This Research Bulletin has been published in Urban Studies, 48 (13), (2011), 2749-2769.
Please refer to the published version when quoting the paper.
Within the literature on world cities and urban networks, considerable empirical attention has been focused on the intra-firm networks of advanced producer services that connect different cities with each other, so enabling production of goods and services on a global scale (Sassen, 2002; Taylor, 2004). The different types of advanced producer services (APS), e.g. finance, insurance, consultancy, are treated as a distinctive sector that serves global production functions and that tends to agglomerate within ‘world cities’ or ‘global cities’. However what is largely ignored, both conceptually and empirically, is the fact these types of firms provide advanced services to other firms that operate within entirely different sectors (cf. Jacobs et al 2010). So, while Sassen (2010) reaffirms that these APS provide ‘organizational commodities’ to the support the command and control functions of their global operating clients whatever the sector in which those clients operate, the possibility that such interactions can lead to sector-specific specialization amongst APS may be under-appreciated. This is problematic to the extent that the ‘sector’ in which global flows of finance and information are organized is somehow regarded as monolithic and hence separate from the resource, manufacturing or consumer services sectors of the ‘real’ economy that it serves. In short, the underlying depiction of the advanced services sector as relatively independent from the economic sectors it serves bears closer examination.
The central question of this paper then becomes to what degree does sector-specific specialization influence the location pattern of advanced producer services and the spatial configuration of inter-city networks? We propose:
H1: Sector specificity of advanced producer services influences the location pattern and the spatial configuration of urban networks.Exploring this hypothesis raises difficult methodological questions about the conceptualization, definition and measurement of the advanced services sector; how narrowly or broadly should the boundaries of the advanced services sector – and specialized portions within it - be drawn? In this study we will contribute to the debate by adding such sector specificity by focusing on those APS involved in the maritime transport sector (Jacobs et al, 2010). The maritime transport sector is an intriguing site for this research because of ongoing debates about the relationship between port activity and the economies of port cities (Grobar 2008; Hall 2009). Although the maritime transport sector is by definition highly mobile, its core activities (transhipment, stevedoring, warehousing, logistics) and the global flows of commodities it facilitate concentrate within seaport nodes located in major metropolitan regions. Maritime APS (hereafter AMPS) do not necessarily share the same need for seaport infrastructure as transportation firms, but location near nodes of transportation activity might be necessary and beneficial to sustain business relations and monitor market demands. Furthermore, to the extent that global command and control functions of the global shipping and port related industry agglomerate in specific places, AMPS may locate in proximity to these localization economies. The empirical question then becomes to what extent these specialized maritime advanced producer services agglomerate nearby major transport nodes such as seaports, and/or in proximity to global shipping and port related localization economies? We propose:
H2.1: Maritime advanced producer services agglomerate nearby seaports and transhipment nodes of commodity flows.
On the other hand, it might be the case that while specialized advanced maritime producer services (AMPS) emerged and historically evolved in the direct spatial proximity of seaports, their direct spatial proximity ceased to be important over time. The spatial and functional-economic relationships between cities and ports have changed fundamentally during the second-half of the twentieth century (see Bird, 1963; Hoyle, 1989; Hall, 2007). Spatially, the increased intensity of port-industrial activity, in combination with urban growth, lack of available land for further expansion, and environmental constraints have led to the relocation of port facilities away from city centers. In economic terms ports have become less and less dependent on the urban labor market due to increased automation and operational rationalization. Cities have also become less dependent on ‘their’ ports for local economic growth, but often struggle - like many industrial centers in the developed world - to upgrade and diversify their economies. There is however recognition that port-city relationships still exist through more sophisticated - albeit less visible - forms within the tertiary sector (O’Connor 1989; Slack 1989; Ducruet & Lee, 2006) in the form of specialized advanced producer services. Some port cities have managed to economically diversify into thriving service based economies even though their initial advantage of deepwater access ceased to be important for growth (Fujita & Mori, 1996). Indeed, the world/global city literature emphasizes the agglomeration effects of advanced service providers. The question then becomes whether specialized maritime advanced service providers (AMPS) favour a location in proximity of other advanced services in world cities. We propose:
H3: Specialized maritime advanced service providers agglomerate nearby other advanced service providers.
Note that H2 and H3 are not alternative hypotheses; it is possible that AMPS locate in proximity to both seaports or maritime localization economies, and other advanced services. For this reason we maintain each as a separate hypothesis in the analysis. The structure of this paper is as follows. In the second section we deal with the issue of sector-specificity of advanced producer services in world city networks. We ask to what extent sector-specific specialization to the maritime transport sector affect their location pattern and network pattern. In the third and the fourth section we address our methodology and results. The conclusion considers the implications of the analysis for urban economic development.
Sector specificity in world city networks: maritime transport sectorWithin world city research two different units of analysis are used. Some studies focus on the location pattern and network configuration of the largest multinationals on the planet, irrespective of the industry they represent (cf. Alderson & Beckfield, 2004; Wall 2009). The other approach focuses on the corporate networks of the largest advanced producer services in line with Sassen’s (2002) work on global cities, often making use of the database constructed by the ‘Globalisation and World City’ (GaWC) research network (Taylor 2004). The services that are considered as advanced and that are included in the analysis vary, but mostly they include finance, insurance, accountancy, law and advertisement. It is argued that it is the very nature of their business, namely to provide and control knowledge and capital intensive inputs for global producers, which makes these services the very ‘heart and soul’ of the capitalist world economy and that their geographical concentration at certain locations constitute ‘world cityness’ (Taylor 2004; Sassen, 2002; Friedmann 1986).
However, what is largely ignored in most studies on world cities (but see: Lambregts, 2008), is that these advanced services are provided to other firms that are often in entirely different sectors. This omission can be explained by the fact that most empirical studies focus on the geography of intra-firm networks at the expense of inter-firm relationships (Jacobs et al 2010; Lüthi et al 2010). The lack of attention to inter-firm relationships in the analysis of advanced service provision obscures two fundamental and related issues. First, by focusing on global hierarchical relationships in the organization of finance and data flows, the potential for sector-specific specialization of advanced producer services and hence the potential localization of these services away from the iconic world cities is obscured. Advanced services may be sector-specific as they are provided to other industries with a specific and distinguished demand for such services. For example, risk management of a fleet of oil tankers is something different than that of a fleet of lease cars. Not every advanced service provider holds the same sector-specific expertise and not every advanced service provider is competing in the same markets or for the same clients. Hence, taking into account a certain sector-specificity may provide a more accurate geographical picture of actual competing advanced producer services than is provided by conventional world city research.
Second, related to the issue of sector-specificity is geographical proximity. In urban and regional economics it is argued that firms benefit from being co-located or clustered in space since this allows the capture of positive economic externalities. Localization economies refer to external economies that are available to all local firms within the same sector or industry, whereas urbanization economies refers to externalities available to all local firms irrespective of sector and which arise as a result of urban size, population density and the location of certain facilities from the public sector such as universities or government administration. A special, dynamic, form of urbanization economies highlighted by Jane Jacobs (1969) refers to external economies available to all local firms in a city or region stemming from a variety or diversity of sectors. Continuous market and non-market (cf Storper 1997) interaction between firms from different sectors within geographical proximity can over time lead to specialization and a ‘related variety’ (Frenken et al 2007). From this argument it follows that specialized advanced producer services may be co-located in close proximity to the industry to which they provide these services, as well as in places with diverse and vibrant urbanization economies.
In this study we focus on the maritime transport sector for a number of reasons. First, the maritime transport sector is a major facilitator of the process of economic globalization (Levinson, 2006) as it links distant production clusters and consumer markets. Second, there has been neglect of physical flows within World City Network research (Derudder, 2006), which has tend to focus on intra-firm networks, air-links (Witlox & Derudder, 2005) and internet connections (Choi et al, 2007) instead. Within the maritime transport sector there is demand for advanced services: to finance ships and port facilities, to insure ships and its cargo, to have a legal representation in case of an incident, to have software solutions in supply chain management, to inspect ships and to provide technical expertise on damages. The geography of those specialized and sector-specific advanced producer services have not received much attention (but see Jacobs et al 2010; Hall et al 2010).
What factors might explain the location and network configuration of AMPS? Based upon on exploratory interviews with senior managers of AMPS firms (including maritime law firms, surveyors, insurance brokers, P&I insurance companies and correspondents, Lloyd’s agents, banks as well with ship owners), we can distinguish a number of factors. Maritime APS functions have differential requirements for geographical proximity to seaports and commodity flows, global shipping and port related industry, and to other APS. For some AMPS direct proximity to seaports and physical goods movement is important: in order to inspect or classify ships when in port, to inspect damage of ship or cargo, to legally represent the ship or cargo owners in case of damage etc. Although these activities imply higher educated professions, their activity tends to be routine and in real-time demand.
On the other hand we have the possibility of geographical proximity to other APS in which direct proximity to seaports and commodity flows is less relevant. An example is the cluster surrounding Lloyd’s market and the Baltic exchange in the City of London where the representatives of ship and cargo owners buy and share risk products from underwriters, fixed premium insurance companies and P&I clubs. These are highly capital intensive products in which proximity to financial services and specialized human capital becomes crucial. Agglomeration based upon this mechanism tends to apply to the global corporate decision-making offices of the APS firms. Here we can expect a considerable overlap with the conventional world city hierarchies (Taylor 2004; Taylor et al 2009). As the director of a major London-based marine insurance company expressed in an interview when asked about the importance of proximity to advanced services:
“This is the most important issue. In London, we have the Lloyd’s and Companies market close-by, the London broking community is very important to our business, there are of course a lot of capital providers like banks in the City and we are close to maritime lawyers, surveyors and other experts. London is currently a very expensive location to have office, but all this proximity to expert services more than compensates for the additional costs”A third form of geographical proximity is with the customers of AMPS, the ship owners and other port related industries. A high concentration of ship owners at a particular location might attract offices of maritime advanced services, which is for exemplified by the relatively large concentration of London-based AMPS in the Piraeus-Athens urban region. Likewise, large ship owners might be inclined to open offices nearby centres of AMPS to facilitate the finance and insurance of their fleet. As expressed in an interview by a senior manager of a London-based marine insurance club:
“The high number of Greek ship owners in the UK P&I Club has indeed been a reason to set up an office in Piraeus. Basically, they demanded that. But we have historically been also doing a lot of Greek business from London. Greek ship owners have historically been strongly represented in London, exactly because of the services here. It’s a bit of a chicken and an egg story”
Institutional and historical factors may also influence the location and constellation of networks of maritime advanced services. The dominant position of London for example can be traced back to the heyday of the British Empire and the port of London in the late nineteenth century (see also Jacobs et al 2010). Although the empire disintegrated and the port declined, the city remained the international center of maritime advanced producer services up till this date. This history also helps explain the strong positions of former British crown colonies of Singapore and Hong Kong and their strong business relationships with London (Jacobs et al 2010). The relatively strong position of Scandinavian cities within maritime advanced services (e.g. Det Norkse Veritas) can also be attributed to a long history of seafaring in which these cities developed business expertise in terms of insurance and surveying early in the nineteenth century (see Johnstad, 2000). Such path dependent institutional evolution is also responsible for the fact that BIMCO contracts (Baltic and International Maritime Council) between ship owners, cargo owners and third parties are based upon English maritime law, with arbitration and hearings more likely to take place in either London or New York (under American maritime law). This provides these places with a considerable ‘jurisdictional advantage’ (Feldman & Martin, 2005) over all other places to attract and develop expertise.
Mapping the location pattern and network formation of advanced maritime producer services
Database Construction and Specification of the Network
Network analysis of nodes and links show us patterns of connectedness that can reveal how the global network of maritime advanced producer services is shaped. Network analysis is widely used in the social and behavioural sciences and is often done on a micro scale. For example, individuals are regarded nodes and their social relations constitute links. In this sense the world city network is rather unusual since the network is analyzed on a world scale. Moreover, the world city network has three levels (Taylor 2001). The first level is the world economy, where the network operates. The second level consists of cities (nodes) where the production of services takes place. The sub-nodal level is constituted by the advanced producer firms who produce the services. Although cities do have decision makers who can influence world city formation (Beaverstock et al. 2002), the prime actors are the firms on the sub-nodal level. Firms have the opportunity to relocate, to establish new relations and make the decisions.
When Taylor et al. (2002) measured the world city network they selected 100 firms from different advanced producer service sub-sectors2 that pursue a ‘global locational strategy’, i.e. firms that have offices in more than 15 cities and at least one office in each prime ‘globalistion arena’ (North America, Western Europe and Pacific Asia). In our measurement of the world maritime network of advanced producer services we do not exclude small firms. The shipping industry is itself highly international, linking distant centres of supply and demand through the transport of goods. This international nature of shipping implies that advanced producer services also need to have an international orientation when providing legal, financial and insurance products. However, advanced producer services for the maritime transport sector constitute a specialized niche, often confined to a certain region. For example, legal conflicts about cargo damage at a certain port may fall within a particular national or state jurisdiction. Hence, we do not want to make a priori assumptions about what are ‘globalisation arenas’ and what defines peripheral areas.
Results of the Locations of Maritime Advanced Producer Services
In figure 1 we present the spatial representation of the locations of the establishments of maritime advanced producer firms. It is not surprising that London ranks first with 385 AMPS firms. Singapore follows with 199 establishments. In the top ten there are a number of cities that are not part of the rankings of different world city studies. Examples are Piraeus (148 establishments), Rotterdam (128 establishments), Hamburg (104 establishments), Houston (96 establishments) and Panama City (95 establishments). All these cities are port cities and accommodate port related industries. For example, many Greek ship-owners base their commercial operations in Piraeus. In Rotterdam and Houston a huge number of port-related industries are located, such as large oil-refining and chemical firms. Tokyo is only ranked 12th and is certainly not one of the most important global maritime cities. However, it is joined by a number of capital cities that do not host a seaport but nevertheless do contain a substantial number of establishments of AMPS. Examples are Madrid, Paris and Moscow which rank respectively 15th, 16th, and 19th.
Figure 1: The locations of AMPS.
Results of the Network Analysis of Maritime Advanced Producer Services
To get a better picture of the constitution of the network we also map the links between cities and the GNC of cities. In figure 2 we can see the results of the network analysis. We only show the links which are more than 5% of the strength of the most important link, Hong Kong – London. Furthermore, we represent only cities which have a GNC of at least 1% of that of London.
Figure 2: The network of AMPS.
This analysis provides some first results which are in line with the findings of Jacobs et al (2010). First, we observe that cities that host many AMPS firms are in a number of cases not part of the traditional top ten of world cities. Second, looking at the constellation of the network, the most important relations are between London, Hong Kong and Singapore. Third, it can be observed that London is very dominant, in terms of connectivity, and number of establishments. With respect to H1, we conclude that sector specificity exerts a noticeable influence both on the locational pattern of establishments and on intra-firm network formation when compared with conventional world city research which only looks at the largest APS-firms regardless of sectoral specialization. However we need to go further to explain what factors influence the location and connectivity of specialized advanced producer services.
Factors explaining location pattern and network constitution of specialized advanced maritime service providers
In the section we investigate which factors might explain why some cities host more AMPS firms than others. Furthermore, we explore the impact of these factors on GNC.
In the absence of panel data, our analysis proceeds with cross-sectional count data on the number of AMPS and general APS, and their respective GNC. Oftentimes, count data is treated as continuous, because then the use of Ordinary Least Squares methods is feasible. However, this may lead to biased and inefficient estimates (Long 1997). The Poisson regression technique is widely used for count data; it is suitable when the dependent variable is a non-negative integer with a Poisson-distribution. However, an important assumption of Poisson regression is equi-dispersion (Cameron and Trivedi 1998). In our sample it appears that all the dependent variables have a significantly higher variance than the mean, in other words, there is over-dispersion. A regression-technique that is more appropriate then, is the negative binomial regression method which allows for over-dispersion. We also determined that we do not have a problem of excess zeros (Long 1997), and hence there is no need for a zero-inflated Negative Binomial regression.
Since it is nearly impossible to collect complete data for all the cities in which maritime advanced producer services are located, we end up with a database of 459 cities representing a balanced sample of both port cities and cities which do not have a seaport. Table 1 contains the descriptive statistics of our sample; details on data sources are provided in Appendix A.
Table 1: Descriptive Statistics
The dependent variables are the location and network connectivity of specialized maritime and general advanced producer services (ESTABLISHMENTS AMPS, GNC AMPS, ESTABLISHMENTS APS, GNC APS). The independent variables are organized into three categories respectively corresponding to H2.1, H2.2 and H3: port-specific variables (CONTAINERS, COASTAL CITY, ISLAND), localization variables (SHIP-OWNERS, PORT-INDUSTRIES), and urbanization variables (POPULATION, GDP per CAPITA, GOVERNANCE, CAPITAL and UNIVERSITIES). While port-specific variables are closely related to the localization variables, they describe physical movement characteristics rather than economic relationships.
For all these independent variables we test whether they influence the location of AMPS and the GNC of AMPS (cf. model 1 for establishments; model 5 for GNC). In order to check whether the influence of the independent variables on AMPS differ substantially from general APS, we also run models with establishments and GNC of APS as dependent variables (cf. models 2 and 6). Since it has been shown in the previous section that some locations that do host a lot of AMPS but do not have a seaport (e.g. Paris, Madrid), we reduce our sample to those 275 cities that actually have a seaport (model 3 and 7).
In order to test hypothesis 2.1 (namely, that unlike general APS, AMPS agglomerate near seaports) conclusively we removed all the port-related variables in the models for APS and looked at whether the results differ significantly for both establishments (model 4) and GNC (model 8). Finally, we investigate whether the presence of general APS is important for AMPS (model 9 for all 459 cities and model 10 for the reduced sample of 274 port cities) and GNC (model 11 and 12) of AMPS in order to test hypothesis 3.
We also ran all these models with total throughput measured in Million Metric Tons instead of using container data. This changes the sample slightly since there are a number of ports from which we have only container data and no throughput tonnage data. However, the results of these regressions are very similar to those presented in tables 2-4.
What immediately becomes clear from models (1) and (2) is that the urbanization variables (CAPITAL, UNIVERSITIES, GDP per CAPITA, POPULATION) are far more important for APS than for AMPS. Being a capital for instance lead to 73.4% more establishments of APS; for AMPS this factor results in a more modest 47.2% increase in the number of establishments. The factors CONTAINERS, SHIP OWNERS and PORT RELATED INDUSTRIES have a somewhat larger effect on the number of establishments of AMPS than on APS, although the differences are small. Possibly, ship-owners and port industries are not only demanding specialised AMPS, but also general APS, which makes it attractive for APS to locate near ship-owners and port-related industries. Because the sample of AMPS cities also include cities that do not have a seaport, we also run the same model (3), but now for only those 274 cities that in fact have a seaport. We observe that the importance of container throughput increases substantially. Surprisingly, the importance of port-related industries and ship-owners is not affected much. In addition we observe in model 3 that the importance of being a capital city or having a top university decreases. Possibly, the maritime services which locate in seaports are less knowledge intensive routine activities, while maritime services which locate in main urban centres are more command and control oriented (Hall et al. 2010).
Table 2: Results of Negative Binomial estimations of the number of maritime and general APS
In model (4) the port and transport related factors have been removed. In comparison with model (2), we observe an increase of the coefficients of all factors except for GOVERNANCE. This points to the presence of an omitted variable bias in model (4) [and also in model (8)] resulting from the exclusion of the variables SHIP OWNERS and PORT-RELATED INDUSTRY which significantly influence both the location and GNC of APS as became clear in models (2) and (6). On the other hand, we observe that all the signs of coefficients remain the same.
Table 3: Results of Negative Binomial estimations of the global network connectivity of maritime and general APS
When we take GNC as the dependent variable, we see the results are largely in accordance with the results of models (1)-(4) where the number of establishments was the dependent variable. However, there are some noteworthy differences. The coastal city variable has a larger (negative) effect, especially on the connectedness by APS while ISLAND has a positive significant impact on the network connectivity. This might be because many island locations, such as for example Hamilton, Bermuda or Limassol, Cyprus, have favourable tax-regimes for international shipping. From our data it is surprising that the presence of SHIPOWNERS and PORT RELATED INDUSTRIES are more important for the GNC of APS than it is for AMPS. For instance: a 1% increase in the number of ship owners will lead to respectively an increase in GNC of AMPS of 0.072% and GNC of APS of 0.194%. Furthermore we observe that GDP per Capita does not influence GNC significantly. Being a capital is even more important for the connectivity of APS than for the locations of APS. For AMPS being a capital is a less influential determinant of connectivity.
Table 4: Results of Negative Binomial estimations
A number of important issues become clear from the models (9)-(12). We observe that AMPS tend to agglomerate near APS: an increase of 1% in APS establishments leads to an increase 0.13% of AMPS. However, the influence of APS on AMPS decreases when we look at the reduced sample of those cities that actually have a seaport (model 10). This is again evidence that the reason of the location of AMPS in cities that not have a seaport is due to urbanization externalities and proximity to APS in general. Furthermore, across all models, maritime localization economies (SHIPOWNERS and PORT RELATED INDUSTRIES) remain significantly correlated with AMPS, but most port-specific variables (CONTAINERS, COASTAL CITY, ISLAND) are not. It is only in port cities that container throughput is correlated with AMPS (models 10 and 12). It is also striking that the effects of universities is much smaller in these models, probably because some AMPS need specialised localised knowledge, which is probably provided by APS in general.
When we exclude this variable, UNIVERSITIES account in part for this effect, so the coefficients of previous models are probably somewhat overstated.
Recent contributions within world city network research have raised the question of the effect of sector-specific specialization of advanced producer services on urban hierarchies and the geography of global networks between cities. This question is related with a more fundamental lacuna in world city network research, namely the limited confrontation of the empirical research with urban economic theory. In this study we have looked at the location and network connectivity of those advanced producer services that are specialized in providing services to the maritime transport industry. In addition we explored what factors influence both the location and connectivity of these specialized advanced services compared with advanced producer services in general.
Alderson, A. S. and J. Beckfield (2004), ‘Power and position in the world city system’,
American Journal of Sociology, 109, 811–51.
Becker, S.O., Ekholm, K., Jäckle, R., Muendler, M.A. (2005). Location choice and employment decisions: a comparison of German and Swedish multinationals. Kiel: Institute for World Economics.
Johnstad, T. (2000), Mutual Maritime Insurance Clubs- Co-operation and Competition, Annals of Public and Cooperative Economics, 71, pp. 525-555.
Wall, R.S. (2009). Netscape: Cities and Global Corporate Networks. Rotterdam: ERIM PhD Series 169.
* Wouter Jacobs, Urban and Regional Research Center Utrecht, Utrecht University, the Netherlands, e-mail: firstname.lastname@example.org
1. Note that our approach is different than that of Verhetsel & Sel (2009). They base the hierarchy of world maritime cities on the location of headquarters and offices of container carriers and global terminal operating companies. We look at the advanced producer services specialized in the maritime transport sector.
2. Respectively accountancy firms, advertising firms, banking & finance firms, insurance firms, law firms and management consultancy firms.
Note: This Research Bulletin has been published in Urban Studies, 48 (13), (2011), 2749-2769