DYNAMICS AND PERFORMANCE DETERMINANTS IN CLUSTERS OF FIRMS: A COMPUTATIONAL APPROACH

Objective of the Department of Computer Science:

The aim of the Computer Science Department is at building computer simulation models for the analysis of emerging cluster dynamics and the investigation of success and growth determinants. The research team in particular focuses upon de-localisation dynamics of Italian firms. Computer simulations support understanding of causes of de-localisation dynamics and anticipate long-term consequences of alternative strategies.

Theoretical Background

It is well-known that clustered networks of small and medium firms constitute the prevalent industrial structure of the Italian Center-North and North-East (Becattini 1975, 1979; Brusco 1979; Lorenzoni 1979), but similar arrangements can be found in productive systems as diverse as Scandinavia (Maskell 1998), the ''Silicon Valley'' as well as the software districts in the emerging countries of East Asia (Saxenian 1996, 2002; Kennedy and von Burg 2001). Obviously, such a wide diffusion comes along with an extreme diversity of organizational and structural forms.Both because of their diffusion and because of their structural variety, networks of firms are key to characterize and improve the performance of national productive systems (Pyke and Sengenberger 1990; Porter 1998). In particular, several authors remarked that since the time when Becattini recognised in Italy the features of the ''industrial districts'' depicted by Marshall, many features have changed.
Those agglomerations of large numbers of minuscule firms specialised in tiny fractions of the productive process often gave way to more complex organizational forms, in which the variety of firms size contributes to the richness and multiplicity of interactions (Park and Markusesn 1995; Nuti and Cainelli 1996; Rabellotti and Schmitz 1998; Lazerson and Lorenzoni 1999). Some attempts have been made to understand in which directions the structures and networks of firms may evolve and, possibly, to identify a lifecycle of industrial clusters (Carminucci and Casucci 1997). Unfortunately, the diversity of industrial clusters and the heterogeneity of research methodolgies have posed huge barriers to systematical investigations (Morrison and Staber 2000).Indeed, although business networks and industrial clusters are high in research agendas we lack tools to evaluate their performance and predict their dynamics.
Essentially, the problem is that the clusters of networked firms exhibit an aggregate behavior that results from self-organization of local decisions. Consequently, the tools of mathematical analysis do not apply. On the contrary, simulations and computational tools provide a promising avenue of research (Swann 1998; Beghella e Becchetti 2000; Clark 2000; Quadrio Curzio e Fortis 2002).Several studies in spatial economics pioneered this methodological shift by making use of simulations in order to investigate the spontaneous aggregation of firms in a particular geographical area (Krugman 1992, 1996, 1997; Krugman, Venables e Fujita 1999). The spectrum of available techniques is growing rapidly. Among simulation methodologies, we may identify at least the two following streams.On the one hand, systems dynamics models are appropriate to highlight the feedbacks that arise in clusters through the interaction of the strategies pursued by the single firms. These feedbacks at the micro level may originate collective macroscopic dynamics of a considerable complexity (Forrester 1961, 1968; Goodman 1983; Sterman 2000).
It is a top-down approach aimed at identifying those relational structures that may trigger explosive and destructive dynamics for the cluster as a whole. Consequently, the systems dynamics approach is able to detect key microstructures for the aggregate behaviour of a cluster ((Delauzun e Mollona, 1999; Mollona, 2001; Marafioti e Mollona, 2000; Coda e Mollona, 2002, Garzia e Mollona, 2002 )).On the other hand, connectionist models lend themselves to a bottom-up modelisation of industrial clusters. After initial attempts with neural networks (Giaccaria 1997) and cellular automata (Brusco, Minerva, Poli and Solinas 2002), agent-based models conquered the scene (Cavezzali and Rabino 2003). Recently, a series of agent-based models of industrial clusters appeared on the Journal of Artificial Societies and Social Simulation, the leading journal in this rapidly growing field (Brenner 2001; Fioretti 2001; Squazzoni and Boero 2002; Albino, Carbonara and Giannoccaro 2003). Just like bottom-up does not oppose top-down, agent-based models are complementary to systems dynamics models. In fact, the reliability of agent-based models is greatly enhanced if systems dynamics macro-equations are available, that anchor the results of agent-based models to a well-known framework.
. The scope of computational approaches is likely to be greatly enhanced by the diffusion of information and communication technologies (ICT). In fact, the need of large amounts of electronically-mediated communication may enhance the need of simulation models that aid coordination between geographically distant productive units.Albeit the international dimension of business networks has always been important (Rullani 1993; Grandinetti and Rullani 1994), their extension has increased dramatically as a consequence of the emergence of competitors from East Asian countries, often in the very industries where Italian firms traditionally specialised. Moreover, the possibility of moving production to neighbouring East European countries with lower labour costs offers other possibilities to the adoption of ICT, that are still waiting to be exploited (Crestanello and Tattara 2004). ICT have upset the previous architecture of businesses networks. Delocalisation of production plants has become a paramount issue for many small and medium Italian firms (ICE 2003). Networks change in shape and function, and ICT allow the existence of electronic markets according to novel institutional arrangements (Lucking-Riley and Spulber 2001; Sabbatini 2003).
This context may pave the way to virtual clusters of firms that are not necessarily geographically proximate, though occasional face-to-face meetings are still necessary. In this respect, many Asian countries are quite advanced (Fioretti 2003).Consequently, several initiatives are under way in order to develop communication protocols for business networks. We may mention the project National Information Infrastructure Protocols in the U.S. (NIIP 1993), mainly for the mechanical industry, and the project Middleware Tools and Documents to Enhance the Textile/Clothing Supply Chain through xML in Italy (MODA-ML 2003).At the same time, the quality of communication protocols is improving very rapidly. Notably, communication protocols are becoming ever more able to take account of the psychological aspects of interpersonal communication. In this respect we may point to the project EMCORE (Fasoli and Messina 2001), which developed group communication software by integrating the competences of psychologists and information scientists. With increasing use of ICT, both systems dynamics and agent-based models may be utilised with greater ease to coordinate firms and help them conceive their strategies. In this respect, dedicated platforms such as the Java Enterprise Simulator (Terna 2002, 2004) offer unprecedented possibilities. Preliminary tests suggest that by integrating computational models with communication protocols it may become possible to reconstruct and to predict the dynamics of clusters of firms in real time. We deem that the amount of data that will be made available by ICT tools will enable performance evaluations to an extent that could not be attempted hitherto. In combination with simulation models, these data may suggest actions for improving unsatisfactory performances.

Detailed description of roles of team members

The research project on the evolutionary dynamics and the performance of clusters of firms is articulated in the following phases:

1. Data gathering and Building of preliminary theoretical frameworks.
2. Building of a ''virtual laboratory'' and Computer simulation experiments.

The Department of Computer Science contributes to both phases.

PHASE I: Data gathering and Building of preliminary theoretical frameworks.

In the first phase, the Department contributes to the research programme by both collecting data and building hypotheses on how software tools may lead to improvement in communication flows among agents in different organisations within a cluster.

1.a - Data gathering.
The Department will develop software to monitor information flows among organisations within a cluster. The main idea is to understand how groupware tools, that is, tools allowing interaction between people even many miles apart, affect a cluster's performance.

1.b - Construct definition.
The Department will elaborate constructs in order to observe and analyse the relationship between characteristics of communication flows among actors within a cluster of firms and emerging performances of the cluster. The objective is to elicit the role of groupware in facilitating information and communication flows. Groupwork and exchanges of information between people belonging to different organisations are key to establish interactions and building synergies within clusters of firms. These activities are usually carried out either face-to-face or by phone, with a marked preference for the first type of interaction at the initial stages of a new relationship. The diffusion of computers should ease this sort of interactions, for instance because it allows both synchronous and asynchronous interaction between people even many miles apart.

1.c - Building of a preliminary frameworks of hypotheses.

The Department will develop a set of hypotheses on how the structure of communication flows among actors within a cluster of firms influences the performances of the cluster. In spite of remarkable efforts by researchers of computer science and software developers, so far no groupware tailored to collaborative worh by the means of computer has spread so rapidly as it was expected. In our opinion, the main reason is that all technical solutions developed hitherto addressed and solved the problems of human-computer interactions and computer-computer interactions, but they are very unsatisfactory as far it concerns human-human interactions. Since the pioneering work of Bion (1961), psychologists know that a group of people assembled to reach a goal - as it is the case of a workgroup - may engender disruptive dynamics that might impair its functioning. In collaboration with a group of psychologists, some of the proponents of this project developed a tool for human-human interaction in a computer-mediated workgroup (EMCORE, Fasoli & Messina 2001). The goal of this project is to extend the applications of EMCORE to the analysis of discussion groups in an industrial cluster and to obtain useful information regarding optimal strategies.

PHASE II: Building of a ''virtual laboratory'' and Computer simulation experiments.

In the second phase, the Department will focus on modelling and computer simulation experiments. Computer simulation models are used as 'laboratories' where unfolding consequences of different policies or effects of alternative environmental scenarios can be tested. Analysis of simulation experiments, thus, supports theory building by providing a virtual environment where set of hypotheses can be tested for robustness and consistency.

Modelling activities are grouped into three steps:

a. Building of computer simulation models on the grounds of data collected in the empirical research phase.
b. Test of simulation models.
c. Simulation experiments, Analysis of results, Validation and Articulation of theoretical framework.

Department of Computer Science will develop computer models to be used by the other research units to articulate and corroborate set of hypotheses, which have been developed in the first phase of the research. The former three modelling steps will apply to two different models, which will be developed employing two different techniques:

a. Models of dynamic systems based on difference equations
b. Agent-based models

(a) Models of dynamic systems based on difference equations

According to this approach, networks of decision processes, information flows and interpersonal relations translate into structures of feedback loops and time lags between decisions, actions and their consequences (Forrester, 1961; Mollona, 2000). These structures can be described by differential equations, eventually approximated by difference equations. Because of their structural complexity, analytical techniques cannot be applied to these systems. On the contrary, computer simulations offer numerical solutions that allow understanding how a system evolves out of different initial conditions and reacts to the impact of different environmental perturbations.
This modelling technique will be applied to a simulation model dealing specifically with the following issues:
1. Feedback processes connecting the cluster originally including firms that moved production to low-wage countries and the clusters that eventually emerged in destination countries
2. Time lags and inertial dynamics due to processes of resource accumulation within clusters.

(b) Agent-based models

Agent-based models assume that economic systems can be modelled and understood as dynamical systems that evolve out of interacting but autonomous learning agents (Lane, 1993; Fioretti, 2001). Thus, agent-based models lend themselves to economic applications of a fundamental paradigm of the theory of complex adaptive systems. In fact, a crucial feature of agent-based models is the emergence of ordered structures independently of top-down planning.
Examples include the spontaneous formation of monetary standards and market protocols. Agent-based models provide virtual laboratories to test the impact of alternative institutions and organisational models for industrial clusters. The normative effectiveness of such 'laboratories' reflects into the explanation of observed behaviour and the formulation of possible development trajectories for clusters of firms.

Agent-based models are particularly well suited to research projects that aim at understanding the processes by which macro-structures at the level of a cluster emerge from local relational micro-structures. Therefore, this approach will be used to address the following issues:
1. Analysis of efficacy and effectiveness of alternative communication and interaction mechanisms among agents within a cluster and between different clusters.
2. Analysis of inter-agents dynamics leading to the emerging of a cluster of firms.
3. Analysis of the role of specific actors that might play as facilitators in establishing and influencing the relationships between other actors.

BRIEF DESCRIPTION OF TEAM MEMBERS AND THEIR ROLES

Edoardo Mollona is in charge of:
1) difference equation modelling
2) agent-based modelling.

Edoardo Mollona is associate professor of Business Economics at the University of Bologna, faculty of Sciences. Since 2002 he is teaching Management Science and Decision-Making in Complex Organisations to the students of Internet Sciences. He obtained a PhD at the London Business School in Strategic Management and Decision Sciences. Edoardo Mollona studies the application of complexity theories, evolutionary approaches and computational techniques to organisation theory. In particular, his research is focused on modelling and simulating complex organisational systems. He authored 2 books on the applications of systems theory to organisations science. Furthermore, he published a number of scientific papers on national and international journals as well as in conference proceedings. Edoardo Mollona has been programme and conference chair in a number of, national and international, conferences dealing with system approach to corporate strategy and organisation. He is founding member of the Italian Chapter of the International System Dynamics Society (SYDIC) and, in 2002, contributed to found the Laboratory of System Modelling at the University of Lugano.

Antonio Messina is in charge of:
1) analysis of groupware impact on cluster performance;
2) agent-based model building.

Antonio Messina is associate professor at the University of Bologna, faculty of Sciences. He teaches Distributed Computation Environments and Coordination to master students in computer science. Thanks to the skill developed in numerical techniques and information technology in general, from 1994 to 1996 Antonio Messina was IT project leader of DREAM (Dynamic Rotorcraft Evaluation by Advanced Modeling), a project supported by the Italian aeronautical industry. In the second half of the 1990s, his academic research addressed the study of coordination languages applied to numerical fluidodynamics simulations and computer supported collaborative work. Under contract of a government organization and in collaboration with the students of his course in computer science, Antonio Messina developed the workflow tools JEPA (1999), as well as a tool for collaborative work to enhance the possibility of including human-human interaction in computer supported collaborative communication. Antonio Messina published a number of papers in national and international journals.

References

Albino V., Carbonara N. e Giannoccaro I. (2003) Coordination Mechanisms based on Cooperation and Competition within Industrial Districts: An agent-based computational approach. Journal of Artificial Societies and Social Simulation, 6 (4).

Becattini G. (1975) Lo sviluppo economico della Toscana, con particolare riguardo all'industrializzazione leggera. Firenze, IRPET.

Becattini G. (1979) Scienza economica e trasformazioni sociali. Firenze, La Nuova Italia.

Beghella M. e Becchetti L., curatori (2000) The competitive advantage of industrial districts: Theoretical and empirical analysis. Heidelberg, Physica-Verlag.

Brenner T. (2001) Simulating the Evolution of Localised Industrial Clusters - An Identification of the Basic Mechanisms. Journal of Artificial Societies and Social Simulations, 4 (3).

Brusco S. (1979) La relazione tra dimensione e saggio di sviluppo nelle imprese industriali: una ricerca empirica. Modena, Stem Mucchi.

Brusco S., Minerva T., Poli I. e Solinas G. (2002) Un automa cellulare per lo studio del distretto industriale. Politica Economica, 18 (2): 147-192.

Carminucci C. e Casucci S. (1997) Il ciclo di vita dei distretti industriali: ipotesi teoriche ed evidenze empiriche. L'industria, 18 (2): 283-315.

Cavezzali A. e Rabino G. (2003) Multi-Agent Systems and Territory: Concepts, methods and applications. Congresso annuale della European Regional Association (ERSA).

Clark, G. L. et al., curatori (2000) The Oxford Handbook of Economic Geography. Oxford, Oxford University Press.

Coda, V. e E. Mollona, Il Governo della Dinamica della Strategia; Finanza. Marketing e Produzione, EGEA, n° 4. 2002.

Crestanello P. e Tattara G. (2004) Connessioni e competenze nei processi di de-localizzazione delle industrie venete dell'abbigliamento e delle calzature in Romania. Mimeo.

Delauzun, F. e E. Mollona. Introducing System Dynamics to BBC World Service. An insider perspective. Journal of Operational Research Society, 50: 364-371, 1999.

Fasoli N. E Messina A. (2001) EMCORE - Emotional cooperative groupware: a groupware tool for collaborative work. International Journal of Modern Physics, 12 (4): 589-600.

Fioretti G. (2001) Information Structure and Behaviour of a Textile Industrial District. Journal of Artificial Societies and Social Simulation, 4 (4).

Fioretti G. (2003) Will Industrial Districts Exploit B2B? A local experience and general assessment. Working Paper, disponibile al seguente indirizzo: http:/ /www.cei.unimo.it.

Forrester J.W. (1961) Industrial Dynamics. Cambridge (MA), Productivity Press.

Forrester J.W. (1968) Principles of Systems. Cambridge (MA), Productivity Press.

Garzia, C. e Mollona, E. Aluminium for the Transportation Industry in Europe. Innovation and Competitive Advantages. A System Dynamics Approach. EGEA, 2002.

Giaccaria P. (1997) Modellizzazione di un distretto industriale mediante simulazioni fondate su agenti. Sviluppo Locale, 4: 131-171.

Goodman M.R. (1983) Study Notes in System Dynamics. Cambridge (MA), The MIT Press.

Grandinetti R. e Rullani E. (1994) Sunk Internationalization: Small firms and global knowledge. Revue d'Economie Industrielle, 67 (1): 238-254.

ICE, Istituto per il Commercio Estero (2003) Internalizzazione dei distretti industriali. Roma, Atti del convegno.

Kennedy M. e Von Burg U. (2001) Paths and regions: the creation and growth of Silicon Valley. In Path Dependence and Creation, a cura di R. Garud e P. Karnoe. Londra, Lawrence Erlbaum Associates.

Krugman P.R. (1992) Geography and Trade. Cambridge (MA), The MIT Press.

Krugman P.R. (1996) Development, Geography, and Economic Theory. Cambridge (MA), The MIT Press.

Krugman P.R. (1997) How the economy organizes itself in space: a survey of the new economic geography. In The Economy as an Evolving Complex System II, a cura di W.B. Arthur, S.N. Durlauf e D.A. Lane. Proceedings Volume XXVII, Santa Fe Institute, Studies in the Sciences of Complexity. Reading (MA), The Advanced Book Program, Perseus Books.

Krugman P.R., Venables A.J. e Fujita M. (1999) The Spatial Economy. Cambridge (MA), The MIT Press.

Lazerson M.H. e Lorenzoni G. (1999) The Firms that Feed Industrial Districts: A Return to the Italian Source. Industrial and Corporate Change, 8 (2): 235-266.

Lorenzoni G. (1979) Una politica innovativa nelle piccole medie imprese. Milano, ETAS.

Lucking-Riley D. e Spulber D. (2001) Business-to-business electronic commerce. Journal of Economic Perspectives, 15: 55-68.

Maskell, P. (1998) Successful low-tech industries in high-cost environments: The case of the Danish furniture industry. European Urban and Regional Studies, 5 (2): 99-118.

Marafioti, E. e E. Mollona. Governare un sistema complesso con il System Dynamics: La nascita di un hub; Economia & Management, .n. 6, Etas Libri, Novembre, 2000.

MODA-ML (2003) Middleware Tools and Documents to Enhance the Textile/Clothing Suypply Chain through xML. Disponibile al seguente indirizzo: http:/ / www. Moda-ml.org.

Mollona E. (2001) Analisi Dinamica dei Sistemi Aziendali. Milano, EGEA.

Morrison C. e Staber U. (2000) The Empirical Foundations of Industrial District Theory. In Innovation, Institutions and Territory: Regional Innovation Systems in Canada, a cura di D. Wolfe e A. Holbrook.Montreal, McGill-Queen's University.

NIIP (1993) National Information Infrastructure Protocols. Disponibile al seguente indirizzo: http/ /niip.org.

Nuti F. e Cainelli G. (1996) Changing Directions in Italy's Manufacturing Industrial Districts. Journal of Industry Studies, 3 (2): 105-118.

Park S.O. e Markusen A. (1995) Generalizing new industrial districts: a theoretical agenda and an application from a non-Western economy. Environment and Planning A, 27: 81-104.

Porter, M. (1998) The Competitive Advantage of Nations. Basingstoke, MacMillan.

Pyke F. e Sengenberger W. (1990) Industrial Districts and Inter-firm Cooperation in Italy. Ginevra, International Institute for Labour Studies.

Quadrio Curzio A. e Fortis M. 2002. Complessita' e distretti industriali: dinamiche, modelli, casi reali. Bologna, Il Mulino.

Rabellotti R. e Schmitz H. (1998) The Internal Heterogeneity of Industrial Districts in Italy, Brazil and Mexico. Regional Studies, 33 (2): 97-108.

Rullani E. (1993) Networks and Internationalization: Managing complexity through knowledge. In Perspectives on Strategic Change, a cura di L. Zen, S. Zambon e A.M. Pettigrew. Boston, Kluwer Academic Publishers.

Sabbatini P. (2003) B2B and the transaction costs paradigm. Working Paper, disponibile al seguente indirizzo: http:/ /papers.ssrn.com/abstract=274974.

Saxenian, A. (1996) Regional Advantage: Culture and Competition in Silicon Valley and Route 128. Cambridge (MA), Harvard University Press.

Saxenian, A. (2002) Transnational Communities and the Evolution of Global Production Networks: The case of Taiwan, China and India. Working Paper, disponibile al seguente indirizzo: http: // ww w.sims.berkeley.edu/~anno/.

Squazzoni F. e Boero R. (2002) Economic Performance, Inter-Firm Relations and Local Institutional Engineering in a Computational Prototype of Industrial Districts. Journal of Artificial Societies and Social Simulation, 5 (1).

Sterman J. (2000) Business Dynamics, Systems Thinking and Modeling for a Complex World. New York, McGraw-Hill.

Swann G. M. P. et al. (1998) The dynamics of industrial clustering: International Comparisons in computing and biotechnology. Oxford, Oxford University Press.

Terna P. (2002) Simulazione ad agenti in contesti d'impresa. Sistemi Intelligenti, 16 (1): 33-51.

Terna P. (2004) Java Enterprise Simulator. Disponibile al seguente indirizzo: http. // web . econ.unito.it/terna/jes.