacles. To illustrate a number of differences, Table 4 presents a comparison of the 20 most connected cities in the airline network as identified (i) in Smith and Timberlake (2001, 2002) and (ii) the MIDT-database. It is important to stress that this comparison is only made for illustrative purposes. The list of ‘most important cities’ in Smith and Timberlake (2001, 2002) is constructed on the basis of a fully fledged centrality analysis, while the MIDT-list merely reflects a ranking based on the total number of passengers boarding on/off. Furthermore, the data in Smith and Timberlake (2001, 2002) refer to passenger flows in the year 1997, while the MIDT data feature statistics for the year 2001. As a consequence, the reader should bear in mind that we merely compare both lists to point to some data-induced differences.

First, as the MIDT-database contains origin/destination information, the overrating of the connectivity of airline hubs and first-tier world cities is minimized, which allows assessing the relational patterns in the lower rungs of the WCN in more detail (e.g. the downsizing of the importance of hub cities such as Amsterdam and Frankfurt in Table 4). Second, the MIDT-based database does not distinguish between national and international flows, and can therefore be thought of as a truly global intercity matrix. The New York–Chicago link is appropriately treated in the same way as the New York–Toronto link, which further reduces the underestimation of second-tier cities in large and/or significant nation-states (e.g. there are 9 US cities in the top-20 for the MIDT-data, while only 5 US cities feature in Smith and Timberlake (2001, 2002), this appearance of US cities is mainly at the expense of Southeast Asian cities). Third, reconfiguring the database by using GaWC’s detailed world city list excludes obvious holiday destinations, which results in a (modest) redirection of plain air flow-centred information to city-centred information. Fourth, this database contains relational information in a single global framework. This allows overall connectivities to be disentangled into spatial patterns, while analyses of area subsets can be carried out in the context of an overarching WCN, as suggested by Taylor and Derudder (2004)8 .

Table 4: Comparison of most important cities in Smith and Timberlake (2001, 2002) and the MIDT-database.

Rank

 

Smith and Timberlake (2001, 2002)

Rank

MIDT-database

1

London

1

New York ↑

2

Frankfurt

2

London ↓

3

Paris

3

Los Angeles ↑

4

New York

4

Paris ↓

5

Amsterdam

5

Chicago ↑

6

Miami

6

San Francisco ↑

7

Zurich

7

Hong Kong ↑

8

Los Angeles

8

Miami ↓

9

Hong Kong

9

Frankfurt ↓

10

Singapore

10

Atlanta ↑

11

Tokyo

11

Madrid ↑

12

Seoul

12

Toronto ↑

13

Bangkok

13

Washington ↑

14

Madrid

14

Rome ↑

15

Vienna

15

Bangkok ↓

16

San Francisco

16

Milan ↑

17

Chicago

17

Amsterdam ↓

18

Dubai

18

Dallas ↑

19

Osaka

19

Boston ↑

20

Brussels

20

Singapore ↓

Although this MIDT-based intercity matrix is able to overcome and/or circumvent some of the problems that have been associated with the use of airline data in WCN analyses, a number of problems remain for future research. The main problem is that it remains largely impossible to discern world city flows from other flows. The importance of the New York–Miami route and particularly the New York–Las Vegas route, for instance, suggests more than a business link (see Table 5). The linkages related to obvious holiday destinations such as Palma de Mallorca have been deleted from the database, but this manipulation only works for airports that are obviously not world-city airports. Furthermore, although problems with the availability of global origin/destination data addresses the undervaluation of second-tier world cities, airline data cannot avoid undervaluing a second-tier city that is close to a major world city. For example, a passenger travelling from Rotterdam to New York is likely to depart from Amsterdam because of (i) the short distance between Rotterdam and Amsterdam (less than 50 miles) and (ii) the importance of Amsterdam’s Schiphol airport. This bias is exacerbated by the overall tendency to underestimate the connectivity between nearby cities because of the availability of other modes of transportation. Elaborate high-speed rail networks, for instance, may be an alternative to short-haul flights, which results in an underestimation of the importance of some short-distance intercity links. An overall solution to this underestimation problem may be to omit some cities from the database. Although this implies a further reduction in the geographical detail of the database, ensuing analyses would be more meaningful and robust9. A final drawback of the dataset described in this paper is that because it only covers a single period, it cannot be used to analyse the evolution of the WCN. Thus, unlike Smith and Timberlake (2002), we cannot track changes over time.

Table 5: Most important North American inter-city links in the MIDT-database.

 

Rank

Connection

Number of passengers

1.

Los Angeles

New York

1697593

2.

Los Angeles

San Francisco

1375660

3.

Chicago

New York

1182326

4.

Boston

New York

988976

5.

Miami

New York

955838

6.

Atlanta

New York

935265

7.

Las Vegas

Los Angeles

924732

8.

New York

San Francisco

909514

9.

Las Vegas

New York

856221

10.

New York

Washington

806785

V. CONCLUSION

In this paper, we argued that air transport data are potentially a prime data source for assessing spatial patterns in the world city network (WCN). We also argued that previous WCN analyses have been hampered by inadequate and/or partial data. The ensuing analytical deficiencies have reduced the overall value of these WCN analyses. Hence, in this paper, we examined how some of these deficiencies might be overcome. We showed how a new intercity matrix complied on the basis of the Marketing Information Data Transfer (MIDT) database is able to overcome a number of these problems and leave only a few minor problems unresolved. In the next phase of our research, we intend to analyse this database along the lines suggested by Shin and Timberlake (2001). This involves a two-step analysis of this 306 x 306 matrix. First, a general analysis of the overall centrality of key cities is to be undertaken. Then, a detailed geographical analysis of these connectivities is to be made by using a number of standard network analytical tools, such as clique and block analyses. In future papers, we hope to report on these findings.

ACKNOWLEDGEMENTS

We would like to thank the editor and two anonymous reviewers for their helpful comments on an earlier version of this paper. Thanks also to Laetitia Vereecken and Anne Bunneghem for part of the MIDT data handling. This research work is funded by the Research Foundation - Flanders, Research Project G.0214.04.

REFERENCES

ALDERSON, A.S. and BECKFIELD, J. (2004) Power and position in the world city system, American Journal of Sociology, 109, pp. 811–851.

ALLEN, J. (1999) Cities of power and influence: settled formations. In: J. Allen, D. Massey and M. Pryke (Eds.), Unsettling Cities, pp. 181–228. London: Routledge.

Association of European Airlines (2002) Yearbook 2002, Brussels: Association of European Airlines.

BEAVERSTOCK, J.V. (2004) Managing across borders: knowledge management and expatriation in professional service legal firms, Journal of Economic Geography, 4(2), pp. 157–179.

BEAVERSTOCK, J.V, HUBBARD, P.J. and SHORT, J.R. (2004) Getting away with it? Exposing the geographies of the super-rich, Geoforum, 35(4), pp. 401–407.

BEAVERSTOCK, J.V., SMITH, R.G. and Taylor, P.J. (2000) World city network: a new metageography?, Annals of the Association of American Geographers, 90(1), pp. 123–134.

BOBERG, K.B. and COLLISON F.M. (1989) International air transportation trends in the Pacific Basin, Transportation Journal, 3, pp. 24–35.

BROWN, E., CATALANO, G. and TAYLOR, P.J. (2002) Beyond world cities: Central America in a space of flows, Area, 34, pp. 139–148.

BURGHOUWT, G. and HAKFOORT, J. (2001) The evolution of the European aviation network: 1990–1998, Journal of Air Transport Management, 7, pp. 311–318.

CASTELLS, M. (1996) The Rise of the Network Society. Oxford: Blackwell.

CATTAN, N. (1995) Attractivity and internationalisation of major European cities: the example of air traffic, Urban Studies, 32, pp. 303–312.

DERUDDER, B., TAYLOR, P.J., WITLOX, F. and CATALANO, G. (2003) Hierarchical tendencies and regional patterns in the world city network: a global urban analysis of 234 cities, Regional Studies, 37, pp. 875–886.

DERUDDER, B. and TAYLOR, P.J. (2005) The cliquishness of world cities, Global Networks, 5(1), pp. 69–89.

FRIEDMANN, J. (1986) The world city hypothesis, Development and Change, 17, pp. 69–83.

FRIEDMANN, J. and WOLFF, G. (1982) World city formation: an agenda for research and action, International Journal of Urban and Regional Research, 3, pp. 309–344.

GODFREY, B.J. and ZHOU, Y. (1999) Ranking world cities: multinational corporations and the global urban hierarchy, Urban Geography, 20, pp. 268–281.

GOETZL, D. (2000) Southwest Airlines takes flight fare fight to the web, Advertising Age, 71(40), pp. 20–22.

GRAHAM, S. and MARVIN, S. (1996) Telecommunication and the City. London: Routledge.

GROSFOGUEL, R. (1995) Global logics in the Caribbean city system: the case of Miami, In: P.L. KNOX and P.J. TAYLOR (Eds.) World Cities in a World-System, pp. 156-170. Cambridge: Cambridge University Press.

HYMER, S. (1972) The multinational corporation and the law of uneven development, In: J. Bhagwati (Ed.) Economics and World Order from the 1970s to the 1990s, pp. 113–140. London: Collier-MacMillan.

IVY, R.L. (1995) The restructuring of air transport linkages in the new Europe, Professional Geographer, 47, pp. 280–290.

JACOBS, J. (1984) Cities and the Wealth of Nations. New York: Vintage.

KEELING, D.J. (1995) Transport and the world city paradigm, In: P.L. KNOX and P.J. TAYLOR (Eds.) World Cities in a World-System, pp. 115–131. Cambridge: Cambridge University Press.

KUNZMANN, K.R. (1998) World city regions in Europe: structural change and future challenges, In: F-C. LO and Y-M. YEUNG (Eds.) Globalization and the World of Large Cities, pp. 37–75. Tokyo: United Nations University Press.

MAREK, S. (1992) Private networks directory, Satellite Communications, September issue.

MATSUMOTO, H. (2004) International urban systems and air passenger and cargo flows: some calculations, Journal of Air Transport Management, 10, pp. 241–249.

MILLER, W.H. (1999) Airlines take to the internet, Industry Week, 248(15), pp. 130–134.

MOSS, M.L. and TOWNSEND, A.M. (2000) The internet backbone and the American metropolis, The Information Society, 16, pp. 35–47.

NIJMAN, J. (1996) Breaking the rules: Miami in the urban hierarchy, Urban Geography, 17, pp. 5-22.

O’CONNOR, K. (2003) Global air travel: toward concentration or dispersal? Journal of Transport Geography, 11, pp. 83–92.

PERKINS, S. J. (1997) Internationalization: the People Business, London: Kogan Page.

RIMMER, P.J. (1998) Transport and telecommunications among world cities, In: F-C. LO and Y-M. YEUNG (Eds.) Globalization and the World of Large Cities, pp. 433–470. Tokyo: United Nations University Press.

SAMERS, M. (2002) Immigration and the global city hypothesis: towards an alternative research agenda, International Journal of Urban and Regional Research, 26, pp. 389–403.

SASSEN, S. (1991) The Global City: New York, London, Tokyo. Princeton: Princeton University Press.

SASSEN, S. (2000) Cities in a World Economy. Thousand Oaks: Pine Forge.

SASSEN, S. (2001) The Global City: New York, London, Tokyo. (2nd edition). Princeton: Princeton University Press.

SASSEN, S. (2002) Global Networks, Linked Cities (Ed.). London: Routledge

SCOTT, A.J. (2001) Global City-Regions: Trends, Theory, Policy (Ed.). Oxford: Oxford University Press.

SCOTT, J. (1991) Social Network Analysis. London: Sage.

SHIN, K.H. and TIMBERLAKE, M. (2000) World cities in Asia: cliques, centrality and connectedness, Urban Studies, 37, pp. 2257–2285.

SHORT, J. R., KIM, Y., KUSS, M., and WELLS, H. (1996) The dirty little secret of world city research, International Journal of Regional and Urban Research, 20, pp. 697–717.

SMITH, D.A. and TIMBERLAKE, M. (1995a) Conceptualising and mapping the structure of the world system’s city system, Urban Studies, 32, pp. 287–302.

SMITH, D.A. and TIMBERLAKE, M. (1995b) Cities in global matrices: toward mapping the world-system’s city system, In: P.L. KNOX and P.J. TAYLOR (Eds.) World Cities in a World-System, pp. 79–97. Cambridge: Cambridge University Press.

SMITH, D.A. and TIMBERLAKE, M. (2001) World city networks and hierarchies 1979–1999: an empirical analysis of global air travel links, American Behavioral Scientist, 44, pp. 1656–1677.

SMITH, D.A. and TIMBERLAKE, M. (2002) Hierarchies of dominance among world cities: a network approach. In: S. SASSEN (Ed.) Global Networks, Linked Cities, pp. 117–141. London: Routledge.

SMITH, N. (2002) New globalism, new urbanism: gentrification as global urban strategy, Antipode, 34, pp. 434–457.

SMITH, R.G. (2003a) World city actor-networks, Progress in Human Geography, 27, pp. 25–44.

SMITH, R.G. (2003b) World city topologies, Progress in Human Geography, 27, pp. 561–582.

TAYLOR, P.J. (1996) Embedded statism and the social sciences: opening up to new spaces, Environment and Planning A, 28, pp. 1917–1928.

TAYLOR, P.J. (1997) Hierarchical tendencies amongst world cities: a global research proposal, Cities, 14, pp. 323–332.

TAYLOR, P.J (1999) So-called “World Cities”: the evidential structure within a literature, Environment and Planning A, 31, pp. 1901–1904.

TAYLOR, P.J. (2001) Specification of the world city network, Geographical Analysis, 33, pp. 181–194.

TAYLOR, P.J. (2004) World City Network: A Global Urban Analysis. London and New York: Routledge.

TAYLOR, P.J. (2005) Parallel paths to understanding global inter-city relations, American Journal of Sociology, forthcoming.

TAYLOR, P.J., CATALANO, G. and WALKER, D.R.F. (2002a) Measurement of the world city network, Urban Studies, 39, pp. 2367–2376.

TAYLOR, P.J., CATALANO, G. and WALKER, D.R.F. (2002b) Exploratory analysis of the world city network, Urban Studies, 39, pp. 2377–2394.

TAYLOR, P.J. and DERUDDER, B. (2004) Porous Europe: European cities in global urban arenas, Tijdschrift voor Economische en Sociale Geografie, 95, pp. 527–538.

WARF, B. (1995) Telecommunications and the changing geographies of knowledge transmission in the late 20th century, Urban Studies, 32, pp. 361–378.

WITLOX, F., VEREECKEN, L. and DERUDDER, B. (2004) Mapping the global network economy on the basis of air transport passenger flows, GaWC Research Bulletin No. 157, available on-line at http://www.lboro.ac.uk/gawc/rb/rb157.html.

 


NOTES

* Derudder B. & Witlox F. ben.derudder@ugent.be and frank.witlox@ugent.be Geography Department, Ghent University, Krijgslaan 281/S8, B9000 Gent, Belgium

1 Castells’s (1996, p. 379) own empirical treatment of world cities is quite conventional. For instance, he identifies Sassen’s (1991/2001) triad (London, New York, Tokyo) as the top of a global urban hierarchy.

2 The gist of this paragraph is that data on the actual origin and destination of flights have some clear advantages over information on individual flights because this latter approach artificially inflates the importance of key airports. However, it can equally be argued that this hub function is a part of how cities become world cities. Miami’s role as a hinge between Anglo-America and Latin America (Brown et al., 2002), for instance, is being reflected in its role as the main airline hub between the two regions. A plain origin/destination dataset will depreciate this essential function, but taken together, we believe that information on actual origins and destinations is able to present a fuller grasp of the spatiality of the WCN.

3 One empirical solution may lie in the use of air transport statistics that differentiate between business-class and economy-class travel. However, given the profitability of business-class travel in a highly competitive market, it is no surprise that airlines are not willing to make such data publicly available.

4 Cliques are cohesive subgroups within a network (Scott, 1991), which implies that world cities that feature in the same clique are cohesively related to each other and weakly related to cities outside the network. Nodes with a comparable level of connectivity can be found in different cliques on the basis of having distinct relational patterns, and the spatiality of a world city’s relations can therefore be captured by the particular nature of its cliquishness.

5 The database is referred to as ‘partial’ because it only covers a limited time span, and does not contain any information regarding hotel bookings or car rentals. Also missing are data relating to flight numbers, class of service, booking date, departure date, agency name, subscriber country, and cancellation indicators. These statistics are however not important for the present analysis.

6 Working with data covering the period before September 2001 has the advantage that it does not include the impact on passenger air transport of the terrorist attacks of 11 September 2001. In the months following this date, many destinations were not serviced as numerous connections were (temporarily) removed, and some carriers disappeared from the market. The Association of European Airlines (AEA) announced an intra-European traffic decrease of 11.6% during the last four months of 2001, compared with the same period in 2000. On the North Atlantic routes, air traffic dropped severely in November 2001, with a 30–35% fall from the previous year. Thereafter, the falls diminished, and had stabilized at 20% by the end of the year. On the Far Eastern routes, the number of passengers decreased to more than 20% in October and November 2001. However, by March 2002, traffic had almost returned to the previous year’s levels (AEA, 2002). Therefore, none of these traffic changes bias our data.

7 In reality, airport codes do not exclusively include airports, but also encompass train and bus stations. Many major airports are indeed directly connected with rail services that provide an alternative to short-haul flights. For example, in addition to the air link between Brussels and Paris Charles de Gaulle, Air France provides the Thalys High Speed train from Brussels South train station. This means that it is possible to ‘board’ at the Brussels station, which has thereby obtained an IATA airport code. However, this implies that because traffic emanating from Brussels’ international railway station has been added to its air transport links, this alternative way of travelling does not distort our analysis (http://www.airfrance.be, last accessed 16/8/2004).

8 These last two advantages over some of the previous analyses do not pertain to the work by Smith and Timberlake (2001, 2002): their database equally (i) excludes holiday destinations and (ii) enables the analysis of cities in a single, relational framework. A more elaborate comparison of the MIDT database with the statistics presented in Smith and Timberlake (2001, 2002) should in any case be based on the results of a detailed network analysis rather than a mere comparison of passenger flows. Such a detailed comparison may very well prove to be useful for estimating some of the data biases addressed in this paper. Such estimations could then, for instance, be used to apply corrections to some of the less-than-ideal databases.

9 This equally implies that concepts measured by using airline data are closer to Scott’s (2001) global city-regions than to Sassen’s (2001) global cities or Friedmann’s (1986) world cities.

 


Edited and posted on the web on 13th October 2004; last update 21st September 2005


Note: This Research Bulletin has been published in Urban Studies, 42 (13), (2005), 2371-2388