Papers Reg. Sci. 83, 165176 (2004) DOI: 10.

1007/s10110-003-0181-z

c RSAI 2004

Agglomeration and networks in spatial economies

Department of Economics, J onk oping International Business School, SE551 11 J onk oping, Sweden (e-mail: [email protected]) Department of Economics, University of California, Berkeley, CA 94720-3880, USA (e-mail: [email protected])

Abstract. We consider the parallel developments in the economics of agglomeration and the economics of networks. We explore the complementarities between the productivity benets of agglomeration and those of network linkages, arguing that networks of actors dispersed over space may substitute for agglomerations of actors at a single point. JEL classication: D20, L14, R30 Key words: Agglomeration, networks, spatial economics 1 Introduction The modern analysis of economic relationships in urban space emphasises the importance of agglomeration. Of course the concept of agglomeration is itself not new. Informal discussions of external economies of scale can be traced back as far as Marshall (1890), and the concept gures prominently in textbooks written more than fty years ago (e.g., Hoover 1948). The novelty of the external economy lies in our newly acquired ability to model it precisely. This owes much to the insight of Fujita (1988) who demonstrated how the monopolistic competition model of Chamberlin (1933) could be adapted to generate spatial agglomeration of economic activities. These models made it clear that standard market processes based on price interaction alone could generate increasing returns, the external economies of agglomeration. The economic advantages of proximity are now quite explicit. The concept of networks in space has undergone a parallel treatment, as modern methods have made the efciency gains from networks quite explicit. At rst glance, it would seem that network developments are unrelated to agglomeration. After all, agglomeration describes a point while networks consist of nodes, and links connecting these nodes, in order to facilitate transactions among agents. Yet

166

B. Johansson, J.M. Quigley

these latter connections may lead to precisely the same external benets that arise from agglomeration, and for precisely the same reasons. Indeed, networks among economic actors dispersed over space may act as a substitute for agglomerations of actors at a single point, providing some or all of the utility gains and productivity increases derived from agglomeration. This essay hints at the complementarities between agglomerations and networks in providing benets to market participants benets that arise from standard market processes, but which are external to individual participants. Insights about networks rst arose from considerations of the nature of the rm and its natural boundaries. At about the same time, insights about agglomeration rst arose from considerations of the spatial linkages among those boundaries. In the spatial context, networks play a role in facilitating exchange both within and between regional agglomerations. One issue is how this role differs in intraregional and interregional contexts. Another issue is the distinction between the tangible physical networks and the intangible economic networks that include transaction agreements and routinised arrangements (Williamson 1975). Our perspective on networks and agglomeration stresses distinctions between public goods and private capital, and among institutional forms. We suggest that the formation and efciency of agglomeration arise from its character as public capital; households and rms in the same agglomeration share its benets in common. In contrast, an economic network is private capital shared primarily by the network participants. Agglomerations also rely on public institutions, which aggregate individual decisions. In contrast, economic networks arise from a collective decision by group members, generating a private institution. Networks are clubs in which exclusion is possible and price discrimination is the norm. Agglomerations cannot exclude economic actors from receiving benets nor can they price these benets efciently.

2 A partial history 2.1 Agglomeration and markets During the decade after World War I, a consortium of economists at Columbia led by Robert Murray Haig (1926) reported the rst extensive analysis of the colocation of rms and industries in New York. They considered, for example, the garment industry. The pattern of interrm contact among designers, fabricators and marketers destined the industry to remain spatially concentrated, according to the authors, so as to rely upon face-to-face, local contact. When the researchers considered other industries (e.g., cooperage), they concluded that standardisation enabled rms to be widely dispersed throughout the metropolitan region. Presumably, a modern interpretation would highlight the fact that transaction costs between agents were sufciently low for these standardised products, so that proximity and local networks of buyers and sellers offered little advantage over anonymous market transactions in space. In contrast, the complex interactions in the un-standardised fashion industry are more efciently accomplished through local networks.

Agglomeration and networks in spatial economies

167

Three decades later a second burst of empirical analyses of large cities New York again, but also London and Stockholm extended these primitive insights. In New York in the 1950s, a group of Harvard economists sought to project economic and demographic conditions three decades into the future, and this practical objective provided academic researchers with a golden opportunity for intensive study of the fundamental factors affecting the development of industry and the location of economic activity. A central nding of this large-scale study was the importance of external economies of scale, that is, the cost advantages some rms can achieve when they operate in the context of a larger local economy. The summary volume of these studies by Raymond Vernon (1962) includes a chapter devoted to the rise and spread of external economies and to the impact of these externalities on rm location and the well being of central cities. Simultaneously, empirical studies of the Stockholm economy emphasised the economies achieved by rms in the location of their activities in proximity to the sites chosen by suppliers of inputs and purchasers of output. Signicantly, these studies also emphasised the economic returns to the co-location of retail establishments, producer and consumer services, in cities (Artle 1959). The current emphasis on externalities and the growth of urban areas can, to a certain extent, be traced to the stylised facts about Stockholm, London and NewYork uncovered in the late 1950s. These developments emphasised the intra-metropolitan location patterns of activity, the intensity of use, and the contribution of these factors to the growth of cities and metropolitan regions. For example, Made in New York, a compendium of descriptive case studies of manufacturing in New York, includes the following passage (Hall 1959):
Rubbing elbows with others of their kind and with ancillary rms that exist to serve them, [rms] satisfy their variable wants by drawing upon common pools of space, labor, materials, and services. In more concise language, they can take advantage of external economies. The economies are external in the sense that the rm obtains them from outsiders, and they are economies in the sense that the rm can satisfy its variable or part-time needs in this manner more cheaply than it could satisfy them from within. The outsider, in turn, can afford to cater to the rms fractional needs because he also caters to many other rms. The external economy may derive from an electrician or a sewing machine repairman or a free-lance photographer, responding to the call of a rm, which does not need him full-time. ... It may even grow out of a revolving supply of specialized labor, such as garment workers accustomed to seasonal cycles, printers, staff writers, editors, or electronic engineers. Such a supply enables a rm to pick up employees quickly and let them go with equal suddenness, and makes it unnecessary to maintain a stable force of workers for an unstable demand. Thus, it is obvious that external economies reduce the cost of doing business just as labor and transport [savings] do. Indeed, there is no real line of demarcation.

This description seems surprisingly similar to recent economic models in which the production of individual rms is competitive, with constant returns to scale, but there are socially increasing returns as aggregate production rises. In the world of these recent models, investment in real capital can have social benets not reaped by private investors. The private investors are guided only by protability, but the economy of the urban area is more productive due to the external effects arising from the location of investment. The original applications of the modern endogenous growth models emphasised the stock of accumulated knowledge. Ideas can clearly benet others as much or

168

B. Johansson, J.M. Quigley

even more than they benet the inventor of the idea. Knowledge or human capital may be the most important example of the application of the theory of endogenous growth. The real capital stock in cities nevertheless has other important attributes that affect the growth of the economy in analogous ways most especially their capacity for differentiation and diversity. As noted above, the standardisation of barrel manufacturing in the 1920s was associated with its decentralisation to outlying parts of the metropolitan area. A central conclusion of the Hall (1959) volume, published almost 40 years later, detailing trends in the apparel, publishing, and electronics industries was the following:
The chief common denominator in these manufacturing operations that were attracted more strongly to other places than to the [New York] region appears to be standardization. The rest of the country gained relative to New York in products whose specications could be planned in advance with reasonable assurance. Large numbers of identical copies house dresses, magazines, radio sets could be poured out of the plants without making any changes in the design ... But the fact remains that the manufacture of standardized products ... has shown pervasive tendencies ... to prefer locations far from New York.

Of course this statement referred only to a single metropolitan region and only to a small number of industries studied intensely. But within these limits, the evidence showed that rms producing non-standardised differentiated output were more strongly attracted to the urban core than those rms producing homogeneous products. Empirical interpretations of agglomeration economies, including the studies mentioned above, are often incapable of distinguishing between different mechanisms or sources of the agglomeration benets described. There are at least three basic cases to be distinguished although they may be interconnected in many concrete situations. In the rst place, an entire industry may benet from agglomeration, since the size of the agglomeration provides sufcient demand to allow individual rms with internal scale economies to develop differentiated products. Second, an individual rm may benet from the option to buy more specialised inputs at lower transactions costs from differentiated input suppliers within the region. Third, an individual rm may benet from information spillovers outside the market that arise from proximity within an agglomeration. The economic consequences in the rst two cases are generated through the market. The cases are symmetrical in the sense that benets (or externalities) are generated simultaneously on the supply and demand side. In the third case, the productivity of an individual rm or of the entire region increases without affecting the availability of inputs or outputs within the region. 2.2 Networks and markets More than sixty years ago shortly after the inuential work of Haig and his associates analysing spatial location in New York Ronald Coase (1937) suggested a set of criteria to indicate when it is efcient to decompose an organisation into subunits that may interact through a market instead of interacting as part of the same entity. Coases insight did not involve space, but it helped explain why organisations under specied conditions develop as distinct rms, and why rms so integrated can

Agglomeration and networks in spatial economies

169

perform more efciently than subunits that interact through a market. The modern answer is that when interaction via the market makes transaction costs too large, it becomes advantageous to organise interactions as processes taking place inside the rm. To reach this modern formulation, economics had to pass through its transaction-costs revolution in the 1970s, with contributions by Alcian and Demsetz (1972) and Williamson (1979). One implication of this new approach to the analysis of market formation is a distinction among types of products: for some products, transaction costs may be inherently high, while for others these costs may be negligible. An obvious question arises: how will the transaction arrangements vary across products for which transaction costs are high and those for which transaction costs are low? A systematic answer to this question brings us to the analysis of transaction networks. The costs of economic transactions may be categorised as those of exclusion and those of interaction. The latter may be related to describing, inspecting and measuring the object of the interaction. Other aspects of a transaction that affect costs include search, negotiation, contract formulation, legal advice, and documentation. It is evident that if the same pair a buyer and a seller is involved in similar transactions regularly and frequently, the pair will have an incentive to organise the transaction procedures and processes so that costs are reduced. They may routinise this interaction, thus forming a transaction link between them. The buyer and seller represent nodes connected by a specied linkage. This, of course, reects Coases insight about the nature of the rm. When a rm is established with Coasean motivation the action is nothing but the formation of a network internal to the organisation for interaction among a set of interdependent subunits. The internal network is one extreme solution to the exchange problem. The other extreme solution is the pure market interaction with no transaction links. The latter has two components of infrastructure: a system of prices and a system of information dissemination about available options and their associated prices. In between these two solutions intrarm and the anonymous market one can nd agreements and established transaction links between and among actors. These nodes and links may form networks that reduce transaction costs. Their continued existence may reect lumpy investments in transaction links. Outside any network, buyers and sellers can nd each other at low transaction costs in proximity markets, i.e., markets where the distances, perhaps spatial, between buyers and sellers are short. Such markets can be found in urban regions, and they become more ubiquitous as the urban size increases. This is a consequence of the structure of space, and it represents one aspect of the spatial externality described in the agglomeration literature. From the transaction costs perspective, the nature of a transaction link or a more complex economic network is an empirical issue. In general, an economic network is an organisation of interlinked agents combining some features of a rm and of the pure market. It internalises some interaction costs and includes, at least implicitly, contingency agreements of the kind we nd in market contracts. The incentive to form such economic networks stems from possibility to reduce transaction costs. When transaction costs are distance dependent, transaction links have the potential

170

B. Johansson, J.M. Quigley

of overcoming distance. At the same time, there are costs of establishing a network, and these transaction costs may be lower inside an agglomeration. 3 Some theory 3.1 Agglomeration theory The heavily empirical research sketched out above emphasises diversity and heterogeneity. The heterogeneity of products and the diversity of consumers lead to increases in well being. Interesting and powerful economic models of diversity and heterogeneity have been around for about fteen years and are still under development. These models are based upon the Chamberlinian perspective on competition and product diversity developed by Dixit and Stiglitz (1977), and rst applied to space by Fujita (1988). This inuential work considered explicitly the trade-off between the output of goods and their variety. Consider consumers: household well being depends on consumption of traded goods, housing, and a variety of local goods. The markets for traded goods and housing are competitive, while the differentiated local goods are sold in a monopolistically competitive market. If there is less differentiation among local goods, then variety is less important in household well being; greater differentiation means that variety improves consumer utility. Under reasonable conditions, the well being of a household in the city will be positively related to the aggregate quantity of local goods it consumes and to the number of types of these goods, which are available in the economy. Consider producers: the importance of a variety of locally produced inputs operates in a parallel fashion. Suppose that the output depends on labour, space, and a set of specialised inputs. Again, the markets for labour and space are competitive, while the differentiated local inputs are purchased in a monopolistically competitive market. If there is less differentiation among inputs, then variety loses its impact on output; greater differentiation means that variety has a greater effect on output. For example, the general counsel of a rm may operate alone. However, she may be more productive if assisted by a general practice law rm, and even better served by rms specialising in contracts, regulation and mergers. Again, under reasonable conditions, output in the city will be related to quantities of labour, space, and also to the number of different producer inputs available in that city. This analysis yields a remarkable conclusion: diversity and variety in consumer goods or in producer inputs can yield external scale economies, even though all individual competitors and rms earn normal prots. The intuition works in this way: The size of the city and its labour force will determine the number of specialised local consumer goods and the number of specialised producer inputs, given the degree of substitutability among the specialised local goods in consumption and among specialised inputs in production. A larger city will have a greater variety of consumer products and producer inputs. Since the greater variety adds to consumer well being, it follows that larger cities are more productive, and the well being of those living in cities increases with their size. This is true even when all rms in these cities earn a normal rate of prot.

Agglomeration and networks in spatial economies

171

The theoretical perspective outlined above includes two principal models. The rst, and most highly developed, is the core of urbanisation economies. In demand, it emphasises diversity and consumers taste for variety; in supply, it emphasises the productivity of specialised inputs in production. The second model is derived from quite another perspective, distinct from models that emphasise diversity. In this latter framework the focus is on rms or producers, and how their efciency is enhanced by proximity and linkages achieved through agglomeration. At the rm level, the perspective is shifted from input diversity to forward and backward linkages among agents. These linkages may be of a pure-market type or may involve transaction links. These models inspired by Marshall are indeed models of agglomeration. However, these agglomerations representing linkages among rms may provide gains to smaller regions, as well as large urban areas. They can be termed localisation economies as accurately as urbanisation economies. If we consider externalities arising from these linkages, two cases can be distinguished: input-cost externalities and delivery-cost externalities. Both cases emphasise the consequences of proximity for transaction costs. A typical backwardlinkage or input-cost externality arises from a rms spatial location relative to input suppliers, providing inputs with lower transaction costs and potentially at lower prices. The same argument applies to the supply of labour inputs to the rm. The essence is that input supply at short distances reduces total costs of producers, and these linkages form an agglomeration. Producers can be better off locating where input suppliers are clustered. Proximity advantages apply also to forward linkages. With proximity, transaction or delivery costs can be lower inside an agglomeration than they are when the product is delivered to buyers outside the region. Again, rms have an incentive to locate in an agglomeration where the demand from input-buying rms is large, and this reinforces the tendencies towards agglomeration. In summary, within a market, agglomeration provides two sources of efciency gains. In the rst place, a diverse set of products is only exchanged inside the agglomeration, i.e., products whose transaction costs increase strongly with distance. Distance-sensitive transaction costs imply that diversity is fostered in agglomerations. In the second place, transaction and transportation costs are lower with proximity. Firms with forward and backward transactions and transportation cost advantages have incentives to co-locate and to form an agglomeration.

3.2 Network theory The agglomeration economies described above emphasise the Chamberlinian model of monopolistic competition in which diversity plays a central role. The size of a market (or city) determines the diversity of inputs available to rms and the variety of consumer products offered to households. In this framework the externality operates through well-dened anonymous market interactions. A more comprehensive set of external economies, following Marshall, includes both pecuniary, market based and non-market based externalities. Spillovers, a prominent example of the latter, may not be reected in market prices. Are networks

172

B. Johansson, J.M. Quigley

and transactions links un-priced spillovers, inefciently supplied in the market? Not at all. The key to pricing is the transaction costs associated with establishing specic links. Absent transaction costs, Radners (1968) results are transparent. A competitive equilibrium with complete contracts exists for a network economy. This result provides the framework for pricing link-specic transaction costs in a network equilibrium (Nagurney 1999). More complex and problematic network pricing issues arise when links involve set-up costs incurred by transacting parties. In this case, markets may remain incomplete and efcient prices need not be linear. The prevalence of xed costs for transaction links has consequences for dynamic adjustments in the market. Transaction patterns may change slowly, or they may even appear rigid; this provides a motivation for interregional input-output models of trade (Isard 1956). Of course, input-output coefcients and interregional trade coefcients are analytically meaningful only when they are xed or when they change according to some systematic evolution. Thus, the path dependencies of trade patterns and regional coefcients (Sonis and Hewings 1998) may reect the structure of transaction costs. What properties make the links of a network different from the market interaction? Link transactions often involve rms both as seller and buyer, making repeated and similar transactions. Importantly, transaction link partners are identiable and distinct. The interaction is not anonymous, and the agents can take prior interactions into account in each new transaction. Anonymity provides less information. Moreover, for a network, the institutional capital is private, distributed among the participants, each of whom may have made specic investments. The existence of this capital has consequences: once the participants have invested in a network, future interactions are affected by the sunk cost. However, before capital has been committed, the decision to form a link will take capital costs into consideration. Following Williamson, the protability of a linkage is more likely if it is expected to be used during a longer time period, or more intensively. Often the relevant transactions are repetitive variants on a specic kind of interaction. Typical examples are modern production systems organised as supply chains, product assembly with a network for timing and delivery of components, wholesale producers who have links to rms supplying products, and to retailers who are distributors of the nal product. Arrangements like these not only facilitate transactions across regional boundaries, but they also make it possible for rms to make location decisions that reect advantages sometimes dispersion in space, sometimes co-location (Polenske 2002). The formalisation of network problems in spatial economics took two basic forms during the 1950s. One is derived from the regional and interregional inputoutput framework, which has the character of a pure network model in which everything including the structure of prices is determined by xed delivery coefcients. But empirical applications of the interregional model have awkward interpretation. In principle, regions specialise in certain types of products and export excess supply to other regions. However, observations on trade ows invariably indicate that cross-hauling of the same products is the rule rather than an exception.

Agglomeration and networks in spatial economies

173

When the input-output model is applied in the study of international trade, the results are invariably similar. Products, even when nely described, are recorded as twoway ows between nations at the same level of development. This embarrassment gave rise to the so-called Armington assumption a product is always differentiated from similar products by its origin. A similar product produced in and supplied in two different regions cannot be identical (Armington 1969). The second network model that was formalised early is the spatial price equilibrium model (Samuelson 1952). For each product in this model, there is a set of supply regions and a second set of demand regions. The regions (agglomerations) are nodes in a trade network. Associated with each trade link connecting two nodes is a product-specic transaction cost. In equilibrium, a product ows only one way on any trade link. As noted above, this contradicts existing empirical observation that two-way ows are a generic phenomenon. But there is an alternative, modern solution the Dixit-Stiglitz model of monopolistic competition with differentiated products developed in an international trade context by Krugman (1979). With this formulation, two trading regions will quite naturally exchange similar but differentiated products. Note the completely parallel way that network and agglomeration models have developed a common platform. Intellectually, the gulf between international economics and urban and regional economics has, until recently, been large. The new ideas needed a decade or more to become intertwined. In retrospect, of course, these things are clear. Interregional and international trade ows are artefacts of an accounting system for ows. The statistical records of ows represent trade between rms located in different agglomerations or between different establishments of a multi-location rm. These ows arise from long distance deliveries that, to a large extent, are organised in networks. A part of these trade ows reects standardised and homogenous products, such as basic foodstuffs, chemicals, oil, and electricity, which are traded anonymously. But these types of products do not dominate the value of interregional and global trade. Paradoxically, international trade is characterised by differentiation, which reveals a taste for variety. We thus observe large agglomerations of diversied economies trading with other diversied urban agglomerations. This suggests two things. First, urban agglomerations with all their diversity still remain specialised vis-` a-vis each other. Second, two-way ows of diversied products between urban regions provide even greater benets to consumers and producers in urban regions. In this sense, the demand for differentiated products shows no sign of saturation. Trade development inside the European Union conrms these trends: intra-industry trade of diversied products continues to expand, even across large agglomerations (Balassa and Bauwens 1988; CEC 1996). The story could end here. We have seen that agglomeration economies and network economies are two different, complementary ways of solving problems of market exchange. There is a clear relationship between the alternates, and diversity plays a key role. There is, however, another relationship between agglomeration and networks, focusing on the role of spillovers or communication externalities. Again, there are two basic approaches to these externalities, the pure market agglomeration and the pure network solution.

174

B. Johansson, J.M. Quigley

In the course of ordinary transactions between rms, information about technical solutions and product attributes can spill over between them. The rms learn from each other. The very fact of existing private information, which can diffuse among rms, is in sharp contrast to a model where all relevant information is already available. The diffusion mechanism is related to proximity. This implies that rms can benet by clustering together in an agglomeration or an industrial district. Empirical testing and verication often focuses on innovations development activities as distinct from production activities. This is again different from the monopolistic competition model discussed earlier, in which new product varieties are available without cost. Information and knowledge diffuse quite easily among rms belonging to the same transaction network. In this case the spillover may be a by-product of transactions between rms in a network. As a consequence, a link or a network can function as a substitute for proximity in the process of knowledge diffusion. Moreover, networks can be designed to include spillover mechanisms. Thus, network links between rms can develop inside an agglomeration, and the existing literature offers one model depicting explicit linkages between rms in the form of so-called industrial complexes, and another model that focus on social networks related to rms in an agglomerations (Gordon and McCann 2002). Two hypotheses are associated with the knowledge-spillover model (Baptista 1998; Glaeser et al. 1992). The rst is the so-called Marshall-Arrow-Romer proposition that knowledge diffuses between rms within the same industry. Hence, this mechanism can operate in smaller agglomerations. The alternative hypothesis presumes that innovation is especially stimulated by spillovers across industries (Jacobs 1969). In this latter case, large urban regions are not only more efcient than smaller regions, they have an advantage in innovation, and their economies can thus grow faster. The analysis of spillovers due to proximity and spillovers in networks suggests that agglomerations will have a more rapid development of technology, and hence faster productivity growth. But the case is unproved. It may be true that innovations occur more frequently in regions allowing spillovers, but the value of an innovation is reduced when knowledge diffuses quickly to other rms. The current literature on innovation and growth in agglomerations and networks thus presents an unnished and inconsistent picture.

4 Integration and communalities How can the benets of agglomeration be achieved? If economic actors are sufciently close in space, the anonymous market may achieve everything: shared inputs purchased on a market; retail rms separately owned, but located together to reduce shopping costs and to increase variety; efcient forward and backward linkages among rms. At a single point, there is of course no distinction between a network relationship and an anonymous market relationship. All actors can make economic decisions with full information about alternatives, and nodes are simply co-located without linkages between them.

Agglomeration and networks in spatial economies

175

But suppose this spatial proximity cannot be achieved for economic, political or technical reasons. Can any of the external benets be realised anyway? The answer, of course, is that for many transactions, an established network reduces the effective distance between nodes, reducing the transaction (or transport) costs that would otherwise be prohibitive. When co-location is infeasible, networks may substitute for agglomeration. This possibility of substitution means that small regions may survive and prosper to the extent that networks can substitute for geographically proximate linkages, for local diversity in production and consumption, and for the spillouts of knowledge in dense regions. The technical developments, which have facilitated networks, are quite impressive and clear examples of these substitution possibilities abound. Consider the diversity in consumption. In the recent past, outside of francophone countries it required a city of reasonably large size to offer a decent selection of French lms. Now the universe of French language lms is only a keystroke away from any isolated consumer as are out-of-print books, Jam on Ib erico, and participation in competitive chess tournaments. Many of the historical advantages in consumption arising from the specialisation afforded by dense agglomerations, can be achieved by networks in smaller urban regions or even rural areas. Technical advance has meant that transaction conditions have improved over time, and this means that market agents can increasingly substitute agglomeration proximity for network contacts and vice versa. Improvements in technology alter the trade-off between agglomeration and network solutions, and current cost conditions provide a stimulus for efforts to develop new routines. How does this substitution function? Technology permits goods and services to become standardised. Standardisation of complex commodities makes it possible to rely upon network solutions to achieve diversity in consumption and production. The potential for substitution between dense agglomeration and network solutions in facilitating diversity is apparent beyond consumption activities. Analogies in the diversity of producer inputs, in linkages among producers and suppliers, and in knowledge spillouts are ubiquitous. Consider knowledge spillouts and externalities, for example. The ftieth North American regional science conference and the Regional Science Association International (RSAI) being celebrated in Philadelphia in 2003 is a concrete example of the scope for substituting networks for physical agglomeration in diffusing useful knowledge. All the articles in this volume have been produced through collaboration, mostly intercontinental, using informal networks and routinised interactions fostered by modern technology. The RSAI, the club facilitating this exchange, is a network, which reduces the transaction and transport costs of producing knowledge. Information generated in networks of collaborators is diffused in face-to-face meetings, and the spillouts from these activities are inputs into further collaborations using network technology. The emergence of agglomerative economies and the spread of these external economies by networks is the hallmark of regional development in the twenty-rst century.

176

B. Johansson, J.M. Quigley

References
Alcian AA, Demsetz H (1972) Production, information cost and economic organization. American Economic Review 62: 777795 Armington PA (1969) A theory of demand for products distinguished by place of production. International Monetary Fund Staff Papers: 159176 Artle R (1959) Studies in the structure of the Stockholm economy. Business Research Institute at the Stockholm School of Economics, Stockholm Balassa B, Bauwens L (1988) The determinants of intra-European trade in manufactured goods. European Economic Review 32: 14211437 Baptista R (1998) Clusters, innovation and growth: A survey of the literature. In: Swann GMP, Prevezer M, Sout D (eds) The dynamics of industrial clustering. Oxford University Press, Oxford Central European Commission (1996) Economic evaluation of the internal market. European economy, Reports and Studies, Luxemburg Chamberlin E (1933) The theory of monopolistic competition. Harvard University Press, Cambridge, MA Coase R (1937) The nature of the rm. Economica 4: 1316 Dixit AK, Stiglitz J (1977) Monopolistic competition and optimum product diversity. American Economic Review 67: 297308 Fujita M (1988) A monopolistic competition model of spatial agglomeration: A differentiated product approach. Regional Science and Urban Economics 18: 87124 Glaeser E, Kallal HD, Scheinkman JA, Shleifer A (1992) Growth in cities. Journal of Political Economy 100: 11261152 Gordon IR, McCann P (2000) Industrial clusters: Complexes, agglomeration and/or social networks. Urban Studies 37: 513532 Haig RM (1926) Towards an understanding of the metropolis. Quarterly Journal of Economics 40: 402434 Hall M (ed.) (1959) Made in New York. Harvard University Press, Cambridge, MA Hoover EM (1948) The location of economic activity. McGraw-Hill, New York Isard W (1956) Location and space-economy. MIT Press, Cambridge, MA Jacobs J (1969) The economy of cities. Random House, New York Krugman PR (1979) Increasing returns, monopolistic competition and international trade. Journal of International Economics 9: 469479 Marshall A (1890) Principles of economics, 8th ed. Macmillan, London Nagurney A (1999) Network economics: A variational inequality approach. Kluwer Academic Publishers, Boston Polenske K (2002) Clustering in space versus dispersing over space: Agglomeration versus dispersal economies. Paper presented at the Uddevalla Symposium, Sweden (June) Radner R (1968) Competitive equilibrium under uncertainty. Econometrica 36: 3158 Samuelson PA (1952) Spatial price equilibrium and linear programming. American Economic Review 42: 283303 Sonis M, Hewings GJD (1998) Economic complexity as network complication: Multiregional inputoutput structural path analysis. Annals of Regional Science 32: 407436 Vernon R (1962) Metropolis 1985. Harvard University Press, Cambridge MA Williamson OE (1975) Markets and hierarchies: Analysis and antitrust implications. Free Press, New York Williamson OE (1979) Transaction-cost economics: The governance of contractual relations. The Journal of Law and Economics 22: 233261