Abstract
The article conducts an explorative research on the competitive success of industrial districts (IDs) based on their capacity to adapt and evolve in response to the environmental changes. The aim is to identify the ID structural features supporting adaptation by using the complexity theory. Thus, IDs are considered as complex adaptive systems (CASs) and the ID features that foster adaptation are identified based on the main CAS properties, namely inter-connectivity, heterogeneity and control. To formulate the theory linking the values of the ID structural features with the ID competitive success, a multiple case study is carried out. Finally, three theoretical propositions are derived.
Introduction
Industrial districts (IDs) are geographically defined production systems, characterized by a large number of small- and medium-sized firms that are involved at various phases in the production of a homogeneous product family. These firms are highly specialized in a few phases of the production process and integrated through a complex network of inter-organizational relationships (Becattini, 1990; Porter, 1998). The literature on IDs is quite rich and involves different streams of research, such as social sciences, regional economics, economic geography, political economy and industrial organization. Referring to this literature, studies have mainly provided key notions and models to explain the reasons of IDs’ competitive success. Examples of such models are: the flexible specialization conceptualized by Piore and Sabel (1984); the localized external economies concept anticipated by Marshall (1920) and further formalized by Krugman (1991); the industrial atmosphere notion conceived by Marshall (1919); and the innovative milieux notion developed by the Groupe de recherche européen sur les milieux innovateurs (GREMI) (see, for instance, Maillat, Lecoq, Nemeti, & Pfister, 1995). The foregoing models show that the main critical factors governing the success of ID firms can be traced back to the following ID features: physical and cultural proximity of many small- and medium-sized firms; division of labour among firms; presence within the area of complementary competencies and skills; high degree of specialization of both firms and workforce; existence of a dense network of inter-firm relationships where firms cooperate and compete at the same time; presence of a dense network of social relationships based mainly on face-to-face contacts; and the easy and fast circulation of knowledge and information in the area. These features appear relevant where the competitive context is characterized by both increasing and not particularly sophisticated demand, but they seem to be insufficient to guarantee the IDs’ success in the recent competitive scenario. In fact, by looking at different cases of IDs and, more generally, of geographical clusters spread all over the world, we observe examples of successful and dynamic IDs that are modifying their structures and strategies, thereby losing some of their traditional features, as well as some IDs that are undergoing a declining phase. Despite the great number of cases presented in the literature (Guerrieri, Iammarino, & Pietrobelli, 2001; Porter, 1998; Rabellotti, 1995; Sammarra & Belussi, 2006), only few studies have investigated the reasons of the IDs’ decline (Bathelt, 2001; De Marchi & Grandinetti 2014; Pyke & Tomaney, 1999; Saxenian, 1996; Sull, 2003), and very few try to explain the differences in how IDs compete in the global arena (De Marchi, Lee, & Grandinetti, 2014). Yet, the foregoing studies do not give generalizable reasons of why some IDs fail when they face the new competitive scenario while others not. Nor do they explain why some IDs evolve by assuming different structures to remain competitive and others not. These studies, in fact, adopt a static perspective to analyze IDs by restricting their analyses to the identification of a set of conditions that explain the IDs’ competitive advantage in a particular context. In addition, they focus on the system as a whole and not on the single components (firms), observe the phenomena when they are already happened at the system level and describe them in terms of cause–effect relationship. Our intention is to overcome these limitations by adopting a different approach. We look at the IDs’ competitive advantage as not the result of a set of pre-defined features characterizing them but as the result of their capability to adapt and evolve with the external environment. In fact, if the IDs possess the conditions that allow them to adapt and co-evolve with the environment, they will then modify themselves so as to keep their competitive advantage even when the environment changes. Thus, IDs have a competitive advantage not because they are characterized by a set of features but because they are able to evolve exhibiting features that are the spontaneous result of the adaptation to the environment. Such a result is not known a priori, but emerges from the interactions among the system components and between them with the environment. In line with this perspective, we move beyond the current economic, organization science and economic/spatial geography approaches, by using the complexity theory, a discipline originating in physics and biology. It is increasingly popular in explaining managerial and organizational dynamics (Goldstein, 1994; McKelvey, 2004), strategic change (Houchin & MacLean, 2005), policy (Allen, 1993), entrepreneurship (Peterson & Meckler, 2001) and organizational information systems (Allen & Varga, 2006). Recently, complexity science has also been applied to economic and spatial geography (Batty, Barros, & Junior, 2006; Iandoli, Marchione, Ponsiglione, & Zollo, 2009; Manson & O’Sullivan, 2006; Martin & Sunley, 2012). Amongst the different streams of study in complexity science, we focus primarily on complex adaptive system (CAS) theory, which studies CASs and explains causes and processes underlying emergence in CASs (Axelrod & Cohen, 1999; Gell-Mann, 1994; Holland, 2002; Lane, 2002). CASs consist of an evolving network of heterogeneous, localized and functionally integrated interacting agents. The latter interact in a non-linear fashion and can adapt and learn, thereby evolving and developing a form of self-organization that enables them to acquire collective properties, which each of them does not have individually. CASs have adaptive capability and co-evolve with the external environment, modifying it and being modified. CAS theory identifies as the main features influencing adaptation the inter-connectivity, the level of control and the heterogeneity (Arthur, Durlauf, & Lane, 1997; Holland, 1995). Specifically, it is proved that a moderate level of inter-connectivity, a high value of heterogeneity and a moderate level of control (Kauffman, 1993) are conditions that foster adaptation.
Once IDs have been recognized as CASs (Albino, Carbonara, & Giannoccaro, 2003, 2006; Boero & Squazzoni, 2002; Fioretti, 2001), the CAS theory is used to define the IDs’ features that allow their adaptability in ‘high-velocity’ environments. In particular, these ID features are identified by translating the three CAS properties fostering adaptation into ID structural features. In the light of this argument, the major contribution of this article is to derive theoretical propositions that link the competitive success of the IDs to specific ID structural features and to their values. Based on the literature, we associate the level of inter-connectivity with the number of links among ID firms, the heterogeneity with the diversity among ID firms and the level of control with the governance of the ID’s organizational structure.
We argue that specific values of these features foster the adaptation of IDs in response to environmental changes and therefore improve their performance. Finally, we conduct an explorative empirical research based on a multiple-case-study design (Yin, 1989) with the aim of formulating theoretical propositions on the relationships between the IDs’ structural features that foster adaptation and their performance. We choose four cases of Italian IDs characterized by different competitive performances and compare their structural features. For the comparison, we adopt methods of social network analysis (Borgatti & Everett, 1999; Wasserman & Faust, 1994). Thus, first, we model each ID as a network of business inter-firm relationships, then we identify three network attributes assumed as proxies of the three ID structural features and finally we compare the network structures of the analyzed IDs.
The article is organized as follows: First, the main theories explaining the sources of IDs’ competitive advantages are briefly reviewed. Then, the complexity theory is presented, and the key CAS properties fostering adaptation are defined. In sections ‘The Cases’ and ‘Result of the Cross-case Analysis’, the case studies are discussed and the theoretical propositions presented. Finally, conclusions, implications for managers, policy makers and further research are presented, along with limitations of the study.
Industrial District Competitive Advantage
Traditional Sources of Competitive Advantage for Industrial Districts
Traditional studies on IDs have analyzed the main advantages of IDs and explained the reasons for their success. In particular, the studies of economic geography have pointed out the benefits associated to the ‘agglomeration external economies’, mainly due to the lower input costs, the development of common suppliers, specialist labour pools, spillover of technical know-how and the development of a greater comprehension of the workings of the particular industry by individuals and firms (Becattini, 1990; Marshall, 1920). In the field of organizational sociology and industrial organization, scholars have highlighted the reduction in the transaction costs due to geographical proximity of firms and informal and face-to-face contacts among them as one of the most important benefits of IDs (Asheim, 1996; Aydalot & Keeble, 1988; Powell, 1987). Other studies have stressed that one of the key sources of the IDs’ competitive advantage is their capacity to develop product and process innovations, which mainly result from the presence of highly specialized technical competencies, the existence of networks of formal and informal relationships and the geographical proximity that creates an environment wherein information, codes, languages, routines, strategies and knowledge are easy to be transferred and shared (Cooke, 1999; Cooke & Morgan, 1998; Henry & Pinch, 2002). Synthesizing the results of these studies, the key source of the IDs’ competitive advantage is the static efficiency, namely cost advantages gained by clustered firms, due to a set of features characterizing them which are as follows: the specialization of firms, the presence of a specialized workforce, the division of labour among firms, the accumulation of specific knowledge in the local area, the networking processes among both the economic and social system, the development of a widespread innovative capacity and the presence in the local area of a common system of social and cultural values. However, in the recent years, these factors determining the success of IDs in a competitive context characterized by both increasing and not particularly sophisticated demand seem to be insufficient to guarantee a sustainable competitive advantage to both the system and its firms. In this new context, new sources of competitive advantage that ensure not only the static efficiency but also the dynamic one, namely the ability to proactively respond to the environmental changes, are needed.
Knowledge-based Competitive Advantage of Industrial Districts
Borrowing from the resource-based view and the knowledge-based view theories (Barney, 1991; Grant, 1998; Hamel & Prahalad, 1994; Prusak, 1997), some scholars started shifting their attention from pecuniary externalities (e.g., labour and supplier pooling) to the knowledge-based benefits of clustering. In line with this shift, numerous studies have analyzed the role of knowledge in IDs and have proposed a knowledge-based theory of IDs (Camuffo & Grandinetti, 2011; Malmberg & Maskell, 2004; Maskell, 2001). These works have investigated the nature of knowledge circulating in IDs (Lawson & Lorenz, 1999; Tallman, Jenkins, Henry, & Pinch, 2004), the frequency and the effectiveness of the knowledge transfer processes among ID firms (Bathelt, Malmberg, & Maskell, 2004; Gordon & McCann, 2000; Mesquita, 2007), the learning processes activated by firms in IDs (Albino et al., 2006; Baptista, 2000; Belussi & Sedita, 2012; Carbonara, 2004a; Sull, 2003), the effect of knowledge spillovers on the ID’s innovative capabilities (Boschma & Ter Wal, 2007; Breschi & Lissoni, 2001) as well as on its structure and evolution (Iammarino & McCann, 2006), the relationship between the knowledge flow in IDs and the structural properties of inter-organizational networks (Cowan & Jonard, 2004, 2009; Iandoli et al., 2009), just to mention a few stream of studies.
According to these studies, the key source of IDs’ competitive advantage is their superior capacity to support processes of knowledge transfer and creation and to facilitate innovation (Lorenzen & Maskell, 2004). Opposite to the traditional studies, where the source of competitive advantage is static, based on the possess of given features, in the knowledge-based theory of IDs, the ID’s competitive advantage results from dynamic processes activated by ID firms, namely the learning and knowledge management processes (Arikan, 2009; Tallman et al., 2004).
According to this new perspective, we seek new dynamic sources of competitive advantage by adopting a different theoretical approach, namely the complexity science.
The Key Properties of Complex Adaptive Systems
Amongst the different streams of study of complexity science, we focus primarily on the American School that consists largely of scholars associated with the Santa Fe Institute (Anderson, Arrow, & Pines, 1988; Arthur et al., 1997; Cowan, Pines, & Jonard, 1994; Kauffman, 1993; Pines, 1988; Stauffer, 1987). Drawing from the life sciences and making extensive use of agent-based computational approaches, the American school complexity literature develops the CAS theory, identifying the main properties of CASs and studying their dynamics. A CAS is a system of heterogeneous and interacting agents that emerges over time into a coherent form and adapts and emerges itself without any singular entity deliberately managing or controlling it (Holland, 1995). The agents interact in a non-linear fashion, can adapt and learn, thereby evolving and developing a form of self-organization that enables them to acquire collective properties, that each of them does not have individually. CASs adapt to changing environmental conditions without any singular entity deliberately managing or controlling them and co-evolve with the external environment, modifying it and being modified (Axelrod & Cohen, 1999; Gell-Mann, 1994; Lane, 2002). Based on the main contributions of CAS theory, three main structural properties of CASs that foster their adaptive capacity can be identified, namely the inter-connectivity, the level of control and the heterogeneity.
Inter-connectivity
CAS theory identifies the number of interconnections within the system as a critical condition for self-organization and emergence. Kauffman (1995) points out that the number of interconnections among the agents of an ecosystem influences the adaptive capacities of the ecosystem. He uses the NK model to investigate the rate of adaptation and the level of success of a system in a particular scenario. The adaptation of the system is modelled as a walk on a landscape. During the walk, agents move by looking for positions that improve their fitness represented by the height of that position. A successful adaptation is achieved when the highest peak of the landscape is reached. The ruggedness of the landscape influences the rate of adaptation of the system. When the landscape has a very wide global optimum, the adaptive walk will lead towards the global optimum. In a rugged landscape, given that there are many less differentiated peaks, the adaptive walk will be trapped on one of the many suboptimal local peaks. By using the concept of tunable landscape and the NK model, Kauffman (1995) demonstrates that the number of interconnections among agents (K) influences the ruggedness of the landscape. As the number of interconnections among agents (K) increases, the ruggedness increases and the rate of adaptation decreases. Therefore, in order to assure the adaptation of the system to the landscape, the value of K should be moderate. This result has been applied in organization studies to modelling organizational change and technological innovation (Kauffman, Lobo, & Macready, 2000; Rivkin & Siggelkow, 2002). In organization studies, K represents the connectivity among the organization’s members, namely the extent to which components of the organization affect each other. According to Kauffman’s NK model, adaptive organizations are those which are able to evolve towards a dynamic equilibrium at the edge of chaos, as such they need to be moderately coupled, since for modest level of connectivity, organizations are in an ordered state, and for large level of connectivity, they remain in the chaotic regime (Kauffman & Macready, 1995).
Heterogeneity
Different studies on complexity highlight that variety destroys variety. For example, Ashby (1956) suggests that successful adaptation requires a system to have an internal variety that at least matches environmental variety. Systems having agents with appropriate variety will evolve faster than those without. The same topic is studied by Allen (2001), LeBaron (2001) and Johnson (2000). Their agent-based models show that novelty, innovation and learning all collapse as the nature of agents collapses from heterogeneity to homogeneity. Dooley (2002) states that one of the main properties of a complex system that supports the evolution is diversity. Such a property is related to the fact that each agent is potentially unique not only in the resources that it holds, but also in terms of the behavioural rules that define how it sees the world and how it reacts. In a complex system, diversity is the key towards survival. Without diversity, a complex system converges into a single mode of behaviour.
Level of Control
The governance of a system is another important characteristic influencing a CAS’s self-organization and adaptive behaviours. Le Moigne (1990) observes that CASs are not controlled by a hierarchical command-and-control centre but instead manifest a certain form of autonomy. The latter is necessary to allow the system evolution and adaptation. A strong control orientation tends to produce tall hierarchies that are slow to respond (Carzo & Yanousas, 1969) and invariably reduce heterogeneity (Jones, 2000; Morgan, 1997). The presence of ‘nearly’ autonomous sub-units characterized by weak but not negligible interactions is essential for the long-term adaptation and survival of organizations (Sanchez, 1993; Simon, 1996). Furthermore, Granovetter’s (1973) research finding is that novelty and innovation happen more frequently in networks consisting mostly of ‘weak ties’ as opposed to ‘strong ties’. The latter tend to produce ‘groupthink’ (Janis, 1972).
The Complex Adaptive System Properties and the Industrial District Features
The basic assumption of this study is that IDs are CASs, since they exhibit different CAS properties (Albino, Carbonara, & Giannoccaro, 2005; Boero & Squazzoni, 2001; Borrelli, Ponsiglione, Iandoli, & Zollo, 2005; Carbonara, Giannoccaro, & McKelvey, 2010; Curzio & Fortis, 2012). Table 1 synthesizes the main CAS properties and the corresponding features of IDs.
Once IDs are recognized as CASs, by exploiting the analogy between CASs and IDs, the ID features that foster its adaptive capacity may be identified. In the following, the ID features that can be associated with the three CAS properties enabling adaptation, that is, inter-connectivity, heterogeneity and level of control, are defined.
Framing IDs as Complex Adaptive Systems
The inter-connectivity property of CASs is associated with the number of business and social linkages among ID firms, namely the inter-connectivity among ID firms. Business linkages among firms of the same ID are due to the horizontal and vertical labour division characterizing the ID production model and these links may occur for any business matters, for example, trade of inputs, participation in the same business association and exchange of information. Social links are established thanks to the face-to-face contacts and the friendship and kinship existing among employees of different firms.
The heterogeneity property of CASs is associated with the diversity among ID firms, which in turn depends on their resources, knowledge, competitive strategies, etc. As such, we consider that the ID heterogeneity is influenced by the structure of the network of relationships among firms. In fact, since the network of relationships acts as a means of diffusion of information, knowledge and competitive strategies (Giuliani, 2007; McEvily & Zaheer, 1999), it tends to increase the homogeneity across the firms.
Finally, we associate the level of control of CASs with the governance of the ID’s organizational structure. Two main alternative ID’s organizational structures, each characterized by a different type of governance, may be identified: the so-called Marshallian ID, where firms are mostly independent of each other, and the ID with the leader firms, characterized by the existence of firms taking a leader position in the district that assume a role of managerial guide and control on their network (Albino et al., 2006).
Aiming at formulating theoretical propositions on the value that the ID’s structural features should exhibit to foster adaptation in response to environmental changes, we develop an explorative empirical research based on multiple case studies.
Methodology
Research Design
The explorative research adopts a multiple-case-study approach (Yin, 1989). This methodology is particularly appropriate when theory building is the main aim of the research and a further exploration on the constructs of the theory is required. The cases are four IDs localized in Southern Italy. The selection was driven by the need to compare polar cases of declining and successful districts. Eisendhardt (1989) in fact argues that theory building from case studies can be enhanced by choosing cases that highlight extreme situations or polar types in which the phenomenon under investigation is observable. We focused on made-in-Italy sectors and chose the four cases on the basis of mainly secondary data and previous studies. At the end of this activity, we identified the four following IDs: (a) Barletta knitwear district, (b) Andria underwear district, (c) Barletta footwear district and (d) Foggia agro-food district. To measure the competitive performance of the four cases, we used two different measures based on firms’ accounting performance: the return on investment (ROI) index, that is, the ratio between the operating profit and the capital employed, and the value added (VA), that is, the difference between the total sales revenue and the total cost of components, materials and services purchased from other firms. These indicators are based on firms’ published profit and loss and balance sheet statements, which are typically created using widely accepted accounting standards and principles. The application of these standards and principles makes it possible to compare the accounting performance of one firm to the accounting performance of other firms, even if those firms are not in the same industry (Barney & Hesterly, 2010). In particular, the performance index of each district based on the ROI is calculated asking each firm of a selected sample of the district firms to indicate whether its ROI is lower, equal or higher than the average ROI of the sector. Then we have assigned the value of 1, 2 and 3 to each firm, respectively. Value 1 means that the firm earns below-industry average accounting performance and corresponds to a competitive disadvantage, value 2 means that the firm earns industry average accounting performance and corresponds to a competitive parity, whereas value 3 indicates that the firm earns above-industry average accounting performance and corresponds to a competitive advantage. The performance index of the district is the average across the performance index of the sampled firms. As for the VA, national statistical data are used. The higher the value of VA, the higher is the competitive advantage of the district, since it measures the economic value created by a firm.
To compare the structural features of the four considered IDs, we used methods of social network analysis. 1 Thus, we modelled each ID as a network of business inter-firm relations and identified a set of network attributes assumed as proxies of the three ID structural features. Specifically, we assume that the network density, the network heterogeneity and the network centrality as proxies of the interconnectivity level of the ID, the diversity among ID firms and the governance of the ID’s organizational structure, respectively.
The network density (ND) is defined as the proportion of possible linkages that are actually present in a graph. ND is calculated as the ratio of the number of linkages present, L, divided by its theoretical maximum in the network, n(n – 1), 2 with n being the number of nodes in the network (Borgatti & Everett, 1997; Wasserman & Faust, 1994):
The network heterogeneity is referred to the ‘coreness’ of each actor in the network, where the coreness is the degree of closeness of each node to a core of densely connected nodes observable in the network. Using actor-level coreness data, we calculated the Gini coefficient, which is used as an index of heterogeneity (Borgatti & Everett, 1999). The index aims at capturing the level of diversity of actors in the network, where the diversity is measured by the different distribution of linkages. The Gini coefficient ranges from a minimum value of zero to a theoretical maximum of one. The Gini coefficient decreases together with the firms’ coreness diversity, for example, a low Gini coefficient indicates that the agents in the network have a great differentiation of their coreness. This, in turn, means that agents are equally connected. On the contrary, a high Gini coefficient means that the actors in the network are characterized by a different distribution of their linkages.
The network centrality is calculated by the normalized average degree centrality of the nodes and the normalized average closeness centrality. The degree centrality of a node DC(i) is defined as the number of edges incident upon that node. Thus, degree centrality refers to the extent to which an actor is central in a network on the basis of the ties that it has directly established with other actors of the network. This is measured as the sum of direct linkages x of node i with other j nodes of the network:
To compare network of different size, it is better to use the average normalized degree centrality, calculated by averaging the normalized degree centrality of each node, where the normalized degree centrality is the degree centrality divided by the maximum possible degree expressed as a percentage (Borgatti & Everett, 1997).
The closeness centrality is most frequently used to measure relative access to network resources and information and can be interpreted as measuring the degree of independence from others in the network (Borgatti & Everett, 1997). It is inversely proportional to the total geodesic distance from the node to all other nodes in the network. Geodesic distance is defined as the length (number of edges) of the shortest path linking two nodes (Freeman, 1979). As for the degree centrality, to compare the closeness centrality of the different networks, we use the normalized averaged closeness centrality. Network measures were evaluated using the network analysis software toolkit UCINET 6.0 (Borgatti, Everett, & Freeman, 2002).
Sampling
A sample of firms has been selected within each ID using a stratified sampling technique rather than a random one. In particular, we have stratified the population on the basis of the production phase that the firm performs (production stages) or the manufacturing specialization of the firm. In this way, the risk that the network of business inter-firm relationships is influenced by a random selection of firms is reduced. In fact, the random sampling technique could select disproportionally firms belonging to the different stages.
Data
Data Collection
The study is based on micro-level network data, collected at the firm level in the four IDs. To collect data, we conducted interviews on the field with managers and owners of the companies. Interviews were carried out on the basis of a structured detailed questionnaire aimed at defining networks of business interactions. In particular, each firm was provided with a list (or roster) on which the sample firm names were already given. The firm had to indicate on the list the firms in the sample with whom they have business exchange. In particular, the respondents were asked to respond to the following question: ‘With which of the firms mentioned in the roster do you interact for business matters?’
The resulting network of business interactions maps the presence of linkages among firms of the same IDs, which may occur for any business matter, for example, trade of inputs, participation in the same business association and exchange of information.
Macro-data on the four IDs have been collected from the 2011 edition of the Italian Service and Industry Census (ISTAT, 2011).
The fieldwork was conducted over a period of about 12 months, starting from 2012.
Data Analysis
Data analysis consists of two steps: first, we carried out the case studies and then, we employed a cross-case analysis aimed at identifying similarities and differences among the analyzed cases and generating hypotheses (Sammarra & Belussi, 2006).
The Cases
The Barletta knitwear district is specialized in the knitwear production and the main products are: casual clothes, women dress, woollens, sweatshirts, etc. Firms are classified into the following five main stages of the production process:
Yarn and fabric producers, Screen printings, which put trimmings, make moulds and print fabrics, Laundries, which dress the clothes and iron them, Knitwear producers, who make the knitted clothes and Garment makers, who assemble the final products and sell them to the distributors.
In order to analyze the whole structure of the local district, we selected a stratified sample of 217 firms (Table 2).
The Andria underwear district is located in the province of Bari. The district is specialized in the underwear production and the main products are as follows:
Ladies underwear and lingerie (e.g., panties, petticoats, bras), men underwear (pyjamas, underpants); Woollen underwear (undershirts, panties, petticoats, etc.).
We chose a stratified sample of 188 firms specialized in five different production phases: yarn and fabric producers, screen printings, woollen underwear producers, laundries and underwear makers (Table 3).
Key Information on the Barletta Knitwear District and the Sampled Firms
Key Information on the Andria Underwear District and the Sampled Firms
The Barletta footwear district is located in the newborn province of Barletta-Andria-Trani, in the Puglia region. Small- and medium-sized firms are concentrated on the production of casual shoes for lower market spheres, mostly provided with so-called injected soles. Some of the larger firms in the district are specialized in the production of more advanced safety shoes to use in dangerous working circumstances. District firms encompass branded and non-branded end-product manufactures, several specialized suppliers (such as shoestring manufacturers, diesinkers, heels manufacturers, etc.) and subcontractors (such as upper and sole producers). In order to analyze the structure of the district, we selected a stratified sample of 179 firms (Table 4).
The Foggia agro-food district is located in the province of Foggia, in the Puglia region. Agriculture and agro-industry represent the mainstay of the economy of the province of Foggia, as confirmed by the total agricultural surface area of the province that exceeds 560,000 hectares, the number of farms (about 61,000) and the number of agricultural employees (the incidence of agriculture employment is 13.9 per cent in the province of Foggia versus the average national value of 4.4 per cent). The area, called ‘Tavoliere’, is among the major Italian producers of tomatoes, olives, wine grapes and vegetables, but also shows high production ratios for vegetable oils, the processing and preservation of fruit and vegetables and the production of corn seed and starch products. Alongside the primary cultivation activity, a reasonably sized satellite district of Small and medium enterprise (SMEs) specialized in the food processing, food packaging and fresh fruit and vegetables conditioning has developed in the last few decades. In the district, we have counted 295 firms out of which we chose a stratified sample of 142 firms specialized in the three main agro-food productions characterizing the district, namely the olive oil, the wine and the fruit and vegetable production (Table 5).
Key Information on the Barletta Footwear District and the Sampled Firms
In order to analyze the structure of the four IDs, on the basis of the information collected through the questionnaires, we have first built the incidences matrix for each case and then represented the district network by using the software UCINET. The network attributes measured for the four IDs are reported in Table 6.
Result of the Cross-case Analysis
The four networks are compared in terms of network attributes and economic performance (Table 7).
Figure 1 plots the values of the network density, used as a measure of the inter-connectivity among firms, and the performance indexes for the four networks. Figure 1 shows an inverse U-shaped relationship between the performance indexes and the network density, in fact, the performance indexes increase when the network density increases, but till a certain value (threshold). When the network density overcomes the threshold, the performance indexes decrease. This result is coherent with what CAS theory affirms on the relation between the level of inter-connectivity of the CAS and its capacity to adapt to the environment and successfully evolve. Too many links make the system of firms more vulnerable. In fact, for example, in the Barletta footwear district, firms are densely connected by buyer–supplier relationships. As a consequence, the decline of some buyer firms as well as the strategic decisions of some other buyers to delocalize great part of production in Albania and Romania have decreased the performance of the connected firms and determined the district crisis. Too few links make the district firms isolated and, in such conditions, they lose the benefits of the information and knowledge sharing. On the basis of the discussion above, we formulate the following proposition:
Proposition 1. The ID performance first tends to increase and then to decline with the number of links among ID firms. It seems that a threshold of the number of links among ID firms exists.
Key Information on the Foggia Agro-food District and the Sampled Firms
Network Attributes
Economic Performance
Figure 2 plots the values of the Gini coefficient, used as measures of the ID heterogeneity, and the performance indexes of the four districts.
As the Gini coefficient increases, the performance indexes also increases. Being the Gini coefficient a proxy of the network heterogeneity, the obtained finding is consistent with the CAS theory that highlights the importance of the heterogeneity as a source of innovation and adaptiveness. A low Gini coefficient characterizes a network where the agents are equally connected and thus have the same network of relationships. This determines an even distribution of information, knowledge and competitive strategies, which flow through the network of relationships, and then increase the homogeneity of the system. On the contrary, a high Gini coefficient means that the actors in the network are characterized by different distribution of linkages. This implies an uneven distribution of information, knowledge and competitive strategies and then a greater heterogeneity in the system.
In the Foggia agro-food district and in the Barletta footwear district, both presenting a low value of the Gini coefficient, firms have adopted similar manufacturing and marketing strategies so as to reduce the variety of the strategic behaviours in the district. In particular, a great part of the firms operating in the Foggia district is mainly involved in the first stages of the agro-food supply chains and they are scarcely distributed on the downstream stages of the supply chain. Such a condition has negatively affected the performance of the district firms, which do not have a direct control on the final market and are unable to add value to their products. As regards the Barletta footwear district, most of the firms have focused on lower market segments, for many years, they have adopted conservative strategies with no propensity to innovate and to upgrade the locally embedded knowledge. This behaviour has guaranteed the survival of the district in a not particularly dynamic market, but recently, when the market has become more competitive, due to a more sophisticated demand of the consumers oriented to brand names and due to the increasing competition from low-cost countries, has led to the district crisis. The Andria underwear district has faced the same change in the competitive context but the differentiation of firms in terms of capabilities and knowledge has favoured the formulation of different competitive strategies, for example, some firms created a brand or acquired licensing exploiting their marketing capabilities, some others focused on niches such as the handmade embroidered products or low-cost products exploiting their manufacturing capability and some others became subcontractors of large firms external to districts exploiting their capacity to provide high logistics performances. Such a differentiation allowed to keep low the rivalry, namely the intensity of competition, among the ID firms and to protect the district from the crisis. Therefore, we posit that:
Proposition 2. The more uneven the distribution of the linkages across the ID firms, the higher is the ID performance.
Figure 3 shows the values of the normalized average degree centrality/closeness centrality, used as measures of the level of control inside the ID, and the performance indexes of the four networks.
As the average degree centrality/closeness centrality increases, the performance indexes first increase and then decline. Being the average normalized degree centrality/closeness centrality a proxy of the level of control/level of independency of the actors in the network, this trend is coherent with the CAS theory which puts in evidence the need to balance hierarchy and autonomy. The Barletta footwear district is characterized by a high level of normalized average degree and closeness centrality. This is due to the presence in the ID of a few large-sized firms that assume a focal position in the network, since they create their own supply network and manage it by using a centralized control. This narrows the information flows inside each supply network established around each large focal firm without any sharing among them. On the other extreme, the Foggia agro-food district is characterized by firms that are scarcely interconnected by both vertical and horizontal linkages. They neither are integrated along the supply chain nor cooperate in the same stage of the supply chain. This reduces the possibility to offer to the market a wider range of products, to implement a system of product traceability that ensures the quality and security of products and to increase the contractual power of the district SMEs towards the big buyers.
The districts with intermediate level of normalized average degree and closeness centrality, namely the Andria underwear district and the Barletta knitwear district, present higher performances. The supply networks inside these IDs are mainly managed by using a decentralized control, where each firm makes independent production and inventory decisions. Thanks to this form of governance, the IDs gain the advantages of the flexible specialization model and keep their performance high.
Therefore, we hypothesize that:
Proposition 3. ID performance first increases with the level of control of the ID organizational structure and then tends to decrease. This suggests the presence of a threshold of the level of control of the ID organizational structure.
Conclusions
This article uses complexity science concepts to offer a new perspective on the theoretical understanding of the IDs’ competitive advantage. In particular, complexity science has been used as a conceptual framework to investigate the reasons of the competitive success of IDs. This approach is particularly valuable since it allows overcoming the limits of traditional studies on IDs. According to this approach, the competitive advantage of IDs is not the result of a set of pre-defined features characterizing them, but it is the result of dynamic processes of adaptability and evolution of the IDs in response to environmental changes. Therefore, the ID’s competitive success is linked to its adaptive capacity that is a key property of CASs. Using the theory of CASs, the key structural features of IDs that give them the adaptive capacity have been identified, namely: (a) the inter-connectivity level of the ID; (b) the diversity among ID firms and (c) the governance of the ID’s organizational structure. An explorative research adopting a multiple-case-study approach has been conducted. It involved four in-depth case studies on Italian IDs selected as polar cases of declining districts and successful ones. The IDs have been compared in terms of CAS properties using the social network theory and measures.
In particular, the networks of the business inter-firm relationships have been mapped and then their attributes of network density, Gini coefficient, degree centrality and closeness centrality have been calculated. These attributes are used as measures of the ID structural features fostering adaptation: the network density is used as proxy of the inter-connectivity level of the ID, the Gini coefficient as proxy of the diversity among ID firms and the degree centrality and closeness centrality as proxies of the governance of the ID’s organizational structure. Comparing the four networks, three theoretical propositions on the relationships between the IDs’ structural features that foster adaptation and their performance are formulated. In particular, the research findings suggest that the ID performance first increases and then decreases as the number of links among firms and the level of control of the ID’s organizational structure increases, suggesting the existence of a threshold. Moreover, a high heterogeneity in the distribution of the links across the ID firms assures higher ID performance.
These findings have broader policy implications. For example, policies addressed to sustain IDs’ competitive advantage should be devoted to avoid the formation of overcrowded networks of links among ID firms, to increase the heterogeneity among firms and to assure the balance between hierarchy and autonomy in the ID’s organizational structure.
More practically, policy makers should promote the formation of hierarchical and structured organizational forms, such as consortia and holdings. Also, the creation of networks controlled by leader firms should not be avoided, even though these networks introduce a top-down control in opposition to bottom-up autonomy. There are examples of IDs in Emilia Romagna that have profoundly changed their organizational structure by moving from the traditional Marshallian model towards configurations characterized by higher levels of hierarchy and by the presence of business groups. The existence of leader firms favours the adoption of best practices and speeds up the change needed to face the competitive scenario. In addition, to keep a high level of heterogeneity, policies should be addressed to limit the imitation, by promoting patents, brand policies, etc.; to favour the admission of firms located outside IDs; and to stimulate the exchange of knowledge, skills and competence with the external environment, by hiring managers coming from different industries and regions, graduates from diverse universities and by creating links between ID firms and external players worldwide, such as universities, research centres and other firms.
We recognize certain limitations in the study. In particular, the conceptual framework driving the field research is built on a limited set of CAS properties. Then, in order to refine the propositions grounded on empirical evidences, further research should be devoted firstly to incorporate in the framework a wider set of CAS properties and then to operationalize them as IDs features. For instance, key constructs such as emergence or co-evolution should be added in the framework and then operationalized in the context of the ID.
The incorporation of longitudinal data would also strengthen the results emerging from this study and validate the ‘coevolutionary view’ of successful IDs that act as CASs. Thus, further researches should be devoted to improve empirical investigation and test extensively the hypotheses.