This Research Bulletin has been published in Tijdschrift voor Economische en Sociale Geografie, 97 (5), (2006), 515-534.
Please refer to the published version when quoting the paper.
INTRODUCTION: THE PROBLEM OF ‘SECOND’ CITIES
The closure of Rio de Janeiro’s Stock Exchange in 2000 consolidated Sao Paulo’s position as Brazil’s financial centre. Although an important national event, in fact this process of financial concentration within countries has been a worldwide phenomenon. ‘Second cities’ have been declining on all continents: for instance, Zurich is now undisputed Swiss financial leader over Geneva, Frankfurt over Hamburg (or Munich), Sydney over Melbourne, Johannesburg over Cape Town, Mumbai over New Delhi, Toronto over Montreal, with Osaka the classic case of such decline, relative to Tokyo (Hill and Fujita 1995). In short, this appears to be a worldwide process. It suggests that contemporary globalization is creating spatial economic concentration within countries through focused linking to the world economy – in Friedmann’s (1995) terms ‘articulation’ of national and regional economies into the world economy.
In fact, the traditional urban geography concept of ‘gateway city’ (see, for instance, Johnston 1982) has been revived to describe such articulations. In particular, Andersson and Andersson (2000) in their Gateways to the Global Economy have brought together twenty studies that focus upon gateway cities and regions. These researches highlight a selection of leading cities in which their city functions are interpreted as gateway processes. In contrast, following Grant (1999) and Nijman (2000), Short et al. (2000) ‘use the term gateway to refer to the fact that almost any city can act as a gateway for the transmission of ... globalization’ (p. 318). In other words, they are using the concept as part of an argument for shifting concern for cities in globalization away from focus on just a few major ‘global cities’ - Sioux Falls is one of seven ‘sample cities’ they discuss. This is consistent with Marcuse and Van Kempen ‘s (2000) shift from exclusive ‘global city’ to inclusive ‘globalizing cities’. Thus, gateway cities have reappeared in the world city literature in quite contrary guises: as a result of concentration processes (Andersson and Andersson 2000), and as an outcome of dispersion processes (Short et al. 2000).
For national ‘second’ cities, these two interpretations have profound implications for whether they can expect to experience relative economic decline or not. And the jury is still out. Thus, although the closing Rio’s stock exchange lends weight to Santos’ argument (1993) that the key process in Brazil has been continual economic concentration in Sao Paulo, this has not gone unchallenged. According to IPEA (et al. 2001), it is a set of medium-sized Brazilian cities, such as Manaus and Fortaleza that are the dynamic success stories of the national urban network. In fact, the ‘comeback’ of ‘medium-sized’ cities is becoming another common story across the world (Geyer 2002): for instance, in Mexico (Aguilar 1999), in the UK (Parkinson et al. 2004), and in the USA (Lang 2003). These studies suggest gateway city functions are not becoming wholly concentrated in a few leading cities, thus lending weight to Short et al.’s (2000) position. In this argument ‘second’ cities have much less to fear from globalization than is often supposed.
This conundrum of contrary predicted urban outcomes under conditions of contemporary globalization is not as problematic as it first appears. Globalization is not a simple singular process; it is a myriad of practices that generate a range of outcomes. Thus, alongside the concentration there are tendencies towards economic dispersion. In fact the ‘combination’ of ‘global integration’ and ‘global dispersion’ was one of the starting points of Sassen’s (1994, p.4) influential writings on world cities. In other words, while we may expect that some globalization processes will promote concentration of gateway functions within a country or region, this does not preclude other processes operating to encourage gateways to be more dispersed.
How this clash between concentration and dispersion processes in globalization operates within a country can only be discerned empirically. In this paper we look at contemporary outcomes in the Brazilian city network. We use this case study as a means to inform debates on changing city network configurations within countries resulting from globalization. The agents of globalization that we study are banks; we look at how both domestic and foreign banks are using Brazilian cities. There are three reasons for this choice of economic sector. First, data availability is important but this pragmatic reason is supported by, second, the fact that banking and finance have been at the cutting edge of globalization, and third, because financial services provide a particularly strong test for the complexity/dispersion gateway thesis given the well known concentration of finance functions, including leading cities identified as ‘financial gateways’ (Andersson 2000). Although we show below that indeed banking functions are more strongly concentrated in Brazilian cities than other economic sectors, and confirm the overall importance of Sao Paulo for financially articulating the Brazilian economy into the world economy, our main finding is that the situation is more complex than a simple, single gateway pattern. Certainly, the contemporary world economy is very complex and requires much more than a limited number of ‘global cities’ to operate successfully, and this is reflected in contemporary Brazilian city network.
The research reported here follows the approach and methods used in the quantitative studies of the Globalization and World Cities (GaWC) network (Taylor 2004) and builds upon a previous related paper that focuses on the connectivities of Brazilian cities (Rossi and Taylor 2005). Both Brazilian city studies are distinctive in being explicitly multi-scalar in design: we consider the top Brazilian cities in an integrated data set with the top world cities. We focus upon leading banks in Brazil and this enables us to search out patterns of operation that are regional, national, transnational and combinations of these scales. In short, we ask questions such as how do these banks use cities like Manaus and Belem, compared to Sao Paulo and Rio de Janeiro, and to London and New York? To answer such questions the argument is organised into four parts. First, we consider further the concept of gateway city and suggest that its use in some recent third world studies is only partial in nature. Second, we describe the methodology, both data collection and analysis, that overcomes the partiality. For the analysis we use principal components analysis and interpret the components as defining banking locational strategies constituting subnets of cities. It is in such subnets that we can expect to find gateway cities. Third, we describe six banking locational strategies that we identify in the analysis. Here we present the substantive findings of the research, identifying gateway cities. As expected, cities of different stature are intermingled within strategies and subnets illustrating our starting point on the complexity of city development under conditions of contemporary globalization. Fourth, we discuss what these results mean for Brazilian cities within globalization, for Sao Paulo and Rio de Janeiro, but also for other cities that feature in our analysis. In a short conclusion we consider our results more generally for the concept of gateway cities in development policy making under conditions of contemporary globalization.
THE CONCEPT OF GATEWAY CITY
‘Gateway’ is as a very common spatial term in the geographical literature, and the gateway concept was discussed for the first time by R. D. McKenzie, in 1933, in The Metropolitan Community1. In the seminal treatment of ‘gateway city’, Burghardt (1971) describes it as a positional concept that characterizes the city as a combined entry/exit for a given area or region (Burghardt 1971). The concept was developed to explain a particular feature of the US frontier cities in the 19 th century, that were simultaneously centres of economic power within a broad area and strongly connected to distant cities2 (Drennan, 1992). Thus in historical studies, “gateway city” is antithetical to the concept of the city in central place theory (Drennan 1991; Burghardt 1971). The latter’s emphasis on local hierarchies in symmetrical spatial patterns are precisely what is not found in gateway city studies.
In 1957, E. L. Ullmann,3 aggregated to the concept the idea that ‘gateway cities’ host the headquarters of the most important national services, in this case: railroad and navigation lines (Burghardt 1971). Also US airlines linked the frontier concept to the ‘gateway city’, defining it as the first/last American city of an international flight. This idea that gateway cities are ‘located on a site of considerable transportational significance’ is emphasized by Burghardt (1971, 270). The ‘crossroads’ idea has been brought up to date by Pred, (1977),4 in terms of the most used paths of information flows through cities that serve their hinterlands and distant foreign markets (Drennan, 1992).
In translating these ideas to conditions of contemporary globalization there has been a tendency to tell only half the story. This stems at least as far back as Friedman’s (1995) identification of world cities as ‘national articulations’ in his world city hierarchy. One side of the articulation – world cities in the world-economy – is described in some detail but the ‘national economies’ being ‘articulated’ are neglected. Translated to gateway city, this is to study relations in the ‘foreland’ while ignoring the hinterland. Two important empirical studies illustrate this limitation of using the gateway concept in globalization. Grant and Nijman (2002) observed that the gateway function of world cities in less-developed countries is played operated through the large port cities ( Accra and Mumbai), and this function affects some parts of the internal spatial structure of the city. Here, the concept was partially applied – there is no mention to the cities’ hinterlands – the focus was on identify the city original function and the gateway city function, and the latter’s effects on the city internal area. In a related paper, Grant (2001) studies the liberalization polices in Ghana and foreign-company activity, more specifically in Accra. Detailed descriptions of foreign markets linkages through the foreign companies are provided but, again, Accra’s hinterland, by implication the rest of Ghana, is not similarly treated. This focus on just half the gateway flows inevitably leaves many questions unanswered. How dominant are these gateway cities within their own ‘national economies’? How are the other cities within the country faring – are they simply dependent on these national articulator gateways? These questions, of course, are precisely those that have stimulated this study. Put another way, are ‘second cities’ inevitably losers in the globalization process? We have noted in the introduction that this is not invariably the case. But studying only the prime city and its ‘external’ links does predispose argument towards economic concentration rather than dispersion. In this study, and related studies (Rossi and Taylor 2005 a; Rossi, Beaverstock and Taylor, 2005 b), we explicitly research both ‘external’ and ‘internal’ financial relations for our case study, Brazil.
The reason why we should take dispersion seriously when researching the gateway city concept is because, as Burghardt (1971) emphasizes, gateway processes are inherently variable: there is a ‘temporal as well as locational aspect to the development of gateway cities’ (p. 272). As transport/communication hubs, the fortunes of gateway cities often change with technological advances. This is, of course, vital in considering economic globalization since this is where the ‘end of geography’ thesis was first pronounced (O’Brien 1992). We don’t subscribe to such an extreme dispersal argument but instantaneous communication, especially in global finance, does require a critical approach be made towards simple application of the idea of a ‘gateway’. In Friedman’s (1995) world city hierarchy, Sao Paulo is one of only two ‘third world’ ‘national articulators’, could it possibly be the case that financial flows in and out of Brazil require and use this one ‘gateway’? Below we provide evidence to assess the degree to which this may be the case.
Since the 1980s, banking markets that were traditionally strongly regulated and protected have liberalized such activities, given rise to a considerably presence of foreign banks. These last have established their services overseas through acquisition of extensive branch networks in the host country or establishing there a single office for specific ends (Cerutti et al. 2005). In this paper, studying globalization as a world city process we focus on the foreign and domestic links of the Brazilian banking system. Foreign banks have operated in Brazil since the capital market reforms of 1964 and 1966, under severe limits, serving firms headquartered in their country of origin (Carvalho, 2000). Against the weak social, economic and political Brazilian environment at the beginning of the 1990s, the foreign banks were ‘rushing to increase business in Brazil’ (Hieronymus 1991). The real entrance of foreign banks as major players in Brazil occurs after 1995, when the domestic banks Nacional and Economico both were liquidated (Carvalho 2000). The result has been that among ‘emerging markets’ (data on 19 countries in Latin America and Eastern Europe, including Mexico, Argentina, and Chile), Brazil is ranked first, hosting 73 foreign banks, followed a long way behind by Poland with 45, and Argentina with 30 (Cerutti et al. 2005). However, given the dimensions of the country, according to analysts, the Brazilian market place has room for just five or six large-scale banks (Caplen 2001). In 2005, seven top banks in Brazil have 78% of all bank branches in Brazil:5 two state banks Banco do Brasil (3993 branches) and Caixa Economica Federal (1854), two Brazilian private banks Bradesco (2914) and Itau (2207), followed by the English HSBC (930), the Brazilian private Unibanco (907), and the Dutch ABN AMRO (775). But, of course, the foreign interlopers have their strengths, by definition, beyond just Brazil in a situation somewhat more spatially complex than a focus just within the territorial bounds that is Brazil.
In a previous study (Rossi and Taylor 2005 a), we selected 30 top Brazilian urban sites (city-regions and detached cities) hosting the biggest clients of banks in Brazil: these were cities hosting headquarters of the top 500 firms in the period 1966-2002. However, the correlation results from city sales of the firms and city banking connectivities were not significant. This implies that the concentration of top firms headquartered in Sao Paulo explains only partially the global banking links of this Brazilian city. On the other hand, the same study identified less well connected cities as sites of important clients for top foreign and domestic banks in Brazil (Rossi and Taylor 2005 a). In this empirical study, we explore these relations further by delineating the configurations of banking links within Brazil and beyond in order to assess Sao Paulo in its gateway role, from both sides of ‘the gate’ as it were.
The related paper (Rossi and Taylor 2005 a) used the interlocking network model (Taylor 2001) to measure banking interlock connectivities of Brazilian and world cities. Here we use the same data as this earlier paper wherein its collection is described in detail. Thus below we provide only a basic description of the data to make this paper freestanding; for more information reference should be made to Rossi and Taylor (2005 a). However the analysis of the data is completely different. Whereas the previous paper produced results summarising the data as a whole, here we are concerned for dividing up the banks into separate clusters based upon their locational strategies. Thus we provide new measures on different subnets of cities.
Data collection on banks in cities - Data has been collected on the office networks of banks, both main offices and branches. The result is a ‘banking activities matrix’ relating 31 banks in Brazil to 54 cities both domestic and foreign. The selection of firms and cities, and derivation of ‘activity scores’ was as follows.
The end-result of this data collection exercise was a 31 banks x 54 cities activity matrix. Each column defines the locational strategy of a bank – which cities it operates in and its levels of operation. Each row defines a city’s mix of banks – in terms of banks present and their levels of activity – available to service clients. It is these 1674 activity values that are analysed to explore the variation in banks’ uses of Brazilian and world cities.
Principal components analysis defining city subnets - Principal components analysis is a member of the factor analytic family of multivariate statistics. These techniques are employed on large data matrices to reduce them to their basic ‘factors’, or in this case, ‘components’. We use the principal components model because it is the simplest (with least theoretical baggage – see Rummel (1970)), and because it has proven to be successful for this type of analysis before ( Taylor and Walker 2001; Taylor et al. 2002b and 2004; Taylor 2004). The specific model we employ is a principal components analysis with varimax rotation, the latter to ensure clear, interpretable results.
The data reduction consists of transforming a large number of variables into a much smaller number of components. For instance, a typical analysis might have 100 variables, which are reduced to, say, 8 components. Each component is a new composite variable that brings together selected original variables on the basis of their inter-correlations. Thus components are clusters of like-variables. In our case, there are 31 variables, the locational strategies of banks. In the analysis reported below we reduce these to just six components, which are clusters of banks with similar office networks. We interpret the components as composite or common locational strategies for the given cluster of banks.
The principal components analysis provides three key pieces of information through which to interpret results: component loadings on variables (banks), component scores on objects (cities), and a measure of the importance of each component.
A key feature of principal components analysis is the choice of number of components. There is no generally accepted way to make this choice; we have used the experimental method to find particularly stable components - ones that exist through many choices of component numbers (see Taylor et al. 2002b).
THE GLOBALIZING CONTEXT OF THE LOCATIONAL STRATEGIES OF BANKS IN BRAZIL
Before we look at the empirical findings from the data analysis, it is necessary to place the banks operating in Brazil into their economic geography context. We are particularly interested in banks servicing transnational capital and therefore we focus upon foreign firms in Brazil, both where they come from and where they are located within Brazil. Only a basic description is provided and the focus is on providing the essential context in which the 31 banks we study are operating.
Considering the top 500 firms in Brazil in 2002, 221 are fully or part foreign owned. In Table 1 we present the distribution of this ownership across the world. As expected the USA dominates but western European countries are also well represented. Outside these two core regions, only Japan has appreciable foreign investment: it ranks seventh for number of firms but note that its sales are actually less than for Portuguese firms. It is these foreign-owned firms that we would expect to likely use transnational banks in Brazil. However, of the 25 countries listed in Table 1, there are only eight countries with banks that feature in the top banks in Brazil (and therefore in our data). These frequencies are also listed in Table 1, again showing the USA ranked first but with much less dominance than for firms in general. With Germany and Japan between them having more banks than the USA, and with both Italy and Portugal without banks, there is by no means a simple match between origins of foreign investment and foreign banks in Brazil.
Within Brazil, foreign-owned firms are located in 21 of the 30 cities in our data. These are listed in Table 2 where the distribution across countries is provided. In this case the most notable feature is the overwhelming dominance of Sao Paulo (36%), led by US firms but with Germany and France well represented. Rio de Janeiro, while not challenging Sao Paulo is the clear second city for foreign investment, again dominated by the USA but this time followed by Spain and the UK. Manaus in third place is a surprise, and is almost wholly due to Korean and Japanese firms. Campinas in fourth place is also unexpected and, in this case, is the result of US firms. Among the major Brazilian cities, the low ranking of Brasilia is noteworthy confirming its specialist Brazilian political role. This city distribution can be compared to the Brazilian headquarters of our 14 foreign banks in Table 3. Here the concentration is stark: 12 foreign banks in Sao Paulo (86%). The interesting feature is the two other foreign banks being located in Curitiba. It is only after adding Brazilian private and state-owned banks that more bank–headquarter Brazilian cities appear, but only six in all. Thus, banking is clearly much more concentrated in general than for other firms, and the concentration in Sao Paulo remains extreme even after adding Brazilian banks (20 out of 31, or 65%).
At this organizational level of headquarters, there is obviously no direct relation between banks and foreign firms because of the former’s more concentrated geography. However our analysis of banking circles is not limited to headquarter functions as the last section made clear. The distribution of branches across both the Brazilian and world cities in our data are also shown on Table 3. Obviously, it is the foreign banks that have the larger transnational pattern of offices, with a majority of Brazilian banks having no foreign offices. It is this mix of national and transnational service distributions that represents the contemporary circles of banks in Brazil. The banking activities data matrix that we have constructed provides the details of these office distributions and it is to the analysis of this matrix that we now turn.
BANKING SUBNETS: LOCATIONAL STRATEGIES OF BANKS IN BRAZIL
The principal components analysis of our 31 banks x 54 cities activity matrix produced a six-component solution that accounted for 76% of the original variability among banks’ locational strategies. These six components are common locational strategies of banks in Brazil that focus on specific subnets of cities. Each component is described in the same form to facilitate comparison: interpretations are based upon listings of relevant loadings and scores. Our thresholds for defining ‘relevant’ are quite low so that we do not miss parts of the configurations. For loading we include all banks above 0.34 (this indicates at least 10% of the bank’s variability is accounted for by the component), and for scores cities over 0.5 are included. In both cases both positive and negative scores reaching the thresholds are included. There is a table for each component and the list of firms indicates those contributing to the common locational strategy, while the list of cities indicates the subnet of cities that is the locational strategy.
The six components fall into three strata in terms of size. There are two very important components that between them very nearly account for half the overall variability: they both account for 23%. The third component accounts for approximately half the variability of the first two (12%) and forms a middle-strata. Finally, there are three small components with one at 8% and two accounting for 5% each. We describe and interpret each component from largest to smallest in turn.
The Brazilian cities hierarchical strategy - This major component defines a common strategy of a large number of Brazilian banks (Table 4). There are a few foreign banks but these tend to be the banks without headquarters in our data’s world cities (Citibank at the bottom of the list is the exception). These banks define a subnet of cities (with the positive scores) that is a hierarchical ordering of leading Brazilian cities (Rossi and Taylor 2005), starting with Sao Paulo and Rio de Janeiro: hence our labelling of this component. This is a national strategy since the leading cities between them service all regions from North ( Belem) to South ( Porto Alegre). However, notice that the national gateway function appears quite diffuse: although Sao Paulo and Rio de Janeiro have the highest scores, there are four other important cities – Belo Horizonte, Salvador, Porto Alegre and Brasilia. Note also that the negative scores are dominated by smaller Brazilian cities plus world cities, including London, also appearing. Hence this is not a simple ‘Brazilian cities versus world cities’ component; it is hierarchically selective within the Brazilian city network.
The world cities strategy - This major component defines a common strategy limited to foreign banks in Brazil (Table 5). This involves both pairs of Japanese and German banks and the French bank in our data. Citibank again appears where it is much more important than for the previous component. These are all banks with large world networks that use Sao Paulo as their Brazilian headquarters. In contrast the banks with negative loadings are all Brazilian banks (GM7 is a special exception) with low presence in the selected set of world cities). Thus it is not surprising that these banks define a subnet of cities (with positive scores) that is a hierarchical ordering of world cities (Taylor 2004) but with Sao Paulo reallocated to the top as Brazil’s global gateway city. The only other Brazilian city to be in this subnet is Rio de Janeiro. The negative scores is a long list of Brazilian cities specifically identified as the cities this common strategy under employs including important cities like Brasilia and Salvador. There are no world cities in this negative score list. Hence, there is a strong tendency for this to be a ‘Brazilian cities versus world cities’ component but Sao Paulo and Rio de Janeiro prevent this simple division interpretation. Brazilian banking circles are not separate from transnational circles and this is the locational strategy where Sao Paulo, in particular, and Rio de Janeiro, operate as financial gateways to the Brazilian economy.
The world cities and southern Brazilian cities combined strategy - The medium component is quite complex as the above label indicates. The common strategy is composed by Brazilian banks largely based in Sao Paulo and Brasilia but with the Spanish bank in the data having the largest loading and two American banks also (Bankboston and GM) near the bottom. The resulting subnet defined by the positive scores combines selected world cities, with Buenos Aires and Madrid and important southern Brazilian cities. The non-southern Brazilian city exceptions are Sao Paulo (note, however, that in this component Sao Paulo does not record the highest score), Sao Jose dos Campos, Pocos de Caldas, and Brasilia. Rio de Janeiro does not make this list. The negative scores include less important world cities plus two non-southern Brazilian cities. Hence, we can interpret this as a particular regional-transnational locational strategy that hints at possibly by-passing Brazil’s gateway cities from the previous component.
The Brazilian core-region strategy - This component, somewhat in between the medium and small components, is a common strategy of Brazilian banks, the only exceptions being the (small) Swiss bank and the unusual American bank (GM) in the data. Thus the positive scores define a Brazilian cities subnet, led by Sao Paulo and neighbouring Campinas as Brazilian core gateway cities. The outer bounds are defined by the bottom two cites among the positive scores list, Belo Horizonte to the north and Curitiba to the south. This defines the economic core-area of Brazil. Although headed by Sao Paulo like the first Brazilian subnet, this is distinctive in its areal definition, including both major and minor cities within its confines. Thus, in the negative scores, the major cities of Brasilia and Porto Alegre are highlighted as beyond the core-area along with other ‘outer’ Brazilian cities and important world cities.
The Curitiba new gateway strategy - This minor component is based upon two foreign banks that are quite different: a small US bank (CNH, whose headquarter city is not a world city and thus is not in our data) with Brazilian interests and HSBC the UK’s largest global bank. They are brought together in this common strategy through their both using Curitiba for their Brazilian headquarters. The result is a subnet centred upon Curitiba that has by far the highest loading recorded in the whole analysis. London has the second highest score with three minor Brazilian cities completing the subnet. The latter are cities serving important agricultural regions (like Curitiba): the CNH bank specialises in providing funds for investment in agricultural machinery. The negative scores are particularly interesting. Composed mainly of Brazilian cities, but with Los Angeles and Chicago having the most negative scores, it is particularly fascinating that the next most negative scores are for Sao Paulo and Rio de Janeiro. Hence, despite its the ‘odd couple’ banking basis, this component most certainly shows an alternative locational strategy that completely bypasses Brazil’s two traditional gateway cities.
The Northern Brazil / European cities strategy - This smallest of components is not unlike the medium component but with the opposite regional emphasis. The common strategy is made up largely of the Brazilian-Dutch bank in our data with only small contributions from two other banks and a negative loading for the largest US bank in our data. The end-result is that the subnet defined by positive scores features largely northern Brazilian cities, especially Belem and Manaus that both score higher than Sao Paulo, plus European cities, with Amsterdam notably scoring higher than London. Belem qualifies as the northern gateway city. The negative scores confirm this pattern where a few Brazilian cities are dominated by non-European world cities including all five US cities in our data. One noteworthy feature is the absence of both Japanese banks and Tokyo from the table given the East Asian firms investments in Manaus reported in Table 2. Clearly the Japanese banks have not followed Japanese firms in their locational strategy towards Brazil: this is not a surprise since it reflects the more general mismatch between Tables 1 and 2 reported above. Whatever the explanation for the latter, this smallest component does show another possible alternative strategy to the traditional, two city gateways into Brazil.
WHAT DO THESE STRATEGIES MEAN FOR BRAZILIAN CITIES?
All Brazilian cities are part of contemporary globalization and our analyses show the banking circles within and without Brazil of the thirty leading cities. In a previous study focusing on interlocking connectivity (Rossi and Taylor 2005) we have shown that seven Brazilian cities appear as the most connected through banks both worldwide and domestically. Overall they were ranked as follows: Sao Paulo, Rio de Janeiro, Curitiba, Belo Horizonte, Porto Alegre, Brasilia, and Salvador. In this section we focus upon these cities to show how our analyses identifies the variety of roles each plays in domestic and transnational banking. Here an important property of our principal components analysis is particularly relevant: the components are orthogonal, which means they are computed to be independent of each other. In other words components define separate and distinctive locational strategies with city subnets.
Three ‘Sao Paulos’ - It will have come as no surprise that Sao Paulo has dominated our analysis. It appears as a relevant city in all locational strategies and is the most important city in three of the resulting subnets. The latter indicates that the city has three distinct important roles for banks in Brazil. Thus we can say that the use of Sao Paulo by banks in Brazil defines three cities:
Although identification of each role is by no means new, they have never previously been specified through the operation of banking practices and shown to be distinct and separate outcomes.
Our findings for Sao Paulo on the other three components show that the city’s role is not just limited to the above. There are secondary roles for Sao Paulo in the two regional-transnational strategies, both southern and northern, implying its overall importance is such that its gateway function is difficult to avoid in banking circles. However, perhaps more interesting is the negative score for Curitiba’s alternative gateway for here we see the one and only example of a by-passed Sao Paulo.
Rio de Janeiro as a lesser Sao Paulo - In this analysis Rio de Janeiro appears clearly as Brazil’s second city but without a distinctive role: it never appears in a component without Sao Paulo and it always scores lower than the latter city. Thus for banks in Brazil, Rio de Janeiro seems to function as a lesser Sao Paulo (Tolosa 1998). In particular, this involves the city in two impressive roles relative to the remaining cities:
Elsewhere it has secondary roles as part of the economic core strategy, features in the northern but not the southern Brazil strategy and, like Sao Paulo is specifically by-passed in the Curitiba strategy.
Curitiba as alternative gateway - Part of the banks’ domestic hierarchical and core-area strategies, Curitiba’s unequivocal importance is the glimpse it gives of a new gateway into the Brazilian economy that by-passes Sao Paulo and Rio de Janeiro. The latter is only a very small component but the significance of its distinctiveness more than makes up for its statistical size.
Belo Horizonte as less important globally - Belo Horizonte is not greatly used by banks in Brazil beyond confirming that it is more important domestically than in global circles (Rossi and Taylor 2005). Thus it appears only as part of the domestic hierarchical and core-area strategies and, curiously appears to be by-passed by banks for their northern strategy.
Porto Alegre as domestic centre - Porto Alegre’s three roles are domestic: near the top of the Brazilian city hierarchy, an important part of the banks’ southern strategy, and most definitely outside the domestic economic core-area strategy.
Brasilia’s neglect in world banking circles - Our analysis confirms that Brasilia has not developed important roles in banking circles: it is part of the domestic hierarchical strategy and southern strategy, and outside the core-area strategy. However the key finding is the city’s negative score for the world cities strategy: it is not making its economic mark on a global stage.
Salvador as a by-passed city - Salvador confirms its place in the Brazilian hierarchical strategy but otherwise is defined negatively in this study: it by-passed by the core-area strategy, the northern strategy and, most important, by the world city strategy .
In terms of the banking transnational relations of these seven cities, we can summarise the situation as follows: two cities appear to have clear worldwide roles – Sao Paulo and Rio de Janeiro –, Curitiba has a possible emergent worldwide role, while the other cities transnational banking links are bound to Sao Paulo, Rio de Janeiro or Curitiba.
There is a tendency in the world cities literature to identify particular important cities in countries and regions as the sole ‘gateway’ from and to the wider world-economy. Using this simplistic geographical model it becomes an easy task to argue that Brazil’s connection to the world-economy through its banks is merely a matter of the rise of Sao Paulo as an international financial centre. While our unique analysis, combining domestic and world cities, does confirm that banking in Brazil is dominated by Sao Paulo, we have shown that the situation is rather more complex than a single, simple gateway function. We have identified six different locational strategies wherein Sao Paulo is the key city in three, including the two most important.
However, Brazil is not just simply ‘ Sao Paulo’s hinterland’, and Sao Paulo is not simply ‘ Brazil’s gateway’. The analysis shows three Brazilian regions explored by foreign banks: Northern, Southern and Rural, covering a huge area beyond Sao Paulo, and indicating that paths of information are being used to link such regions to New York, Buenos Aires, London, Madrid, Paris, Frankfurt, Tokyo, Milan, Miami, Amsterdam and Toronto. And these are both serviced through three Brazilian gateway cities bringing together Brazilian hinterlands and distant foreign markets in the manner of the original conception of gateway city.
This study has been conducted as extensive research that provides patterns and contexts in banking configurations but it cannot inform and evaluate the actual “entry” and “exit” processes operating through the gateway cities. However, our broad brush analyses can contribute to debates on development polices. The gateway city concept encompassing hinter-and forelands may possibly recover the lack of interaction between the different territorial scales in globalization. A core debate on national development strategies is sometimes presented as choice between locational concentration and dispersion. Our analysis confirms the more sophisticated world cities’ position that both are occurring simultaneously. This means that sophisticated arguments such as Scott’s (2002) promoting ‘regional pushes’ as pathways to development via agglomeration economies and externalities should not be used to promote gateway thinking that results in policies of singular concentration. Historical experience tells us that successful economic development has been a process of transformation from relatively simple economies to much more complex patterns of production and relations. This requires multiple economic niches in numerous cities. Mumbai is certainly India’s major world city and leading gateway but its economic successes are certainly insufficient to account for the recent impressive growth of the Indian economy. In many ways it is the smaller city of Bangalore that is at the cutting edge of India’s relatively successful engagement with contemporary globalization. In other words, there can, and should, be several ‘regional pushes’ in a given development strategy. Over-focus on the Sao Paulo core area in Brazilian development thinking might just stymie the rise of ‘ Brazil’s Bangalore’.
AGUILAR, A. G. (1999), Mexico City growth and regional dispersal: the expansion of largest cities and new spatial forms. Habitat International 23, pp. 391-412.
ANDERSSON, A. E. (2000), Financial gateways. In A. E ANDERSSON & D. E. ANDERSSON, eds., Gateways to the Global Economy pp. 31-48. Cheltenham, UK: Edward Elgar.
ANDERSSON, A. E. and ANDERSSON, D. E., eds., (2000), Gateways to the Global Economy. Cheltenham, UK: Edward Elgar.
BURGHHARDT, A. F. (1971), A hypothesis about gateway cities. Annals of the Association of American Geographers, 61, 2, pp. 269-285.
CAPLEN, B (2001), Banks fight for a future. The Banker, 151, 899, p. 97.
CARVALHO, F. J. C (2000), New competitive strategies of foreign bank in large emerging economies: the case of Brazil. Banca Nacionale del Lavore Quarterly Review, 213, p. 135-169.
CERUTTI, E., DELLARICCIA, G., PERIA, M. S. M. (2005), How banks go abroad: branches or subsidiaries?. World Bank Policy Research Working Paper 3753, October.
DRENNAN, M. P. (1992), Gateway cities: the metropolitan sources of US producers service exports. Urban Studies, 29, 2, pp. 217-235.
FRIEDMANN, J. (1995), Where we stand: a decade of world city research. In P. L. KNOX & P. J. TAYLOR, eds., World Cities in a World-System. Cambridge: Cambridge University Press .
GEYER, H. S., ed., (2002), International Handbook of Urban Systems. Cheltenham, UK: Edward Elgar.
GRANT, R. (1999), The Gateway City: Foreign Companies and Accra, Ghana. Paper delivered to the Third World Studies Association Meeting, San Jose, Costa Rica, 21 November.
GRANT, R. (2001), Liberalization polices and foreign companies in Accra, Ghana. Environment and Planning A, 33, pp. 997-1014.
GRANT, R. and NIJMAN, J. (2002), Globalization and the corporate geography of cities in the less-developed world. Annals of the Association of American Geographers, 92, 2, pp. 320-340.
HILL, R. C. and FUJITA, K. (1995), Osaka’s Tokyo problem, International Journal of Urban and Regional Research 19, pp. 181-91.
HIERONYMUS, B. (1991), Sensing the birth of a boom, foreign bankers pour cash into Brazil. American Banker, 156, 180, p.6.
IPEA, IBGE and UNICAMP (2001), Caracterização e tendências da rede urbana do Brasil. Brasilia: IPEA, V. 1-6.
JOHNSTON, R. J. (1982), The American Urban System: A Geographical Perspective. London: Longman.
LANG, R. E. (2003), Edgeless Cities – Exploring the Elusive Metropolis. Washington: Brookings Institution Press.
MARCUSE, P. and VAN KEMPEN, R. (2000), Globalizing Cities. Oxford: Blackwell.
NIJMAN, J. (2000), The Paradigmatic City. Annals of the Association of American Geographers 90, pp. 135-45.
O’BRIEN, R. (1992) Global financial integration: the end of geography. London: Pinter.
PARKINSON, M., HUTCHINS, M., SIMMIE, J., CLARK, G. and HERDONK H. (2004), Competitive European cities: where do the Core Cities stand? London: ODPM.
ROSSI, E. C. and TAYLOR, P. J. (2005 a), Banking networks across Brazilian cities: interlocking cities within and beyond Brazil. Cities 22, 5, pp. 381-393.
ROSSI, E. C., BEAVERSTOCK, J. V. and TAYLOR, P. J. (2005), Transaction Links through Cities: ' Decision Cities' and ' Service Cities' in Outsourcing by Leading Brazilian Firms. GaWC Research Bulletin 175.
RUMMEL, R. J. (1970), Applied Factor Analysis. Evanston: Northwestern University Press.
SANTOS, M. (1993), A Urbanização Brasileira. Sao Paulo : Hucitec.
SASSEN, S. (1994), Cities in a World Economy. Thousand Oaks, CA: Pine Forge.
SCOTT, A. J. (2002), Regional push: towards a geography of development and growth in low- and middle-income countries. Third World Quarterly, 23, 1, pp. 137-161.
SHORT, J. R., BREITBACH, C., KUSS, M., and WELLS, H. (2000), From world cities to gateway cities. City, 4, pp. 317-40.
TAYLOR, P. J. (2001), Specification of the world city network. Geographical Analysis 33, pp. 181-94.
TAYLOR, P. J., CATALANO, G. and WALKER, D. R. F. (2002a), Measurement of the world city network. Urban Studies 39, pp. 2367-76.
TAYLOR, P. J., CATALANO, G. and WALKER, D. R. F. (2002b), Exploratory analysis of the world city network. Urban Studies 39, pp. 2377-94.
TAYLOR, P. J. and WALKER, D. R. F. (2001), World cities: a first multivariateanalysis of their services complexes. Urban Studies 38, pp. 23-47.
TAYLOR, P. J. (2004), World City Network – A Global Urban Analysis. London: Routledge.
TAYLOR, P. J., CATALANO, G. and WALKER, D. R. F. (2004), Multiple globalisations: regional, hierarchical and sectoral articulations of global business services through world cities. The Services Industries Journal 24, pp. 64-82.
TOLOSA, H. C. (1998), Rio de Janeiro as a world city. In F.-C. LO & Y.-M. YEUNG, eds., Globalization and the World of Large Cities, pp. 203-27. Tokyo: United Nations University Press.
Eliana Consoni Rossi was supported by CNPq, a Brazilian governmental institution that supports Scientific and Technological development. We acknowledge Prof. Elton Fernandes (Production Engineering, COPPE, UFRJ, Brazil) for his help with the preliminary dataset. We thank the referees for their helpful comments.
1. New York, Russel & Russel, p. 4.
2. EATON, L. (1989), Gateway cities and other essays. Ames, IA, Iowa State University Press.
3. American Commodity Flow: a Geographical Interpretation of Rail and Water Traffic Based on Principles of Spatial Interchange, Seattle, University of Washington Press, pp. 13 and 18.
4. City Systems in Advanced Economies, New York, Halsted Press .
6. The full reference for our source is: Brazil. Ministerio da Fazenda. Banco Central do Brasil. Sistema Financeiro Nacional. Informacoes cadastrais e contabeis. Informacoes Cadastrais. Instituicoes Financeiras. Bancos Comerciais Multiplos e Caixa. Ranking dos Bancos – Posicao dos Balancetes de Junho ou Dezembro: (documento 4010). Ativo Total. http://www.bcb.gov.br/?IF (3 May 2004).
7. GM bank is a special case because it belongs to General Motors in Brazil. Its world headquarters is in Detroit, but its branches are found throughout Brazil within General Motors’ dealerships. Hence, although it is a foreign bank, it has a locational strategy similar to many domestic banks.
Table 1: Worldwide distribution of foreign firms in Brazil's top 500 firms by domestic sales (2002) compared to national origins of banks in our sample.
Sources : Brazil. Ministerio da Fazenda. Banco Central do Brasil . Sistema Financeiro Nacional. Informacoes Cadastrais e Contabeis. Instituicoes Financeiras. Bancos Comerciais, Multiplos e Caixa. http://www.bcb.gov.br/?IF (03 May 2004); EXAME Melhores e Maiores . http://portalexame.abril.com.br/ (Oct-Nov 2003), and EXAME Melhores e Maiores , printed edition (2003).
1 Firms which shareholding control has two different origin (being foreign or Brazilian the second shareholding control) were counted as half (0.5), the same was done to their sales.
2 Firms which shareholding control has three different origin (being foreign or Brazilian the third shareholding control) were counted as one third (0.33…), the same was done to their sales.
3 The shareholding control of the ABN-AMRO in Brazil is Brazilian/Dutch. For this reason it appears in this table as being half foreign capital.
Table 2: The locations of the foreign Top 500 firms in Brazil by Country, across the top 30 Brazilian city-regions/detached cities (2002)
Source: EXAME Melhores a Maiores , printed edition (2003).
4 The only Israeli firm in Brazil is in Londrina, one of the cities that do not belong to the sample of 30 Brazilian City-Region/Detached Cities which comprehends more than 96% of the total sales of the Top 500 firms in Brazil, from 1996-2002 (Rossi and Taylor, 2004).
Table 3: Characteristics of the 31 selected banks used in our analysis
Sources : Brazil. Ministerio da Fazenda. Banco Central do Brasil . Sistema Financeiro Nacional. Informacoes Cadastrais e Contabeis. Instituicoes Financeiras. Bancos Comerciais, Multiplos e Caixa. http://www.bcb.gov.br/?IF . Bank web sites.
5 Sao Paulo is an hybrid city in this study. It belongs to the top 25 world cities and at the same time, it is the most important Brazilian city. When linked to the world cities, it operates as an world city. When linked to Brazilian cities, it operates as a Brazilian city.
Table 4: Domestic leading cities concentration strategy (Component I, 23%)
Table 5: World cities strategy (Component II, 23%)
Table 6: Transnational centre/Southern strategy bypassing Brazilian economic core cities (Component III, 12%)
Table 7: Domestic core area strategy (Component IV, 8%)
Table 8: Curitiba new gateway strategy (Component V, 5%)
Table 9: North strategy (Component VI, 5%)
Table A1: Top 25 world cities defined by global network connectivity
Figures in brackets are actual rankings
Table A2: Top 30 Brazilian cities by Top 500 firms' sales (2002)
Figures in brackets are actual rankings
Edited and posted on the web on 21st September 2004; last update 3rd March 2006
Note: This Research Bulletin has been published in Tijdschrift voor Economische en Sociale Geografie, 97 (5), (2006), 515-534