SME Clusters, Acquisition of Technological ... - Springer Link

Journal of Industry, Competition and Trade, 3:3, 187±210, 2003 # 2003 Kluwer Academic Publishers. Manufactured in The Netherlands.

SME Clusters, Acquisition of Technological Capabilities and Development: Concepts, Practice and Policy Lessons MARJOLEIN C.J. CANIEÈLS [email protected] Faculty of Management Sciences (MW), Open University of the Netherlands (OU), P.O. Box 2960, 6401 DL Heerlen, The Netherlands HENNY A. ROMIJN [email protected] Eindhoven Centre for Innovation Studies (ECIS), Faculty of Technology Management, Eindhoven University of Technology (TUE), DG 1.21, P.O. Box 513, 5600 MB Eindhoven, The Netherlands Abstract. The paper contributes to the current policy debate about promoting regional economic growth and competitiveness of small and medium industrial enterprises in development. An analytical framework is elaborated, which integrates complementary insights from existing approaches that have been used as a basis for policy design in this area. These are: the meso-level collective ef®ciency approach and the micro-level technological capability approach. The new framework gives insight into the different ways in which ®rm-level technological learning could be fostered through geographical clustering. It also provides new directions for policy. An empirical case study of farm equipment manufacturing in Pakistan's Punjab Province is used to illustrate the added value of the new approach. Keywords: industrial agglomerations, technological capabilities, small and medium enterprise development, knowledge spillovers, developing countries, regional industrial policy

1.

Introduction

In many parts of the developing world, small and medium industrial enterprises (SME) are currently being confronted with formidable competitive challenges. Few are still able to insulate themselves from the pervasive effects that economic liberalisation and deregulation are having on their local economies, particularly through international trade and foreign direct investment. For some ®rms, new business opportunities emerge as international production and trade chains extend into far-¯ung places in continuous search of cheap sources of supply. Many others are challenged to defend their traditional home marketsÐthe mainstay of the great majority of small ®rms in less developed economiesÐagainst new dynamic competitors. Recent research has begun to throw light on how SME are responding to these processes, and to identify suitable types of assistance. One salient ®nding is that their competitiveness could be boosted by being part of regional agglomerations of ®rms

188

CANIEÈLS AND ROMIJN

engaged in similar and complementary activitiesÐcommonly denoted as ``clusters''.1 It is claimed that SME clusters may be able to boost regional development by creating possibilities for accumulating capital and skills through ``collective ef®ciency'' (Schmitz, 1995; Schmitz and Nadvi, 1999).2 However, so far the collective ef®ciency studies have concentrated mainly on the economic bene®ts to which clusters may give rise, while the technological factors that supposedly underpin these bene®ts have been given rather cursory treatment. The main analytical focus of studies in this line is on the meso-level, while the internal functioning of the individual ®rms that make up a cluster remains largely a black box (Bell and Albu, 1999). Yet, development of technological competence by small ®rms is arguably crucial for them to be able to hold their own in fast-changing and ®ercely competitive markets. Flexible and quick adjustment and adaptation requires technical skills, knowledge and capacityÐ technological capability for shortÐto make the right investment choices, increase productive ef®ciency, meet tight deadlines, and engage in continuous upgrading of quality and design (Romijn, 2001; Bell and Albu, 1999). Since only a very small minority of SME in developing countries are already well equipped for these tasks, we pose that regional industrial growth and competitiveness in these countries ultimately cannot be sustained without advancement of the technological capability-base of individual ®rms. Likewise, effective policies aimed at promoting competitiveness of regional clusters cannot be based solely on insights about cluster-dynamics. In addition, they need to build on knowledge about the internal technological functioning of the companies operating in those clusters, and they have to address the main bottlenecks that occur at that level. The aim of this paper is to contribute to the current policy debate about promoting regional economic growth and SME competitiveness in developing economies by opening up the black box of the ®rm in the Collective Ef®ciency (CE) approach. We do this by linking that approach to the so-called ``Technological Capability'' (TC) literature, which sheds light on the micro-level technological underpinnings of industrial competitiveness in a developmental context. The meso-level CE approach and the micro-level TC approach yield complementary insights into the determinants of long run competitiveness of SME in a developing economy context. While CE studies analyze the effects of clustering on economic performance and competitiveness of clusters and countries, the TC approach studies how intra-®rm learning processes affect long-term competitiveness. The uni®ed conceptual framework that results from joining the two approaches is then used to shed new light on the forces that determine competitiveness and growth of industrial SME clusters in development, and to derive new guidelines for policy. The paper is thus positioned within an academic debate among researchers writing about problems of less developed economies; and the approaches used in this particular

1 2

Some authors de®ne clusters more widely as mutually interdependent and interacting actors in the value chain e.g., OECD, 2001. In contrast, we are concerned strictly with clusters in a territorial sense. See especially the case studies in the special issue of World Development on industrial clusters in developing countries (vol. 29 no. 9, 1999).

SME CLUSTERS, ACQUISITION OF TECHNOLOGICAL CAPABILITIES AND DEVELOPMENT

189

debate constitute the point of departure for the analysis. At the same time, we take account of voluminous bodies of literature pertaining to regional and ®rm-level innovation and -learning in economically advanced countries, assessing their relevance for solving the observed meso±micro-gap observed in the research about lesser developed economies. However, we will show that these literatures do not yield a straightforward blueprint solution for the problem posed in this paper. In Section 2, we show the gap left between the micro and meso literatures and then proceed to review in more detail the key contributions made to the debate about industrial growth by the ``micro'' approach. In addition to discussing the Technological Capability studies pertaining to developing countries, we explore the relevance of related literatures about technological learning in economically advanced economies. In Section 3, we do the same for the ``meso'' literature about regional agglomeration. Again we start by highlighting the literature pertaining to developing countries i.e., the Collective Ef®ciency approach, and then assess the relevance for our paper of a broader set of studies pertaining to economically advanced countries. Armed with insights from this review, we proceed to introduce the new conceptual framework in Section 4. In Section 5, we spell out the main new policy insights that result from adopting the framework, showing its added value by contrasting them with the sort of policy implications that are typically derived by the two partial approaches on their own. Section 6 contains an empirical case study of farm equipment manufacturing in Pakistan's Punjab Province, which illustrates the additional value of the new approach for policy design in practice. Section 7 presents the conclusions.

2.

Micro approaches to industrial competitiveness: A review

The TC literature puts intra-®rm knowledge accumulation processes centre stage. The term ``technological capability'' was ®rst coined in the early 1980s by researchers probing intra®rm technological dynamics in developing countries, where ®rms typically operate far from the world's technological frontier.3 Inspired by upcoming evolutionary theories of technological change (later culminating in Nelson and Winter, 1982; and Dosi, 1988), and following many disappointing experiences with technology transfer projects, they showed importation of new technologies from advanced countries by itself to be insuf®cient for enhancing productivity and inducing self-sustaining industrialization. Mere access to foreign technologyÐwhether in the form of plant and machinery or documentation and blueprintsÐdoes not imply mastery over it (Dahlman et al., 1987). Tacitness associated with new knowledge, and the fact that foreign technologies are less than perfectly suited to speci®c local needs and conditions, can be powerful barriers to the effective implementation of new technologies in a new setting. Therefore, accumulating technological capability requires considerable technological effortsÐinvestment in time and resources aimed at 3

There are many contributions to the capability literature. Lall (1992), UNCTAD (1996) and Herbert-Copley (1990) are good reviews.

190


assimilating, adapting and improving known technologies, and (ultimately) creating new technologies in-house. A lengthy learning process is usually involved. The capabilities that ®rms acquire help to improve their economic performance, and (by assumption) regional and national economic performance as well. A ®rm's technological improvement efforts are seen to be somehow pushed and enhanced by various external stimuli. For instance, Lall (1992) and Cassiolata and Lastres (2000) have drawn attention to the importance of being part of larger `ìnnovation systems'' or networks composed of actors and institutions involved in complementary activities. However, there has been no systematic conceptual treatment of how and why spatial proximity between the actors in such a system or network could somehow contribute to technological capability acquisition within individual ®rms (Bell and Albu, 1999; CanieÈls and Romijn, 2003). The TC approach originated in a development context. Hence, its main focus has been on the assimilation of existing technologies created elsewhere. Capabilities that initially need to be mastered in this setting are engineering knowledge and skills to select, install and deploy new hardware. The learning process mainly involves trial and error and experimentation on the shop ¯oor, activities which have little in common with formal R&D. The main policy instruments that can induce (or constrain) the extent to which ®rms invest in learning are regulation of trade and ®nance, and public investments in literacy and technical education. In contrast, innovation policies of the sort commonly deployed in advanced countries e.g., R&D subsidies and patenting are of limited use and may even be counterproductive. Despite the radically different setting, it is of some value to examine recent theories of ®rm behavior in economically advanced countries, to ®nd out whether these offer any insights which can help us to bridge the micro±meso gap in a development context. Studies about organizational learning, core competences, and knowledge accumulation and competitiveness have gained increasing prominence over the past decade or so. This literature is vast and wide-ranging, including contributions from strategic management, organizational science and evolutionary theory of the ®rm e.g., Dodgson, 1993; Dosi et al., 2000; Teece et al., 1997; Prahalad and Hamel, 1990; Cohendet et al., 1998; Nelson and Winter, 1982; Eliasson, 1990; Foss, 1993, 1998. Although diverse, the various strands share a common set of core ideas. In much of the literature the emphasis is on leading-edge corporations whose R&D processes are central to sustained competitiveness. On the face of it, these corporations are far removed from the typical developing country ®rm that is the central focus of this paper. However, just like in the TC literature, a ®rm's economic performance is conceptualised as the consequence of a continuous learning process. This process is driven by a ®rm's resource baseÐa stock of human skills and knowledge, physical assets, and organizational routines. Routines are ``. . . a set of ways of doing things and ways of determining what to do'', which are built into organizations at any one time (Nelson and Winter, 1982, p. 400). They have the function of coordinating the other resources of the ®rm in particular ways, leading to their productive utilization (Dosi et al., 2000, p. 5). The economic environment generates continuous pressures on ®rms to subject their routines to evaluation, to ensure that the ®rm's competitive position is maintained.


191

Routines change in response to trial and error, experimentation, and organizational search (Eliasson, 1990; Nelson and Winter, 1982; Radner, 1986; Teece et al., 1997), implying intentional activities to improve routines for better economic performance. Hence, as in the TC literature, trial and error, experimentation and search are seen as activities that require investment of resources. However, in advanced economies investment in R&D obviously constitutes the bulk of such activities e.g., Dodgson, 1993, p. 388. In contrast, the learning in companies in less developed economies is predominantly non-R&D driven, including information search, debugging, incremental adjustment and the like. This is especially true for the small and medium sized companies that are the central focus of this paper. This issue will come back in our analysis. The learning process causes ®rms to accumulate so-called capabilities, bundles of related routines governing the exploitation of their resources. Capabilities are resident in a particular function (Javidan, 1998). Examples are marketing-, production-, and human resource management capabilities. Capabilities that are cross-functionally integrated and coordinated are denoted as competencies Ibid.. Competencies express what a ®rm is able to do well (Prahalad and Hamel, 1990). A subset of such competencies are the basis for a ®rm's unique competitive advantage at a given point in time. These distinctive competencies are called core competencies. They encompass what the ®rm is able to do better than others (Lawson and Lorenz, 1999, p. 306). The ability to adapt core competences quickly to changing opportunities is what ultimately drives competitiveness over time. In the words of Prahalad and Hamel, `Ìn the long run, competitiveness derives from an ability to build, at lower cost and more speedily than competitors, the core competencies that spawn unanticipated products'' (1990, p. 81). Teece et al. (1997) refer to this ability as the dynamic capabilities of a ®rm ( p. 516). Compared to the TC approach that we take as our point of departure, the literature reviewed above has a broader focus. Technological capabilities form just a sub-set of a larger array of organizational and managerial competences. The difference in orientation is understandable. In a context of underdevelopment, the lack of purely engineering skills and knowledge needed to operate a given plant ef®ciently often constitutes a ®rm's primary bottleneck. While capabilities of a non-technical nature are usually also weak, in all likelihood non-technical problems cannot be overcome successfully without ®rst addressing the underlying immediate technological constraints. There seems to be a general agreement in the literature that the latter constitute the foundation upon which the former have to be built. More speci®cally, acquisition of speci®c ``how to do'' skills (also termed ``single-loop learning'') is a necessary starting point for the ultimate development of the dynamic organizational-managerial capabilities for continuous technological absorption and organizational improvement that Teece and colleagues refer to. Dodgson (1993) states that dynamic capabilities arise from ``double-loop learning'', which in turn results from repeated cycles of ``single-loop learning''.4

4

Several writers have referred to the same phenomenon in different terms. In the words of Stiglitz, `Òne learns to learn at least partly in the process of learning itself'' (1987, p. 130). Cohen and Levinthal refer to learning-to-learn as acquisition of `àbsorptive capacity'' (1989, p. 569).

192


This sequential nature of learning is the reason why technological competenciesÐas distinct from organizational onesÐare given special importance in an advanced economy context as well (Dosi and Teece, 1993; Malerba and Marengo, 1995). All being considered, we have concentrated the analysis in this paper mainly on the acquisition of technological capabilities, while acknowledging that the relationship between technological and non-technological capabilities is important. Recapitulating, the literature about capabilities, competences and knowledge accumulation in advanced economies yields a number of insights about the nature and driving forces of learning at the ®rm level, with which our insights into learning processes in economically less developed settings are enriched and put into broader context. Like the TC literature in less developed economies, the learning literature in advanced countries sees the individual ®rm as the prime actor in the generation of knowledge. At the same time, it has been widely recognized that ®rms do not learn in isolation. The importance of being part of strategic R&D networks and alliances features strongly (e.g., Gancel et al, 2002; Duysters et al., 1999). However, the dimension of geographical closeness of the actors in these networks has not been explicitly considered. Several other writers have emphasized the importance of being embedded in an innovation systemÐa range of interconnected actors and institutions each of which plays its own unique role in the national knowledge accumulation and diffusion process. Thus, the learning process in the ®rm is assisted and facilitated by complementary knowledge inputs emanating from other actors in the system. When discussing policy, writers about innovation systems emphasize the importance of strengthening the quality and intensity of the interactions between the actors in the system, in addition to measures that facilitate knowledge accumulation in individual actors (Metcalfe, 1995; Lundvall, 1988, 1993; Edquist, 1997; Freeman, 1995; Nelson, 1993). However, as in the strategic network literature, the role which spatial proximity between actors in an innovation system might play in promoting innovation and learning has not been analysed systematically and exhaustively. The key to well-functioning innovation systems is seen to lie primarily in high interconnectivity and an effective division of labor in the production and transfer of new knowledge (e.g., Metcalfe, 1995, p. 464). Some time after the birth of the national system of innovation approach, there has been a distinct shift in focus from the national to the regional level, giving rise to the concept of Regional Innovation System (RIS). This happened under the in¯uence of a rising wider interest in the role of regions for economic dynamism and competitiveness, a debate in which geographers, sociologists and researchers from related social disciplines have taken a lead role. At ®rst glance, the RIS approach does make a link between the micro-level literature about ®rm knowledge accumulation with meso-level factors that drive innovation and learning in an advanced economy setting. However, in order to accurately assess the value of the RIS literature in this respect, we ®rst need to have a better understanding of meso-level approaches, in addition to the micro approaches discussed so far. Therefore we ®rst discuss relevant meso approaches. Subsequently, we will assess the relevance of studies that have tried to link the two levels of analysis so far.


3.

193

Meso approaches to industrial competitiveness: A review

The studies in the CE line of thinking put strong emphasis on the advantages for ®rms that accrue from being part of a geographical industrial cluster. These gains tend to boost the economic performance of the cluster and therefore the region as a whole. At the same time, the implications of clustering for technological progress have remained rather peripheral. Studies that shed light on inter-actor relations (rather than intra-actor dynamics) have formed the main sources of inspiration, including transaction cost theory, socio-geographical studies dealing with regional dynamics and sociological approaches. This focus on inter-actor relations is clearly re¯ected in the way clustering is seen to impact on economic growth. Two mechanisms are distinguished: Marshallian externalities and cooperation, also termed ``passive CE'' and `àctive CE''. Passive CE includes cost advantages due to agglomeration, including availability of a pool of specialized workers; easy access to suppliers of varied and specialized inputs; and quick dissemination of new knowledge and ideas. All these bene®ts ``. . . fall into producers' laps without deliberate efforts to bring them about'' (Schmitz and Nadvi, 1999, p. 1505). This contrasts with `àctive CE'', which materializes only as a result of purposive actions aimed at generating them, involving deliberate cooperation and collaboration between actors. Strong institutions are seen to be crucial for active CE to occur, particularly trust. CE writers put great store by active CE, which they consider to be the chief driving force for attaining sustained competitiveness. In short, active inter-actor collaboration and networking, rooted in a common culture and supported by common institutions is seen to be the main driving force in boosting competitiveness in the CE approach. The role played by intra-®rm factors in this process have received scant attention. As noted in the Introduction, although most CE studies do address some aspects of `ùpgrading'' in local ®rms, this concept is not rooted in an analytical framework in which ®rm-level technological change takes central stage. At ®rst sight, several approaches studying regional economic dynamism in economically advanced countries would appear to have adopted a more explicit innovation and learning perspective than the CE approach. Some have used the notion of the ``learning region'', in which institutional actors are seen to play a central role in promoting and facilitating regional innovative behavior (Morgan, 1997; Florida, 1995). A related concept is the `ìndustrial district'' (Scott, 1988; Storper, 1995), a highly geographically concentrated group of companies that `èither work directly or indirectly for the same end-market, share values and knowledge so important that they de®ne a cultural environment, and are speci®cally linked to one another in a complex mix of competition and co-operation'' (Rosenfeld, 1995, p. 13). Yet, despite the innovation and learning terminology used in these approaches, one will search in vain for a clear conceptualisation of what learning (and innovation) at the ®rm level actually involves. As in the CE approach, the prime engine for competitiveness is seen to lie in inter-actor relations rather than within individual companies. This is evident from the emphasis on elements like trust, solidarity and co-operation between ®rms, the

194


result of a close intertwining of economic, social and community relations e.g., Harrison, 1992. These are seen to play a key role in overcoming tacitness of knowledge, thereby facilitating knowledge access. However, just how this impinges on the knowledge production function of the individual ®rm remains obscure. Furthermore, since there is no clear understanding of the way in which the ®rm goes about accumulating new knowledge, other than collaboration-based mechanisms through which agglomeration may impinge on that process are easily overlooked. From this point of view, the earlier-mentioned RIS approach is more promising. Isaksen (2001), a leading RIS proponent, observes that the innovation performance of a region ``. . . depends to a large extent on how ®rms utilize the experience and knowledge of other ®rms, research organisations, government sector agencies, etc., in innovation processes, and how they blend this with the ®rms' internal capabilities'' ( p. 108, emphasis added). In the same vein, Asheim and Isaksen argue that the key driving force behind regional innovative capacity is the proximity between different actors, which ``makes it possible for them to create, acquire, accumulate and utilize knowledge a little faster than ®rms outside of knowledge intensive, dynamic regional clusters'' (2002, p. 83, emphasis added). The underlying reason lies in the tacit features of knowledge. Local spaces facilitate trust and common conventions and institutions, which are considered to be important vehicles for its effective diffusion e.g., Cooke, 1998. The emphasis on social-cultural and institutional aspects in the RIS approach closely resembles the industrial district and CE approaches, with the de®ning difference that the RIS has a clear evolutionary underpinning. Evolutionary features such as inter-®rm heterogeneity and complementarity, and intensive local competition are considered to be the key factors that drive innovation in the regional system. Yet, upon close inspection, even in the RIS approach one is still left wondering as to exactly how regional proximity is supposed to enable the speeding-up of the intra-®rm knowledge accumulation processes that Asheim and Isaksen (2002) referred to. Which are the spatial proximity advantages that play a role here, and how do these advantages impinge on ®rms' internal knowledge accumulation processes? The RIS writers have not really examined these questions. Their main aim has been to understand the dynamics of learning of the region as a whole, and they explain this in terms of the nature of interactor relations, albeit from an evolutionary perspective. Studies similar in orientation to the RIS approach focus on ``collective learning'' (see, e.g., Keeble and Wilkinson, 1999; and Lawson and Lorenz, 1999), and local knowledge spillovers (Malmberg and Maskell, 2002). Here, too, it is not the intra-®rm learning processes as such that are the central focus, but rather the learning of the region as a whole. Important principles originally used by evolutionary theorists to explain competence building at the level of the individual ®rm are `ùpscaled'' to the regional level. In this way, learning is seen to depend on tacit knowledge being shared amongst actors in a region (rather than amongst the employees of one ®rm). Moreover, generating new knowledge within the region depends on combining diverse kinds of knowledge, signifying the importance of complementarities between regional actors (rather than between individuals within a single ®rm).


195

This approach does lead the authors to identify an important knowledge-related advantage of geographical proximity, namely spillovers of tacit knowledge that are facilitated by trust and shared routines. This is a valuable input for ®lling the meso-micro gap identi®ed in the introduction of this paper. However, in our view the `ùpscaling'' approachÐwith the region as the unit of analysisÐstill falls short of systematically identifying all sorts of agglomeration advantages and tracing their impact on ®rm-level learning. Doing this requires that we map out the interaction between two separate levels of analysis, region and ®rm, each with their own theoretical concepts. This is the subject of the following section.

4.

Joining the collective ef®ciency and the technological capability approaches

The key question is now, in which different ways the acquisition of technological capabilities at the level of the individual ®rm (TC) could be enhanced by co-location in a regional industrial agglomeration (CE). Figure 1 shows how we intend to join the CE and TC approaches. The analytical perspective of the CE approach is represented in the lefthand side of the ®gure. The right-hand side of the ®gure presents the analytical perspective taken in the TC literature, which we shall use to open the black box left in the CE approach.

Figure 1. Integrating meso with micro

196


Looking from left to right in Figure 1 along the dotted lines that connect the left-hand side to the right-hand side, we zoom in on the ®rm level. The crux of the integration of the two approaches then revolves around the question as to how exactly the connection between the meso- and the micro-level is made. In other words: What are the mechanisms by which the different agglomeration advantages that are commonly observed in an industrial cluster translate into increased and/or more ef®cient technological effort (or organizational search) within individual ®rms in that cluster? Once this is known, we are able to understand the different ways in which clustering may enhance ®rm-level technological learning, leading to more advanced capabilities and better economic performance, and (ultimately) how it contributes to regional and national economic growth. We need two ingredients in order to answer this question. First, on the micro-side we need to know which are the different types of technological efforts commonly undertaken by ®rms. A useful classi®cation has been developed by Bell, based on a survey of a range of empirical capability studies, including: staff training, staff hiring, in-house technological improvement (including R&D) and external search for information about new technologies and markets (Bell, 1984). The second ingredient consists of the main types of agglomeration advantages emanating from geographical clustering by ®rms. Information about this is provided from literature on the meso-side of Figure 1. There are two main types. The ®rst category comprises economies of scale, scope and transaction, which are all kinds of cost advantages (``pecuniary economies'') that accrue from ®rms locating close to each other e.g., Marshall, 1920; Richardson, 1978. The second category consists of direct knowledge inputs. These are the technological or knowledge spillovers as identi®ed by RIS researchers and associated studies (Section 3).5 Since these two main groups are still rather heterogeneous, it is useful to make some further sub-divisions. Within the category of scale, scope and transaction economies, we identify two groups according to the two main types of economic activity that ®rms pursue: (I) economies of scale, scope and transaction in the production of goods and services; and (II) economies of scale, scope and transaction in undertaking technological effort. Further, we distinguish three groups of knowledge spillovers, following a classi®cation adopted by Stewart and Ghani (1991), namely, spillovers emanating from: (III) changing attitudes and motivations; (IV) human capital formation through informal learning-by-doing; and (V) transfer of technological information. We now proceed to examine the different ways in which these ®ve agglomeration advantages may contribute to the four types of ®rm-level technological efforts identi®ed by Bell (1984). The discussion is summarised in Table 1. The ®ve rows in the table 5

We follow Jaffe's de®nition of knowledge spillovers as intellectual gains through exchange of information for which a direct compensation for the producer of the knowledge is not given, or for which less compensation is given than the value of the knowledge (Jaffe, 1996, p. 5). Hence, spillovers are costless by de®nition. However, the existence of a certain minimum level of absorptive capacity on the part of the receiving ®rm is a precondition for spillovers to occur. Thus, in clusters where ®rms with low technological capabilities dominate, spillovers are unlikely to be a major innovation-enhancing mechanism.

D. R&D

III. Knowledge spillovers: Changing motivation and attitudes IV. Knowledge spillovers: Human (a) Exposure/demonstration effect/contagion stimulate demand for TE. capital formation through informal (b) Direct free input learning-by-doing through industry-wide accumulation of skills. V. Knowledge spillovers: Direct free input through inter-®rm movement of (a) Direct free knowledge input through trade journals, meetings, Technology transfer trained labor. fairs, etc. (b) Direct free input through user±producer interaction.

(a) Large local market gives rise to critical minimum demand for innovations, inducing technological efforts to develop them. (b) Presence of specialized suppliers lowers transaction costs, which facilitates easy and cheap access to specialised inputs needed for technological effort. (c) Low transaction costs facilitate joint undertaking of technological efforts, thus leading to cost-savings. (d) Low transaction costs stimulate additional technological effort in joint lumpy and complementary projects, which in turn facilitates access to, and leads to generation of new information and knowledge. Exposure/demonstration effect/contagion stimulate demand for TE.

Lower unit cost due to large market size leaves more resources for technological effort.

C. Information search

I. Economies of scale, scope and transaction in production II. Economies of scale, scope and transaction in knowledge accumulation

B. Training

A. Hiring

Agglomeration advantages

Table 1. Direct effects of agglomeration advantages on the technological efforts (TE) of the ®rm.


197

198


represent the types of agglomeration advantages introduced above, while the four columns, labeled A through D, represent the types of technological effort. The contents of the cells describe the (sub-) mechanisms through which the agglomeration advantages affect these technological efforts. Some types of agglomeration advantages, particularly II and V, can in¯uence technological effort in several distinctly different ways. This is shown in the table by means of a ®ner subdivision within the main rows. We discuss the table row-wise. Row I indicates a mechanism associated with direct cost advantages in production obtained by clustered ®rms. One such cost saving emanates from high demand (Swann, 1998). Since clustered ®rms reap more economies of scale in production compared to non-clustered ones, they are left with more ®nancial resources to invest in technological effort. Another source of cost savings lies in the fact that clustering may induce more intensive competition among input suppliers, which reduces input costs for user ®rms (Nadvi, 1999b). Both types of cost-savings may affect all types of technological effort (columns A through D). Row II indicates that economies of scale, scope and transaction in knowledge accumulation itself may have four signi®cant effects on technological effort. First, clusters can generate a critical minimum demand for new, specialized products or services that cannot be produced pro®tably elsewhere. In turn, this stimulates investment in efforts to master the production of these new items (IIa) (Stewart and Ghani, 1991). This effect may apply to all kinds of technological effort (columns A through D). A second important link deals with the local presence of suppliers of specialized inputs who are attracted by large local demand. This may lower transaction costs associated with procurement of specialized inputs. Thereby, clusters act to reduce costs of specialized inputs needed to undertake investments in technological effort (IIb) (Porter, 1998). This mechanism may again in¯uence all kinds of efforts, because there are manifold actors offering specialized services, including workers with specialized skills and technical consultants (A), institutions providing training courses (B), government extension services (C&D), sourcing agents looking for suitable suppliers (C), suppliers of machinery, materials and components (C&D), and so on. A third important link in row II operates by offering possibilities for ®rms to join networks of innovators because of low transaction costs associated with local interaction (IIc) (Freeman, 1991; DeBresson and Amesse, 1991). This leads to cost-advantages from sharing costs and risks. Existing literature pointing to this mechanism relates primarily to economically advanced countries, so that the focus has been primarily on formal R&Dtype efforts (D), but it is no less likely to work with respect to more informal types of effort with scope for collective investment. In economically less developed settings, these activities are more likely to take the form of training (B) and search (C) than R&D. In another study pertaining to advanced countries (Baptista, 1998), it has been pointed out that pooling R&D resources will induce more R&D investment as well, as it becomes feasible to embark on large, costly projects that are beyond the capacity of individual investors (IId). A variation on this theme is the case where proximity allows parties to invest in technological effort that requires mutual commitment, since they need to supply complementary inputs for it. As in the case of row IIc, this mechanism might operate just as well in the form of more informal technological effort in less developed economies. In


199

particular, training (B) and search (C) are two non-R&D-based efforts for which collective investments are also feasible. Rows III, IV and V indicate that knowledge spillovers from other ®rms may complement a ®rm's own efforts and thereby increase the ef®ciency of those efforts. Implementing knowledge from outside the ®rm increases its chances of success. Studies such as Malmberg and Maskell (2002) and recent RIS studies (Section 3) have pointed to the possibility of ®rms bene®ting from complementarity and synergy effects that arise from the technological improvement activities undertaken by other ®rms in the cluster. Spillovers are facilitated by opportunities for ®rms to establish direct contact with each other in a cluster, such as through inter-®rm labor mobility and formal and informal exchange of information and ideas (Nelson, 1993; Feldman, 1994; Von Hippel, 1988; Baptista, 1998). Changing attitudes and motivation (III) primarily work by exposing people to new ideas and artifacts in a particular environment (Stewart and Ghani, 1991). These act on people's mental predisposition in such a way that they will begin to favor change over stability, and thereby stimulate investment in the technological efforts needed to bring it about. These advantages affect ®rms' efforts in a broad manner. For example, changing attitudes happen through exposure to new information, ideas and products, which generally stimulates demand for technological improvement efforts of all kinds (A through D). Human capital formation through informal learning-by-doing (IV) likewise acts through changing attitudes, in this case attitudes towards work (IVa) Ibid.. Like mechanism III, it is a broad effort-inducing mechanism (A through D). In addition, learning-by-doing entails assimilation of a basic body of more speci®c production-related technical knowledge and skills that are common in a local industrial environment (IVb) Ibid.. This constitutes a direct free input complementing a ®rm's own investments in staff training (B). Thus, this spillover not only affects the demand for technological effort, but also the supply of inputs for it. Technological transfer (V) acts entirely on the supply side. It operates through three channels: inter-®rm movement of trained labor (A); trade journals, meetings, trade fairs and various other fora for inter-personal exchange (C&D); and user±producer interactions which often occur in the course of implementing and perfecting innovations in iterative fashion (also C&D) Ibid..6 Inter-®rm movement of trained labour boosts skill levels through hiring of new staff; while communication fora and user±producer interactions are primarily sources of free new information and knowledge about technologies and markets, which complement the ®rm's own search and research efforts. Technology transfer spillovers often interact with economies of scale, scope and transaction. Low transaction costs in clusters directly facilitate business interaction, joint projects, and labor mobility, which are the main vehicles through which skills, knowledge and ideas travel across ®rms. Furthermore, we have seen that economies of scale, scope

6

The importance of this spillover has also been documented in many empirical studies, e.g., Johnston and Kilby, 1975; Fransman, 1982; Cortes, 1979.

200


and transaction boost technological effort in various ways. Clearly, the more actively ®rms are engaged in learning, the more spillovers to neighboring ®rms are likely to result. The recipients essentially receive free inputs that complement their own efforts and in this way increase the effectiveness of their learning processes. In sum, when economies of scale, scope and transaction work in tandem with knowledge spillovers, both the amount and effectiveness of intra-®rm technological effort will receive a boost. All mechanisms discussed so far referred to positive effects of agglomeration on technological effort. However, positive forces may be counteracted to some extent by negative effects from clustering. This cannot be ignored, particularly when it comes to deriving policy lessons. In clusters where secrecy is hard to maintain and legal protection of innovations is non-existent, knowledge spillovers may also have drawbacks as they reduce innovation incentives for the party that generates them. Further, considerable social barriers may impede technological improvement efforts in change-resistant traditional communities. Third, excessive competition among small producers who are incapable of differentiating their products substantially from those of others may squeeze margins, leaving fewer resources for technological improvement. This can hold back technological progress for several decades. Seen from a long-term perspective, however, this effect may not be so negative. Competition ultimately contributes to a healthy ``shake-out'', in which a few technologically most advanced companies survive and grow (Amsden, 1977; Cortes, 1979). 5.

Policy implications from the integrated approach

Our integrated framework has distinct added value for policy making, in comparison to the two partial approaches. The main policy prescriptions of the CE literature hinge upon the distinction between ``passive CE'' and `àctive CE''. The CE writers expect a lot from policies aimed at promoting active CE, through strengthening institutions that promote joint action, inter®rm collaboration and horizontal and vertical networking (Schmitz and Nadvi, 1999; Ceglie and Dini, 1999; UNCTAD, 1998). Collaboration with buyers is seen to be especially important, as understanding customers' needs helps producers to tackle key competitiveness problems. In contrast, the passive, non-collaborative externalities are essentially taken as given in the CE approach. The integrated framework in Table 1 suggests a broader range of policy instruments than what has been offered by the CE studies. It is of course possible, on the basis of the table, to make a similar distinction between agglomeration advantages that need active collaboration and networking to bring them about (notably IIc and IId), and others that occur without purposive action of this sort. However, there are no a priori reasons why active collaboration should be singled out in our framework as the only feasible way to intervene. Indeed, many of the spontaneous, non-collaborative mechanisms could be amenable to further stimulation too. Possibilities in this direction appear to exist especially in the sphere of knowledge spillovers. A relevant example would consist of measures to upgrade basic physical infrastructure facilities such as roads and telephone


201

networks, which may indirectly induce more inter-®rm spillovers by facilitating tradebased interactions and exchange and by increasing labour mobility through the mechanisms listed in row V of Table 1. Another example would be the establishment of education and training facilities that help to make people mentally more receptive to the adoption of new production techniques by exposing them to the same. Once a few ®rms have adopted a new technique, in due course it is bound to be adopted more widely and contribute to the spread of knowledge through demonstration effects and other mechanisms identi®ed in rows III and IV. The broader policy focus suggested by our framework has particular advantages in those situations where horizontal inter-®rm cooperation in clusters has proved dif®cult to get off the ground, and/or hard to sustain. This has indeed been the case in many of the clusters examined in the CE studies (see, e.g., Nadvi, 1999a, b; Visser, 1999). In these kinds of environments, measures aimed at facilitating individual knowledge absorption and subsequent spontaneous diffusion among ®rms may turn out to be a more effective and sustainable policy strategy. Our framework further suggests cost-effective ways of providing some of these types of support, for example, in the sphere of education and training. Targeting a small number of progressive companies with programs to improve work practices, upgrade quality, and so on, may be suf®cient. Spillovers will ensure the gradual diffusion of the new knowledge and skills throughout the cluster, through demonstration effects (III & IVa), industry-wide accumulation of skills (IVb±B); inter-®rm movement of labor (V±A); and circulation of information and knowledge (Va&b±C&D). An important example of this approach is provided by programs involving technology transfer from foreign TNC subsidiaries to local supplier ®rms that show potential for technological upgrading. Countries like Singapore and Malaysia have socalled ``buyer-mentoring'' programs in which foreign electronics assemblers are compensated for coaching a selected number of local components makers in quality upgrading, process engineering and industrial engineering through on-site visits, workshops, consultancy activities, and so on (UNCTAD, 2001; Meyanathan, 1994). The framework laid out in Table 1 also suggests a somewhat more focused approach to intervention compared to the CE approach. Given the central policy goal as stated in the IntroductionÐnamely, the structural improvement of regional competitivenessÐ interventions need to be directed as closely as possible towards the creation of dynamic learning economies. This is achieved by directing assistance towards stimulating technological effort. However, the sort of collaboration-focused support advocated in CE studies does not appear to be tightly focused on those dynamic effects. Some of the policy instruments suggested in the CE approach seem to be geared towards generating static production economies instead. For example, strengthening collaborative institutions may help producers to lobby for common access to credit, joint purchasing of inputs and tax relief, but there is no direct link between such activities and technological learning. A similar policy orientation is also evident in the literatures that establish a partial integration between meso- and ®rm-levels of analysis, notably the RIS studies. Their main policy directive boils down to strengthening local resources, organizations and institutions, as well as upgrading the quality and intensity of their interactions for the purpose of innovation (Cassiolata and Lastres, 2000, p. 51; Metcalfe, 1995, pp. 467±468).

202


However, to the extent that the CE studies do address dynamic `ùpgrading'' issues, they lack speci®c policy guidelines. For example, in their recent summary paper about the CE literature, Schmitz and Nadvi state that ``. . . In more advanced clusters, policy measures have to be geared to promoting upgrading, in particular in areas of technical learning and innovation . . .''. A ®rst step is ``. . . benchmarking local practices against those of global market leaders'' (1999, pp. 1509±1510). Yet, no details are given about the types of learning, innovation and work practices actually involved, and the policy instruments that could be used. RIS studies, although clearly focused on learning and innovation in regions, likewise remain rather vague about the speci®c policies that should be applied to stimulate institutional upgrading and networking among parties in the local system. The application of our integrated framework affords more speci®c policy conclusions in this direction, since it delves more deeply into ®rm-level processes of technological improvement. To do this, we used an established analytical approach (TC), which has come up with a range of policy lessons that are in principle applicable to clusters as well as to ®rms individually. Its policy ®ndings rest on the recognition of pervasive market failures in the development process. On the one hand, market-oriented interventions to stimulate demand for innovations are considered essential, as these create incentives for ®rm-level investment in technological effort. Relevant instruments include trade and industrial policies that regulate market structure and competition; ®scal, legal and investment policies that regulate access to ®nancial capital and foreign technologies, and general economic policies that promote a stable macro-economic climate and economic growth. These demand-stimulating interventions are complemented by supply-side assistance aimed at supporting ®rms' learning processes through the provision of critical missing inputs. Important supply-side measures are the creation of an information-rich environment, the building of science and technology support institutions, education and training (Lall, 1992; UNCTAD, 1996). However, since the TC approach has not paid speci®c attention to geographical agglomeration, its policy guidelines do not take account of any special advantages for learning and growth that obtain in clusters. Our integrated approach does make this possible, through the insights gained from the meso-literature on the left-hand side of Figure 1. First, conventional TC measures are potentially more effective when supplemented by policies to induce ®rms engaged in similar and complementary activities to co-locate. Second, the framework provides speci®c handles for enhancing capability accumulation in ®rms that are already located in clusters. In particular, several of the learning-enhancing mechanisms identi®ed in Table 1 can be leveraged through targeted interventions. Relevant policies include, for example, collective institutionbuilding initiatives that promote inter-®rm trust. These may lead to cost-savings and may induce new large knowledge-building projects that are beyond the scope of individual ®rms. They may also indirectly trigger various inter-®rm knowledge spillovers. Other policies could aim at stimulating spillovers directly, by concentrating new technological and market information, training, and extension services locally in clusters. Third, the framework suggests once again that such interventions do not need to aim at directly targeting all potential bene®ciaries in a locality, as the presence of local spillovers ensures quick diffusion of knowledge and skills (see above).


203

The additional value of these policies could be outweighed to some extent by various negative effects that potentially also emanate from clustering (see Section 2). These effects are likely to play a role of some importance, especially in clusters whose ®rms are still functioning at a very low technological level. Producers in that situation usually lack the skills to diversify their products substantially from each other, so that keen competition is hard to avoid. Copying is also bound to be rampant in this kind of environment. Furthermore, in traditional communities the threat of social sanctions on deviant behavior may pose a considerable constraint on investment in technological innovation. It is important that policy makers be aware of these kinds of problems, but they cannot be remedied quickly and easily. Sometimes, measures to foster inter-®rm collaboration and trust may help, but in general clusters simply need time to outgrow this stage.

6.

An empirical illustration

Some of the principles discussed in the previous section will be illustrated with some case-study material. The focus is on farm equipment manufacturing, a highly important industry in developing countries when it comes to initial absorption and accumulation of knowledge about basic mechanical engineering principles. The industry essentially functions as a springboard from where capabilities subsequently diffuse to other industries as industrialization gets under way (Johnston and Kilby, 1975). The Pakistan Punjab is a fertile agricultural region known for its widespread adoption of modern cultivation practices since the late 1950s. Agricultural modernization has been supported by a sizeable farm equipment industry that is organized into approximately eight major clusters, each consisting of ®fty to sixty small-scale ®rms.7 The industry emerged in the early 1960s in response to surging demand for irrigation equipment by farmers who were increasingly beginning to switch over to irrigation-based cultivation methods Ibid.; Aftab and Rahim, 1986. By the 1990s, the industry had evolved to the extent that it employed about 5,000 people and was capable of manufacturing well over ®fty different items (Romijn, 1999). These included a number of relatively complex pieces of equipment with moving parts and high-grade steels, along with the simple rigid structures (such as mould board ploughs) that were already being produced in the early stages of the industry's existence. Incremental design improvements had also been made to increase capacity, sturdiness, safety and ef®ciency, although quality problems in production were still persistent Ibid.. We ®rst bring the CE and TC perspectives to bear to re¯ect on the evolution of the industry's performance, and to identify the best ways forward. Then we identify new policy insights by imposing our integrated approach on the case. The purpose of this exercise is not to come up with detailed policy prescriptions based on in-depth assessments according to the three approaches. Rather, the goal is to illustrate the broad 7

Small-scale in this context means that ®rms employ roughly between ®ve and ®fty workers.

204


differences in analytical perspectives taken, and derive some important policy directions from them. Starting with the CE perspective, the available evidence suggests that active networking and collaboration can be at best a partial explanation of the industry's growth. There is no tradition of active cross-®rm horizontal collaboration in the production chain, as there is a fundamental lack of trust among competitors. Competition and associated secrecy is severe, even among people who belong to the same bradri (caste) (Nabi, 1988, p. 123). Cooperation has worked somewhat better across different stages of the production chain, because these vertical relations are more complementary than competitive. For example, skilled machinists employed by parent ®rms reportedly pay frequent visits to subcontractors to ensure that new components are manufactured according to speci®cation, and that materials are selected according to the parent ®rms' requirements Ibid., pp. 121±122. There is also some interaction with buyers of farm equipment. Some farmers provide performance feedback to the producers and suggest further improvements (Romijn, 1999). However, since the buyers are not themselves technologically knowledgeable, the impact of these linkages for ®rms' upgrading may not be all that dramatic. Also, the networking relations are not institutionalized in any formal way. There do not seem to be any bodies, such as industry associations or governmental agencies, which command suf®cient trust and that are organizationally capable enough to be able to play a notable promotional role in this respect.8 If the past cooperation record is anything to go by, in the sort run there is unlikely to be much scope for stimulating the industry's development through the kind of policy approach advocated in the CE literature. In the prevailing social-cultural environment it will doubtlessly prove hard to initiate and sustain active cooperative behavior, especially among close competitors. The TC approach draws attention to a rather different set of determinants to explain the industry's past growth. First, strong demand-pull effects arising from agricultural modernisation provided incentives for innovation (Child and Kaneda, 1975). The industry emerged in the proximity of agricultural areas with fast increasing crop yields. Rising purchasing power of farmers combined with increasing seasonal labour shortages fuelled massive investments in mechanization, which in turn induced a critical minimum local demand for many new types of farm equipment. There was an obvious impetus for producers to introduce new product technologies by reverse engineering prototypes imported by large engineering ®rms and wealthy farmers (Johnston and Kilby, 1975). Supply-side factors were also important. The industry is endowed with a somewhat better institutional science and technology infrastructure than other sectors of small-scale industry. This has to do with its link with agriculture, a priority sector in the economy. There are national and provincial institutions employing engineers who are involved in adapting foreign farm equipment designs to local conditions. Their outreach to local ®rms has not always been effective, but some positive results have been noted (Romijn, 1999). 8

Romijn, own ®eldwork observations, 1994.


205

For example, several ®rms in Daska, one of the biggest clusters, had been approached by extension of®cers from the national Farm Machinery Institute in Islamabad, who were looking for suitable partners with whom they could commercialize farm machinery prototypes developed by them. Interested workshops would receive in-house technical training and counselling and would embark on a process of collaborative effort to iron out teething problems of new equipment during the stage of ®eld trials with local farmers Ibid., p. 243. Firms in another major cluster, Mian Channun, had bene®ted from the establishment of a Dutch-funded training and common facility project on the local industrial estate, run under the aegis of the provincial Punjab Small Industries Corporation. Local ®rms received short training courses in heat treatment, properties of different metals and their uses, and use of jigs and ®xtures. Moreover, they had access to specialised machining services Ibid., p. 243. Diffusion of skills was further assisted by technical and vocational training centres set up by the central and provincial governments (Aftab and Rahim, 1986). From the perspective of the TC approach, policies for further upgrading would focus on the strengthening of these institutional features. For example, one could suggest to improve the outreach and content of some of the existing technical extension and training programmes in order to make them more effective. Furthermore, the TC assessment would point to the importance of stimulating demand for the industry's products through measures supporting the continued advancement of the agricultural sector (see, e.g., Romijn, 1999, pp. 257±263). Additional insights about the industry's development can be gained by applying the combined meso-micro framework outlined in Table 1. One clear example relates to the favorable effect on ®rm-level technological effort emanating from transaction cost advantages and high spillovers in the local labor market. In particular, clustering of the farm equipment ®rms facilitated the establishment of specialised training suppliers who were attracted by a large local demand for their services. This did not just involve the government-run training services already mentioned above. Some reputable private ®rms in the industry also began to assume the status of training institutes. They even issued certi®cates as evidence that apprentices had completed their training there. The authenticity of these documents and the reputation of the ®rms in question could be veri®ed easily in the local community, thereby reducing transaction costs for workshops looking to hire competent workers in the local labor market (mechanism IIb, A-D in Table 1).9 These specialized training suppliers also contributed to signi®cant local labor-market spillovers (mechanism V, A) due to high inter-®rm movement of trained labor (Stewart and Ghani, 1991, p. 585). High spillovers in the product market constitute another example. Progressive farmers who import new farm equipment models are sources of information about new product designs. Manufacturers generally try to reverse-engineer foreign prototypes when they are passed on to them for repair and maintenance (Romijn, 1999, pp. 196±197). Knowledge about new designs spreads rapidly due to co-location (mechanism Va, C&D). 9

Romijn, own ®eldwork observations, 1994.

206


Technology transfer appears to result predominantly from informal contact and observation, although marketing lea¯ets and the annual Horse and Cattle Show in Lahore (which features new locally produced farm equipment) may be of some importance as well Ibid., p. 226. These observations illustrate some of the growth mechanisms that were identi®ed in our framework analyzing ®rm-level technological learning in clustered settings. Having these clearly in view enables us to derive more speci®c policy conclusions than would be possible on the basis of either one of the two partial approaches. In particular, the examples suggest that while the above-mentioned science- and technology policies emanating from the TC approach already point in the right direction, these policies could be made more effective. This could be done, for example, by taking account of the local diffusion mechanisms and low transaction costs discussed above. Speci®cally, singling out progressive companies in the delivery of technical extension, consultancy services and training programs could be a more ef®cient way to deliver assistance than a broader-based effort aiming for a direct coverage of large numbers of companies. This approach avoids the problem of having to mobilize large numbers of rather conservative entrepreneurs, often observed in more broad-based assistance programs aimed at small-scale enterprises in poor countries. Many small entrepreneurs are dif®cult to convince about the bene®ts of participating in such formal assistance programs, and are mistrustful of government involvement. Well-targeted interventions can help to create forerunners who could play the role of catalysts in spreading knowledge and progressive attitudes. 7.

Conclusions

The analytical framework elaborated in this paper extends the collective ef®ciency approach that has been commonly used to analyze regional economic growth and SME competitiveness in development. This is done by opening the black box of the ®rm, using well-known literature on ®rm-level technological capability acquisition in development. The resulting conceptual taxonomy sheds light on how ®rm-level capability building could be fostered through geographical clustering. The main new policy insights that can be derived from our framework are summarized as follows: First, the policies that are already part of the conventional collective ef®ciency approach, namely promotion of active networking and cooperation, are bound to become more effective when they are speci®cally aimed at stimulating technological learning, rather than covering all kinds of production-focused activities as well. Dynamic learningfocused collaborations are likely to have the most durable impact on industrial competitiveness. Second, the demand- and supply-policies to stimulate technological innovation that are typically part of the conventional technological capability approach, become more effective when closely targeted at clusters rather than industries in general. The operation of the various agglomeration effects on ®rm-level technological effort and learning


207

identi®ed in this paper enhances the impact of these policies. Moreover, the mechanisms underlying technological learning in clusters could be taken into account in the design of the policies themselves. Notably, selective targeting of support to progressive ®rms is likely to pay dividends due to high spillovers effects. In sum, the integrated approach suggests three different ways to provide more focus in policies to promote SME competitiveness. One lies in the object of the policies themselves (stimulating technological learning); the second is the geographical coverage (concentrating on industrial clusters); and the third is the implementation modality selected (targeting a few progressive ®rms). The paper should be seen as a ®rst step to design more effective policies promoting regional economic growth and SME competitiveness in development. The policy conclusions listed above clearly do not pretend to be more than broad guidelines for this purpose. The precise details of an SME assistance strategy in a speci®c setting will obviously have to depend on the relative importance of the different mechanisms listed in Table 1 in that particular context. The taxonomy in the table can serve as a guideline for identifying these. Further work should be aimed at operationalising the mechanisms in the taxonomy, so that it can be used as a tool kit for empirical research about ®rm learning in clusters in practice. Acknowledgments The authors would like to thank two anonymous referees for their useful comments on an earlier version of the paper. References Aftab, K. and Rahim, E., ``The emergence of a small-scale engineering sector: The case of tubewell production in the Pakistan Punjab,'' The Journal of Development Studies, vol. 23 no. 1, pp. 60±76, 1986. Amsden, A.H., ``The division of labour is limited by the type of the market: The case of the Taiwanese machine tool industry,'' World Development, vol. 5 no. 3, pp. 217±233, 1977. Asheim, B.T. and Isaksen, A., ``Regional Innovation Systems: The integration of local `sticky' and global ùbiquitous' knowledge,'' Journal of Technology Transfer, vol. 27 no. 1, pp. 77±86, 2002. Baptista, R., ``Clusters, innovation and growth: A survey of the literature.'' In G.M.P., Swann, M., Prevenzer, and D. Stout, (eds), The Dynamics of Industrial Clusters: International Comparisons in Computing and Biotechnology. Oxford University Press: Oxford, pp. 13±51, 1998. Bell, R.M., ```Learning' and the accumulation of industrial technological capability in developing countries.'' In M. Fransman, and K. King, (eds), Technological Capability in the Third World. Macmillan: London, 1984. Bell, R.M. and Albu, M., ``Knowledge systems and technological dynamism in industrial clusters in developing countries,'' World Development, vol. 27 no. 9, pp. 1715±1734, 1999. CanieÈls, M.C.J. and Romijn, H.A., `Àgglomeration advantages and capability building in industrial clusters: The missing link,'' The Journal of Development Studies, vol. 39 no. 3, pp. 129±154, 2003. Cassiolata, J.E. and Lastres, H.M.M., ``Local systems of innovation in Mercosur countries,'' Industry and Innovation, vol. 7 no. 1, pp. 33±53, 2000. Ceglie, G. and Dini, M., ``SME cluster and network development in developing countries: The experience of UNIDO,'' Paper presented at the International Conference on Building a Modern and Effective Development

208


Service Industry for Small Enterprises, March 2±5, Committee of Donor Agencies for Small Enterprise Development: Rio de Janeiro, 1999. Child. F.C. and Kaneda, H., ``Links to the Green Revolution: A study of small-scale, agriculturally related industry in the Pakistan Punjab,'' Economic Development and Cultural Change, vol. 23 no. 2, pp. 249±277, 1975. Cohen, W.M. and Levinthal, D.A., `Ìnnovation and learning: The two faces of R&D,'' Economic Journal, vol. 99 no. 397, pp. 569±596, 1989. Cohendet, P., Llerena, P., and Marengo, L., ``Theory of the ®rm in an evolutionary perspective: A critical assessment,'' Paper presented at the 2nd Annual Conference of the International Society for New Institutional Economics, Paris, 18±19 September, 1998. Cooke, P., `ÌntroductionÐOrigins of the concept.'' In H.-J., Braczyk, P., Cooke, and M. Heidenreich, (eds), Regional Innovation SystemsÐThe Role of Governances in a Globalized World. UCL Press: London, pp. 2± 25, 1998. Cortes, M., ``Technical development and technology exports to other LDCs,'' Annex I in ArgentinaÐStructural Change in the Industrial Sector. The World Bank, Development Economics Department: Washington, D.C., 1979. Dahlman, C.J., Ross-Larson, B., and Westphal, L.E., ``Managing technological development: Lessons from the newly industrializing countries,'' World Development, vol. 15 no. 6, pp. 759±775, 1987. DeBresson, C. and Amesse, F., ``Networks of innovators: A review and introduction to the issue,'' Research Policy, vol. 20, pp. 363±379, 1991. Dodgson, M., `Òrganizational learning: A review of some literatures,'' Organization Studies, vol. 14 no. 3, pp. 375±394, 1993. Dosi, G., ``The nature of the innovative process.'' In G. Dosi, et al. (eds), Technical Change and Economic Theory. Pinter Publishers: London, pp. 221±238, 1988. Dosi, G., Nelson, R.R., and Winter, S.G. (eds), The Nature and Dynamics of Organizational Capabilities. Oxford University Press: Oxford, 2000. Dosi, G. and Teece, D., ``Competencies and the boundaries of the ®rm,'' Center for Research in Management, CCC Working Paper No. 93-11, University of California: Berkeley, 1993. Duysters, G.M., Kok, G., and Vaandrager, M., ``Crafting successful strategic technology partnerships,'' R&D Management, vol. 29 no. 4, pp. 343±351, 1999. Edquist, C., Systems of Innovation: Technologies, Institutions, and Organizations. Pinter: London and Washington, 1997. Eliasson, G., ``The ®rm as a competent team,'' Journal of Economic Behavior and Organization, vol. 13, pp. 275±298, 1990. Feldman, M.P., The Geography of Innovation. Kluwer Academic Press: Boston, 1994. Florida, R., ``Toward the learning region,'' Futures, vol. 27 no. 5, pp. 527±536, 1995. Foss, N.J., ``The competence-based approach: Veblenian ideas in the modern theory of the ®rm,'' Cambridge Journal of Economics, vol. 22, pp. 479±495, 1998. Foss, N.J., ``Theories of the ®rm: Contractual and competence perspectives,'' Journal of Evolutionary Economics, vol. 3, pp. 127±144, 1993. Fransman, M., ``Learning and the capital goods sector under free trade: The case of Hong Kong,'' World Development, vol. 10 no. 11, pp. 991±1014, 1982. Freeman, C., ``Networks of innovators: A synthesis of research issues,'' Research Policy, vol. 20, pp. 499±514, 1991. Freeman, C., ``The national system of innovation in historical perspective,'' Cambridge Journal of Economics, vol. 19 no. 1, pp. 5±24, 1995. Gancel, C., Raynauld, M., and Rodgers, I., Successful Mergers, Acquisitions And Strategic Alliances: How To Bridge Corporate Cultures. McGraw-Hill Professional: London, 2000. Harrison, B., `Ìndustrial districts: Old wine in new bottles?,'' Regional Studies, vol. 26 no. 5, pp. 469±483, 1992. Herbert-Copley, B., ``Technical change in Latin-American manufacturing ®rms: Review and synthesis,'' World Development, vol. 18 no. 11, pp. 1457±1469, 1990.


209

Isaksen, A., ``Building regional innovation systems: Is endogenous industrial development possible in the global economy,'' Canadian Journal of Regional Science, vol. 24 no. 1, pp. 101±120, 2001. Jaffe, A.B., `Èconomic analysis of research spillovers. Implications for the Advanced Technology Program,'' National Institute of Standards and Technology: Gaithersburg, MD, 1996. Javidan, M., ``Core competence: What does it mean in practice?'' Long Range Planning, vol. 31 no. 1, pp. 60± 71, 1998. Johnston, B.F. and Kilby, P., Agriculture and Structural Transformation. Economic Strategies in LateDeveloping Countries. Oxford University Press: New York, 1975. Keeble, D. and Wilkinson F., ``Collective learning and knowledge development in the evolution of regional clusters of high technology SMEs in Europe,'' Regional Studies, vol. 33 no. 4, pp. 295±303, 1999. Lall, S., ``Technological capabilities and industrialisation,'' World Development, vol. 20 no. 2, pp. 165±186, 1992. Lawson, C. and Lorenz, E., ``Collective learning, tacit knowledge and regional innovative capacity,'' Regional Studies, vol. 33, no. 4, pp. 305±317, 1999. Lundvall, B.-A., `Èxplaning inter®rm cooperation and innovation: Limits of the transaction-cost approach.'' In G. Grabher, (ed.), The Embedded FirmÐOn the Socio-economics of Industrial Networks. Routledge: London, pp. 52±64, 1993. Lundvall, B.-A., `Ìnnovation and an interactive process: From user±producer interaction to the national system of innovation.'' In G. Dosi, et al. (eds), Technical Change and Economic Theory. Pinter Publishers: London, pp. 349±369, 1988. Malerba, F. and Marengo, L., ``Competence, innovative activities and economic performance in Italian hightechnology ®rms,'' International Journal of Technology Management, vol. 10, pp. 96±112, 1995. Malmberg, A. and Maskell, P., ``The elusive concept of localization economies: Towards a knowledge-based theory of spatial clustering,'' Environment and Planning A, vol. 34, pp. 429±449, 2002. Marshall, A., Principles of Economics. MacMillan: London, 1920. Metcalfe, S., ``The economic foundations of technology policy: The equilibrium and evolutionary perspectives.'' In P. Stoneman (ed.), Handbook of the Economics of Innovation and Technological Change. Blackwell: Oxford, pp. 409±512, 1995. Meyanathan, S.D., `Ìndustrial structures and the development of small and medium enterprise linkages. Examples from East Asia,'' EDI Seminar Series, The World Bank: Washington, D.C., 1994. Morgan, K., ``The learning region: Institutions, innovation and regional renewal,'' Regional Studies, vol. 31 no. 5, pp. 491±503, 1997. Nabi, E., Entrepreneurs & Markets in Early Industrialization. A Case Study from Pakistan. ICS Press: San Francisco, Cal., 1988. Nadvi, K., ``Shifting ties: Social networks in the surgical instrument cluster of Sialkot, Pakistan,'' Development and Change, vol. 30 no. 1, pp. 141±175, 1999a. Nadvi, K., ``The cutting edge: Collective ef®ciency and international competitiveness in Pakistan,'' Oxford Development Studies, vol. 27 no. 1, pp. 81±107, 1999b. Nelson, R.R. (ed.), National Innovation Systems. AComparative Analysis. Oxford University Press: Oxford, 1993. Nelson, R.R. and Winter, S.G., An Evolutionary Theory of Economic Change. The Belknap Press of Harvard University Press: Cambridge Mass, 1982. OECD, Innovative Clusters. Drivers of National Innovation Systems. OECD: Paris, 2001. Porter, M.E., ``Clusters and the new economics of competition,'' Harvard Business Review, Nov.±Dec., pp. 77± 90, 1998. Prahalad, C.K. and Hamel, G., ``The core competence of the corporation,'' Harvard Business Review, May± June, pp. 79±91, 1990. Radner, R., ``The internal economy of large ®rms,'' Economic Journal, vol. 96 (Supplement), pp. 1±22, 1986. Richardson, H.W., Regional and Urban Economics. Dryden Press: Hinsdale, 1978. Romijn, H., ``Technology support for small-scale industry in developing countries: A review of concepts and project practices,'' Oxford Development Studies, vol. 29 no. 1, pp. 57±76, 2001. Romijn, H., Acquisition of Technological Capability in Small Firms in Developing Countries. Macmillan: London, 1999.

210


Rosenfeld, S.A., Industrial Strength Strategies: Business Clusters and Public Policy. Aspen Institute: Washington, DC, 1995. Schmitz, H. and Nadvi, K., ``Clustering and industrialization: Introduction,'' World Development, vol. 27 no. 9, pp. 1503±1514, 1999. Schmitz, H., ``Collective ef®ciency: Growth path for small-scale industry,'' Journal of Development Studies, vol. 31 no. 4, pp. 529±566, 1995. Scott, A.J., New Industrial Spaces: Flexible Production Organisation and Regional Development in North America and Western Europe. Pion: London, 1988. Stewart, F. and Ghani, E., ``How signi®cant are externalities for development?'' World Development, vol. 19 no. 6, pp. 569±594, 1991. Stiglitz, J.E., ``Learning to learn, localised learning and technological progress.'' In P. Dasgupta, and P. Stoneman, (eds), Economic Policy and Technological Performance. Cambridge University Press: Cambridge, 1987. Storper, M., ``The resurgence of regional economics, ten years later: The region as a nexus of untraded interdependencies,'' European Urban and Regional Studies, vol. 2 no. 3, pp. 191±221, 1995. Swann, P., ``Towards a model of clustering in high-technology industries.'' In G.M.P. Swann, M. Prevenzer, and D. Stout, (eds), The Dynamics of Industrial Clusters: International Comparisons in Computing and Biotechnology. Oxford University Press: Oxford, pp. 52±76, 1998. Teece, D.J., Pisano, G., and Shuen, A., ``Dynamic capabilities and strategic management,'' Strategic Management Journal, vol. 18 no. 7, pp. 509±533, 1997. UNCTAD, World Investment Report 2001. Promoting Linkages, United Nations: Geneva and New York, 2001. UNCTAD, ``Promoting and sustaining SME clusters and networks for development,'' Paper presented at the Expert Meeting on Clustering and Networking for SME Development, September 2±4 UNCTAD Secretariat: Geneva, 1998. UNCTAD, Fostering Technological Dynamism: Evolution of Thought on Technological Development Processes and Competitiveness. A Review of the Literature. United Nations: Geneva and New York, 1996. Visser, E.-J., `À comparison of clustered and dispersed ®rms in the small-scale clothing industry of Lima,'' World Development, vol. 27 no. 9, pp. 1553±1570, 1999. Von Hippel, E., The Sources of Innovation. Oxford University Press: New York, 1988.