He gives examples of Italian ceramics and gold jewellery ...... vested interests, supporting the political elite rent-seeking instead of the common well-being, while.
Agglomeration and regional growth policy: Externalities vs. comparative advantages
1. Introduction Supported by many empirical studies (Enright, 1996; Isaksen, 1997; Cooke, 2001; Rosenthal and Strange, 2004; Spencer et al., 2009) both policymakers and development agencies usually assume that industrial clusters play a key role in regional development (European Commission, 2008; Brachert et al, 2011). In fact, the research in economic geography (Rosenthal and Strange, 2003) and regional science has empirically shown that agglomeration has been positively associated with local productivity level, not only in the US (Ciccone and Hall, 1996) but also in Europe (Ciccone, 2002). In spite of the positive correlation between agglomeration and productivity, the rhetoric about regional clusters has been widely adopted in policy circles only since the early 1990s. This rhetoric can be viewed as a mixture of Michael Porter's point of view about what creates competitive advantage for firms and nations, and regional theories on localisation advantages and industrial districts. Although one can trace the origin of this rhetoric in Porter's diamond model1 (originally developed to analyse competitive advantages of nations in international markets) Porter’s idea that ‘competitive advantage is created and sustained through a highly localised process’ (Porter, 1990: 19) has determined a refocusing of competitive advantage from nations to regions. Consequently, in line with the deep discussion that has characterised the literature on agglomeration externalities, Porter’s competitiveness concept has also come to be used to examine regional performance. In the course of this change, the argument that competitive advantage results from characteristics about entire industries in their ‘home region’ has evolved into a universal policy therapy that assures sustained growth to any locality or region (Perry, 2010a, 2010b). In fact, the Porter-inspired cluster developments (e.g., Porter, 1998, 2000) gave policy professionals a rationalization for local intervention and often for building ‘Silicon Somewhere’ (Hospers et al, 2008). Policy-makers in many countries and regions view this validation as advice to combine cluster support with any type of intervention, and Porter’s consultant work, alone or in association with his Monitor Company, has contributed to the wide 1
Porter's diamond model considers the following as the most important factors for explaining the competitive advantage of nations: i) the context for firm strategy and rivalry; ii) demand conditions; iii) factor conditions; and iv) related and supporting industries.
1
diffusion of cluster strategies in many European countries (Benneworth et al., 2003; Rosenfeld, 2005). In this paper, we use Porter's definition of clusters as ‘geographic concentrations of interconnected companies, specialized suppliers, service providers, firms in related industries, and associated institutions (e.g. universities, standards agencies, trade associations) in a particular field that compete but also cooperate’ (Porter, 2000: 15). This concept doesn’t incorporate all the richness that has emerged in the two last decades about clusters (see, for instance, Saxenian, 1994; Markusen, 1996; Cooke, 2002; Martin and Sunley, 2003; Brachert et al, 2011). But, even though simplistic, it is sufficiently broad and pragmatic for including the main characteristics of the three ‘ideal-typical models’ of industrial clusters presented by Gordon and McCann (2000: 513)2. The essentially deductive approach of Gordon and McCann (2000) complements the Markusen’s (1996) inductive research about ‘slippery space’. Both approaches show that clusters reflect not only economic responses to the pattern of available opportunities and complementarities, but also the level of embeddedness and social integration. So, while different in method and focus, both the ‘ideal-typical models’ of the former and the four types of ‘sticky places’ of the latter are difficult to be conceived without understanding the characteristics of the regions (and the concept of region), where the cases studied are embedded. In regional science, there are many concepts of region (Bailly, 1998): natural region, homogeneous region, historical region, functional region, etc., but a quick look at the long run evolution of geography (Freeman, 1961) reduces the plurality of concepts to three distinct perspectives. The first view is formal. It considers regions as organic entities, i.e. as natural phenomena representing the territorial expression of long-standing interaction between particular human populations and the lands they occupy. Such a view was dominant before the 1950s and provided the basis for the so-called regional geography (Freeman, 1961). The second and more recent view sees regions mainly as a descriptive tool defined according to specific criteria. In this scheme, a particular approach is to identify regions according to their function, thus distinguishing functional regions from the formal regions mentioned above. A functional region is one that displays a certain functional coherence, or a system of interactions when defined against certain
2
Gordon and McCann (2000) suggest three basic models of cluster processes: agglomeration economies, industrial complex, and social networks. Each of these models produces cluster benefits in very different ways. But, as these authors have recognized (2000: 528), ‘actual clusters may contain elements of more than one type’.
2
criteria. They are often described as nodal regions composed of heterogeneous units and populations (typically a network of towns and dependent smaller communities) often identified, or confined territorially, by the pattern of flows of goods, services and people (Karlsson, 2008a). Both the functional coherence and the system of interactions, keys in the functional region, are significant for delimiting (and distinguishing between) different types of clusters as well as for defining the commonalities where a particular cluster is based on. So, it seems that the functional concept of region is the most appropriate for examining the geographical dimension of clusters, and this is even more evident if the definition of clusters depends on a limited number of characteristics. For instance, if we consider that the crucial characteristic of a cluster is the bulk of advantages that spill out from a common labour market, it is not unrealistic to consider that this cluster will be limited by the same frontiers that confine the (functional) region centred on this labour market. However, real-world clusters are not based on only a particular characteristic, and the broad cluster concept we use in this paper points to another direction. In fact, somewhat intermediary between formal and functional regions is the notion of programming, or administrative regions (Stilwell 1992). This concept provides a more pragmatic view of regions, recognising the fact that economic and social institutions tend to operate within administrative boundaries. These represent the limits of governance, and habitually enclose the structure for strategic decisions and for distributing services. It is in this pragmatic perspective that the term region is used in this paper. Although the admiration for clusters, devoted by policy circles, may not be fully shared by the research community, in which some authors consider the popularity of the cluster policy essentially as a result of the use of techniques of ‘brand management’ rather than as a genuine intellectual discourse (Martin and Sunley, 2003, Perry, 2010b), an increasing interest in clusters is undeniable as two handbooks on the matter demonstrate (Karlsson, 2008a, 2008b). This interest was supported not only on some literature that has highlighted and developed a set of decade-old ideas, but also on the discovery by the new (or endogenous) growth theory (Romer, 1986) of how to deal with externalities and increasing returns in economic growth models. The extension of the endogenous growth theory to regional economics (Johansson, Karlsson and Stough, 2001) as well as the appearance of the new economic geography and its developments (Krugman, 1991a, 1991b, 2011; Scott, 2004) contributed to giving an additional force to a regional cluster growth policy, which emphasizes agglomeration advantages over alternative modes of industrial organization. While there is evidence that such advantages exist
3
(Rosenthal and Strange, 2004), from a regional policy standpoint a question is mandatory: what is the appropriate policy for enforcing the clustering development and thus promoting regional economic growth3? The traditional answer corresponds to the classical optimal-policy perspective (Rodriguez-Clare, 2007), i.e., to provide a production subsidy to firms generating externalities, with the subsidy adjusted for equalizing the strength of the externality. However, this is a very demanding solution since it is extraordinarily difficult, or even impossible, to compute the exact power of externalities. Given this difficulty, a mitigated version is usually followed: the presence per se of externalities is taken as a good and enough indication to advise public intervention and, accordingly, to support the industries that are likely to produce positive external economies. Another possible answer to the above question is based on Porter’s policy prescriptions4. But, as Woodward and Guimaraes (2009) point out, there are no known cases where regions or countries have explicitly followed these principles, instead of the industrial targeting associated to the classical optimalpolicy perspective. This paper deals with the above-mentioned traditional answer and adds to the existent literature in three ways. First, it contributes to a better characterization of dynamic externalities by extending the ‘advantages of backwardness’ approach, formerly developed to explain the catching-up of national economies, to the regional context. Second, the paper presents a new model that relates static and dynamic externalities to the clustering mode of production in the regional context. Third, the paper contributes to giving a more accurate theoretic basis to the regional policy and to helping policy makers in choosing the right policy for increasing the well-being of depressed regions. This is also noteworthy because although policy measures have frequently been applied to clusters, the literature about cluster policies is controversial, with its conclusions oscillating between superfluous (Maskell and Kebir, 2006), negligible (van der Linde, 2003) and playing a crucial role (Brenner and Mühlig, 2007).
3
Of course, cluster development is not the only option for local and regional development (Karlsson, 2008b). There are other alternatives to cluster policies for stimulating regional economic growth and it has been alleged that many of them are also responsible by positive externalities: investments in machines and equipment (DeLong and Summers, 1991), infrastructures (Barro, 1990), human capital (Lucas, 1988) and R&D (Romer, 1990). The question of finding the most effective of these policies in pulling the economic growth is empirical in nature, and is therefore beyond the scope of this paper. 4 These prescriptions are synthesized in Woodward and Guimaraes (2009): i) supporting the development of all clusters, not choosing among them; ii) reinforcing established and promising clusters rather than attempting to create entirely new ones; iii) cluster initiatives are advanced by the private sector, with government as facilitator; iv) development should not be guided by top-down policy strategies.
4
The results of this paper have clear implications in regional development, by guiding policy in choosing the most appropriate strategy for facing some frequent dilemmas, as for instance the one embodied in the following question: must policy be focused on the creation of new clusters in activities that have proven to have large positive external effects elsewhere or, conversely, must policy be focused on developing the traditional activities in the region, which allegedly have shown lower externalities? So, the concept of externality is key in the architecture of the model used. Accordingly, the remainder of this paper is organized as follows. Section 2 deals with the externalities associated to industrial agglomerations. Section 3 considers the existence of dynamic externalities and relates them to both localized learning and the advantages of backwardness. Section 4 uses a model that includes static and dynamic externalities associated to agglomeration in order to draw lessons for guiding clustering policy. Section 5 discusses some results of the model. Finally, section 6 concludes.
2. Agglomeration and intra-regional externalities David and Rosenbloom (1990) ask why people and firms tend to congregate spatially. The answers can be done following many theories and approaches ranging from the ‘classical’ location theory to more heterodox views, such as biological and ecological models5, or the agent-based model (Otter, van der Veen and de Vriend, 2001). The latter model is different from the neoclassical model of studying human behaviour, where individuals (in firms and households) are assumed to show optimising rational behaviour based on perfect information. In agent-based models, individuals do not display optimising behaviour, but satisfying behaviour (Simon, 1997). This implies that economic agents have preferences for specific locations due to their perception of reality, as well as due to expectations, ambitions, beliefs and hopes. In this perspective, decision rules are based on preferences for specific locations, and these preferences reflect the characteristics of the economic agents. Respecting to households, the characteristics of their agents, which can change over time, include not only household income, age structure, presence of children, etc., but also perceptions, myths and meaning. For each household agent, these characteristics are translated into both preference for employment, neighbours, service levels, environment and visibility. In the same way, each firm has its own characteristics by which it can be 5
For a review of the theory of firm location decisions and the dynamic of industrial clusters, see Maggioni (2002, especially chapter 3). Trying to give some order to the ‘conflicting conceptualisations’ and ‘generated ambiguity’, Karlsson (2008a) reviews the models and origins of clusters.
5
recognized. This approach differs from the traditional models in which firms locate where expected profits are maximal, or where costs are minimised. In the agent-based model (Otter, van der Veen and de Vriend, 2001), the main characteristics of the firm are visibility and the sector to which it belongs, although other factors related to concepts of attraction and repulsion also contribute to the firm location behaviour6. Whatever the approach, a fundamental reason is recognized: individual companies and households benefit from the high spatial concentration of people, businesses and infrastructure, and such benefits usually have the form of positive externalities, although there are examples where agglomeration can cause negative effects (Feldman, 2000). The external effects (positive or negative) may arise because of both pecuniary and technological externalities (Scitovsky, 1954), the former operating either via the market through price formation, or by means of transaction links (Johansson, 2005). Technological externalities are also known as knowledge spillovers (Krugman, 1991a; King et al, 2003) or communication externalities (Fujita and Thisse, 2002). While technological externalities are often considered black boxes that aim at capturing the key role of complex nonmarket institutions (Saxenian, 1994), the origin of pecuniary externalities is clearer because they focus on economic interactions mediated by the market7. In particular, ‘their impact can be traced back to the values of fundamental microeconomic parameters such as the intensity of returns to scale, the strength of firms’ market power, the level of barriers to goods, and factor mobility’ (Fujita and Thisse, 2002: 9) Since recent research puts knowledge and related spillovers on the forefront, it is essential to discriminate between trade in knowledge and knowledge spillovers. While trade in knowledge refers to formal flows of knowledge through market transactions, spillovers denote the free flow of knowledge. If any embodied knowledge (whether in labour, or in some good or service) is traded consciously between the two sides of the market; as long as it is not entirely paid for, some externality may exist, but of a pecuniary kind. In this case, knowledge flows as a by-product of normal commercial transactions. ‘Pure’ knowledge spillovers can take place only within the dominion of trade-unrelated personal
6
A convenient assumption used in monopolistic competition models consists of considering that firms prefer localizations near households and vice-versa, as in Fujita (1988). 7 For a deeper discussion of the character and role of pecuniary and technological externalities, see Johansson (2005). David and Rosenbloom (1990, p. 349) explore, in a stylised approach, agglomeration pecuniary externalities ‘that tend to reduce the prices at which primary inputs can be purchased as more and more of those inputs come to be assembled at the locale in question’.
6
communication, or through reverse engineering whatever its object is (i.e., reverse engineering of manufactured goods, studying technical documents, such as patents, etc.). Among theories that highlight, in different ways, the importance of the local environment for economic transformation and growth, some can be catalogued as belonging to ‘regional cluster theory’. However, a number of basic concepts are common to most other theoretical perspectives, which have a past well before the cluster’s fashion. The agglomeration externalities constitute the basis of one of such perspectives8 since Alfred Marshall coined the ‘industrial atmosphere’ concept (Marshall, 1920), by describing how companies get advantages as a result of being located in close geographic proximity to other businesses. From Ohlin (1933) on, benefits derived by firms in a particular industry from locating close to each other are termed localisation economies, while the gains obtained by firms of several industries by locating in the same area or city are named urbanisation economies. Localization economies are advantages that result from a spatial concentration of companies operating in the same industry or conducting similar types of activities. Literature usually emphasizes the following positive effects of localization: companies in similar industries benefit from co-location due to the creation of a regional pool of specific inputs as, for instance, creation of a regional pool of specialised and experienced manpower; exchange of knowledge, and collaboration between companies along a product's value chain; better access to the market for goods and to suppliers, and easy flow of technology know-how (Marshall, 1920). Additionally, small firms localized in the same territory can achieve economies of scale that would otherwise only be accessible to large organisations; transforming a large lump-sum investment (Schmitz 1995) into many small investments and, thereby, lowering capital entry barriers (Ruan and Zhang, 2009). But the literature also recognizes that localization can be a source of lock-in effects, which may happen, for example, as a result of a too introverted aptitude (Grabher, 1993). On the other hand, urbanization economies are agglomeration advantages that arise in large cities as a consequence of their rich economic environment, or simply because of their size. In these milieus, a multiplicity of different actors can share access to advanced infrastructure, highly skilled workers or specialised services, which are all to the benefit of businesses in many different industries. However, negative effects have also been identified: e.g., higher property and land prices, pollution and congestion from the use of infrastructure, as well as the higher cost of living that boosts salaries (Glaeser and Maré, 2001).
8
For a discussion of agglomeration economics, and specifically of the importance of clustering and agglomeration, see McCann (2008).
7
The main source of the above agglomeration externalities is proximity (Johansson, 2005). More recently Jacobs (1969) added variety to proximity as a new source of benefits, giving rise to the concept of Jacobs’s externalities. These are economies external to the firm coming from the local variety of industries (Jacobs, 1969, 1984). They are a variant of urbanisation externalities, which place the focus on a region's economic variety (the presence of many different industries, for example) showing that different industries complement each other in the creation of innovations. Many different industries in one region may benefit young companies in their ability to innovate (Duranton and Puga, 2001), as young companies can gain inspiration from other industries for solving their problems. Speeding the flow of ideas (Glaeser and Gottlieb, 2009) and increasing innovation, which results from technology linkages among related industries (Scherer 1982; Feldman and Audretsch, 1999). On the other hand, in cases of too fragmented small sectors, there is a risk that the support functions, such as specialised services, targeted infrastructure initiatives or business policy, will also become too fragmented to be effective. Because the external economies often result from commonalities in inputs (including technology and human capital) rather than from similarities in products, researchers usually include in localization externalities the cases where industries are not only simply similar but also related. However, from a policy point of view it is helpful to distinguish the purely same-industry from related industries with different final markets. This is the main reason for considering a fourth type of agglomeration externalities: ‘the related-variety benefits’ (Frenken et al, 2007). This specification builds on the distinction originally made by Richardson (1972) between similar vs. complementary products — the latter being goods produced for a different market but sharing similar production processes, or other forms of common input markets. The related variety benefits9 are a mix of Jacobs and localisation externalities. In fact, related variety stimulates Jacobs’s externalities and thus fosters economic growth and employment. If industries are related, the likelihood of a successful cross-pollination of ideas increases. If a region is home to many actors in related industries, this can lead to more ideas being spread among industries than if they are unrelated (Frenken et al, 2007). Firms within the industries use similar types of knowledge, or similar types of production technology (or both). For example, companies in the chemicals and pharmaceuticals sectors may
9
While the expression ‘related industries’ is used by many authors (Scherer 1982; Morris, 1990; Feldman and Audretsch, 1999; Porter, 2000) with different meanings, the ‘related variety benefits’ is used in this paper as derived from ‘industry relatedness’ with the meaning given by Neffke and Henning (2009): the extent to the knowledge and skill base of two industries overlap.
8
largely use workers with the same skills and so a region with tradition in chemistry can have an advantage in establishing a pharmaceutical cluster. However, if prevalent knowledge is used routinely there are scarce hypotheses for the emergence of new types of knowledge. Concerns with ‘related variety benefits’ have contributed not only to the emergence of the evolutionary economic geography but also to a better understanding of the benefits of agglomeration, by focusing on both cognitive proximity and complementarity (Boschma and Frenken, 2011). According to Boschma (2009), related variety is crucial to regional growth because it induces knowledge transfer between complementary sectors at the regional level. When Porter (1990) adds competition to proximity another source of intra-regional externalities becomes visible. Porter states that knowledge spillovers in geographically concentrated industries stimulate growth. He insists, however, that local competition, as opposed to local monopoly, fosters the pursuit and the rapid adoption of innovation. He gives examples of Italian ceramics and gold jewellery industries, in which hundreds of firms are located together, and fiercely compete to innovate since the alternative to innovation is end. Porter's externalities are maximized in environments with geographically specialized, competitive industries (Glaeser, et al., 1992). Although localization and urbanization economies are at the basis of posterior developments (Jacobs’s externalities, the related variety benefits and Porter’s externalities), there is a significant difference in the way they affect regional economy. While localization and urbanization externalities explain how natural resources or transport advantage may determine the location of industries, Jacobs’s externalities, as well as related variety benefits and Porter’s externalities explain how regional economy can evolve. Table 1 Sources and types of externalities Spatial origin
Intra-regional
Type
Effects
Proximity
Localization Urbanization
Static
λi
Jacobs Related variety Porter
Dynamic
Zijt
‘Advantages of backwardness’
Dynamic
zˆ
Proximity + variety Proximity + competition
Interregional
Symbol in the model
Source
External
9
Table 1 presents types of externalities classified according to other three criteria: spatial origin, source and effects. Certainly, there are many other types of externalities and many ways of classifying them, but we are not concerned with an exhaustive classification of external economies10. Our focus is only on the distinction between static and dynamic economies, in order to justify the way they will enter in the model developed in section 4. Although there is a consensus about the existence of externalities and that their leading role for industry-level agglomeration, agreement as to the interpretation of empirical results is less visible. While some researchers regard externalities as impressive and abundant (Rosenthal and Strange, 2004), others are apprehensive about the possibility of measuring the effects of different externalities. The disagreement has many reasons: a) differences in methodological design and in the type of selection (De Groot et al., 2008); b) companies benefit from some type, but not from every type, of externalities (Neffke et al., 2008); c) the impact of different types of externalities seems to change with the development phase of the industry (Potter and Watts, 2011), for example, localisation externalities look more valuable for mature and well-established industries while Jacobs’s externalities are more beneficial for young industries in dynamic development stages (Henning et al., 2010). While more empirical research is needed in order to disentangle the relative effects of different sources of externalities, however, from the standpoint of the theoretical foundations of cluster growth policy, all externalities associated to industrial agglomerations must be considered. This does not mean that all externalities must enter autonomously in the economic model; some should enter under some aggregate concept. For the sake of simplicity, the types of externalities presented in table 1 will enter in the model of section 4 as aggregates of SE (static externalities) and DE (dynamic externalities). The division between SE and DE has some similarity with the one presented in Johansson (2005) under the consequence criterion. While Johansson distinguishes between efficiency and development externalities, we name them as static and dynamic externalities, respectively. We consider SE the externalities that affect the TFP (total factor productivity) of firms only by increasing the efficiency of the technologies already in use. This increase in efficiency usually comes from a reduction of costs caused by concentration, such as reduced transportation and transaction costs for intra-organisational exchange and access to external markets. But external economies can influence the development and the relative well
10
This also explains why the relevant distinction, at the micro level, between external economies of scale and scope (Goldstein and Gronberg 1984; Parr 2002) is not considered in table 1.
10
being of different regions, by affecting TFP in another way: by increasing the innovative capacity of firms. That is, they can act dynamically. When this happens, literature usually names them as dynamic externalities (e.g., Glaeser et al., 1992; Ketelhöhn, 2006) or development externalities (Johansson, 2005). From a policy perspective, it is necessary to distinguish between SE and DE.
3. Dynamic externalities Although a consensus exists around the idea that DE are strongly associated with knowledge spillovers (Rosenthal and Strange, 2003), the literature about DE is markedly imprisoned in controversies and conflicting results (Feldman, 2000; Beaudry and Schiffauerova, 2009; van der Panne, 2004). Beginning with the MAR11 (Marshall-Arrow-Romer) vs. the Jacobs’s externalities controversy (Glaeser et al., 1992; Henderson et al., 1995), it soon derived to the specialization vs. diversity debate (Harrison et al., 1996; Henderson et al., 1995; Baptista and Swan, 1998; Feldman and Audretsch, 1999; Neffke et al., 2011), and to the competition vs. monopoly question (Porter, 1990; Glaeser et al., 1992; King et al., 2003; Ketelhöhn, 2006). This paper takes a different perspective, considering, on the one hand, the sources and effects of knowledge transmission and, on the other, distinguishing between the intra-regional and the interregional origin of externalities.
3.1. Intra-regional links, localised learning and competition Many studies about clusters state that they generate knowledge spillovers and enhanced innovation (Camagni, 1991; Cooke, 2002; Maskell and Malmberg, 1999; Porter, 1998). Although Tödtling et al. (2009) consider the channels of knowledge transmission as often remaining unclear, there is evidence that they are multiple and varied: the study of patents or scientific articles (Jaffe et al., 1993), the observation and imitation of competitors (Malmberg and Maskell, 2002), the development of spin-offs or the relocating of qualified labour (Keeble and Wilkinson, 2000). Ketelhöhn (2006) classified the sources of knowledge spillovers in four main categories: buyers (Leslie, 2000; Morris, 1990; Smith, 2000),
11
The term MAR (Marshall-Arrow-Romer) externalities, which relates to technological spillovers between the firms within an industry, was coined by Glaeser et al. (1992) when they added the Arrow’s (1962) formalization of learning and the Romer’s (1986) contributions regarding the impact of dynamic knowledge accumulation to the Marshall’s (1920) seminal idea of localization economies. Storper (2009) disputes the MAR concept, arguing that the true Romer’s sources of increasing returns are not in essence local.
11
suppliers and related industries (Macher et al., 1998; Morris, 1990), competitors (Braun and McDonald, 1978), and academic institutions (Braun and McDonald, 1978; Morris, 1990; Saxenian, 1994). All the above channels and sources have a common trace: the learning capability of persons and firms. Since the learning capability is not randomly distributed across territories12, it is possible to explain DE as resulting from processes of localised learning (Maskell et al., 1998; Maskell and Malmberg, 1999; Malmberg and Maskell, 2006). So, for designing and implementing an effective regional policy, it is necessary to take into account not only realities described by abstract concepts such as Jacobs’s externalities and related variety benefits, but also the specificities of knowledge dissemination. Particularly in this context, policymakers must consider the relative importance of the two dimensions of knowledge: the tacit and the codified (Polanyi, 1967). The difference between these two types of knowledge is well known. While tacit knowledge is based on practice, and so it is difficult or impossible to codify, codified knowledge can be easily formulated, for instance, in designs, texts, or mathematical formulae. In regional innovation research, the tacit dimension of knowledge is usually emphasized, as it may explain how firms can benefit from regional co-location (Gertler, 2003). In fact, proximity and face-to-face contacts are particularly important in transmitting tacit knowledge. Where the proximity is high, it is likely that tacit knowledge spills over to neighbours. Furthermore, proximity increases the strength of linkages between people and organizations13. However, the key point here is that these dimensions are not substitutable. On the contrary, they are complementary: in principle, the application of any codified knowledge requires a degree of tacit knowledge, although the proportions between them can vary according to the regional specificities, as the knowledge bases approach has highlighted (Asheim and Gertler, 2005). Localized learning can occur between many entities and through multiple processes, such as learning by doing, learning by using, and learning by interacting. It involves three facets: horizontal, vertical and multi-level learning. The horizontal learning is performed between firms mostly belonging to the same industry and providing similar goods and services. At this level, the relationship between firms is a 12
The reasons why some regions have more learning capability than others are varied: a) some types of activities can be more prone to knowledge spillovers than others; b) regional clusters with strong industry-institutional linkages can increase the region’s capacity to absorb knowledge spillovers; c) the embodied capabilities of the labour force are not the same across regions, given not only different levels of formal education but also the industry-specific tacit knowledge; e) different regional endowments of entrepreneurial talent. 13 These are some reasons why one can consider the ‘death of geography’ thesis as ‘exaggerated’ (Morgan, 2004), although proximity must not be considered only in terms of geographical distance (Boschma, 2005).
12
composite of competition and cooperation. While competition focuses on the common raw material, labour force and goods market, cooperation focuses on the creation of a common market, the establishment and maintenance of a common brand, and so on. As the similar production conditions and the existence of a ‘common language’ benefit communication and knowledge transfer, the learning process consists of comparing, observing and imitating each other. Vertical learning refers to forward and backward interactions between firms located in different points of the value chain. Clustering firms can acquire timely market information by keeping a close contact with users (Malmberg and Power 2005). Sophisticated buyers especially ask for superior quality and high reliable production, which contributes to designing and upgrading production. On the other hand, backward learning helps firms to acquire complementary technology to upgrade their design and to profit from the effects of their provider’s R&D. In clusters, owing to long-term cooperation and high trust, the clients and suppliers are able to communicate with each other widely and freely, which facilitates the exchange of open information and helps to solve common problems. Multi-level learning refers to the interactive learning between firms and other local actors, such as the local governments, universities, public research institutes and other organizations. These organizations provide local firms with all kinds of services and infrastructures, which promote knowledge-sharing and cooperation. This paper terms ‘localized learning externalities’ all knowledge spillovers that result from the relationship between a specific firm and any other entities (firms, universities, public R&D institutes, etc.) belonging to the same cluster within a region. This concept includes not only learning that takes place through inter-organizational relations but also other pertinent aspects as learning through labour mobility (Almeida and Kogut, 1999; Cooper, 2001; Maskell and Malmberg, 1999; Power and Lundmark, 2004), spin-offs (Boschma and Wenting, 2007) and local buzz (Bathelt et al., 2004; Bathelt and Schuldt, 2010). Another type of intra-regional DE is associated to competition through the concept of Porter’s externalities, which are growth-stimulating knowledge spillovers occurring in geographically concentrated industries. But there is a fundamental difference between the ‘localized learning externalities’ and Porter’s externalities: the instrument for increasing productivity. While the former rely on learning, the latter consider innovation as the driver for increasing productivity14.
14
According to Porter (1990), it is local competition, as opposed to local monopoly, that promotes the pursuit and rapid adoption of innovation.
13
3.2. Inter-regional links and the advantages of backwardness Since people and firms can learn from many sources, it is probable that in open economies regional or national borderlines be no absolute hindrance to learning. Of course, the result and the speed of learning vary with the type of knowledge: while the flows of codified knowledge are often deferred, tacit knowledge faces more crucial impediments. But, in open economies, there is always the opportunity for learning from other sources: learning by using, voyaging, etc. That is, the existence of interregional or international spillovers must also be considered (Rosenthal and Strange, 2003; Coe and Helpman, 1995; Coe et al., 1997): although the economy (national or regional) where the knowledge originates is on the point of benefiting more and sooner, other economies are likely to benefit from spillovers, too. Indeed, the process of economic development can be analysed by focussing on changes occurring in the economy’s industrial structure at the same time as its GDP increases. This is the driving idea of the structural approach to economic development, as well as also of the approach known as the ‘advantages of backwardness’, following the leading work of Abramovitz (1979, 1986) and Abramovitz and David (1995). This approach, also known as the ‘catching-up hypothesis’, states in its simplest form an inverse association between the initial productivity level of an economy and its productivity growth rates in the long run15. It is the existence of a technology gap between economies that allows profiting from advanced technologies without the cost of inventing them (Nelson and Phelps, 1966; Fagerberg, 1987; Fagerberg and Verspagen, 2002) and, since laggards can have access to technologies that have already been employed by the technological leaders they can make a larger productivity jump. Although Abramovitz’s (1986) analysis goes beyond this simplest version, one can take the ‘technology gap’ approach as a basis to show how knowledge spillovers can occur in an international (and interregional) context, and to introduce them in the model of the next section. The ‘advantages of backwardness’ theory was initially developed for a national situation inserted in an international background characterized by a ‘leader’ country and other ‘follower’ countries. Yet one can adapt it to a regional context with few small adjustments16. Indeed, if a laggard region is not totally
15
The negative correlation between the initial productivity level and productivity growth rate is also stated in the neoclassical growth theory (see Barro and Sala-I-Martin, 1992), however the ‘advantages of backwardness’ approach calls attention to other factors that are absent from the neoclassical theory. 16 One of such adjustments is related to the mobility factor: labour mobility is higher across national regions than across countries.
14
closed, it can enjoy four advantages in the growth potential. First, differently from a leader region, which already uses state-of-the-art technology, in a laggard one the tangible capital is likely to be technologically obsolete and so, when the latter expands or replaces its capital stock, the new equipment can embody up-to-date technology. A similar rationale applies to potential advances in disembodied technology, and to the non-technological innovations (new forms of industrial organization and managerial practices, routines of purchasing, production and merchandising, etc.), of a laggard region. Knowledge spillovers play a decisive role here. Furthermore, in laggard regions the low levels of capital per worker tend to increase marginal returns to capital and, so, to promote higher rates of capital accumulation, especially if the modernization of capital stock is considered. Additionally, because laggard regions often maintain relatively large numbers of redundant workers in farming and petty trade with particularly low levels of productivity, the productivity growth can occur by shifting labour from agricultural to industrial jobs and from selfemployment and family shops to business firms, even taking into account the cost of the additional capital that might be necessary to maintain productivity levels in the new occupations. Finally, the relatively rapid growth, resulting from the first three sources, moves towards fast growth in the scale of markets. This promotes technical progress, especially one that is dependent on larger-scale production. Inside the boundaries of laggard regions this sort of technical progress can disguise the absence of technological efforts to create new knowledge through R&D activity. This is also a fertile soil for knowledge spillovers to play a significant role. To sum up, if a poor region trades with a rich economy, economic growth in the poor one can increase, not only through the effect of the production factors and technology used, but also through the occurrence of static and DE, the latter operating through two basic ways: by localized learning and competition, which are external to the firm but internal to the region, and as a result of the ‘advantages of backwardness’, which increase the propensity for profiting from the knowledge spillovers, external to both the firm and the region. Of course, the poor region cannot benefit from knowledge spillovers from the rich region if the former does not have an adequate level of (or does not invest in building) relevant absorptive capacity (Cohen and Levinthal, 1990; Nooteboom, 2000), which firstly is human capital (Nelson and Phelps, 1966).
15
4. The model This section presents a model that demonstrates the action of different types of externalities, and that allows to determine the equilibrium conditions in a poor open region. It is inspired in the standard model of a small open economy that seeks a development strategy (see Ros, 2000) and in which one of the sectors exhibits externalities. Our model also follows the standard result of Corden (1997) by questioning the appropriateness of import substitution strategies in the presence of externalities. However it goes beyond that standard literature in two aspects. First, as in Rodriguez-Clare’s (2007) model, it assumes that externalities are not intrinsic to specific sectors, but rather that they arise from the way in which production is organized: externalities are related to the mode in which goods are produced. This captures the idea that what matters is not what firms are producing, but how they use specific factors in production (Porter, 1990, 1998). Second, the model adapts the above mentioned standard literature, developed for explaining international trade policy, to the relationship between two regions with different levels of development. The modelling follows the stylized form used by several authors in different fields (e.g., Murphy, Shleifer and Vishny, 1989; Rodrik, 1996; Rodriguez-Clare 2007) and is based on the following assumptions: 1) There are two regions, R (rich) and P (poor), indexed by j, one factor of production, L (labour), in fixed supply, and two sectors each one producing only one good i, with i = 1, 2. 2) Both goods can display SE (static externalities), not necessarily in the same degree. This captures the idea that SE are not an automatic result of the type of sector (advanced or backward) but that they depend on the characteristics of the regional milieu. This permits to focus attention on modes of production as the crucial sources of externalities, instead of on the characteristics of goods or sectors. Accordingly, each good can be produced using two possible MoP (modes of production), which are called Cl and Is, ‘clustering’ and ‘isolation’, respectively. These MoP differ in the extent to which they generate externalities: Only the Cl mode produces externalities. These can be localization externalities or / and urbanization externalities, as defined in table 1. 3) There may be exogenous productivity differences across R and P regions in the production of good i (controlled by the productivity parameter
yij ). This exogenous productivity parameter yij is independent
of the mode of production used and, consequently, independent of the SE. The Is mode of production has
16
labour productivity yij. That is, if there are no aggregate externalities, good i is produced with constant returns to scale: a unit of labour produces
yij . So, yijIs = yij .
4) Although good i is produced with constant returns to scale at the firm level, the use of the Cl mode of production makes SE appear and, consequently, an increase in the labour productivity, which, in steady state, equals to:
yijCl = yij λi With the term
λi >1
(1)
representing the maximum benefit of clustering in sector i. It captures the static
external economies in the form of both localization and urbanization externalities. 5) As in similar models, no labour mobility and no transport costs are assumed. The habitual assumptions respecting to the microeconomic behaviour of households and firms are followed: preferences are complete, reflexive and transitive. Since it is supposed that preferences satisfy the Inada conditions, any equilibrium must have positive production of both goods. 6) For convenience, goods are ordered in such a way that
y 2 R / y 2 P ≥ y1R / y1P , so that R has a natural
comparative advantage in good 2. To simplify the exposition, the possibility that the static benefits of clustering are decreasing in i is excluded. This means that
( y iR / yiP )λi is higher for good 1 than for
good 2. That is, the possibility that the sector in which R has a comparative advantage be the one with much lower capacity of generating externalities is excluded. 7) Because it is assumed that P is ‘open’ and ‘small’, the international prices can be derived from the equilibrium of R region, as if the latter were a single economy. If labour in R is chosen as the numeraire, international prices will simply be given by the requirements of labour units in R. So the prices will be:
pi* =
1 1 * if there is no (static) externality and pi = , if there is a SE. The latter shows that y iR λi y iR
clustering causes a lower international price of good i, eroding in this way the benefits of the Cl MoP. Theoretically, the R economy can opt either by allocating17 all labour to the Cl MoP, thus producing each good with productivity (
yiR λi ) higher than the one associated with the Is mode of
production ( y iR ), or by placing all labour on the Is mode of production. In order to simplify the analysis,
17
The concrete process of allocation is not important. Given the general assumptions formulated, the conclusions will be the same if there is a central planning authority or if allocation is done by market.
17
if one assumes that there are clusters in all sectors of R, equilibrium prices can be written as
pi* =
1 . λi y iR Table 2 summarizes possible equilibria in a (small) poor region, in the presence of SE. For an
easy assessment of equilibria the last column of table 2 presents the income gap, i.e. the income difference between R and P. If labour is the only production factor, all income is translated into wages. Since we use labour in R as numeraire, the income in R is represented by 1, while income in P is represented by w.
Table 2 Possible equilibria in P (poor region) considering static externalities Specialization of P Income gap 1) No comparative advantages and •
λi = λ > 1
P specializes in a sector with a cluster (it could be either good 1 or good 2)
No income gap between R and P
1 = 1 )* w λ >1
(i.e., •
No trade and no clusters in P
2) Comparative advantages and
λ1 ≠ λ2
•
Complete specialization in sector 1 with Cl MoP
•
P specializes completely in sector 1, which uses the Is MoP
•
P specializes entirely in sector 2. This would only happen if good 2 was produced with Cl MoP and if
y1R / y1P
λ1 ( y1R / y1P ) < λ1 ( y1R / y1P )
y1P / y1R < λ1 y2P / y2R *
Where w denotes the wage in P.
Table 2 embodies two situations. First, the non-existence of comparative advantages, together with equal externality intensity across sectors, is considered. In this case, there are two possible equilibria: one equilibrium where P specializes in a sector with a cluster from which no income gap between R and P results, since SE have equal intensity in both sectors; the other equilibrium corresponds to an isolation situation, no clusters and no trade in the P economy. Respecting to the first equilibrium, if there are no comparative advantages, productivity is given by the exogenous parameter
yij which is equal in R and P ( yij = 1 ). If P specializes in a sector with a
18
cluster, and externalities are equal in both sectors, then the income gap between R and P is
1 yiR λ 1 = = =1 w yiP λ 1 But if there are no trade and no clusters in P, the ratio of productivities can be represented by
1 yiR λ = λ , which corresponds to the income gap = λ , with λ > 1 . yiP w The second situation represented in table 2 assumes different comparative advantages and different externality intensity across sectors. In this context, there are three possible equilibria in P. If there are comparative advantages, the productivity given by the exogenous parameter from 1 and not the same in both regions ( yiR
yij is different
≠ yiP ).
One equilibrium entails both complete specialization in the sector with the highest relative productivity (i.e., sector 1, by assumption) and clustering in this sector. But, for this to be an equilibrium it is necessary that
w 1 = p1* = λ1 y1P λ1 y1R
18
. That is, the income gap between R and P is
1 λ1 y1R y1R = = w λ1 y1P y1P
Another equilibrium entails complete specialization in this same good but without a cluster. In this case the equilibrium requires
w 1 1 λ1 y1R = p1* = with an income gap of = . y1P λ1 y1R w y1P
Finally, there is the possibility of another equilibrium with complete specialization in sector 2. This equilibrium happens if, and only if, nobody wants to deviate and produce good 1 with the Is MoP. What are the conditions needed to this equilibrium? In order to derive conditions for this equilibrium, let’s consider that if P specializes in sector 2 with a cluster, then it must be that the unit cost of good 2 produced in P, with a cluster, be equal to the international price of good 2, which is determined by the conditions in R. That is:
w 1 = p2* = λ2 y2 P λ2 y2 R And, consequently,
18
Recall that we use labor in R as the numeraire.
19
(2)
w = ( y2P / y2R )
(3)
On the other hand, the unit cost of good 1 produced in P without a cluster would be
y /y w = 2 P 2 R . For complete specialization in good 2 with a cluster to be an equilibrium, it is y1P y1P necessary that its unit cost be higher than the international price of good 1,
p1* , or:
y2P / y2R 1 > y1P λ1 y1R
(4)
And, after rearranging condition (4), we have:
y1P / y1R < λ1 y2P / y2R
(5)
Condition (5) shows that in equilibrium, for complete specialization in good 2 with the Cl MoP, it is necessary that comparative advantage in good 1 relatively to good 2 be weaker than the benefits of the externality in production of good 1. In this situation, the income gap can be obtained by combining equation (3) with condition (4), from which results:
1 < λ1 ( y1R / y1P ) . w
As it is visible from table 2, equilibrium with specialization in good 2 has an intermediate level of income while the highest income level is associated to good 1 produced with the Cl MoP, where P has a comparative advantage. Of course, equilibrium with specialization in a sector without clustering generates the lowest income level. It is also apparent in table 2 that
λ2
does not affect income when P
specializes in good 2 with Cl MoP. The reason for no interference is that the higher productivity generated by the stronger SE in sector 2 is exactly compensated by a lower international price. So, one can take a first conclusion: the power of SE is not significant for the choice among equilibria. If policy tries to maximize welfare, it must choose, among the possible equilibria, an equilibrium with clustering, but the choice must not be exclusively guided by the size of external economies. The target would be the sector with highest externalities if, and only if, the strongest comparative advantage corresponds to this sector. Otherwise, policy makers must opt for the sector with the strongest comparative advantage, irrespective of the dimension of positive externalities. However, this conclusion was based only on the presence of SE. Can DE, and especially interregional (and international) spillovers, alter these results? To answer this question, one needs to introduce some
20
additional assumptions. The key one is that production with the clustering MoP generates both static as well as dynamic externalities. To take the effects of dynamic externalities into account, an additional productivity variable, Zijt, is introduced. This variable increases with time, t, thanks to localized learning and competition externalities. This type of DE is produced only with the Cl and their amount depends on the type of economy. Accordingly, labour productivity across sectors is now also multiplied by this variable, Zijt. So, if there is no trade between regions, production with the Is MoP generates no DE whatever the sector, whereas production with Cl generates external but sector-specific dynamic economies, which lead to increasing productivity in manufacturing. For instance, in R economy, in steady state, with the Cl
MoP, Zijt grows at an x rate (i.e.,
Z&iRt = x , where a dot above the variable means the time derivative of Z iRt
the same variable). However, if one considers interregional trade, the poor region may also be able to benefit from the advantages of backwardness, as exposed in the previous section. In this case, in steady state, the Zijt growth in P is added with
zˆ (i.e.,
Z&iPt = x + zˆ , if zˆ is defined as Z iPt / Z iRt ). Z iPt
In other words, if there is interregional trade, productivity increases caused by localized learning in one economy will eventually diffuse to the other economy, even if there is no cluster there. Thus, in this model, clusters are necessary to generate knowledge spillovers, even though benefiting from knowledge spillovers is independent of the MoP. Profiting from these benefits only depends on the technology gap between P and R19. Supposing, for concreteness, that R has a cluster in sector i, but that P does not, then the rate of growth of the productivity variable ZiPt is governed by:
Z&iPt = zˆ Z iPt
(6)
However, if besides the technology gap, there is also a cluster in P:
Z&iPt = x + zˆ Z iPt
19
(7)
Of course, the amount of benefits depends also on the absorptive capacity of the P economy. However, for simplicity, we have opted for not introducing a new variable for controlling this effect. Besides, the inclusion of such variable would not change the main model conclusions.
21
Focusing on the right hand side of equation (7), the first term captures localized learning externalities, whereas the second term captures the spillovers coming from the advantages of backwardness. Given these assumptions governing DE and catching up, if P does not have a cluster in sector i, its labour productivity in steady state would be
yiP Z iPt = yiP zˆZ iRt , at time t. In contrast, the R’s
productivity in sector i, where it is assumed that a cluster exists, would be
yiR λi Z iRt . Thus, the ratio of
productivities in R versus P in sector i under these circumstances would be
( yiR / yiP )λi (1/ zˆ) . The first
term captures the comparative advantage (i.e., pure productivity differences) whereas the second and third terms capture the impact of the static and dynamic benefits of clustering, respectively. In order to draw lessons from this model, one has to analyse the steady state equilibrium in P, considered as a small open region. Since P is a small region, prices are derived from the equilibrium in R as if this were an isolated economy. Assuming for simplicity that R has clusters in all sectors, the steady state equilibrium in R has productivity given by
yiR Z iRt λi in sector i at time t. Thus, steady state
international prices are:
pit* =
1 yiR ZiRt λi
(8)
Considering the poor region, several situations can occur. Table 3 summarizes these possible situations where both static and DE are present.
Table 3 Possible equilibria in a Poor region considering dynamic externalities Specialization Income gap 1) Without comparative advantages and with
λi = λ > 1
• P specializes in a sector with a cluster • P has no clusters, there is no trade, and
No income gap Income gap is given by
(λ / zˆ ) > 1
Z iPt = zˆZ iRt 2) With comparative advantages and with λ1 ≠ λ2 • P is specialized in good 1 with a cluster •
P is specialized in good 1 with no cluster
•
P is specialized in good 2 with a cluster
22
y1R / y1P
(λ1 / zˆ )( y1R / y1P ) < (λ1 / zˆ )( y1R / y1P )
Imagining first that there are no comparative advantages, i.e.,
λi = λ > 1
yij = 1 for all i, j and also that
for i=1, 2. In this case, two equilibria are possible. First, there is an equilibrium where P
specializes in a sector with a cluster (similarly to the first case of table 2, there would be no income gap). Second, an equilibrium where P has no clusters, there is no trade, and
Z iPt = zˆZ iRt for all i, t. Thus, if
there is no trade and no clusters in P (i.e., in the Is MoP equilibrium) the income gap between R and P can be represented by
1 yiR λZ iRt = = λ / zˆ , with λ / zˆ > 1 . The term λ captures the benefits of SE while w yiP zˆZ iRt
1 / zˆ captures the benefits of DE (although restricted by the advantages of backwardness). If the P region moves from the isolated MoP to equilibrium with a cluster in sector i, then productivity would jump instantaneously thanks to the SE, and there would also be a dynamic effect, reflected in a temporary increase in the growth rate of eventually converge to
Z iPt above x. Clearly, in this case, Z iPt would
Z iRt and the income gap would disappear.
If comparative advantages and differences in the intensity of SE across sectors are assumed (i.e.,
λ1 ≠ λ2 ), the set of equilibria is analogous to
the set of equilibria derived in the model without DE.
First, there is an equilibrium where P is specialized in good 1 with a cluster, and the income gap corresponds to the ratio between exogenous productivity in the production of good 1 (
y1R / y1P ).
Second, there is another equilibrium where P is specialized in good 1 with no cluster. If P does not have a cluster in sector 1, its labour productivity in steady state would be
y1P Z1Pt = y1P zˆZ1Rt , at time t. In
contrast, the R’s productivity in sector 1, where it is assumed that a cluster exists, would be
Thus, under these circumstances,
the income gap would be
y1R λ1Z1Rt .
1 y1R λ1Z1Rt = = ( y1R / y1P )(λ1 / zˆ ) ; w y1P zˆZ1Rt
accordingly, both R and P grow at the same rate, so there is no convergence between the two regions. Convergence would occur if P managed to develop clusters so that it could generate both static and dynamic externalities. Finally, there is another equilibrium where P is specialized in good 2 provided that it satisfies a condition equivalent to condition (5) above derived for the model without DE. In order to derive conditions for this equilibrium, let’s consider that if P specializes in sector 2 with a cluster, then it must
23
be that the unit cost of good 2 produced in P, with a cluster, would be equal to the international price of good 2,
w 1 = p2* = λ2 y2 P Z 2 Pt λ2 y2 R Z 2 Rt
(9)
w = ( y2 P Z2 Pt / y2 R Z2 Rt )
(10)
And, consequently,
On the other hand, the unit cost of good 1 produced in P without a cluster would be
w y Z /y Z = 2 P 2 Pt 2 R 2 Rt . For complete specialization in good 2 with a cluster to be an equilibrium, it is y1P y1P necessary that its unit cost be higher than the international price of good 1,
y2 P zˆZ 2 Rt / y2 R Z 2 Rt 1 > y1P λ1 y1R
p1* , or: (11)
And, after rearranging condition (11), we have:
y1P / y1R λ1 < y2 P / y2 R zˆ
(12)
That is, if condition (12) is satisfied, the ratio of comparative advantages is less than the combined effect of externalities. In this situation, the income gap can be obtained by combining equation (10) with condition (11), from which results:
1 < (λ1 / zˆ )( y1R / y1P ) . w
5. Discussion The results obtained with the SE remain valid when a more realistic setting with DE, including the external economies derived from the ‘advantages of backwardness’ is considered. Regions with no clusters suffer from the lack of both static and dynamic externalities of agglomeration. There are multiple equilibria, and the equilibrium with the highest welfare in a poor region is the one where there is clustering in the sector with the strongest comparative advantage. Policy should focus on promoting clustering in this sector and avoid price distortions. There are significant implications of these results regarding the income gap of the different equilibria. As expected, if the government could choose the equilibrium, it would always choose
24
equilibrium with clustering, but it would also choose equilibrium with specialization in the sector with the strongest comparative advantage; the power of externalities is not relevant for the choice among equilibria. To sum up, when a region has a comparative advantage in producing a good, a policy that promotes an entirely new cluster in another sector may be worse than non-intervention. Besides, this type of intervention is always dominated by the promotion of a cluster in sectors where the region is already showing comparative advantage. Additionally, the model uncovers a paradox: externalities are the reason that advises public intervention, but their power is not a sufficient condition for choosing the sector to be supported. This paradox is also explained in the model: stronger externalities lead to higher productivity level and therefore to lower international prices i.e., the poor region loses via prices what it gains with externalities. The model was developed as if the world was composed by only two regions, each one provided with only a factor of production, L (labour), in fixed supply and related only by trade in goods. However, in the real world there are more than two regions and production factors can move from one region to another. What happens if a third region is included? While a full answer to this question implies the building of another model, which is beyond the scope of this paper, we can speculate that the introduction of a third region will change both the assumptions and the results of the model, particularly if there is mobility of firms and labour between regions. In fact, the results of the model can be altered if labour mobility, and particularly the selective migration of human capital across regions, is considered. If we take into account that embodied human capital may help explain differences in regional learning capacity, then the initial wage inequality between regions can lead to movement of (skilled) workers from the poor region to the more technologically advanced and if it is so, this decreases the absorptive capacity of the former relative to the rich region and acts against the effects of the ‘advantages of backwardness’. Of course, the selective migration issue is of critical importance to development policy (at least in countries where labour is rather mobile across regions) and is essential to any realistic policy discussion of uneven regional development patterns. This phenomenon can have two contradictory results: as a sort of ‘brain drain’, it seems to put poor regions perpetually at the back of the more techdriven economies. But, on the other hand, it is possible to admit that the out-migration can play a role in raising backward regions through several mechanisms, such as: a) increasing the level of regional productivity by forcing firms to change technology; b) helping households that stay in the region with
25
remittances, and so increase the aggregate demand and the structural change of the backward region. To know whether the final effect of selective migration is positive or negative is an empirical question that is beyond the scope of this paper. Modelling entails simplification, and the mathematical model of the previous section is short of other noteworthy aspects related to cluster policies20. One of such missing aspects is the possibility of discussing cluster polices in a system of regions involving what can be called a cluster race problem. When different regions compete by attracting economic activities, giving subsidies to firms for entering in their clusters, there is always a risk that competition does increase the level of subsidies beyond the value of externalities. In this case two negative effects can emerge. First, the race for clustering will uselessly absorb a share of the available economic resources, making the occurrence of a zero-sum game, or even a negative-sum game, more likely. Second, since subsidies are obtained from taxes, a distortion in economic sectors will be unavoidable. Other omitted aspect relates to the costs of cluster policies. Although the motivation for cluster policies is market failure in face of externalities, the main criticism against cluster policy arises from the concept of government failure. Cluster policy is seen as harmful, as governments lack the required information, capabilities and incentives to successfully determine whether the benefits of promoting clusters exceed the costs of implementing such policies. According to the public choice perspective, governments in making decisions, with electoral or personal incentives in mind, can be captured by vested interests, supporting the political elite rent-seeking instead of the common well-being, while distorting the efficient allocation of resources by market forces at the same time. If it is so, the advice of Duranton makes sense: instead of dreaming about clusters, policy-makers ‘should focus their attention away from the local production structure and aim instead at a more efficient provision of public goods that serve the needs of both residents and a broad range of local producers’ (Duranton, 2011: 40). But even if these public choice considerations do not hold, there are large uncertainties about computing externalities. Although this problem affects other policies, one can consider that pragmatically, policy makers do not truly think in terms of abstract ‘high externalities’ so much as they attempt to target hightech and emerging industries, often by trying to strengthen or establish the institutional support mechanisms that might lead to higher spillovers. For example, this might entail working with the local University to
20
For a discussion about how cluster strategies work and the implications for theory and policy, see Karlsson (2008b).
26
establish a ‘clean energy’ research centre in order to support the possible emergence of a clean-tech cluster in the region. But the basic lesson of ‘building from your strengths’ still holds and is probably wise advice in most situations. Yet, the pursuit of emerging sectors — if often driven by concerns of having existing specializations in the declining phase of their life cycle — should consider targeting ones that will put them on a different development trajectory. To consider the ‘related variety benefits’ is crucial for concretising such trajectory.
6. Conclusion Research in economic geography and regional science has empirically shown the advantages of industrial agglomeration over the isolated mode of production. These advantages have been theorized as positive externalities and are in the origin of virtuous circles of growth with implications in regional policy: positive externalities are the source of increasing returns, but they are also causing market failures. As is well known, when the market fails, government should get involved, implementing in this case a policy for promoting the source of increasing returns, which may imply choosing between different clusters. For instance, given the scarcity of resources, policy can be forced to choose between supporting either a wine cluster where the region has comparative advantages, or a health cluster which has shown higher externalities elsewhere. So, in a regional policy standpoint a question is mandatory: what is the appropriate policy for profiting from the greatest benefits of agglomeration and thereby promoting regional economic growth? In order to answer this question, this paper analyses possible static and dynamic externalities and includes them in a model that allows for trade with other regions. Often externalities are associated to specific characteristics of sectors, considering some sectors more externality-friendly than others. Consequently policy should allegedly target these sectors in order to benefit from stronger externalities. In contrast, this paper takes a different path: the model considers externalities as not associated to the intrinsic characteristics of a sector but, on the contrary, originated from the way production is organised. Externalities are considered as arising from the clustering process, and this is assumed as territorially grounded, given the abundant evidence on localization and urbanization externalities. Considering two different modes of production — isolation and clustering — existent in two regions with different levels of development, the model presented in section 4 shows that it is the traditional comparative advantage approach that must go on guiding policy. The main policy implication
27
of this finding is that the strength of expected positive externalities does not matter in choosing which clusters to promote. This conclusion is consistent with the advice to policymakers given by Hospers et al (2008): move away from beloved 'Silicon Somewhere' and embrace a more humble approach. However, modelling involves simplification, and this may cover the qualitative richness of different types of externalities and particularly the ones referred to ‘related variety benefits’. If the emerging and the traditional industries are related, the likelihood of a successful cross-pollination of ideas increases. Also, if a region is home to many actors in related industries, this can lead to more ideas being spread between traditional and emerging industries. These can be some ways of fostering economic growth and employment, and of ultimately changing the regional comparative advantages. This is why the most noteworthy for cluster policy is perhaps to identify the ‘related variety benefits’ of agglomeration. If we recognize that spillovers often exist between industries in seemingly different product markets, we might be able to use this understanding to promote emerging sectors that may still benefit from spillovers generated by traditional specializations.
References Abramovitz M (1979) Rapid Growth Potential and Its Realisation: The Experience of Capitalist Economics in the Post-war Period, in Malinvaud E. (ed.), Economic Growth and Resources: Vol. I: The Major Issues, pp. 1-30. London: Macmillan. Abramovitz M (1986) Catching-up, Forging Ahead and Falling Behind, Journal of Economic History 46(2), 386-406. Abramovitz M, David P (1995) Convergence and Deferred Catch-up Productivity Leadership and the Waning of American Exceptionalism, in Landau R., Taylor T., Wright G. (Eds.), Growth and Development: The Economics of the 21st Century. Stanford CA: Stanford UP. Almeida P, Kogut B (1999) Localization of knowledge and the mobility of engineers in regional networks, Management science, 45, 905-17. Arrow K (1962) The economic implications of learning by doing. Review of Economic Studies 29, 155– 173. Asheim B, Gertler M (2005) The Geography of Innovation in J Fagerberg, DC. Mowery, and RR. Nelson (eds) (2006), The Oxford Handbook of Innovation, Ch. 11, 291-317. Bailly A (1998) The Region: A basic Concept for understanding local Areas and Global Systems, Cybergeo: European Journal of Geography [on line] (http://cybergeo.revues.org/333; DOI: 10.4000/cybergeo.333). Baptista R, Swann P (1998) Do firms in clusters innovate more?, Research Policy 27, 525–540.
28
Barro R, Sala-i-Martin X (1992) Convergence Journal of Political Economy, 100, 223-51. Barro R (1990) Government Spending in a simple model of Endogenous Growth, Journal of Political Economy, 98(5), S103-25. Bathelt H, Malmberg A, Maskell P (2004) Clusters and knowledge: Local buzz, global pipelines and the process of knowledge creation. Progress in Human Geography, 28, 31-56. Bathelt H, Schuldt N (2010) International trade fairs and global buzz, part I: Ecology of global buzz. European Planning Studies, 18, 1957-1974. Beaudry C, Schiffauerova, A (2009) Who’s right, Marshall or Jacobs? The localization versus urbanization debate, Research Policy 38, 318–337. Benneworth P, Danson M, Raines P, Whittam G (2003) Confusing clusters: making sense of the cluster approach in theory and practice, European Planning Studies 11(5), 511–520. Boschma R (2005) Proximity and Innovation: A Critical Assessment, Regional Studies 39(1), 61-74. Boschma R (2009) Evolutionary economic geography and its implications for regional innovation policy. OECD, Paris. Boschma R, Frenken K (2011) The emerging empirics of evolutionary economic geography, Journal of Economic Geography 11(2), 295-307. Boschma R, Wenting R (2007) The spatial evolution of the British automobile industry. Does location matter? Industrial and Corporate Change 16(2), 213-238. Brachert M, Titze M, Kubis A (2011) Identifying industrial clusters from a multidimensional perspective: Methodical aspects with an application to Germany, Papers in Regional Science 90(2), 419-40. Braun E, MacDonald S (1978) Revolution in Miniature: The history and impact of semiconductor electronics. Cambridge: Cambridge University Press. Brenner T, Mühlig A (2007) Factors and Mechanisms Causing the Emergence of Local Industrial Clusters – A Meta-Study of 159 Cases. Papers on Economics & Evolution No 0723. Max Planck Institute of Economics, Jena. Camagni R (1991) Local ‘milieu’, uncertainty and innovation networks: Towards a new dynamic theory of economic space, in: Camagni R. (Ed.) Innovation Networks: Spatial Perspectives, pp. 121–144. London: Belhaven Press. Ciccone A (2002) Agglomeration Effects in Europe, European Economic Review, 46, 213-227. Ciccone A, Hall R (1996) Productivity and the Density of Economic Activity, American Economic Review 86(1), 54-70. Corden M (1997) Trade Policy and Economic Welfare, 2nd. edition. Oxford University Press. Coe D, Helpman E (1995) International R&D Spillovers, European Economic Review 39(5), 859-87. Coe D, Helpman E, Hoffmaister A (1997) North-South R&D Spillovers, Economic Journal 107, 134-49. Cohen MD, Levinthal DA (1990) Absorptive capacity. A new perspective on learning and innovation, Administrative Science Quarterly, 35, 128-152. Cooke P (2001) Clusters as key determinants of economic growth: The example of biotechnology, in Mariussen A. (ed) Cluster policies – cluster development? Nordregio Report 2001/2, Stockholm.
29
Cooke P (2002) Knowledge Economies: Clusters, Learning and Cooperative Advantage. London: Routledge. Cooper D (2001) Innovation and reciprocal externalities: information transmission via job mobility, Journal of Economic Behavior and Organization, 45(4), 403–25. David P, Rosenbloom J (1990) Marshallian factor market externalities and the dynamics of industrial location, Journal of Urban Economics 28, 349–370. de Groot H, Poot J, Smit M (2008) Agglomeration externalities, innovation and regional growth: Theoretical perspectives and meta-analysis, Department of Economics Working Paper Series, No 1/08. Hamilton, New Zealand: University of Waikato. DeLong B, Summers L (1991) Equipment Investment and Economic Growth, Quarterly Journal of Economics 106(2), 445-502. Duranton G (2011) California Dreamin': The Feeble Case for Cluster Policies, Review of Economic Analysis 3(1), 3-45. Duranton G, Puga D (2001) Nursery Cities: Urban Diversity, Process Innovation, and the Life Cycle of Products, American Economic Review 91(5), 1454-1477. Enright M (1996) Regional clusters and economic development: A research agenda, in Staber U., Schaefer N. V., Sharma B. (eds) Business networks: Prospects for regional development. De Gruyter, New York. European Commission (2008) The Concept of Clusters and Cluster Policies and their Role for Competitiveness and Innovation: Main Statistical Results and Lessons Learned, Commission Staff Working Document No. SEC 2637, Europe INNOVA/PRO INNO Europe Paper No. 9. European Communities, Luxembourg. Fagerberg J (1987) A technology gap approach to why growth rates differ, Research Policy, Supplement no. 16, 87-89. Fagerberg J, Verspagen B (2002) Technology-gaps, innovation-diffusion and transformation: an evolutionary interpretation, Research Policy, Supplement no. 31, 1291-1304. Feldman M (2000) Location and Innovation: the new economic Geography of innovation, spillovers, and agglomeration, in Clark G. L., Feldman M. and Gertler M. (eds.) Oxford Handbook of Economic Geography, pp. 373-394. Oxford: Oxford University Press. Feldman M, Audretsch D (1999) Innovation in cities: science-based diversity, specialization and localized competition, European Economic Review 43(2), 409–429. Freeman T (1961) A Hundred Years of Geography, Gerald Duckworth, London. Frenken K, Van Oort F, Verburg T, (2007) Related variety, unrelated variety and regional economic growth, Regional Studies 41(5), 685-697. Fujita M (1988) A monopolistic competition model of spatial agglomeration: a differentiated product approach, Regional Science and Urban Economics 18, 87–124. Fujita M, Thisse J-F(2002) Economics of Agglomeration: Cities, Industrial Location, and Regional Growth, Cambridge University Press, Cambridge, UK
30
Gertler M (2003) Tacit knowledge and the economic geography of context, or the undefinable tacitness of being (there), Journal of economic geography 3(1), 75-99. Glaeser E, Gottlieb J (2009) The Wealth of Cities: Agglomeration Economies and Spatial Equilibrium in the United States, Journal of Economic Literature 47, 983-1028. Glaeser E, Kallal H, Scheinkman J, Shleifer A (1992) Growth in Cities, Journal of Political Economy 100(6), 1126-52. Glaeser E, Maré D (2001) Cities and Skills, Journal of Labor Economics 19(2), 316–342. Goldstein, G. S. and Gronberg T.J. (1984) Economies of Scope and Economies of Agglomeration, Journal of Urban Economics 16, 91-104. Gordon IR, McCann P (2000) Industrial Clusters: Complexes, Agglomeration and/or Social Networks? Urban Studies 37(3), 513–532. Grabher G (1993) The weakness of strong ties: the ‘lock-in’ of regional development in the Ruhr area. In Grabher G. (ed.) The Embedded Firm: On the Socio-economics of Industrial Networks, pp. 255– 277. London: Routledge. Harrison B, Kelley M, Gant J (1996) Specialization versus diversity in local economies: the implications for innovative private-sector behavior. Cityscape: A Journal of Policy Development and Research, 2, 61–93. Henderson V, Kuncoro A, Turner M (1995) Industrial development in cities. Journal of Political Economy 103, 1067–1085. Henning M, Moodysson J, Nilsson M (2010) Innovation and regional transformation: from clusters to new combinations. Malmö: Region Skåne. Hospers G-J, Sautet F, Desrochers P (2008) Silicon Somewhere: Is There a Need for Cluster Policy? In Karlsson C (Ed) (2008b) Handbook of Research on Innovation and Cluster: Cases and Policies, chap. 24, Edward Elgar Publishing Limited, Cheltenham. Isaksen A (1997) Regional clusters and competitiveness: The Norwegian case. European Planning Studies 5, 65–76. Jacobs J (1969) The Economy of Cities, Penguin, London. Jacobs J (1984) Cities and the Wealth of Nations: Principles of Economic Life, Vintage, New York. Jaffe A, Trajtenberg M, Henderson R (1993) Geographic localization of knowledge spillovers as evidenced by patent citations, Quarterly Journal of Economics 108(3), 577–598. Johansson B (2005) “Parsing the Menagerie of Agglomeration and Network Externalities” in Karlsson C, Johansson B, Stough R (Editors) Industrial Clusters and Inter-Firm Networks, chapter 5. Edward Elgar Publishing, Cheltenham, UK. Johansson B, Karlsson C, Stough R (eds) (2001) Theories of Endogenous Regional Growth: Lessons for Regional Policies (Advances in Spatial Science), Springer-Verlag: Berlin-Heidelberg. Karlsson C (ed) (2008a) Handbook of Research on Cluster Theory (Handbooks of Research on Clusters series, 1) Edward Elgar Publishing Limited, Cheltenham. Karlsson, C (ed) (2008b) Handbook of Research on Innovation and Cluster: Cases and Policies (Handbooks of Research on Clusters Series, 2) Edward Elgar Publishing Limited, Cheltenham.
31
Keeble D, Wilkinson F (eds.) (2000) High-Technology Clusters, Networking and Collective Learning in Europe. Ashgate, Aldershot. Ketelhöhn N (2006) The role of clusters as sources of dynamic externalities in the US semiconductor industry, Journal of Economic Geography 6(5), 679-699. King C, Silk A, Ketelhöhn N (2003) Knowledge spillovers and the growth in the disagglomeration of the US advertising-agency industry. Journal of Economics and Management Strategy, 12, 327–362. Krugman P (1991a) Geography and trade. Cambridge, MA: MIT Press. Krugman P R (1991b) Increasing returns and economic geography, Journal of Political Economy, 99, 483–499. Krugman P (2011) The New Economic Geography, now middle-aged, Regional Studies, 45, 1–7. Leslie S (2000) The biggest “Angel” of them all: the military and the making of Silicon Valley, in Kenney M. (ed.) Understanding Silicon Valley: The Anatomy of an Entrepreneurial Region pp. 48–67. Stanford, CA: Stanford University Press. Lucas REJr (1988) On the Mechanics of Economic Development, Journal of Monetary Economics, 22(1), 3-42. Macher J, Mowery D, Hodges D (1998) Reversal of fortune? The recovery of the US semiconductor industry, California Management Review 41, 107–136. Maggioni M (2002) Clustering Dynamics and the Location of High-Tech-Firms (Contributions to Economics), Springer-Verlag Berlin-Heidelberg. Malmberg A, Maskell P (2002) The elusive concept of localization economies: towards a knowledgebased theory of spatial clustering. Environment and Planning A 34(3), 429–449. Malmberg A, Maskell P (2006) Localized Learning Revisited, Growth and Change 37(1), 1-18. Malmberg A, Power D (2005) (How) do (Firms in) clusters create knowledge? Industry and Innovation 12, 409–431. Markusen A. (1996) Sticky Places in Slippery Space: A Typology of Industrial Districts, Economic Geography, vol. 72(3), 293-313. Marshall A (1920) Principles of Economics (8th ed.) London: Macmillan. Original work published 1890. Martin R, Sunley P (2003) Deconstructing Clusters: Chaotic Concept or Policy Panacea? Journal of Economic Geography 3(1), 5-35. Maskell P, Eskelinen H, Hannibalsson I, Malmberg A, Vatne E (1998) Competitiveness, localised learning and regional development. Specialisation and prosperity in small open economies, London: Routledge. Maskell P, Kebir L (2006) What qualifies a cluster theory? in Asheim B, Cooke P, Martin R, (Eds) Clusters and Regional Development. Critical Reflections and Explorations, pp. 30-49. Routledge, New York, NY. Maskell P, Malmberg A (1999) Localised learning and industrial competitiveness, Cambridge Journal of Economics 23(2), 167-185. McCann P (2008) Agglomeration Economics, in Karlsson C (2008a) Handbook of Research on Cluster Theory pp. 23-38, Edward Elgar Publishing Limited, Cheltenham.
32
Morgan K (2004) The exaggerated death of geography: learning, proximity and territorial innovation systems, Journal of Economic Geography 4, 3-21. Morris P (1990) A History of the World’s Semiconductor Industry. London: Peter Peregrinus Ltd. Murphy K M, Shleifer A, Vishny R W (1989) Industrialization and the Big Push, Journal of Political Economy, 97(5), 1003-26. Neffke F, Henning M, (2009) Skill-relatedness and firm diversification, Papers on Economics and Evolution 2009-06, Max Planck Institute of Economics, Evolutionary Economics Group, revised Oct 2010. Neffke F, Henning M, Boschma R (2008) Surviving in agglomerations: Plant evolution and the changing benefits of the local environment, Papers in Evolutionary Economic Geography PEEG. 0820, Utrecht University, Section of Economic Geography. Neffke F, Henning M, Boschma R, Lundquist K-J, Olander L-O (2011) The dynamics of agglomeration externalities along the life cycle of industries, Regional Studies 45, 49–65. Nelson R, Phelps E (1966) Investment in Humans, Technological Diffusion, and Economic Growth, American Economic Review 61, 69-75. Nooteboom B (2000) Learning and innovation in organizations and economies, Oxford University Press, Oxford. Ohlin B G (1933) Interregional and international trade, Cambridge: Harvard University Press. (Revised edition published in 1967). Otter HS, van der Veen A, de Vriend HJ (2001) ABLOoM: Location behaviour, spatial patterns, and agent-based modelling Journal of Artificial Societies and Social Simulation vol. 4, no. 4 Parr JB (2002) Agglomeration economies: ambiguities and confusions, Environment and Planning A 34, 717-731. Perry M (2010a) Controversies in Local Economic Development: Stories, Strategies, Solutions. Abingdon: Routledge. Perry M (2010b) Controversies in Local Economic Development, Local Economy, 25(7), 527-534. Polanyi M (1967) The Tacit Dimension. London, Routledge. Porter M (1990) The Competitive Advantage of Nations. New York, Free Press. Porter M (1998) Clusters versus Industrial Policy, in Porter, M. (ed.) On Competition. Cambridge, US, Harvard Business School Press. Porter M (2000) Location, competition and economic development: local clusters in the global economy, Economic Development Quarterly 14, 15-31. Potter A, Watts H (2011) Evolutionary agglomeration theory: increasing returns, diminishing returns, and the industry life cycle, Journal of Economic Geography 11(3), 417-455. Power D, Lundmark M, (2004) Working Through Knowledge Pools: Labour Market Dynamics, the Transference of Knowledge and Ideas, and Industrial Clusters. Urban Studies, 41, 1025-44. Richardson G B (1972) The Organisation of Industry. The Economic Journal Vol. 82 (327): 883-896.
33
Rodriguez-Clare A (2007) Clusters and comparative advantage: Implications for industrial policy, Journal of Development Economics, 82(1), 43-57. Rodrik D (1996) Coordination failures and government policy: a model with applications to East Asia and Eastern Europe. Journal of International Economics 40 (1-2), 1-22. Romer P (1986) Increasing returns and long-run growth, The Journal of Political Economy 94, 1002– 1037. Romer P (1990) Endogenous Technological Change, Journal of Political Economy, Vol. 98(5), pp. S71102. Ros J (2000) Development Theory and the Economics of Growth. University of Michigan Press. Rosenfeld S (2005) Industry Clusters: Business Choice, Policy Outcome, or Branding Strategy? Journal of New Business Ideas and Trends 3(2), 4-13. Rosenthal S, Strange W (2003) Geography, Industrial Organization, and Agglomeration, Review of Economics and Statistics 85(2), 377-393. Rosenthal S, Strange W (2004) Evidence on the Nature and Sources of Agglomeration Economies, in Henderson J.V. and Thisse J.F. (eds.) Handbook of Urban and Regional Economics. Volume 4 pp. 2119-2171. Amsterdam, The Netherlands: Elsevier Science Publishers. Ruan J, Zhang X (2009) Finance and cluster-based industrial development in China, Economic Development and Cultural Change 58, 143–164. Saxenian A (1994) Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge, MA, Harvard University Press. Scherer F (1982) Demand-Pull and Technological Innovation: Schmookler Revisited, Journal of Industrial Economics 30(3), 225-238. Schmitz H (1995) Collective efficiency: Growth path for small-scale industry, Journal of Development Studies 31(4), 529–566. Scitovsky T (1954) Two concepts of External Economies, Journal of Political Economy 62: 70-82. Scott AJ (2004) A Perspective of Economic Geography, Journal of Economic Geography 4, 479-99. Simon HA (1997) An Empirically Based Microeconomics, Cambridge, UK: Cambridge University Press. Smith T (2000) Semiconductors. Standard & Poor’s Industry Surveys, New York. Spencer G, Vinodrai T, Gertler M, Wolfe D (2009) Do clusters make a difference? Defining and assessing their economic performance, Regional Studies 34, 1-19. Stilwell, F. 1992, Understanding Cities and Regions: Spatial Political Economy, Pluto Press, Leichardt, NSW. Storper M (2009) The economics of context, location and trade: another great transformation? in Becattini, G., Bellandi, M., De Propris, L. (eds.) The Handbook of Industrial Districts, pp 141– 157. Cheltenham: Edward Elgar. Tödtling F, Lehner P, Kaufmann,A (2009) Do different types of innovation rely on specific kinds of knowledge interactions? Technovation 29, 59–71.
34
Van der Linde C (2003) The demography of clusters – findings from the cluster meta-study, in Bröcker, J., Dohse, D., Soltwedel, R., (eds.) Innovation Clusters and Interregional Competition, pp. 130– 149. Springer, Heidelberg. Van der Panne G (2004) Agglomeration externalities: Marshall versus Jacobs, Journal of Evolutionary Economics 14, 593–604 Woodward D, Guimaraes P (2009) Porter’s Cluster Strategy and Industrial Targeting in Goetz SJ, Deller S, Harris T (eds.) Targeting Regional Economic Development. pp. 68-83. Routledge, New York.
35
