21
The evolutionary complexity of complex adaptive supply networks: A simulation and case study Gang Li a, , Hongjiao Yang a , Linyan Sun a , Ping Ji b , Lei Feng a a The Management School of Xi’an Jiaotong University, The State Key Lab for Manufacturing Systems Engineering, The Key Lab of the Ministry of Education for Process Control and Efficiency Engineering, Xi’an 710049, PR China b Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong article info Article history: Received 17 July 2006 Accepted 30 September 2009 Available online 3 December 2009 Keywords: Supply chain management Operation strategy Case study research Simulation Complex adaptive system Complexity abstract A supply chain should be treated not just as a supply chain but also as a complex adaptive supply network (CASN). However, the literature on supply chain management has given little attention to the evolutionary complexity of the network structure and collaboration mechanism of CASNs. In this paper, we first model and simulate the evolution of CASNs based on complex adaptive system and fitness landscape theory. The simulation results indicate the evolutionary complexities such as emergence, quasi-equilibrium, chaos, and lock-in of CASNs. Then, a case study of the evolution of the LVEA (low voltage equipment apparatus) supply network in the emerging Chinese market has been explored to validate the findings from the simulation and develop a better understanding of the general principles influencing the emergence, adaptation and evolution of CASNs in the real world. Based on the simulation and the case study, we propose some propositions about the factors and principles influencing the evolutionary complexity of CASNs. The external environment factors and firm-internal mechanisms appear to be the dominant forces that shape the gradual evolution of CASNs. Factors in the external environment, such as government regulation, market demand and market structure appear to have a long-term impact on the evolution, while a firm’s strategies, product structure, technology, and organization appear to be the internal factors that exert an immediate influence on the evolution of CASNs. Among these factors, cost and quality considerations appear to be the primary forces that influence the structure complexity, centralization and formalization of CASNs. & 2009 Elsevier B.V. All rights reserved. 1. Introduction A supply chain is a network of autonomous or semiautono- mous business entities collectively responsible for procurement, manufacturing and distribution activities, which create value for final customers in the form of one or more families of related products or services (Christopher, 1992; Min and Zhou, 2002; Swaminathan et al. 1998). Firms with highly synchronized supply chains have a number of advantages in today’s highly competitive, fast-changing environment. One of the major challenges for supply chain managers is to develop a network structure and collaboration mechanism that can facilitate adaptive, flexible and synchronized behaviors in a dynamic environment. However, researchers are still in the early stages of investigating the general principles that govern the birth, growth and evolution of supply networks with complex network structure and mechanisms for collaboration. A key to tackling this problem successfully is the realization that supply chain should be treated as a complex adaptive supply network (CASN). Pathak et al. (2007a) proposed that a CASN be viewed as a CAS (complex adaptive system) consisting of interconnected autono- mous entities that make choices concerning adaptation and survival. And as a collective, the system evolves and self-organizes over time, in response to changes in its environment. The concept of CASN allows us to understand how supply chains, considered as living systems, adapt to, and co-evolve with, the rugged and dynamic environment in which they exist, and to identify patterns that arise in such a condition of co-evolution (Surana et al., 2005). In a CASN, different entities operate subject to different sets of constraints, each with their own local objectives, and each, with different local views of the environment. With their interaction, these entities sense, learn, and adapt to the environment. The CASN is a highly nonlinear system, which shows complex multi-scale behavior, and has a dynamically evolving organizational structure and collaboration pattern for a given product. A similar viewpoint has been presented by Choi et al. (2001), who sought to demonstrate how supply chains should be managed if we recognize them as CASs. However, no ARTICLE IN PRESS Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/ijpe Int. J. Production Economics 0925-5273/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpe.2009.11.027 Corresponding author. Tel.: + 86 13096915852; fax: + 86 29 82664643. E-mail addresses: [email protected] (G. Li), [email protected] (H. Yang), [email protected] (L. Sun), [email protected] (P. Ji). [email protected] (L. Feng). Int. J. Production Economics 124 (2010) 310–330

The evolutionary complexity of complex adaptive supply networks: A simulation and case study

  • Upload
    gang-li

  • View
    215

  • Download
    1

Embed Size (px)

Citation preview

Page 1: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

Int. J. Production Economics 124 (2010) 310–330

Contents lists available at ScienceDirect

Int. J. Production Economics

0925-52

doi:10.1

� Corr

E-m

(H. Yan

fenglei@

journal homepage: www.elsevier.com/locate/ijpe

The evolutionary complexity of complex adaptive supply networks:A simulation and case study

Gang Li a,�, Hongjiao Yang a, Linyan Sun a, Ping Ji b, Lei Feng a

a The Management School of Xi’an Jiaotong University, The State Key Lab for Manufacturing Systems Engineering, The Key Lab of the Ministry of Education for Process Control and

Efficiency Engineering, Xi’an 710049, PR Chinab Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong

a r t i c l e i n f o

Article history:

Received 17 July 2006

Accepted 30 September 2009Available online 3 December 2009

Keywords:

Supply chain management

Operation strategy

Case study research

Simulation

Complex adaptive system

Complexity

73/$ - see front matter & 2009 Elsevier B.V. A

016/j.ijpe.2009.11.027

esponding author. Tel.: +86 13096915852; fa

ail addresses: [email protected] (G. Li), recha

g), [email protected] (L. Sun), mfpji@ine

sxfda.gov.cn (L. Feng).

a b s t r a c t

A supply chain should be treated not just as a supply chain but also as a complex adaptive supply

network (CASN). However, the literature on supply chain management has given little attention to the

evolutionary complexity of the network structure and collaboration mechanism of CASNs. In this paper,

we first model and simulate the evolution of CASNs based on complex adaptive system and fitness

landscape theory. The simulation results indicate the evolutionary complexities such as emergence,

quasi-equilibrium, chaos, and lock-in of CASNs. Then, a case study of the evolution of the LVEA (low

voltage equipment apparatus) supply network in the emerging Chinese market has been explored to

validate the findings from the simulation and develop a better understanding of the general principles

influencing the emergence, adaptation and evolution of CASNs in the real world. Based on the

simulation and the case study, we propose some propositions about the factors and principles

influencing the evolutionary complexity of CASNs. The external environment factors and firm-internal

mechanisms appear to be the dominant forces that shape the gradual evolution of CASNs. Factors in the

external environment, such as government regulation, market demand and market structure appear to

have a long-term impact on the evolution, while a firm’s strategies, product structure, technology, and

organization appear to be the internal factors that exert an immediate influence on the evolution of

CASNs. Among these factors, cost and quality considerations appear to be the primary forces that

influence the structure complexity, centralization and formalization of CASNs.

& 2009 Elsevier B.V. All rights reserved.

1. Introduction

A supply chain is a network of autonomous or semiautono-mous business entities collectively responsible for procurement,manufacturing and distribution activities, which create value forfinal customers in the form of one or more families of relatedproducts or services (Christopher, 1992; Min and Zhou, 2002;Swaminathan et al. 1998). Firms with highly synchronized supplychains have a number of advantages in today’s highly competitive,fast-changing environment. One of the major challenges forsupply chain managers is to develop a network structure andcollaboration mechanism that can facilitate adaptive, flexible andsynchronized behaviors in a dynamic environment. However,researchers are still in the early stages of investigating the generalprinciples that govern the birth, growth and evolution of supplynetworks with complex network structure and mechanisms for

ll rights reserved.

x: +86 29 82664643.

[email protected]

t.polyu.edu.hk (P. Ji).

collaboration. A key to tackling this problem successfully is therealization that supply chain should be treated as a complexadaptive supply network (CASN).

Pathak et al. (2007a) proposed that a CASN be viewed as a CAS(complex adaptive system) consisting of interconnected autono-mous entities that make choices concerning adaptation andsurvival. And as a collective, the system evolves and self-organizesover time, in response to changes in its environment. The conceptof CASN allows us to understand how supply chains, considered asliving systems, adapt to, and co-evolve with, the rugged anddynamic environment in which they exist, and to identifypatterns that arise in such a condition of co-evolution (Suranaet al., 2005). In a CASN, different entities operate subject todifferent sets of constraints, each with their own local objectives,and each, with different local views of the environment. Withtheir interaction, these entities sense, learn, and adapt to theenvironment. The CASN is a highly nonlinear system, whichshows complex multi-scale behavior, and has a dynamicallyevolving organizational structure and collaboration pattern for agiven product. A similar viewpoint has been presented by Choiet al. (2001), who sought to demonstrate how supply chainsshould be managed if we recognize them as CASs. However, no

Page 2: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 311

concrete framework has been suggested under which suchconjectures could be verified and generalized (Surana et al., 2005).

The goal of this study is to identify the general principles andthe salient factors that govern the evolution of CASNs. Based on anunderstanding of the operation of CASNs in the real world, thispaper proposes a model for CASN evolution using the principles ofCAS and fitness landscape theory. Based on a simulation of themodel and a case study, some general principles concerning theevolutionary complexity of CASNs are proposed, together withtheir managerial implications.

The remainder of this paper is organized as follows. Section 2reviews the relevant literature. Section 3 presents the evolutionmodel of CASNs. Section 4 presents the simulation of theevolution model and identifies some salient factors and principlesinfluencing the evolution. In Section 5, a case study is conductedto validate the findings of Section 4. In Section 6, we present somepropositions concerning the evolutionary complexity of CASNs. InSection 7, the main contributions of this study are summarized,and some of its implications for management are identified.Finally, concluding remarks and future research directions are setforth in Section 8.

2. Literature survey

A CASN is a distributed network which is inherently difficult tounderstand due to its evolving structure and functions, thediversity of its connections, the dynamic complexity of itsconstituent entities, and the interactions among all these factors.In order to study a CASN as a whole, it is critical to understand thenature of the interplay between its structure and its functioning(Surana et al., 2005). The structural issue focuses on the decisionabout the nodes in the network and the related linkages. Thefunctional issue concerns the ‘‘collaboration’’ between the nodes.Effective CASN management requires ‘‘consistency’’ or ‘‘fit’’between network structures and collaboration patterns (Huret al., 2004; Fisher, 1997; Ramdas and Spekman, 2000; Lee,2002; Stock et al., 2000).

Over the past decade, numerous studies have enriched ourunderstanding of the issues related to CASNs (Beamon, 1998).Analytical models, simulations methods, and empirical ap-proaches have all been used to probe supply networks from bothstrategic and operational level. Most of the analytical studies havefocused on the design and optimization of the operation decisionof supply networks based on the assumption that a supplynetwork is an integrated, static organization (Min and Zhou, 2002;Gunasekaran and Ngai, 2005; Whang, 1995). Although specificresults can be obtained from analytical models via proofs orbounds, the analytical models for supply networks are oftenlimited in their ability to map the dynamics of the system andobtain solutions for problems of reasonable size (Pathak et al.,2007a). Most supply networks are large-scale systems consistingof numerous entities and inter-connections among entities, andwhere the entities and connections are continuously evolving.

Computer-based simulations have been used both to investi-gate the complexity of CASNs and to support real world decision-making for actual supply networks. Simulation is a powerful toolfor investigating the behavior of large-scale systems which areanalytically intractable, and for examining various decisions forthe improvement of a given supply network. Forrester (1961) wasthe first to use system dynamics-based simulation to examinedynamic behavior of a supply chain. Since then, illuminatingresults have been generated by this line of research (e.g. Berryet al., 1995; Larsen et al., 1999; Marquez and Blanchar, 2004;Swaminathan et al., 1998; Towill, 1996; Towill et al., 1997;Wikner et al., 1991). Although system dynamics-based simulation

can be used effectively to examine the internal complexity of asupply network for a given structure, it is difficult to model thedynamic evolution of such a network because it is still based onthe assumption that the structure of a supply network is static orfixed. In the real world, however, as a result of changes inmarkets, technology, and products, among other things, thestructures of CASNs are always evolving. Therefore, regardless ofthe increasing amounts of time and money spent, efforts based onthe static structure assumption and using the analytical modelsand system dynamics simulation approaches, often lead tofrustration and helplessness (Choi et al., 2001).

To better understand and manage CASNs, researchers andmanagers need to examine their complex nature from anevolutionary perspective. They need to address questions suchas: how does a CASN emerge, adapt and evolve over time? Whatgeneral factors and principles govern the evolution of thestructure and the collaborative mechanisms of CASNs? Since theseminal contribution on supply chains as CAS by Choi et al.(2001), other studies on the evolution of supply networks haveappeared. Choi et al. (2001) put forth various conjecturesregarding how the patterns of behavior of individual entities ina supply network relate to the emerging network and theevolutionary complexity of a CASN. Surana et al. (2005) proposedthat the concepts-tools and techniques used in the study of CASscould be exploited to characterize and model supply networks.Wilkinson and Young (2002) argued that network structure andbehavior emerge through the local interaction of networkmembers in a bottom-up self-organizing way. Pathak et al.(2007a) provided a good review of CAS studies, raising somecritical issues and posing some important challenges for CASNs.Although a number of conjectures were proposed, no concreteframework or method was suggested for verifying them.

In recent years, a number of simulation-based and empiricallybased investigations of the evolutionary dynamics of CASNs haveappeared, some of which have focused on the operational level.For instance, Adamides and Pomonis (2007) suggested that afirm’s manufacturing strategy may emerge as a result of acoordinated search in the three correlated fitness landscapes ofproduct, production, and supply chain decisions. Alfaro andSepulveda (2006) found that flexible production systems maydemonstrate chaotic behavior. In linear supply chains, there aresigns of non-linear behavior such as phase transitions and chaos(Nagatani and Helbing, 2004). The deterministic chaos in a three-tier supply chain emerges from the interaction between custo-mers and suppliers (Wu and Zhang, 2007). Holweg and Bicheno(2002) used a participative simulation model to demonstratesupply chain dynamics and to model possible improvements to asupply network. However, none of these studies provides anyinsight into the understanding of the dynamics of the networkstructure and collaboration pattern of a CASN from the strategicperspective.

In fact, only a few studies have been carried out on thedynamic structure, evolution and behavior of CASNs and theyhave proposed inconsistent conclusions. Some of those studieswere based on multi-agent modeling and simulation techniques.Agent-based models can be used to explore various emergentnetwork phenomena, such as, how the behavior of firms and thesupply network evolve under different conditions. For example,Eymann et al. (1998) developed a multi-agent based system forthe simulation of the trade network and proposed that the tradenetwork grew out of the developing cooperative relationshipsbetween firms. Pathak and David (2002), Pathak et al. (2002, 2003,2007b) developed a multi-agent based simulator and showed thatcertain environmental and firm-level factors have an influence onthe eventual evolution of CASN structures. In addition to thestudies of the dynamics of network structure, there has also been

Page 3: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330312

some work on collaboration strategy in CASNs (e.g. Landry, 1998;Akkermans, 2001; Schieritz and GroBler, 2003), but so far, fewgeneral conclusions can be drawn. Although simulation-basedapproaches are powerful, their capability may be limited whendefinitively extrapolating the inner workings of large-scalesystems to the overall system behavior.

There are a few empirical studies which address the strategy,managerial implications, and measurement of key operationalissues in CASNs (Pathak et al., 2007b). Structural complexity andcollaboration complexity are the typical measures of complexityrelied on in empirical studies of CASNs (Daft, 1989; Gibson et al.,1997; Choi and Hong, 2002). More frequently, the emergenceof new organizational structures depends on the rise and declineof industries rather than on the emergence of a new type oforganization (Lewin et al., 1999). Cooperation networks emergefrom the complex and dynamic interplay between institutions,products, technologies, markets and innovative actors (Bruce,2000). The control structures employed by CASNs to deal with theinteraction of these factors are various and evolving (Choi andHong, 2002; Hur et al., 2004; Potter et al., 2004). A major obstaclefor these empirical studies is the difficulty of access, andsometimes the complete lack of access to proper data on theevolution of a given supply network over the long term. It isextremely difficult to collect sequence data consisting of firm-specific and inter-firm specific organizational adaptation eventsover an extended period of time. As a result, these studies aregenerally limited to examining the factors that influence thestructure and the collaboration of CASNs at a give point of time.With changes in the environment and the firms, do the factorsidentified in this literature change over time? Do the inter-connections between these factors evolve? And if so, how do thechanging factors and connections affect the CASN? No one cananswer these questions with any degree of certainty (Harlandet al., 2002).

Most analytical models, simulation models and empiricalstudies only address a firm’s adaptation to the environment, butdo not investigate the issue of how firms in a CASN influence theenvironment that they populate. Firms not only adapt to theenvironment, but also have the power to reshape that environ-ment (Volberda, 1998). Adaptation and environmental selectionare not wholly opposed forces; rather, they are fundamentallyinterrelated. In the real world, the complex web of changes in theenvironment, the firms, and the inter-relationships within aCASN, coupled with a firm’s adaptation to the environment and itsinfluence on the environment, are the primary source of thecomplexity of the co-evolution of CASNs and their environments.Given the evolving nature of CASNs, the challenge is to develop aneffective analytical tool to explore the dynamic behaviors of firms,the CASN and the environment.

The complexity of the evolutionary changes affecting CASNrules out the use of a single approach. Instead, a combinationapproach is necessary to adequately explore difficult issues suchas simultaneous and time-lagged effects among variables, non-linearities, cyclical feedback mechanisms, and path dependencies,from both the theoretical and the practical perspectives (Pathaket al., 2007a). The combination of simulation-based methods andcase studies can not only show the long-term, large-scalebehavior of a CASN in various environments, but also provideuseful descriptions of the great variations in a CASN’s structureand collaboration mechanism. In that way, they can contribute toour understanding of both the objective phenomena and thesubjective experience of CASNs, and in so doing, constitute animportant supplementary methodological tool for deepening ourunderstanding of the evolution of CASNs (Koza and Lewin, 1999).

Based on Choi et al.’s (2001) claim that internal mechanisms,the environment, and co-evolution are the three key foci for

supply network research, this paper proposes a model of CASNevolution based on CAS theory and fitness landscape theory, andsimulates that model using multi-agent technology (Li et al.,2009; Pathak and David, 2002; Pathak et al., 2003). The simulationresults indicate some characteristics of the evolutionary complex-ity of CASNs. Further, these characteristics are validated by a casestudy. Based on the simulation and the case study, somepropositions on the salient factors and principles governing theevolution of CASNs are proposed. Finally, some managerialimplications and research directions are proposed.

3. Modeling the evolution of CASN

We model the evolution of CASNs in terms of three aspects:the environment, the firm, and the supply network. The environ-ment is populated by firms and co-evolves with the firms. At thefirm level, each firm is modeled as an independent agent, whichhas its own internal mechanism and populates in the environ-ment. On the other hand, the interactions among the firms in theenvironment are modeled at the supply network level. We use‘‘fitness’’ to model a firm’s ability to fit to the environment and thenatural selection process of the environment. The fitness func-tions of our model are different from those used in the modelsemployed by Pathak and David (2002), Pathak et al. (2003) and Liet al. (2009), which could not represent the fact that a firm’sfitness is not affected only by its business strategy, capability,product, fixed cost, and technology, but also by the environmentfactors, such as culture, market uncertainty, etc. As theseelements change, a firm’s fitness changes. The firm needs toinvent new supply network structures and collaboration patternsin order to adapt to the environment.

3.1. Modeling the environment

Environment (E): The environment includes factors which areoutside a firm’s boundaries, but which can impact, constrain, orplace requirements on the firm and are affected by theirinteractions with the firm. These factors include demand, supply,technology, market structure, the economy, laws and regulations,and politics. They can be classified into three groups:

(1) The macro-institutional constraints (MIC), which includethe economy, politics, culture, laws, regulations, and marketstructure. They are understood in this study as the rules of a gamein which firms exist. They are exogenous to all firms andconstitute the space within which firms and their supplynetworks operate (Guisinger, 2001; Kinra and Kotzab, 2008).These institutional constraints are stable in a short-term, butcould change in a long run with the dynamic interactions amongthem, and their interactions with the micro-operational con-straints.

With these macro-institutional constraints, the environmentplaces expectations on firms, which is represented by the ‘‘fitnesslandscape’’: Fc (0oFcr1) (Kauffman, 1993). A firm shouldbalance the environmental expectations placed on it with itsresources, capabilities, and products (McCarthy, 2004). If a firmcannot fulfill these expectations, it cannot survive in the market.Different environments place different expectations on firms. Inthis model, a higher Fc indicates a higher degree of difficulty infitting to the environment.

(2) The micro-operational constraints (MOC), which includedemand, supply, price, cost, lead-time, service and the competi-tors for each individual firm. Different from the macro-institu-tional constraints, the micro-operational constraints could bechanged rapidly in a short time. Although the environment places

Page 4: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 313

the same institutional constraints on different firms, it placesindividual firms with different operational constraints.

In summary, the environment is modeled as:

E¼/MIC;MOCS ð1Þ

MIC ¼/Economy; Politics; Laws;Regulations;Culture;Market; FcjFc A ð0;1ÞS

ð2Þ

MOC ¼/Demand; Supply; Price;Cost; Lead Time; Service;CompetitorsS

ð3Þ

3.2. Modeling the firm

The model of a firm includes seven elements: strategy,organization culture, resource constraint, organization structure,business process, product and fitness to the environment(Melody, 1994; Li, et al., 2009), which is represented as a seven-tuple:

Firm¼/S;OS;RC;BP;OC;P; FS ð4Þ

Strategy (S): A firm’s strategy is the collection of decisionsabout what the firm desires to accomplish and how resources andcapabilities will be utilized to accomplish the desired results. Inthe present model, a firm’s strategy is represented as a five-tupleS=/V, M, G, P, ORS, consisting of vision (V), mission (M), goals (G),plans (P), and operation rules (OR).

Organizational structure (OS): The organizational structure of afirm consists of the organizational arrangements (formal andinformal, centralized and decentralized) that are created toaccomplish the overall mission of that firm.

Resource constraints (RC): The resource constraints on a firmare related to people, material, money, technology, knowledge,capability, and information. Firms utilize resources to fulfillcustomer orders.

Business processes (BP): Business processes are the flow ofactivities that enable a firm to utilize resources and produce itsproducts or provide its services.

Organizational culture (OC): Organizational culture is a complexset of explicit and tacit norms, values, management styles, andpatterns or behaviors that can have a positive or negative effect ona firm’s structure and function.

Products (P): Products are what a firm provides for itscustomers. Different products have different structures andtechnologies, which drive and shape the firm’s structures,functions, business process, and required resources. Therefore,the organizational structures, strategy, culture, resources, andbusiness processes of a firm must be such as to enable theeffective production of its products. The success of a firm isdetermined by how well it meets customer demand for itsproducts.

Fitness (F): Firms populate in the environment and need to fitthe environment to survive. Fitness is a process of matching theenvironmental fit and the internal fit, where all the elements in afirm sense, respond to and adapt to the environment, and eachelement of the firm fits with, reinforces, or is consistent with,other elements in the environment (Hamel and Prahalad, 1994;McCarthy, 2004; Miller, 1992; Nadler and Tushman, 1980). Thisprocess involves identifying and realizing appropriate strategies,organizational structures, organizational culture, resources, busi-ness processes, products and services, which are aspects of afirm’s internal mechanism. A firm’s fitness is denoted by thefunction fitnessi(t), representing the capability of firm i to survivein one or more populations at time t, by imitating and/orinnovating combinations of internal elements which will satisfythe firm’s objectives and the market’s needs, and be consistent

with the desires of competing firms (McCarthy, 2004). For thesake of simplicity, fitnessi(t) is usually measured by customersatisfaction and is evidenced in profit or loss. If a firm can fulfilldemand, its fitness and profit will increase. Otherwise, it willdecrease. The value of fitnessi(t) is represented as a real numberbetween zero and one [0,1]. The higher the value, the higher afirm’s ability to adapt to the environment.

A firm’s fitness evolves with the evolution of the supplynetwork. A firm with higher capacity and lower cost has morecompetence to fulfill the demand and improve its fitness. There ispositive feedback between a firm’s fitness and the probability thatit will win the competition. A firm with higher fitness has moreopportunities to win the competition. As a result, the firm whowin the competition has more opportunity to improve its fitness.Therefore, the goal for a firm is to improve its fitness byresponding to the challenges and opportunities posed by theenvironment. Besides, a firm’s fitness decays over time. If a firmcannot get any orders in a demand cycle, its fitness will decreasebecause of the fixed cost. Furthermore, fitness decay is influencedby culture. For instance, some cultures prefer long-term colla-borative relationships, while others prefer short-term relation-ships. The decay rate of a firms’ fitness is higher for firmspopulating an environment that prefers short-term relationships.In addition, a fast changing world prefers short-term perfor-mance, while a stable world prefers long-term performance.Therefore, the greater the environment fitness Fc, the greater thedecay.

In summary, the fitness evolution function of a firm is:

fitnessiðtþ1Þ ¼ fitnessiðtÞþ f ðDiðtÞ;QiðtÞ;CiðtÞ;RiðtÞÞ�gðcfiÞ�diðtÞ ð5Þ

where f ðDiðtÞ;QiðtÞ;CiðtÞ;RiðtÞÞ is fitness variation caused by thefirm’s fulfillment of the demand at time t. f ðDiðtÞ;QiðtÞ;CiðtÞ;RiðtÞÞ isa function of the demand DiðtÞ, the fulfilled demand QiðtÞ, the costof fulfilling the demand CiðtÞ, and revenue RiðtÞ, at time t,respectively. gðcfiÞ represents the decay of a firm’s fitness, whichis a function of the fixed cost cfi. diðtÞ represents the decay of afirm’s fitness, which is related to macro-institutional constraints.If a firm can fulfill the demand, diðtÞ is positive. Otherwise, it isnegative.

With the evolution of a firm’s fitness, the internal elements ofthe firm evolve, too. In the present model, we take account of theevolution of a firm’s capacity and its manufacturing costs. If thefirm fulfills an order (the demand) as required, it can make aprofit. Therefore, it can invest more in equipment and facilities,which help to improve manufacturing capacity, and to furtherreduce manufacturing cost. These changes are represented as:

MCPiðtþ1Þ ¼MCPiðtÞþkðDiðtÞ;QiðtÞÞ ð6Þ

MCiðtþ1Þ ¼MCiðtÞ�hðDiðtÞ;QiðtÞÞ ð7Þ

where kðDiðtÞ;QiðtÞÞ represents the change of manufacturingcapacity, which is a function of DiðtÞ and QiðtÞ. hðDiðtÞ;QiðtÞÞ

represents the change of manufacturing cost at time t, which is afunction of DiðtÞ and QiðtÞ. If a firm fulfills an order, a firm’smanufacturing capacity and manufacturing cost are improved. Onthe other hand, if a firm cannot get any orders in a demand cycle,capacity decreases and costs increase.

The interaction between firms and the environment involvesnatural selection and co-evolution (Li et al., 2009). Firms have torecognize which phenomena they should respond to, and the co-evolution takes place over time (Luhmann, 1995; McCarthy,2004). If fitnessi(t) is higher than Fc, firm i survives in theenvironment. Otherwise, the firm is eliminated from the environ-ment.

Page 5: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330314

3.3. Modeling supply network evolution

Firms collaborate with others to fulfill the demand generatedby the market where they exist. From the topology perspective,the collaborative relationships among firms can be modeled as aweighted bi-directional graph, where the firms are the nodes andthe relationships are the edges, and the weight of an edgerepresents the collaboration preference between two firms. Giventhe interaction among firms and the co-evolution between firmsand the environment, those firms that fit the environment surviveand the supply network emerges and evolves. And those norms,rules, and procedures that facilitate collaboration among firmsemerge and evolve, too. In other words, the weighted bi-directional graph evolves over time.

CASNðtÞ ¼ f ðE; firmi; . . . ; firmnÞ ¼ GðtÞ ¼/NtðGÞ; EtðGÞ;WtðGÞS ð8Þ

NtðGÞ ¼ ffirm1; firm2; . . . ; firmkjkrng ð9Þ

EtðGÞ ¼ fE1; E2; . . . ; Ekg ¼ f/firmi; firmjSjsuch that

ðfirmi; firmjANtðGÞÞ \ firmia firmjg ð10Þ

WtðGÞ ¼ ðwijÞn�n such that wijA ½0;1Þ ð11Þ

where firms is modeled as the nodes, and the evolving CASN ismodeled as a bi-directional graph G(t). Nt(G) is the collection ofnodes at time t, Et(G) is the collection of edges between nodes attime t, and Wt(G) is the weights of the edges at time t, whichpresents how many times that each node has been collaboratedwith others in the past time. The higher the weight of an edge is,the higher the collaboration preference of the two firms is.

4. Simulation of CASN evolution

4.1. The multi-agent architecture

To simulate the evolution of a CASN, this paper utilizes adistributed methodology based on multi-agent technology, wherethe environment and the firms in a CASN are modeled asintelligent agents (Akanle and Zhang, 2008). We identify threekinds of agent in our simulation model.

The environment agent: The environment agent consists of Fc,and a set of operation rules that define its relationship with otheragents in the CASN. It generates the global market demand at thebeginning of each demand cycle, and evaluates the fitness of eachfirm agent at the end of each demand cycle. If the firm agent’sfitness is below the Fc, the environment agent eliminates theagent from the environment.

The blackboard agent: The blackboard agent is a communica-tion center, which contains the knowledge about the environmentand the other firm agents. Through the blackboard agent, a firmagent has the ability to communicate with the environment agentand the other firm agents. Each time a global market demand isgenerated, firm agents interact with each other, and with thedemand, through the blackboard to form a supply network tomeet the global market demand.

The firm agent: Each firm agent consists of the fitness, theproduct, the manufacturing capacity, the manufacturing cost, anda set of rules that define its operational strategy to compete forthe market demand and collaborate with other firm agents tofulfill the demand. At the beginning of each demand cycle, thefirm agents communicate with the blackboard agent to getthe global market demand. They bid for the global demand andcollaborate with others to fulfill the global market demand. This

process is controlled by firm’s collaboration strategy, which willbe reported later.

The analysis of the dynamic evolution of a CASN is based ondiscrete-event simulation of the evolution model with variousenvironmental and operational strategies attributed to the agents.The flow of the simulation is shown in Appendix A.

4.2. Interaction of agents

In the simulation, all the firm agents interact with each otherto obtain the demand generated by the environment agent andfulfill the demand. In each demand cycle, the interaction isdivided into two stages (See Appendix A).

(1)

In the first stage, the market is a monopoly. As global marketdemand is generated at the beginning of each demand cycle,the blackboard agent collects the demand information andsends a message about the demand to all the firm agents.Then, all the firm agents bid for the global demand. In thisprocess, they compete for the demand without consideringcapacity constraints. The environment agent selects the firmagent whose fitness is the highest among the bidders as thesupplier.

(2)

In the second stage, the market is a Bertrand competitionmarket. The winner in the first stage collaborates with othersto execute the order. If the global demand is less than thewinner’s manufacturing capacity, the winner delivers all theproducts by itself. Otherwise, it manufactures the product upto its manufacturing capacity and subcontracts the remainingdemand to one or more of the other firms. The remaining firmagents, who were not selected in the first stage, read theinformation and compete for the remaining demand. Thewinner firm agent selects its subcontractor in conformancewith its collaboration strategy (e.g., price priority, short-termstrategy, and long-term strategy). In these two stages, eachfirm agent competes and collaborates with all the others toimprove its fitness.

At the end of each demand cycle, each firm agent evaluates itsprofits or losses and updates its fitness, manufacturing capacity,and manufacturing costs. The blackboard agent collects each firmagent’s fitness and sends the information to the environmentagent, which evaluates firm agents’ fitness and eliminates thefirms whose fitness is less than Fc.

4.3. Experimental design and analysis

Our focus in this study is on the evolution of the structure andfunctioning of CASNs. To this end, two experiments: the evolvingstructure and the fitness evolution were performed. Repeatedsimulations of each experiment were performed in order to obtainrobust output. The parameter settings of these simulationexperiments are shown in Appendix B. These parameters wereextracted from the investigation of the evolution of the ZTCorporation’s supply network over the past 20 years.

4.3.1. Experiment 1: structure dynamics of CASNs

In experiment 1, there are 8 firms which adopt a cost-prioritystrategy, in which the firm with the lowest bid wins theremaining demand at the second stage. The experiment has twovariants. In experiment 1(a), the 8 firms make a single productwithout any subparts. In experiment 1(b), they make a productwhich has two parts. Each firm can make the two parts andassemble them into the final product. When the winner in thefirst stage subcontracts its remaining demand to one of the

Page 6: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 315

remaining firms, it has two options for selecting its subcontrac-tors: (1) by subcontracting the entire production process to onefirm; or (2) by subcontracting the production of the individualparts to two firms and then assembling the parts into the finalproduct by itself (it is assumed that the assembly cost is 0). Theremaining firms randomly decide to compete for the wholeproduct or for one of the two parts. The selection of the winner isdetermined by cost: the lowest cost option wins.

(1) Emerging and evolving structure: As expected, what emergefrom the interaction of the 8 firms in experiment 1(a) are linear-structured supply networks. As shown in Fig. 1, a linear structuredsupply network emerges in the first demand cycle, which includesall the 8 firms. As a result of changes in the environment and thecompetition among the firms, some firms whose fitness is lessthan Fc are eliminated by the environment and the structure ofthe supply network changes. As a result, the vertical complexity(i.e., depth) of the supply network evolves over time. In 100 runsof the experiment, the mean vertical complexity of the emergedCASN was 3.81, and the mean lifetime of died firms was 109.6demand cycles.

However, the supply networks which emerge in experiment1(b) exhibit various kinds of vertical (i.e., depth) and horizontal(i.e., width) complexities (see Fig. 2). Some of them have a linearstructure, while others are binary-tree structured. There is nofixed type of network structure over the total 240 demand cyclesof a single experiment.

In order to learn more about the dynamics of networkstructures, and to insure robust output, we repeated experiment

4

t=1

6

5

1

3

2

0

7

7

5

6

0

4

2

1

5

6

1

2

0

1

6

5

2

0

5

6

1

2

0

t=26 t=229 t=239 t=240

Fig. 1. Emergence and evolution of CASNs (Product 1).

1

t=1

0

2

5

4

3

0

2

t=121 t=240

0

3

2 1

t=240t=1

7

5

6

1

4

0

2

3

t=126

5

7

4

1

3

6

1

5

7

4

3

Fig. 2. The emerging CASNs with various structures (Product 2).

1(a) 100 times, in the course of which a wide range of structuresappeared, from the simplest linear-structured networks involvingonly one firm to the most complex linear-structured networksinvolving all the 8 firms. The mean vertical complexity of thesupply network in the experiment was 3.89, meaning that thetypical network consisted of about 4 firms. In addition, we foundthat there are three possible patterns for the evolution of thestructure of a supply network: (1) All the firms survive; (2) Allthe firms die; (3) Some firms survive and some die. Each of thesepatterns is associated with a different set of evolutionarybehaviors in terms of network structure and firm fitness, whichwill be detailed in the next section.

In experiment 1(a) and 1(b), various supply networks emergeand evolve as a result of the dynamic interaction among the firms,and between the firms and the environment. As the environmentgenerates demand at the beginning of the first demand cycle, supplynetworks with various structures emerge as the firms compete forthe demand and collaborate to fill it. In the following demand cycles,with the changes in the environment (e.g., demand) and the firms(e.g., products, fitness, capacity and costs), the supply networks (e.g.,memberships, linkage relationships, and the depth) evolve overtime. The emergence and evolution of supply networks is related tothe internal mechanisms of firms (e.g. product structure, fitness,costs, capacity, strategy), and the environment (e.g. marketstructure, demand). This process is not under anyone’s control;rather, they are the result of self-organization.

(2) Quasi-equilibrium: As mentioned in the previous section,we find that there are three prominent patterns which appear inthe evolution of the network structure, and which are dramati-cally different from one another (See Fig. 3 and Table 1).

The first pattern is shown in Fig. 3(a). In the 100 iterations of theexperiment, there are 23 in which the fitness of the 8 firmsdecreases gradually and all the firms are eliminated by theenvironment. The mean depth of the emerged supply network is2.42, which indicates that most of the firms have not participated inthe collaboration. Without a win–win collaboration pattern to helpthe firms to improve their fitness, capacity and cost structure, all ofthem are gradually eliminated by the environment.

The second evolutionary pattern is shown in Fig. 3(b). Incontrast with Fig. 3(a), all eight firms survive in the competitionand collaboration. In 100 runs of the experiment, all the firmssurvive 34 times, and the mean depth of the supply network is4.345. Within those 34 runs, there are 14 in which the fitness ofall the firms is improved, and the mean depth of the supplynetwork is 4.332. In addition, there are 20 runs in which thefitness of some of the firms is improved, and that of othersdecreases. The mean depth of the supply network is 4.355. Theseresults indicate that in this pattern there emerge a win–wincollaboration schema among the firms and this collaborationresults in a superior equilibrium.

In the third evolutionary pattern, shown in Fig. 3(c), somefirms survive, and some are eliminated. The structure of thenetwork and the fitness of the firms evolve in a dynamic way. Forsome firms, fitness increases gradually, for others, it decreasesgradually. Those firms with the highest fitness evolve into thefocal firms of the supply network. As a result, they get control ofthe global demand and subcontract some of it to other firms. Inthe 100 runs of the experiment, there is 43 times when theevolutionary pattern resembles that in Fig. 3(c), in which a focalfirm emerges. The mean death rate of the firms is 23.33%, and themean depth of the supply network is 4.15.

In the 100 runs of the experiment, there emerge twointeresting phenomena. Sometimes, the gradual decrease in afirm’s fitness may lead it to fall into a declining path. But it mayalso happen that, as result of disturbances provoked by somerandom events, the firm may escape from the declining path and

Page 7: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

0 20 40 60 80 100 120 140 160 1800.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

Demand Cycle

Fitn

ess

Firm0 Firm1 Firm2 Firm3 Firm4 Firm5 Firm6 Firm7

0 50 100 150 200 2500.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

Demand Cycle

Fitn

ess

Firm0 Firm1 Firm2 Firm3 Firm4 Firm5 Firm6 Firm7

0 50 100 150 200 2500.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

Demand Cycle

Fitn

ess

Firm0 Firm1 Firm2 Firm3 Firm4 Firm5 Firm6 Firm7

0 50 100 150 200 2500.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

Demand Cycle

Fitn

ess

Firm0 Firm1 Firm2 Firm3 Firm4 Firm5 Firm6 Firm7

0 50 100 150 200 2500.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

Demand Cycle

Fitn

ess

Firm0 Firm1 Firm2 Firm3 Firm4 Firm5 Firm6 Firm7

Fig. 3. Quasi-equilibrium and chaos in the evolution of CASNs.

Table 1Summary of 100 runs of experiment 1(a).

Evolution pattern Percentage (%) Mean

depth

Mean lifetime

of died firms

Death

rate (%)

Remark

Pattern 1 (All firms die) 23 2.42 103.6 100 All the firms are eliminated by the environment

Pattern 2 (All firms survive) 14 4.332 Null 0 All the firm’s fitness is improved

20 4.355 Null 0 Some firms’ fitness is improved

Pattern 3 (Some firms survive) 43 4.15 112.8 23.33 Some firms survive, and some are eliminated by the environment

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330316

gradually improve its fitness (see Fig. 3(d)). It is also the case thata firm can escape from a superior equilibrium and fall into aninferior equilibrium (see Fig. 3(e)).

The transfer from one kind of equilibrium to another indicatesthat the supply networks are not always stable. A firm witha disadvantageous position in the market could grow into aleadership firm by modifying its collaboration pattern andforming a win–win supply network. On the other hand, aleadership firm could fall into an inferior position and be

eliminated by the market if it does not form a win–win supplynetwork.

(3) Chaos: The experiments just described indicate that mostcommon evolutionary pattern resembles that in Fig. 3(c).However, if we repeat the experiment, it is difficult to predictwhich pattern will emerge at the beginning of each experiment. Itis possible for firms to escape from one equilibrium state and fallinto another. Such transformations are dependent on a firm’sinternal mechanisms as well as on random events from the

Page 8: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

Table 2Independent sample t-test of the mean fitness variance ratio in different

environments.

Environment Sample size Mean Standard deviation P-value

Fc=0.25 30 0.10128377 0.369759892 0.457a

Fc=0.35 30 0.03284694 0.337982951

a The significance is 0.01.

Table 3Independent sample t-test of the mean fitness variance ratio with different

strategies.

Firms’ strategy Sample size Mean Standard

deviation

P-value

Long-term strategy 30 0.04546450 0.139334698 0.043a

Short-term strategy 30 �0.03937803 0.176734934

a The significance is 0.05.

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 317

environment. The diversity of the evolutionary patterns encoun-tered indicates that the evolution of CASNs is randomly, andhighly sensitive to the initial conditions and has multipleequilibrium states. There is non-deterministic chaos in theevolutionary process, and with that comes unpredictability.Certain initial conditions and random disturbance may drive aCASN to evolve into a superior equilibrium state and achieve ahigh level of system efficiency. Other initial conditions andrandom disturbance may cause it to fall into an inferiorequilibrium state to operate inefficiently.

(4) Lock-in: In these experiments, we found that firms’collaboration preferences evolve over time. At the beginning ofthe experiments, each firm had no preference concerning whichfirms to collaborate with. But once collaboration and competitionamong firms begins, it was found that most of the firms do notchange their partners very often, and each firm prefers tocollaborate with just a few partners. Once a firm has collaboratedwith partner firms and filled the demand, the fitness, manufac-turing capacity, and manufacturing costs are all improved. Theyhave more opportunities to continue their collaboration in thefollowing demand cycles. As a result, they lock in each other andtheir collaborative relationship becomes more and more stableover time. In other words, the collaboration among firms is self-reinforcing, locking-in and path dependent.

4.3.2. Experiment 2: dynamic evolution of firm’s fitness

(1) Hypothesis: The environment is an important factor thatinfluences the fitness evolution of the firms. In different environ-ments, firms learn, adapt to and react. Once firms form a mutuallycompatible supply network with stable collaborative relation-ships, they can survive and thrive. Germain et al.’s empirical study(2008) proposed that environment uncertainty affects therelationship between firm’s organizational structure and perfor-mance. Li et al. (2009) studied the relationship between thestability of supply networks and the environment, and concludedthat stability of the collaborative relationships in a supplynetwork is of no difference in different environment. Otherstudies explored the relationship between the stability and theperformance of supply networks. Relational stability is conduciveto improve supply network performance because it providesopportunities for learning, acquiring, sharing, and innovating overtime (Krogh et al., 2001; Madhavan and Grover, 1998). A stablesupply network gives the parties involved a better ability tooutperform their competitors (Dyer and Singh, 1998; Fiala, 2005;Krogh et al. 2001; Li, et al., 2006), and relational stabilitypositively affects a firm’s fit to the environment (Yang et al.,2008). Following this line of research, we hypothesize that:

Hypothesis 1. In different environment, the evolution of firm’sfitness is of no difference.

Except for the environment, firm’s collaboration strategy isimportant to cultivate firm’s performance in an environment.Firms need to fit among their strategy, their environment andperformance (Anand and Ward, 2004; Swamidass and Newell,1987; Ward and Duray, 2000; Williams et al., 1995). A well-performing firm has more resources to improve its fitness to theenvironment. A firm’s collaboration strategy impacts its fitness tothe environment. Firm’s performance in an environment can beaffected by the quality of relationships formed with partners andsuppliers (Hsu, 2005). In the real world, many firms prefer long-term collaboration strategies, while others prefer short-termcollaboration strategies. Peleg et al. (2000) compared short-termstrategies with long-term strategies and suggested that nostrategy is generally superior to the others for firms to improvetheir performance. However, Akkermans (2001) showed that the

performance of firms with short-term strategies is better thanthat of firms with long-term relationships. In contrast, Schieritzand GroBler (2003) proposed that the firms with long-termstrategies can achieve higher performance than firms with short-term strategies. Song and Noh (2006) suggested that long-termsupply relationships are more conducive to the development ofnew products. Relational stability positively affects a firm’s fitwith the environment (Yang et al., 2008). Studies have alsorevealed that through long-term relationships, suppliers becomepart of a well-managed chain, which can have a lasting effect onthe competitiveness of the entire supply chain (e.g., Ellram, 1991;De Toni and Nassimbeni, 1999; Kotabe et al., 2003). Following thisline of research, we hypothesize that:

Hypothesis 2. Long-term collaboration strategy is better for afirm to improve its fitness with the evolution of CASNs.

(2) Experiment setting and data collecting: Experiment 2(a) wasperformed to test Hypothesis 1, and experiment 2(b) wasperformed to test Hypothesis 2. Most of the parameter settingsof experiment 2(a) and 2(b) are same as those of experiment 1(a),except that firms’ collaboration strategies (e.g., long-term strategyversus short-term strategy) is different for experiment 2(a) and Fc

is different (Fc=0.25 and Fc=0.40) for experiment 2(b).Each experiment was run 30 times, and the data for each firm’s

fitness value for the 240 demand cycles was collected. TheVariance ratio was used to evaluate the evolution of fitness.A higher Variance ratio indicates that the fitness of the 8 firmsgreatly improved the 240 demand cycles. The data fromexperiments 2(a) and 2(b) are collected in Tables 2 and 3,respectively.

Variance ratio¼X8

i ¼ 1

ðfitnessið240Þ�fitnessið0ÞÞ=X8

i ¼ 1

fitnessið0Þ ð9Þ

(3) Analyzes and results: SPSS was used to analyze the data. Wenote first, that the data in Tables 2 and 3 exhibit a normaldistribution. An independent samples t-test was used to test themeans of the data. The results are shown in Tables 2 and 3.

Hypothesis 1 is accepted. It indicates that no matter howvolatile the environment is, once firms form a mutuallycompatible supply network, they can achieve a win–win perfor-mance. In different environment, firms need to sense changes inthe environment, and adjust their behavior to adapt to thosechanges. Once firms form a mutually compatible supply network,they can survive and thrive in the environment.

Page 9: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330318

Hypothesis 2 is accepted. The data shows that the long-termcollaboration strategy is better for improving a firm’s fitness. Inpractice, many firms achieve best performance by adopting thelong-term collaboration strategy. For instance, Foxconn, whichwas the No. 1 EMS (electronic manufacturing service) supplier inthe world, collaborated with its suppliers and customers (e.g.,Apple, Dell, etc.) in the long-term. Foxconn helped its customersto design and manufacture new products, and improve theirproduct quality, cost, and delivery. As a result of the long-termcollaboration, Foxconn and its partner firms survived and thrivedin the highly uncertain market.

5. Case study

System simulation allows researchers to identify and analyzethe most likely outcomes in the evolution of CASNs. Butsimulation is not sufficient to prove anything. For theory building,it is important to validate and refine the findings from simulation.Empirical methodologies are likely to be important contributorsto the study of supply networks as they can establish links toindustrial realities, providing validation and ensuring the practi-cality of model prescriptions (Pathak et al., 2007a).

A case study approach was chosen to investigate the evolutionof supply networks in the real world. A case study offers theopportunity to study a phenomenon in its own natural settingwhere complex links and underlying meanings can be explored,while also enabling the researcher to study whole supply chains(Miles and Huberman, 1984; Yin, 1994). A case study is alsoappropriate where existing knowledge is limited because itgenerates in-depth contextual information which may result ina superior level of understanding. Given these considerations, theuse of a case study research strategy is deemed appropriate in thepresent study (Oke and Gopalakrishnan, 2008).

5.1. Methods

5.1.1. Background

In this case study, we explore the evolution of the LVEA (low-voltage electric apparatus) supply network of ZT Corporation, acompany headquartered in the city of Wenzhou, in southeasternChina, in order to validate and refine the findings obtained in thesimulation experiments. The LVEA supply network of ZT Corpora-tion was selected due to two reasons: (1) Over the past 30 years,as a result of changes in laws, policies and regulations, Wenzhouhas evolved into one of the most free markets in China. (2) TheLVEA supply network has undergone all the life cycles of CASNevolution (emergence, adaptation, and evolution into new forms).

Wenzhou began its free market reforms surreptitiously in themid 1970s, in advance of the market reforms undertaken by thecentral government of China at the end of 1970s. As a resultof those reforms, many privately-owned firms in the LEVAindustry were established in the region. Over the past 30 years,an extensive LVEA supply network has emerged, adaptedand evolved as a result of the dynamic interaction amongthe firms and the environment. ZT, which was founded in1984, evolved into the focal manufacturer of the supply network.Many characteristics of the evolutionary complexity of CASNshave been identified in the evolution of the LVEA supplynetwork.

The LVEA supply network of ZT Corporation includes thesupply base network and the distribution base network. As of2005, the supply base network consisted of the final assembler(ZT), 786 first-tier suppliers, and more than 3000 second-tiersuppliers. The distribution network had three tiers: 11 first-tierregional distribution centers, 31 second-tier provincial sales

offices, and more than 2000 third-tier retailers located in varioussmall counties around China. The suppliers provided more than10,000 kinds of parts, components and sub-assemblies to ZT. Inaddition, ZT had 5 ZT-owned sales offices and 30 non-ZT-ownedsales agencies outside China (see Fig. A1).

ZT’s main product is the whole set LVEA. In 2005, ZT’s annualsales were RMB 15 billion. Among the 786 first-tier suppliers,there are 7 key-part suppliers were owned by ZT. The 115suppliers whose total annual sales were between RMB 10 millionand 50 million, and other first-tier suppliers’ sales were less thanRMB 10 million. The key components and sub-assemblies, whichconstituted about 10% of the total components used in the finalproducts, were made by the 7 key-part suppliers; the other 90% ofthe parts and components were purchased from the remainingsuppliers. In 2005, ZT spent about RMB 5.5 billion for purchasingof parts, components and sub-assemblies. More than 95% of totalannual sales of ZT were made through the domestic distributionnetwork; the remaining 5% of sales were made through the 5oversea sale offices and 30 sales agencies.

5.1.2. Organizations and informants

Because the purpose of this case study was to investigate theevolutionary complexity of the network structure and collabora-tion mechanisms of supply networks, we focused on theevolutionary process of the structure (horizontal complexity,vertical complexity), and the internal mechanisms (products,technology and collaboration strategies) of the focal manufactur-ing firm (ZT) and its major partner firms.

The drivers of the evolution of CASNs are, as stated in thesimulation section, the environment and the internal mechanismsof the member firms. Therefore, the regulators, such as thegovernment and the LVEA industry association of Wenzhou, aswell as firms involved in the LVEA supply networks, were allinvestigated. Data were collected through on-site visits andinterviews with the appropriate managers of the participatingfirms, and with the regulators.

At the beginning of the study, the research objectives and typeof information needed were explained to ZT, and ZT managerswere asked to assist in this effort of selecting upstream anddownstream partner firms based on three criteria. First, they hadto be major partner firms with strategic importance. Second, thesupply management approach for these partner firms shouldreflect the general supply chain management practices of ZT.Third, the partner firms should be willing to take part in theinvestigation.

Overall, the final assembler (ZT), 7 firms in the supply basenetwork, and 6 firms in the distribution base network wereselected. The 7 supply base firms included 2 first-tier keysuppliers owned by ZT, 2 first-tier independent suppliers, and 3second-tier suppliers. The distribution base firms included 1 first-tier regional distribution center, 2 independent dealers, and 2retailers. The partner firms in the first tier of the supply basenetwork and the distribution base network required multiplevisits. In each tier of the network, more than one firm was visitedfor the purpose of data triangulation. Since it was not possible tovisit all the firms in the supply network, the data from first-tiersuppliers and first-tier regional distributors regarding the evolu-tion of the LVEA supply networks were given higher weight. Theindustrial development office of the Wenzhou government, andthe LVEA industry association of Wenzhou also participated in thestudy.

After obtaining the initial approval from the participating firmsand the regulators, the research objectives and type of informa-tion needed were explained to each firm and the regulators wereasked to assemble basic data before the visits. At the same time, a

Page 10: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 319

contract that assured the confidentiality of the gathered data andinformation was signed with the participating firms.

5.1.3. Data sources

All data and information were gathered between July, 2001,and September, 2005. In the first stage, from July, 2001, toSeptember, 2003, we visited the selected firms and regulators togather the raw data and information. In the second stage, from2003 to 2005, e-mails, telephone conversations, and internaldocuments were used for clarification and additional datacollection. The data and information came primarily from threesources: semi-structured interviews, documents, and observa-tions.

(1) In-depth semi-structured interview: When conducting inter-views at the firms, we visited the following individuals: thechairman of the board or the president of the firm who wasfamiliar with the development history of the firm, the salesmanager who interface with the customers, the operationsmanager who was responsible for manufacturing and deliveringproducts to customers, the purchasing manager who procuredparts and materials for the product, and the research anddevelopment manager who was responsible for new productdevelopment. Before the interviews with individual managers, thepresident of the ZT organized a meeting, attended by themanagers involved in order to coordinate the interview process.When visiting the Wenzhou industrial development officer, andthe president of the Wenzhou LVEA industry association, thedevelopment history of the LVEA industry, and the evolution ofthe market, and of the relevant policies and regulations over thepast 30 years were addressed. The intent was to collect data onthe evolution of the upstream and downstream supply-networkstructure and collaboration mechanisms, as well as the externalenvironment, and the internal operations interface with the focalfirm, its suppliers and customers.

The interview questions were semi-structured and an inter-view framework was developed to conduct the interview (seeAppendix C). The interview instrument addressed the firm’sgeneral background and development history, as well as theevolution of its internal mechanism, and the evolution of theenvironment over the past 30 years. There was no fixed wordingused to ask questions. In keeping with inductive methodology(e.g. Sutton and Callahan, 1987), the conversation was allowed toproceed at the interviewee’s pace; the interviewer simply had theresponsibility to make sure that all the questions on the interviewtool were addressed (Choi and Hong, 2002). The interviews wererecorded and supplemented with company documents, observa-tion data, and illustrations provided by the firms.

For triangulation purposes, the same questions were addressedto multiple informants. Relevant documents were collected tovalidate the informants’ responses. Each interview lasted from 3to 4 h. Interviews were transcribed and verified by the informantsto increase construct validity (Yin, 1994). Repeat visits were madeto finish the interview with the same person or to pick upadditional interviewees. The general rule used in stoppinginterviews was to stop when no more new information wasforthcoming—this marks what Glaser and Strauss (1967) andEisenhardt (1989) refer to as the data-saturation point.

(2) Documents: Five kinds of documents were collected from ZTand its partner firms. (1) A list of ZT’s main products of over thepast 20 years was requested. ZT provided a documented list ofthe products’ names, total annual sales of each product, and thelifecycle of each product. (2) A list of the first-tier suppliers anddistributors of the main products in the past 20 years wasrequested. A list of the supplier names, locations, part names, andtotal annual purchases from each supplier was collected. (3) The

BOM (Bill of Materials) for the main products and theirmanufacturing process planning documents over the past 20years were requested from both ZT and the suppliers. Thesuppliers provided a documented list of part names, numbers,prices, manufacturing processes, and corresponding suppliernames. (4) The supplier selection criteria utilized by ZT and thefirst-tier suppliers performance over the past 20 years werecollected. (5) Other documents related to supplier management.These kinds of documents were asked to be kept in confidence.

Three kinds of documents were collected from the localgovernment and the LVEA industry association of Wenzhou. (1)The development history of the LVEA industry in Wenzhou. Theregulators provided documents that described the emergence andthe development of the LVEA industry in the past 30 years. (2) Theofficial statistical data on the number of firm, firm sizes, totalworkers, and total annual sales of the LVEA industry in the past 30years. (3) The process of economic reform, as well as the evolutionof the policies, laws and regulations that influenced the LVEAindustry in Wenzhou.

(3) Observation data: A plant tour was requested and offered atall firms visited. During the tour of the supply base network, themain parts, components, sub-assemblies and final products wereidentified. Also, the manufacturing process of the main productsof each firm was observed. During the tour of the distributionbase network, we visited the final customers, and observed thebusiness transaction process between the customers and thedistributors or retailers. The purpose of these observations wasprimarily to verify the information collected from interviews anddocumentation.

5.1.4. Data analysis

Since a supply network is essentially an organizational form inlarger context or a system of firms, we can capture the qualitativetraits of the evolution of a supply network by examining thedimensions of the physical structure (network, product), thetechnology (product, process, information sharing), the strategy(business objectives, operation strategy), and the organization(formalization, centralization) (Choi and Hong, 2002; Carbonaraet al., 2002). The methodology of our research was to searchinductively for evidence of the evolution of the LVEA supplynetwork that would validate the findings from the simulationsdescribed above.

The unit of analysis is a single informant within the ZT supplynetwork. As in previous qualitative research (Choi and Hong,2002; Harris and Sutton, 1986; Oke and Gopalakrishnan, 2008;Van Maanen, 1983), the present analysis sought common patternsacross the multiple interviews carried out. Any differences foundwere noted and reconciled (Poole and Van de Ven, 1989). Withinthe context of this study, we focused on the supply networks ofthe LVEA product family produced by the focal manufacturingfirm (ZT). In line with the approach suggested by Miles andHuberman (1984), we began with a within-case analysis in whichwe presented the interviewees’ response to the evolution of theLVEA supply networks that were unique to the informants.Following this we compared and contrasted the responses of theinformants and as a result of this comparison, we developed a setof principles that eventually lead to some general propositionsregarding the evolution of CASNs (Choi and Hong, 2002; Oke andGopalakrishnan, 2008).

5.2. Evolution of the LVEA supply network

5.2.1. Emergence

The emergence of the ZT LVEA supply network can be tracedback to 1970. In that year, while visiting friends in Anhui Province,

Page 11: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330320

a local farmer named Chen heard about the lack of ACC(alternating current contactors) for the coal mining industry. In1970s, the Chinese economy was a ‘‘planned economy’’, highlycontrolled by the government. Privately owned firms wereprohibited. Nevertheless, Chen founded a clandestine workshopin 1972 to make ACC. The workshop was quite successful.Unfortunately, it was closed by the local government in 1975.However, this closing could not dampen people’s aspirations tomake profit. In the years that followed, the former employees ofthe workshop founded many new clandestine workshops in theWenzhou area.

The products produced by those early workshops were of lowtechnological complexity and their structures were very simple.They were hand crafted and characterized by low levels of processtechnology complexity. To protect their secrecy, the workshopsdid not exchange any information or products. There was littlecollaboration and the level of division of labor among them wasvery low. Instead, the individual workshops were virtuallyautonomous. They purchased raw materials, carried out theentire production process, made all the parts, assembled theminto final products, and sold the products by themselves. As aresult, the supply network only included the raw materialsuppliers and the final products manufacturers.

But as the demand from the coal industry grew, and thenumber of local workshops multiplied, some of them began tospecialize and to divide up responsibility for various parts of theproduction and distribution processes. The first distributoremerged in 1978, and a full-scale supply network emerged inWenzhou in 1981, at which time there were more than 300workshops specialized in supply, manufacturing, and distribution.However, the supply network suffered a fatal blow in 1982. Thelocal government closed about 200 firms for the same reason as in1975, resulting in a 53.8% decline in total annual sales of the LVEAindustry. Fortunately, the industry got another chance in 1984, forit was in that year that the Chinese central government, togetherwith the local government of Wenzhou, issued laws andregulations to encourage the foundation of private firms. Morethan 1,000 firms emerged and the parent firm of ZT, called QJ, wasfounded.

Initially, QJ’s main products were low voltage switches, whichwere of low complexity, both in terms of product structure andtechnology. QJ copied the product design from other firms. As aresult, the products of QJ were no different from those of otherfirms. The manufacturing process was very simple. QJ procuredraw materials to make some parts through a simple mechanicalprocess. In addition, QJ procured some parts from suppliers andassembled them, with in-house parts, into the final product.Consequently, there were only a few stages in the supply network.

By the end of the 1980s, there were more than 1000 LEVAindustry firms located in Wenzhou. Their products were verysimilar, and as a result, the competition among them was intense.The only way to increase demand was to decrease the price. As aresult, cost reduction became one of their primary operationalobjectives. QJ was the firm which put the most effort intofounding a distribution network in order to increase demand.From 1984 to 1991, QJ collaborated with more than 1000salesmen all over China to sell its products. They not only soldQJ’s products, but also sold products from other competing firms.Their collaboration with QJ was based on oral and informalagreements.

With regard to strategy, QJ adopted the ‘‘make-to-order’’ and‘‘buy-to-order’’ operational strategies at the emergence stage. Thecollaboration with more than 1000 salesmen helped QJ secure alarge number of orders and resulted in a shortage of capacity. Tofill the demand, QJ either procured parts then assembled into finalproducts subject to its in-house capacity (the ‘‘make-to-order’’

strategy), or subcontracted the surplus demand out to other firms(the ‘‘buy-to-order’’ strategy). However, QJ never disclosedinformation about its demand or its production plans to itssuppliers or subcontractors. Its selection of suppliers andsubcontractors was always based on random market opportu-nities and was short-term oriented. Moreover, QJ always selectedthe lowest bidders as its suppliers and subcontractors. To reducethe operating cost of operations, QJ frequently changed suppliersand subcontractors, which contributed to increasing the numberof suppliers and subcontractors in the same tier and to the varyingstructure of the supply network.

With regard to the structural and organizational dimensions,the supply network at the emergence stage was decentralized,dynamical and loosely controlled. The geographic dispersion ofthe supply network was high; some suppliers were far away fromWenzhou. The horizontal complexity of the supply network wasalso high; there were many suppliers and subcontractors at thesame stage. However, the vertical complexity was low, in that thesupply network had few stages.

At the emergence stage, the collaboration mechanismsof the supply network were characterized by the ‘‘casual-vertical-dependent collaboration schema’’. The suppliers and thesubcontractors were only viewed as passive doers (or capacitybuffers) for supplying QJ’s internal capacity, although they alsohelped QJ reduce the cost of operations and improve capacityflexibility. Their collaboration was based on random opportunitiesand there was no information sharing among them. To maintainflexibility and to insure the most advantageous pricing, thecollaboration between QJ and partner firms was based on oralagreements, which typically set the price, the quantity, and thequality, where the price and the quantity was explicit, and thequality was implicit. As a result, the relationship between QJ andits partner firms was unstable, and their inter-dependence waslow.

5.2.2. Adaptation

The intensive price competition and the implicit agreement onquality at the emergence of the LVEA supply network contributedto a rapid decline in both the quality and the price of the finalproducts. As a result, QJ fell into a declining path where it sufferedgreat losses from price competition and saw its profits graduallydecreased. To survive, QJ adopted the high-quality productionstrategy. It collaborated with technology specialists, researchinstitutes and universities to improve product design and quality.In its partner selection, QJ no longer put the attention on pricealone, but also began to pay attention to quality. In 1988, QJreceived a production license from the government, whichattested that its products were of high quality. The improvedquality made QJ’s products different from those of the other firms.As a result, QJ began to receive more orders, even though itsproduct price was higher than that of the other competing firms.At the end of 1988, QJ escaped from inferior equilibrium andearned RMB one million in profit.

The high quality strategy and the production license enabledQJ to survive in 1990. In that year, the government began toregulate quality standards in the LEVA industry and issuedproduction licenses to firms whose products were deemed tomeet the standards. As a result, 1268 firms with poor qualityproducts were eliminated from the market. QJ and six other finalassembly firms survived and grew rapidly during that year. QJ’stotal sales in 1991 reached RMB 10 million. Also in that year, QJsplit into 2 firms, one of which was ZT.

The establishment of ZT Corporation marked the evolution ofthe LVEA supply network into the adaptation stage. During its firsttwo years, ZT continued the operational strategy of QJ, and

Page 12: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 321

continued its collaboration with the former partner firms of QJ.In 1993, ZT’s total annual sales reached RMB 50 million.

The adaptation of the LVEA supply network from 1994 to 2000was characterized by the ‘‘vertical-dependent collaborationschema’’, i.e., the horizontal and vertical complexity of thenetwork was increasing, while the geographic dispersion of thepartner firms was decreasing. The supply network was becomingmore centralized and formalized than had been the case in theemergence stage.

With respect to its product selection, ZT adopted thedifferentiation strategy. It began to make whole sets of LVEAswhich were of higher structural and technological complexitythan those of the other manufacturers. Both the width and thedepth of the BOM tree had increased.

With regard to the structure of the supply network, theincreasing complexity of the product structure had the result ofinvolving many new suppliers at the beginning of the adaptationstage; at the same time, the width and the depth of the networkincreased. However, as noted above, the geographic dispersion ofthe supply network was decreasing. With the increasing demandfrom ZT, some suppliers established their distributors, salesagents and warehouses in Wenzhou in order to reduce costsand improve service quality. About 70% of the raw materials andparts could be purchased in the local market. Also, ZT strength-ened and enlarged its distribution network in order to increasedemand and market share. One thousand former salesmen wereupgraded into franchised retailers. More than 800 retailers, and21 exclusive distributors organized a formal collaborative rela-tionship with ZT. A three-stage distribution network emerged,involving exclusive regional distributors, distributors in middlesize cities and retailers in small cities and counties. Thisdistribution network got increasing demand of LVEA in themiddle of 1990s, due to the acceleration of China’s marketeconomy reforms and the rapid growth of the Chinese economyduring this period.

1st tier supplier

779 parts suppliers

Plastic Parts Co.

Electronic Parts Co.

Jointing Parts Co.

Punching Parts Co.

Mould Parts Co.

Shell Parts Co.

Case-frame Parts Co.

7 key- part supplier owned by C

HIN

T

……

More than 3000

Second-tiersuppliers

loca

l sup

plie

rs

out-

of-t

own

supp

liers

Purc

hasi

ng

(Inb

Supply base network

Fig. 4. ZT LEVA su

With regard to technology, the increasingly complex productstructure led to an increase in the complexity of product and processtechnology. Some new materials, technology and processes wereintroduced into product development. More than 30 patents wereissued to ZT by the end of 2000. ZT introduced new productiontechnology such as computer numerical control and flow productionlines to improve production efficiency and flexibility.

With regard to strategy, ZT’s business objectives included notonly cost reduction, but also higher quality and increaseddifferentiation. ZT moved its operations strategy from the ‘‘buy-to-order’’ scheme to a combination of the ‘‘make-to-order’’ and‘‘assemble-to-order’’ schemes in order to deliver the increasingnumber of orders at low cost and with high quality. ZTreorganized the supply network and enhanced the division oflabor. The criteria for supplier selection not only emphasizedprice, but also took into account quality and collaboration history.Forty-eight local firms were acquired and merged into the ZTGroup Co. Those firms within the same product families weremerged into 7 key parts suppliers (see Fig. 4). They carried most ofthe key processes to ensure cost and quality. ZT subcontractedabout 90% of the parts to the suppliers, and obtained theremaining 10% from the 7 key parts suppliers. Finally, itassembled the parts and sub-assemblies into the final products.To facilitate collaboration, ZT began to exchange its demand andproduction planning with its key suppliers. The subcontractinghelped ZT reduce manufacturing costs by about 30%. In 1994, ZTpassed the ISO9000 quality certification and implemented the‘‘Second Party Quality Certification Plan’’ which ZT helpedsuppliers to improve their production quality and certificate thequality of their products in order to further ensure quality.

The adoption of the differentiation and high quality strategiescontributed to ZT’s progress in 1998. In that year, with the Asianfinancial crisis and the associated weakening of the Chineseeconomy, market demand decreased significantly. In an attempt tostimulate demand, more than 1000 firms with similar products

ZT Plant

Man

ufac

turi

ng

Sale

s

Logistics ound, Outbound)

6 Overseas Sales Offices

2000

Ind

epen

dent

Dea

lers

12 D

omes

tic R

egio

nal D

istr

ibut

ion

Cen

ters

(9

5% o

f sa

les)

30 Sales Agencies

Retailers

Ret

aile

rs

5% of total sales

95% of total sales

Distribution base network

pply network.

Page 13: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330322

engaged in a price war which resulted in great losses for all of them.In that year, more than 300 firms in Wenzhou went bankrupt, andthe supply network fell into chaos. However, ZT managed to escapethe damage, and achieved 30% growth in annual sales in 1998.

On the organization side, the centralization and formalization ofthe supply network was enhanced. The relationships between ZTand its partner firms moved to a long-term horizon and the level ofcooperation improved. The company implemented the ‘‘Grand ZTPlan’’ in order to share the benefits of its well-known trademarkwith partner firms, and to provide financial and price support for itsprincipal distributors. In turn, the suppliers permitted ZT to delaypayment on orders. In addition to price and quantity, qualitystandards were explicitly specified in oral or written agreements.Also, ZT routinely exchanged information about the demand, theproduction plan, the competition, and the technology with suppliersby telephone, fax and mail. Benefiting from ZT’s support, manysuppliers which had made simple parts in the emergence stageevolved into suppliers of complex components and sub-assemblies.Overall, ZT exerted considerable influence on its partner firms, andthe inter-dependence among them increased.

With the evolution of the ZT supply network into the adaptationstage, the complexity of its structure, its degree of centralization andformalization, and the level of cooperation among partner firms allincreased. Path dependence and a win–win schema emerged out oftheir collaboration. In the year 2000, ZT had about 1200 first-tiersuppliers. And the distribution network had 800 partner firmsorganized into three stages. With the growth of demand, the growthrate of ZT’s annual sales had been kept in 50% for 6 years. In 2000,ZT’s total annual sales reached RMB 4.2 billion. The company’ssuccess contributed to the success of its partner firms, andconsolidated their inter-dependence and the stability of the supplynetwork. ZT’s LVEA supply network of evolved into the ‘‘vertical-dependent collaboration schema’’, which gave it more control overcost and quality.

5.2.3. Maturity

Since the early 2000s, ZT’s LVEA supply network has evolvedinto the stage of maturity, which is characterized by the‘‘horizontal-vertical-dependent collaboration schema’’. With re-gard to the mechanisms of collaboration, the level of centraliza-tion and formalization of the supply network has increased. As tothe structure of the supply network, the horizontal complexity,and the geographic dispersion have decreased, but the verticalcomplexity of the supply network remains high.

With regard to its product line, ZT has intensified theproduction differentiation strategy. It continues to make LEVAs,but has also begun to produce whole set high-voltage electricapparatus (HVEAs). Many new intelligently controlled compo-nents are used in the LEVA products to improve their ‘‘smartness.’’The increasing complexity of LEVA products and the introducingof HVEAs result in a more complex supply network in bothhorizontal and vertical dimensions.

With regard to technology, the technological complexity of theproducts has increased. Many advanced ICTs (information andcommunication technologies) are used extensively in the productdesign and manufacturing processes. Auto-control technology isused in LVEA products, making possible various forms of remotecontrol by computers. Through the use of auto-control technol-ogies, the products have become ‘‘smarter’’ and ‘‘smarter’’, thequality and performance of the products have been improvedsignificantly, and after-sales service costs have been reduced. Thetechnological complexity of the production process has increasedas well. Increasing demand has led to the establishment of morethan 150 new computer controlled automated flow lines in themanufacturing processing. Digitally controlled machining centers

and auto-testing machines are extensively used in the manufac-turing process. The high level of automation of the manufacturingprocess has greatly increased output volume while improvingquality. To facilitate the collaboration between ZT and its partnerfirms, an ERP (Enterprise Resource Planning) system has beendeveloped to manage collaboration activities. Through the ERPsystem, ZT shares demand information, procurement and produc-tion plans with suppliers on a monthly basis.

With regard to strategy, ZT’s business objectives emphasize notonly quality, cost reduction, and differentiation, but also flexibility.Regarding differentiation, ZT has not only developed many newproducts, but has also turned its attention to LVEA system services.The expansion of its product line gives ZT the capability to provideits customers with a complete electric solution. It not only providesLEVA and HVEA products to its customers, but also provides themwith product related services including demand investigation,design services, installation, and maintenance. In the area ofoperations, ZT has been using the ‘‘assemble-to-order’’ and ‘‘de-sign-to-order’’ strategies since 2003. Not only does it produce finalproducts itself, but it also designs products and then subcontractsthe manufacturing of these products to its first-tier suppliers. Toimprove flexibility, the subcontractors are in turn permitted tosubcontract some of these orders to other suppliers in the same tierof the supply network. This horizontal subcontracting enhancesinter-reliance among suppliers at the same tier. Furthermore, ZT hasadopted an international marketing strategy. It has establishedoverseas sales offices and distribution agencies in East Asia, Europeand the USA, which has increased the depth complexity of thedistribution base network.

Organizationally, the centralization of the supply network andthe cooperation among the firms has been reinforced. To limit thecomplexity of the supply network, which resulted from thecomplexity of the product structure, technology and horizontalsubcontracting, ZT has adopted a supplier-bidding mechanism inorder to reduce the number of suppliers. From 2000 to 2005, thenumber of first-tier suppliers decreased from more than 1200 to786. ZT has forged formal long-term collaboration contracts withthese suppliers, and implemented a price protection mechanism inorder to insure that they could make a profit from their collaborationwith ZT. Meanwhile, ZT has implemented some incentive mechan-isms for suppliers, such as technology and information sharing, andinternal transfer pricing, which have boosted technological innova-tion, cost reduction and quality improvement on the part ofsuppliers. Also, lean production and Just-In-Time have beenimplemented in first-tier suppliers. To decrease inventory costsand improve service levels, most of the key suppliers have beenrequired to implement VMI (vendor managed inventory) with ZT. Asto the distribution network, a web-based e-commerce system hasbeen developed between ZT and the distributors. These measureshave helped the partner firms to enlarge their production andreduce their costs. They have also increased partner firms’ relianceon ZT and reinforced the centralization of the supply network.

With the centralization of the supply network, the formalizationof the supply network has also been enhanced. In order to controlthe costs and to improve quality and service levels, ZT has increasedits control and influence over the entire supply network. A suppliermanagement office was established in 2004 to manage thecollaboration with suppliers. It exchanges operational informationand supplier’s performance information with suppliers, and imple-ments many supplier support plans to help them improve theirmanufacturing processes and management practices. ZT has set up ajoint team with suppliers in order to coordinate activities ondemand forecasting, supply and demand information sharing,inventory, and production decisions. More than 50 industrialengineers from ZT work with suppliers to develop new products,improve quality, and control costs. ZT also organizes training

Page 14: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTIC

LEIN

PRESS

Table 4Summary of the evolution of the LVEA supply network.

Dimensions Emergence (Casual-vertical-dependence) Adaptation (Vertical-dependence) Maturity (Horizontal-vertical-dependence)

StructureProduct structurecomplexity

Low horizontal complexity Low-High: horizontal and vertical complexity High horizontal and vertical complexity

Low vertical complexityNetwork structurecomplexity

High horizontal complexity (Many homogeneous

suppliers and distributors were at the same tier)

High: Horizontal complexity (Many new suppliers

involved in the supply network)

High horizontal complexity and verticalcomplexity (The horizontal and vertical complexity

of the supply base network was decreasing, and the

complexity of the distribution base network was

increasing)

Low vertical complexity (Only few stages of the

supply network)

Low-High: vertical complexity (The stage of the

supply base network and distribution base network

increased)

High geographic dispersion (Many suppliers were far

away from the manufacturer)

High-Low: the geographic dispersion wasdecreased (Some suppliers developed their

distributors, warehouses or sales agencies in Wenzhou)

Low geographic dispersion (95% suppliers built their

plants and warehouses in Wenzhou)

TechnologyProduct technologycomplexity

Low Low-High High

Process technologycomplexity

Low Low-High High

ICTs No Yes (Manufacturing process ICTs; ManagementInformation Systems)

Yes (ICTs for intelligent products and auto-control

manufacturing process; ERP)

StrategyBusiness objectives Cost reduction Quality, cost reduction, differentiation Differentiation, quality, cost reduction, and

flexibilityOperation strategy Make-to-order and Buy-to-order Make-to-order and assemble-to-order Design-to-order and assemble-to-orderCompetition strategy Price Quality, price Differentiated product, quality, price, and serviceProduct differentiation No Yes (Differentiation based on high quality) Yes (Product innovation)

Supply chain planningactivities

No Decentralized-Centralized Centralized

OrganizationFormalization Oral agreement (Explicit price and quantity, and

implicit quality)

Formal contracts and oral agreements (Explicit price,

quantity and quality)

Formal contracts, supplier management office;

supplier support plans

Centralization Decentralized Decentralized-Centralized (Trademark sharing;

second party quality certification plan; franchised

retailer.)

Decentralized-Centralized (Supplier reduction; JIT;

VMI)

Level of cooperation Low Low-High HighTime horizontal ofrelationship

Short-term Short-term-Long-term Long-term

Information sharing No Yes (Feedback on performance; information sharingon demand, production plan and technology bytelephone, fax and mail)

Yes (Routinely operation information sharing on

demand forecast, production plan, inventory, new

product development, technology, and competition

by IT systems)

Coordination Market price (Competition on price) Market price Incentive mechanism (Bidding price protection;

internal transfer price; finance support, etc.)

Finance and price support for distributorsPostpone of payment for goods

Structural flexibility High High High

G.

Liet

al.

/In

t.J.

Pro

du

ction

Eco

no

mics

12

4(2

01

0)

31

0–

33

03

23

Page 15: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330324

programs on quality and cost control for suppliers. As a result of thissupport, the quality, cost, flexibility and service level of suppliers hasimproved greatly.

In the maturity stage, with the evolution of the firm’s internalmechanism, the width, depth and the geographic dispersion of thesupply base network is decreasing. There are fewer suppliers in eachtier and fewer stages of the supply base network. More than 95%suppliers built their plants, distributors, sales agencies or warehousein Wenzhou. Furthermore, the centralization and formalization ofthe supply network was enhanced, which helped ZT to better controlthe cost, quality, flexibility and service level, and reinforced thepath-dependence among firms and the stability of the supply basenetwork. However, the horizontal and vertical complexity of thedistribution base network was increasing with the implementationof the internationalization strategy. With the collaboration, ZT andits partner firms reached to superior equilibrium and thrived.

In summary, we have outlined what we have observed in thewithin-case analyzes of the evolution of the LVEA supply networkin Table 4. The observations indicate that the LVEA supplynetwork has emerged from the dynamic interaction amongmember firms and the emerging market. As a result of changesin the environment and in the internal mechanisms of the firms,the LVEA supply network evolves. In the evolution process, manycharacteristics of CAS (e.g., emergence, lock-in, path-dependence,quasi-equilibrium, and chaos) emerge. It seems that thecomplexity of the product and the technology are what havegiven rise to the complexity of the LVEA supply network. Cost andquality considerations drive the LVEA supply network to evolvetoward more formalization and centralization. Long-termrelationships lead to a better fit with the environment, resultingin greater stability and better performance.

6. Results of the study: propositions

The simulation and the case study show how CASNs emergeand evolve as a result of the dynamical interaction among firms inthe environment. With the changes of firms’ internal mechanismand the environment, the structure and the collaborationmechanism of CASN evolves. Our findings are consistent withearlier studies which have demonstrated that the supply networkis a CAS (Choi et al., 2001; Li et al., 2009).

Proposition 1. The CASN emerges and evolves as a result of the

dynamic interaction among firms in the environment.

Within the case study, different factors that influence theevolution from the external environment perspective are identified.First, the government regulations are the prominent factor thatgoverns the emergence of LVEA supply network in a long-term. Theshortage of LEVAs which resulted from the highly controlled Chineseeconomy in the 1970s gave rise to the growth of the LVEA industryin Wenzhou. With the emerging market economy and the loosecontrol of private firms, the LVEA supply network emerged, adaptedand evolved into a huge network, within which ZT evolved into thefocal firm. Large demand stimulated ZT to build its extensivedistribution network. Government regulation of quality supportedZT and led to enhanced formalization and centralization of thesupply network. The fluctuation of the LVEA industry in Wenzhoucan also be traced to the government regulation. With the freemarket reforms of the economy, direct government control of theLVEA industry decreased and the LVEA supply network evolvedindependently. Second, the market structure influences the evolu-tion of the structure and the collaborative mechanisms of LVEAsupply network. The great demand fostered the emergence of manyfirms in the local market and resulted in intense competition amongthose firms. To get more orders, firms have to expand their

distribution base network. The more the firms there are in the localmarket, the faster the decline in price and quality, and the moredisorder there is in the market. Those firms with differentiatedproduct lines and high quality are the ones most likely to succeed inthe market and evolve into the focal firms of the supply network.

Proposition 2. From the macro-institutional constraints perspec-

tives, government regulations, market demand, and market struc-

tures have a long-term impact on the gradual evolution of CASNs.

Proposition 3. The more government regulation there is, the less

likely it is that a CASN will emerge.

Proposition 4. High demand fosters the emergence of the supply

base network and the expansion of the distribution base network.

Proposition 5. The more firms there are in a local market, the

greater the likelihood of chaos in the market and the greater the

instability of any CASNs.

From the simulation experiments, we have learned that a firm’sinternal mechanisms (e.g. product structure, collaboration strat-egy) influence its partner selection. In addition, the case studyindicates that product structure, technology, strategy and orga-nizational considerations influence a firm’s responses to themarket and its collaboration relationships with other firms. Thefocal firm’s business objectives affect its preferences on compe-titive priorities (e.g. cost, price, quality, service), which furtheraffect its products, technology and operations strategy.

The existence of similar products in the local market results inintensive price competition and chaos in the market. In order tosurvive, the final assembler, as the focal firm in the supplynetwork, can move to the product differentiation strategy,producing new products with higher levels of technologicalcomplexity. A higher level of technological complexity of thefinal product results in the need for more parts and manufactur-ing processes, requiring more suppliers and more stages in thesupply network. High value-added processes are always con-trolled by the focal firms. To produce the complex products, costcontrol, quality control, more formal collaboration and intensivecollaboration among firms is required, which further reinforcesboth the dependence among partner firms and the stability of thesupply network.

Proposition 6. The strategies, product structure complexity, tech-

nological complexity, and organizational considerations involved in

the internal mechanisms of firms are the internally dominated forces

that shape the gradual evolution of CASNs.

Proposition 7. Higher structural complexity and technological

complexity in final products results in more suppliers and more

stages in CASNs.

The final assembler’s operational strategy, and preferences onthe time horizon of relationships, influences its selection ofpartner firms and the stability of the CASN. The make-to-orderand buy-to-order strategies at the emergence stage result in moresuppliers and greater instability of the LVEA supply network,while the assemble-to-order strategy results in fewer suppliersand a more stable supply network. The short-term collaborationstrategy and low cost orientation results in more suppliers at thesame stage and the instability of the supply network at theemergence stage. The preferences for long-term relationships, lowcosts and high quality result in fewer suppliers, higher stabilityand better performance of the supply network.

Proposition 8. Long-term collaborations and formal agreements

enhance the stability of CASNs and the performance of the partner

firms.

Page 16: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 325

Proposition 9. Short-term collaborations and informal agreements

result in the instability of CASNs and higher flexibility in the

collaborations.

At different stages of the evolution, the final assembleremphasizes different internal factors in order to adapt to andrespond to changes. Cost and quality pressures affect how firmsselect suppliers and their collaboration patterns thereafter. Withregard to environment, government regulations are the salientfactors that affect a firm’s birth and the emergence of CASNs. Withrespect to the internal mechanism of firms, cost is the mostsalient factor that shapes emergence of CASNs (Choi and Hong,2002). In the adaptation stage, cost and quality considerationleads to the formalization and centralization of CASNs. Theincreasing demand of the final assemblers and its impact on costand flexibility during the mature stage leads to a reduction of thegeographic dispersion of CASNs.

Proposition 10. Cost and quality considerations of the final

assembler enhance the centralization and formalization of CASNs.

Proposition 11. The increasing demand of the final assembler and

its attention to cost and flexibility results in the reduction of the

geographic dispersion of CASNs.

The simulation and the case study shows that the evolution ofCASNs is a complex process, which is affected by the environmentalfactors and firms’ internal factors. The positive and negative feedbackfrom these diverse factors makes evolution unpredictable (Li, et al.,2009). Different initial factors or slight interruptions may drive theevolution into different patterns, where some may be cases ofsuperior equilibrium, and some may be cases of inferior equilibrium.Only the mutual compatible collaboration pattern among partnerfirms can boost path-dependence among firms and enhance thestability of CASNs. The win–win collaboration pattern for partnerfirms in a CASN results in them thriving at the edge of chaos. Theevolution of CASNs is dynamic and cannot be precisely predicted.

Proposition 12. The evolution of CASNs is highly sensitive to the

initial conditions, locking-in and path-dependent. Slight random

disturbance from the environment or from the internal mechanism of

firms can drive the evolution into chaos.

7. Discussion and managerial implications

Based on this study, government regulations, demand andmarket structure are identified as the external environmentalfactors that have the most impact on the evolution of CASNs. Atthe emergence stage, the external factors have tremendousimpact on the birth of CASNs. Great demand may foster theemergence of a new industry and, together with that, theemergence and adaptation of CASNs. The increasing demandfrom the growing final assembler at the adaptation and maturitystages leads to the reduction of the geographic dispersion ofCASNs. But the government’s intensive planning and inappropri-ate regulation may defer the development of the industry andwith it, the emergence and adaptation of a high-performancesupply network. Many firms in a local market may drive themarket competition into chaos and result in the instability andpoor performance of the CASN. But as the structure of the CASNinfluences the behavior of firms, appropriate external interrup-tions can be used to manipulate the structure and drive the CASNto escape from chaos and achieve the desired goals.

This study has identified firm strategies, product structurecomplexity, technological complexity, and organizational considera-tions as the predominant internal factors that influence theevolution of CASNs. Among these factors, cost considerations are

important for firms in developing collaboration relationships at theemergence stage of CASNs. However, not only cost, but also qualityaffects the adaptation and maturity of CASNs. With the evolution ofCASNs, the final assembler’s considerations on cost and quality boostthe centralization and formalization of CASNs. Theses findings differfrom those of the existing literature, which states that costconsiderations are the most salient factor shaping the emergenceof supply networks, and leading to the creation of the centralizedsupply chain (Choi and Hong, 2002).

Based on our findings, we identify the technology complexity ofthe final product and the manufacturing process as new factors thatenhance the structural complexity, centralization, and formalizationof CASNs. Products with higher levels of technological complexityrequire more parts and manufacturing processes, and requireintensive quality control, which results in higher horizontal andvertical complexity of CASNs. To control the complexity, the finalassembler has to enhance the control of the partner firms. As aresult, it intensifies the centralization and formalization of CASNsand makes use of the long-term contract with partner firms. Withthese activities, the stability of the collaboration and the perfor-mance of partner firms are improved. These findings support Hur etal.’s (2004) view that technology influences the complexity of supplynetworks and disclose the relationship between technology and theorganizational form of supply networks.

Furthermore, this study reinforces the view that long-termrelationships are better for the stability of supply networks andthe performance of partner firms. It indicates that the explanationof the stability and performance of CASNs as deriving from boththe strategic and technological dimensions is better than anexplanation based on one dimension only. For final assemblerswho make products with high technological complexity, it isbetter for them to form long-term collaborations with partnerfirms in order to control quality, cost and risk. However, forproducts with low technological complexity, short-term colla-boration and implicit agreements may lead to greater flexibility inrelationships and heighten the sense of equity. This could be usedto explain different views on the influence of a firm’s collabora-tion strategy on the performance of supply networks. Thecollaboration strategy for supply chain management should fitto the technological complexity of products.

This study contributes to the literature on operations manage-ment in several ways. First, it proposes an evolutionary model ofsupply networks based on CAS and fitness landscape theory. Thisapproach underscores the importance of models in which differententities in a supply network operate subject to their own localstrategies, constraints and objectives. With the simulation of theevolutionary model, firms’ dynamic behaviors can be analyzed froma variety of organizational perspectives. Many characteristics of CASare identified from the simulation experiments of CASNs. Theevolutionary mechanism are important issues that help us to betterunderstand the antecedent factors and mechanisms affecting theevolution of network structure and collaboration mechanisms andthe consequent outcome in the performance of CASNs. Second,many predominant factors from the external environment and afirm’s internal mechanism dimensions are unveiled. Governmentregulation, market demand and market structure are identified asthe predominant external environment factors that shape theevolution of CASNs in the long run. And a firm’s products, strategy,technology and organizational considerations are identified as theinternal mechanism factors that influence the evolution immedi-ately. In different stages of the evolution, the predominant factorsevolve. The dynamic interaction of the internal mechanism factorsand the environmental factors makes it difficult to precisely predictthe evolution. Furthermore, some propositions are proposed tobetter understand the evolution. The results of this study shed lighton the importance of balancing the external factors and internal

Page 17: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330326

mechanism factors. To get higher performance, firms need to matchtheir products, strategy and organization considerations with theenvironment. Cost and quality are the most important factors thatinfluence the evolution.

The managerial implications of understanding the evolution-ary complexity of CASNs are many. Perhaps the chief insight isthat the evolution is a self-organizing process. A CASN is able torespond in more ways to its environment than a controlling actorcan manage or even envisage. The order of the evolution is theresult of the process of interaction among firms in a dynamicalenvironment, and not the careful deliberation of the regulators orany single firm. Any incorrect assumptions about a CASN bymanagers and regulators can lead to unintended consequencesdue to the continuously evolving nature of CASNs. While the self-organizing evolution of CASNs helps us to understand some of theemergent properties of the supply networks in the dynamicenvironment, it also provides some guidelines for understandingand evaluating the merits of external and internal attempts byregulators and managers to intervene in the evolution.

At the external environment level, regulators of the environmentand firms should be aware of the fact that any planning andregulation of the market and firms will be undermined by the factthat the outcomes are both open and unknowable. Perhaps moreimportantly, the regulator’s influence on the evolution of CASNs caninterfere with the natural process of self-organization by creating anenvironment that is benefit to the emergence and adaptation ofdiversified organizations, innovation and learning, rather than focuson static objectives in a dynamic environment. The managingmethods of CASNs should be adjusted with the changes of theenvironment. It is more powerful to utilize the varied principalfactors that govern the dynamic evolution in different stage tomanaging CASNs. From the evolutionary perspectives, competitionand cooperation among firms facilitates innovation, learning anddiversification of firms, which further boosts the adaptation ofCASNs. Therefore, regulators should support the competition andcooperation among firms and adjust the regulations with theevolving environment. An institutional environment that promoteslabor division, resource utilization, knowledge sharing and businessinnovation of firms should be established. On the contrary, thecircumstance that jeopardizes firms’ adaptation to the environmentand substantial development in the long-run should be prohibited.Based on the establishment of the institutional environment, theevolution of CASNs can be better managed. For example, when theevolution falls into inefficiency or disorder (e.g. intensive competi-tion with low price and quality), regulators can intervene in theevolution by appropriate methods such as selective taxes, subsidiesfor technological innovations, production licenses, and so on.

At the firm level, managers can use the result of this research todevelop their business objectives, strategies, supply networkplanning activities and product planning to enhance the adaptationand the performance of supply networks. Instead of the dominantsupply chain management paradigm of explicit planning in advanceas the most effective way of achieving goals, more adaptive planningis called for where there is a process of continuous modificationbased on current conditions, feedback and improvisation. Thedecision-making for managers should be changed from the tradi-tional deterministic paradigm to the non-deterministic paradigm.Forecast, planning and control should be replaced by learning andinnovation. An open environment, flexible organization structure,and appropriate incentive mechanism should be established tostimulate innovation behavior of different agents in firms. Commu-nication and interaction should be encouraged to foster theinnovation of new products, new business models, and newcollaboration mechanisms, etc. With the emergence of innovations,old orders in firms and CASNs may be broken. A dynamicmechanism needs to be established in order to control the conflict

between innovations and old orders of firms and CASNs, in order toavoid chaos of CASNs.

Managers are advised to match firm’s internal mechanism to theenvironment in different stages in the evolution. To match theevolution of complexity on environmental, a firm’s businessobjectives, products, strategy, technology, culture, organization andbusiness process should be adjusted with changes in the environ-ment. Also, the elements of a firm’s internal mechanism should keepin consistency. The changes of anyone element of a firm’s internalmechanism should be accompanied with the adjustment of the otherelements to keep internal fit. For example, the competitive prioritiesof a firm’s manufacturing strategy should be changed with thedynamic evolution. In different stages of the evolution, the priority ofquality, cost, time, service and environment considerations should beadjusted to keep the internal fit. A high level internal fit of a firm’sinternal mechanism improves the firm’s fitness to the environment.

The findings of this study also suggest that managers who aim toenhance the performance of CASNs should focus on developing long-term collaboration to make the collaborative relationship stable andsustainable. Formalization and centralization are helpful for finalassemblers to control costs and quality, and to reduce thecomplexity of the structure of supply networks. Furthermore, lock-in and path-dependence in collaboration helps to explain why newsuppliers have a difficult time being adopted in a supply network.Many advantages of the former supply network are self-reinforcing,which means that the collaboration schema could not be brokeneasily. Therefore, new suppliers should understand and be comfor-table with various anxieties and frustration in the collaboration.However, long-term collaboration does not equal to the fixed andunchangeable supply network. The former collaborative relationshipneeds to be more open about outcomes and opportunities, ratherthan relying on mechanistic blueprints and ‘‘how-to’’ manuals.Partner firms which are not fit to the competition and cooperationshould be removed, and new firms which are of higher performanceon product innovation, cost reduction, service improvement, etc.

In a dynamic environment, managers and regulators shouldinvestigate the CASN as a whole and learn how to balanceemergence and control of the dynamical interaction among firms.Too loose control of the market place by regulators and of partnerfirms by focal firms may drive the CASN into disorder. A looselycontrolled CASN and market may provide much flexibility intransactions, but it may lead to intensive price competition andlow quality of the final products, and defer the upgrade of the CASN.While, too rigid control may lead to less effective and innovativenetworks. For example, IBM’s tight control of its reseller networkresulted in an efficient but static distribution base network thateventually lost out to the rapidly changing and more innovativesupply networks of other companies. Furthermore, the greater thepower and influence exercised by a firm on its relations andnetworks, the more its partner firms’ sense of inequity. The pressurefrom price bidding and cost-reduction (e.g. 5%) for suppliers yearover year may impede the suppliers from developing innovativeproducts and improving manufacturing processes in the long run,and may result in the instability of the supply network. Thesefurther impede the cost reduction of partner firms and result in alow level of effectiveness and efficiency in the collaboration. Amutually compatible network structure and collaboration mechan-ism that takes more account of the needs of other participantswould seem to play an essential role in bringing about theemergence of more complex, effective and performance CASNs.

8. Conclusion

The evolution of CASNs provides a fascinating area of research.If we are to truly practice management of CASNs, we need to

Page 18: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

Begin

Initiate the environment agent, and the blackboard agent

Initiate the firm agents, and set the total demand cycles

Start the demand cycle

The environment agent generates the global market demand, and the blackboard agent reads the demand information and announces it on the blackboard.

The firm agents compete for the global market demand.

The environment agent selects the firm agent whose fitness is the highest as the winner to execute the global market demand.

The subcontractor makes the order subjected to its capacity, and subcontracts its remaining demand to the other firm agents, until the order can be fulfilled.

Whether the global demand is less than the winner’s capacity?

No

Yes

The winner manufactures the order subjected to its capacity, and announces the remaining demand on the blackboard.

The remaining firm agents compete for the remaining demand

The winner selects its subcontractors according to its collaboration strategy

The subcontractors make the product subjected to its capacity, and subcontract its remaining order to the remaining firms.

Does simulation time exceed the total demand cycles?

The remaining firm agents compete for the remaining demand, and the subcontractor selects its subcontractor.

Firm agents deliver the finished products to their customers.

Firm agents evaluate their profit and update its fitness, capacity and cost.

The environment agent evaluates firm agents, and eliminates those agents whose fitness is less than Fc.

No

End

Yes

Output the simulation report.

Fig. A1. The interaction of agents and the simulation process flow.

Table A1parameters setting of the experiment 1.

Parameters Values Remarks

Environment fitness landscape (Fc) 0.35 Fc is generated heuristi

The global market demand (D) D�U[Dmin, Dmax] Dmin=5000, Dmax=8000

Number of firm agents (N) N=8

Firm’s fitness (Fitness) Fitness�U

[0.40,0.75]

It indicates that the en

Cmin=100, Cmax=1800

Firm’s capacity (CP) CP�U[Dmin, Dmax] Cmax is calculated such

Firm’s cost (Ci) Ci�U[50,150] Ci is the cost of the wh

Firm’s bidding price (pi) pi ¼ ð1þbiÞCi bi �U½0;1�

g(cfi) 0.75% In China, the average fi

Ddi(t) 1.1% In China, the mean life

Products One parts; two

parts

In experiment 1(a), firm

parts

Simulation time (the total demand

cycles)

240 One demand cycle repr

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 327

understand the evolution of the structure and collaborationmechanisms of CASNs and be able to build theories of CASNs.We have taken a small step toward this end.

In this paper, we have proposed evolutionary model of CASNsbased on CAS and fitness landscape theory. The simulation of theevolutionary model indicates many characteristics of CAS.Furthermore, the evolution of the LVEA supply networks in theemerging market of China was explored to validate these findingsand get more knowledge of the evolution in the real world. Thecase study analyzed the evolution of the LVEA supply networkover the past 30 years and showed how the network structure andcollaboration mechanisms operate and behave in the fact ofinfluence from the external environment and firms’ internalmechanisms. Several propositions about the dominant factors andprinciples of the evolution have been constructed based on thesimulation and the case study to guide the future theorydevelopment and empirical studies.

Acknowledgments

This paper was partially supported by The Natural ScienceFoundation of China (Project no. 70701029), the National SocialScience Foundation of China (no. 08XJY016), and The Hong KongPolytechnic University (Project no. A-PG64). The authors thankDr. Pathak, professor David and the anonymous referees for theirgood advices through the work.

Appendix A

Fig. A1.

Appendix B

See Table A1.

Appendix C. Interview instrument

A.1. General information

Visit the general manager, the founders of ZT and its majorpartner firms to get general information about ZT.

cally

vironment is of high volatility

that the mean cumulative firm capacity is 95% of Dmax

ole product. If firms make one of the individual parts, the cost is 50% of Ci

xed cost of manufacturing firms are between 0.5% and 1%

of a firm is 7.3 years

s make product with one part. In experiment 1(b), firms make product with two

esents one month in the real world

Page 19: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330328

(1)

Get the firm’s background which includes history, annualsales, number of employees, organization structure, mainproducts, main customers and suppliers, the related laws, andregulations that have influence on the firm.

(2)

Ask the founder of ZT to describe the milestones of thedevelopment history. Try to divide the development intosome stages.

A.2. Environment information

Visit the general managers of ZT and its major partner firms,and ask them to list government departments have influence onthe visited firms. And then visit the officers of these departmentsto get information about their influence on the firms.

Visit the president of the LVEA industry association ofWenzhou to get information of the development of the LVEAindustry and the evolution of the environment for LVEA industry.

Visit the managers of ZT and its major partner firms to getinformation about the relationship between ZT and the regulators.

(1)

Ask them to describe the information about the developmenthistory of the LVEA industry in Wenzhou, which includes theannual market demand, total number of firms, total annualsales, market competition, and technology development inthe past 30 years.

(2)

Collect information about the government’s regulation on theLVEA industry such as laws and regulations on the foundationof firms, investment, market regulations, tax and privilegesubsides, quality certification, production license, etc.

(3)

Collect information about the evolution of the relationshipbetween the visited firm and the regulators. How do theregulators influence on the firm?

A.3. Firm’s internal mechanism

(1) Product information: Visit the manager of the R&DDepartment and the Manufacturing Department of ZT and itsmajor partner firms (e.g., suppliers, distributors, and retailers) toget product information.

(1)

Ask the managers to depict the main products in each stage oftheir development, and make a list of these products.

(2)

Select one or two products in each stage from the list and askthe managers to show the BOMs of these products. Collectinformation from BOMs about the total number of parts,product structure, width, and depth of each product.

(2) Technology information: Visit managers from the R&DDepartment and operation related departments of ZT and itsmajor partner firms (e.g., suppliers, distributors, and retailers) toget technology about the products, process and ICTs.

(1)

Collect information about the technology of the mainproducts. The information includes characteristics of producttechnologies, patents, know how, product development time,cost, and total number of engineers involved in the develop-ment, etc.

(2)

Visit R&D managers and manufacturing managers to collectinformation about the machines, production planning, processplanning, and manufacturing technology of main products ineach stage of the development.

(3)

Ask appropriate managers (e.g., manager of informationmanagement department, manufacturing managers, etc.)

about the application of ICTs in business management,operation and manufacturing process? Collect the mainapplications (e.g. MIS, DSS, ERP) of ICTs.

(3) Strategy information: Visit the general manager, the foundersand operation managers of ZT and its major partner firms (e.g.,suppliers, distributors, and retailers) to collect information aboutthe firm’s strategy.

(1)

Get information about the firm’s business objectives in each ofits development stage (e.g., profit, quality, cost, flexibility, ordifferentiation). If not these objectives, ask them to describetheir objectives.

(2)

Get information about the firm’s operation strategy in each ofits development stage. What is the main manufacturingparadigm in each stage? For example, all the products weremade in-house, make-to-order, buy-to-order, assemble-to-order or design-to-order? Introduce the definition of theseoperation paradigms at first.

(3)

Get information about the firm’s competition strategy in eachof the development stage. What is the main competence of itsproducts? For example, price, quality, service, differentiatedproduct or lead-time, etc.?

(4)

Do your firm’s products have great difference to competitors’products in each stage? If yes, what are the differences?

(5)

Get information about the firm’s collaboration mechanismwith the suppliers, distributors and customers in each of itsdevelopment stage. The working relationship? How about thefirm’s influence on partner firms? How does the firm controlthe collaboration with partner firms in order to achieve itsdesired goals? Do the partner firms always fulfill the firm’srequirement?

(4) Organization information: Visit the sales manager, themarketing manager, the procurement managers, the suppliermanagement manager, the R&D manager, and the chief informa-tion officer of ZT and its major partner firms to get informationabout the collaboration activities among the visited firm and itspartner firms (e.g., suppliers, distributors, and retailers) in eachdevelopment stage.

(1)

Please depict the supply network in each development stageand draw a picture of the supply network in each develop-ment stage. How many suppliers are there in the first-tier ofthe supply base network in each stage? How manydistributors, retailers, and sales agencies in each stage ofthe distribution base network in each stage?

(2)

Please depict the structure and collaboration mechanism ofthe supply base network in each stage. Select main suppliersand collection the following information of them. Confirmthe part(s) that the firm supplies to this customer for theLVEA products. Who are the suppliers for the part(s)? Annualsales? Percentage sales to the present firm? Please show thestatistic of the annual procurement value from the mainsuppliers.

(3)

Please depict the structure and collaboration mechanism ofthe distribution base network in each stage. Confirm thebuying company (customer). Locations? Annual revenue?What are the percentage orders of these firms? Please showthe data of the annual sales to the main distributors andretailers. Are there any distribution partners had beenremoved from the partner list? If yes, how many and whatare the main reasons?

(4)

Please describe the form of the collaboration with mainpartner firms (e.g. suppliers, distributors and retailers) in
Page 20: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330 329

each development stage, and describe the main content ofthe agreements. How does the firm select its suppliers anddistribution partners? Are there any partner firms had beenremoved from the supplier list? If yes, what are the mainreasons? What is the process of awarding a contract to yoursuppliers? How do additional contracts/works awarded tosuppliers? What are the main forms of the collaborationagreement? How much percentage of the formal andinformal agreements in the collaboration? How explicit arethe rules/norms of operation with your suppliers anddistributors?

(5)

How can your firm manage the collaboration with suppliers,distributors and retailers? Does your firm have investmentin these firms? Work process/procedure with these firms?How influenced and influence in decision-making? Do youhave any plans to support the collaboration with suppliers,distributors and retailers? How/to what extent does yourfirm participate in the operation of suppliers and distribu-tors? Collect the methods, documents, rules and businessprocess for the management of suppliers and distributors.

(6)

What type of performance data do you maintain on yoursuppliers and distributors? How measured? Is there anyestimation on how well they perform (e.g. poor to excellenton cost, quality, and delivery dimensions) based on yourperception?

(7)

How about your firm’s collaboration with the partner firms?What kind of relationship do you have with the partnerfirms? Which suppliers are managed directly? How/in whatcapacity? What kind of relationships do you have with thesesuppliers on a partnership scale? Do you think your firm hasa close relationship with partner firms? Please show themain methods to control the collaboration. Give someexamples of a ‘‘close’’ relationship and a ‘‘not close’’relationship and additional examples of relationships thatmight fall between the two extremes.

(8)

Discuss the mode of communication with your partner firms(on a regular basis and, if so, how often, and on ad hoc basisand, if so, when). Does your firm share information withpartner firms? If yes, what kind of information has beenshared? By what way do they share the information?

(9)

Describe the collaboration history with the main suppliers,distributors and retailers. Working relationship with partnerfirms? How many years have your firm collaborate withthese partner firms? What is the length of contracts?

(10)

Supplier locations? Major warehouses? How many procure-ment could be fulfilled in the local market? Identify thephysical distance to each first-tier supplier and the mainsecond-tier suppliers.

(11)

Identify some second-tier suppliers for the main first-tiersupplier. What parts do they supply? Discuss their workingrelationship with the first-tier suppliers and ZT.

References

Adamides, E.D., Pomonis, N., 2007. The co-evolution of product, production andsupply chain decisions, and the emergence of manufacturing strategy.International Journal of Production Economics, doi:10.1016/j.ijpe.2006.11.025.

Akanle, O.M., Zhang, D.Z., 2008. Agent-based model for optimising supply-chainconfigurations. International Journal of Production Economics, doi:10.1016/j.ijpe.2008.02.019.

Akkermans, H., 2001. Emergent supply networks: system dynamics simulation ofadaptive supply agents. In: Proceedings of the 34th Hawaii InternationalConference on System Sciences, Hawaii, pp. 1–5.

Alfaro, M.D., Sepulveda, J.M., 2006. Chaotic behavior in manufacturing systems.International Journal of Production Economics 101 (1), 150–158.

Anand, G., Ward, P.T., 2004. Fit flexibility and performance in manufacturing:coping with dynamic environments. Production and Operations Management13 (4), 369–385.

Beamon, B.M., 1998. Supply chain design and analysis: models and methods.International Journal of Production Economics 55 (2), 281–294.

Berry, D., Naim, M.M., Towill, D.R., 1995. Business process re-engineering anelectronic products supply-chain. IEE Process Science Measurement Technol-ogy 142 (5), 395–403.

Bruce, K., 2000. The network as knowledge: generative rules and the emergence ofstructure. Strategic Management Journal 21 (3), 405–425.

Carbonara, N., Giannoccaro, I., Pontrandolfo, P., 2002. Supply chains withinindustrial districts: a theoretical framework. International Journal of Produc-tion Economics 76 (2), 159–176.

Choi, T., Dooley, K., Rungtusanatham, M., 2001. Supply networks and complexadaptive systems: control versus emergence. Journal of Operations Manage-ment 19 (3), 351–366.

Choi, T.Y., Hong, Y., 2002. Unveiling the structure of supply networks: case studiesin Honda, Acura, and DaimlerChrysler. Journal of Operations Management 20(5), 469–493.

Christopher, M., 1992. Logistics and Supply Chain Management: Strategies forReducing Costs and Improving Service. Pitman, London.

Daft, R.L., 1989. Organization Theory and Design. West Publishing, New York.De Toni, A., Nassimbeni, G., 1999. Buyer–supplier operational practices, sourcing

policies and plant performance: result of an empirical research. InternationalJournal of Production Research 37 (3), 597–619.

Dyer, J.H., Singh, H., 1998. The relational view: cooperative strategy and sources ofinter organizational competitive advantage. Academy of Management Review23 (4), 660–679.

Eisenhardt, K.M., 1989. Building theories from case study research. Academy ofManagement Review 14 (4), 532–550.

Ellram, L.M., 1991. Supply chain management: the industry organizationperspective. International Journal of Physical Distribution and LogisticsManagement 21 (1), 13–22.

Eymann, T., Padovan, B., Schoder, D., 1998. Avalanche—an agent based value chaincoordination experiment. Working Notes in Autonomous Agents’98 Workshopon ‘‘Artificial Societies and Computational Markets’’, Minneapolis, MN,pp. 48–53.

Fiala, P., 2005. Information sharing in supply chains. Omega 33 (5), 419–423.Fisher, M., 1997. What is the right supply chain for your product? Harvard

Business Review 75 (2), 105–116.Forrester, J.W., 1961. Industrial Dynamics. MIT Press, Cambridge, MA.Gibson, J.L., Ivancevich, J.M., Donnelly Jr., J.H., 1997. Organizations: Behavior,

Structure, Processes. Irwin, Chicago.Glaser, B., Strauss, A., 1967. The Discovery of Grounded Theory: Strategies for

Qualitative Research. Aldline, Chicago.Guisinger, S., 2001. From OLI to OLMA: incorporating higher levels of environ-

mental and structural complexity into the eclectic paradigm. InternationalJournal of the Economics of Business 8 (2), 257–272.

Germain, R., Claycomb, C., Droge, C., 2008. Supply chain variability, organizationalstructure, and performance: the moderating effect of demand unpredictability.Journal of Operations Management 26 (5), 557–570.

Gunasekaran, A., Ngai, E.W.T., 2005. Build-to-order supply chain management: aliterature review and framework for development. Journal of OperationsManagement 23 (5), 423–451.

Hamel, G., Prahalad, C.K., 1994. Competing for the Future. Harvard Business SchoolPress, Boston, MA.

Harland, C.M., Zheng, J., Lamming, R.C., Johnsen, T.E., 2002. A taxonomy of supplynetworks. IEEE Engineering Management Review 30 (4), 12–20.

Harris, S.G., Sutton, R.I., 1986. Functions of parting ceremonies in dyingorganizations. Academy of Management Journal 29 (1), 5–30.

Holweg, M., Bicheno, J., 2002. Supply chain simulation-a tool for education,enhancement and endeavour. International Journal of Production Economics78 (2), 163–175.

Hsu, L.L., 2005. SCM system effects on performance for interaction betweensuppliers and buyers. Industrial Management and Data Systems 105 (7),857–875.

Hur, D., Hartley, J.L., Hahn, C., Hahn, C.K., 2004. An exploration of supply chainstructure in Korean companies. International Journal of Logistics: Research andApplications 7 (2), 151–166.

Kauffman, S.A., 1993. The Origins of Order: Self Organization and Selection inEvolution. Oxford University Press, New York.

Kinra, A., Kotzab, H., 2008. A macro-institutional perspective on supply chainenvironmental complexity. International Journal of Production Economics,doi:10.1016/j.ijpe.2008.05.010.

Kotabe, M., Martin, X., Domoto, H., 2003. Gaining from vertical partnerships:knowledge transfer, relationship duration, and supplier performance improve-ment in the US and Japanese automotive industries. Strategic ManagementJournal 24 (4), 293–316.

Koza, M.P., Lewin, A.Y., 1999. The coevolution of network alliances: a longitudiualanalysis of an international professional service network. Organization Science10 (5), 563–638.

Krogh, G.V., Nonaka, I., Aben, M., 2001. Making the most of yourcompany’s knowledge: a strategic framework. Long Range Planning 34 (4),421–439.

Landry, J.T., 1998. Supply chain management: the case for alliances. HarvardBusiness Review 76 (6), 24–25.

Larsen, E.R., Morecroft, J.D.W., Thomsen, J.S., 1999. Complex behaviour in aproduction–distribution model. European Journal of Operational Research 119(1), 61–74.

Page 21: The evolutionary complexity of complex adaptive supply networks: A simulation and case study

ARTICLE IN PRESS

G. Li et al. / Int. J. Production Economics 124 (2010) 310–330330

Lee, H.L., 2002. Aligning supply chain strategies with product uncertainties.California Management Review 44 (3), 105–119.

Lewin, A., Long, C., Carroll, T., 1999. The coevolution of new organizational forms.Organization Science 10 (5), 535–550.

Li, G., Ji, P., Sun, L.Y., Lee, W.B., 2009. Modeling and simulation of supply networkevolution based on complex adaptive system and fitness landscape. Compu-ters & Industrial Engineering 56 (3), 839–853.

Li, S., Ragu-Nathan, B., Ragu-Nathan, T.S., Rao, S.S., 2006. The impact of supplychain management practices on competitive advantage and organizationalperformance. Omega 34 (2), 107–124.

Luhmann, N., 1995. Social Systems. Stanford, Stanford University Press, CA.Madhavan, R., Grover, R., 1998. From embedded knowledge to embodied

knowledge: new product development as knowledge management. Journalof Marketing 62 (4), 1–12.

Marquez, A.C., Blanchar, C., 2004. The procurement of strategic parts. Analysis of aportfolio of contracts with suppliers using a system dynamics simulationmodel. International Journal of Production Economics 88 (1), 29–49.

McCarthy, I.P., 2004. Manufacturing strategy: understanding the fitness landscape.International Journal of Operations & Production Management 24 (2),124–150.

Melody, A.R., 1994. Enterprise architecture: definition, content, and utility. In:Proceeding of the Third Workshop on Enabling Technologies: Infrastructure forCollaborative Enterprises. IEEE Computer Society Press, Washington, pp. 106–111.

Miles, M.B., Huberman, A.M., 1984. Qualitative Data Analysis: A Source Book ofNew Methods. Sage, Newbury Park, CA.

Miller, D., 1992. Environmental fit versus internal fit. Organization Science 3 (2),159–178.

Min, H., Zhou, G.G., 2002. Supply chain modeling: past, present and future.Computers & Industrial Engineering 43 (1–2), 231–249.

Nadler, D.A., Tushman, M.L., 1980. A model for diagnosing organizational behavior:applying the congruence perspective. Organizational Dynamics 9 (2), 35–51.

Nagatani, T., Helbing, D., 2004. Stability analysis and stabilization strategies forlinear supply chains. Physica A 335 (3–4), 644–660.

Oke, A., Gopalakrishnan, M., 2008. Managing disruptions in supply chains: a casestudy of a retail supply chain, International Journal of Production Economics,doi:10.1016/j.ijpe.2008.08.045.

Pathak, S.D., David, M.D., 2002. Simulating supply networks using complexadaptive systems theory. In: Proceeding of IEEE International EngineeringManagement Conference (IEEE catalog No. 02CH37329), Cambridge, UK.

Pathak, S.D., David, M.D., Gautam, B., 2003. Multi-paradigm simulator forsimulating complex adaptive supply chain networks. In: Proceedings of the2003 Winter Simulation Conference, Washington, pp. 808–816.

Pathak, S.D., Day, J., Nair, A., Sawaya, W.J., Kristal, M., 2007a. Complexity andadaptivity in supply networks: building supply network theory using a complexadaptive systems perspective. Decision Sciences Journal 38 (4), 547–580.

Pathak, S.D., Dilts, D.M., Biswas, G., 2007b. On the evolutionary dynamics of supplynetwork topologies. IEEE Transactions on Engineering Management 54 (4),662–671.

Peleg, B., Lee, H.L., Hausman, W.H., 2000. Short-term e-procurement strategiesversus long-term contracts. Journal of Operations Management 11 (4),458–479.

Poole, M.S., Van de Ven, A.H., 1989. Using paradox to build management andorganization theory. Academy of Management Review 14 (4), 562–578.

Potter, A., Mason, R., Naim, M., Lalwani, C., 2004. The evolution towards anintegrated steel supply chain: a case study from the UK. International Journalof Production Economics 89 (2), 207–216.

Ramdas, K., Spekman, R., 2000. Chain or shackles: understanding what drivessupply-chain performance. Interfaces 30 (4), 3–21.

Schieritz, N., GroBler, A., 2003. Emergent structures in supply chains—a studyintegrating agent-based and system dynamics modeling. In: Proceedings of the36th Annual Hawaii International Conference on System Sciences, Hawaii,pp. 1–9.

Song, M., Noh, J., 2006. Best new product development and management practicesin the Korean high-tech industry. Industrial Marketing Management 35 (3),262–278.

Sutton, R.I., Callahan, A.L., 1987. The stigma of bankruptcy: spoiled organizationalimage and its management. Academy of Management Journal 30 (3),405–436.

Stock, G.N., Greis, N.P., Kasarda, J.D., 2000. Enterprise logistics and supply chainstructure: the role of fit. Journal of Operations Management 18 (5), 531–547.

Surana, A., Kumara, S., Greaves, M., Raghavan, U.N., 2005. Supply chain networks: acomplex adaptive systems perspective. International Journal of ProductionResearch 43 (20), 4235–4265.

Swaminathan, J.M., Smith, S.F., Sadeh, N.M., 1998. Modeling supply chaindynamics: a multi-agent approach. Decision Sciences 29 (3), 607–632.

Swamidass, P., Newell, W., 1987. Manufacturing strategy, environmental un-certainty and performance: a path analytical model. Management Science 33(4), 509–524.

Towill, D.R., 1996. Industrial dynamics modeling of supply chains. LogisticsInformation Management 9 (4), 43–56.

Towill, D.R., Evans, G.N., Cheema, P., 1997. Analysis and design of an adaptiveminimum reasonable inventory control system. Production Planning & Control8 (6), 545–557.

Van Maanen, J., 1983. Epilogue: qualitative methods reclaimed. In: Van Maanen, J.(Ed.), Qualitative Methodology. Sage, Beverly Hills.

Volberda, H.W., 1998. Building the Flexible Form: How to Remain Competitive.Oxford University Press, Oxford.

Ward, P.T., Duray, R., 2000. Manufacturing strategy in context: environment,competitive strategy and manufacturing strategy. Journal of OperationsManagement 18 (2), 123–138.

Whang, S., 1995. Coordination in operations: a taxonomy. Journal of OperationsManagement 12 (3), 413–422.

Wikner, J., Towill, D.R., Naim, M.M., 1991. Smoothing supply chain dynamics.International Journal of Production Economics 22 (3), 231–248.

Wilkinson, I., Young, L., 2002. On cooperating firms, relations and network. Journalof Business Research 55 (2), 123–132.

Williams, F.P., D’Souza, D.E., Rosenfeldt, M.E., Kassaee, M., 1995. Manufacturingstrategy, business strategy and firm performance in a mature industry. Journalof Operations Management 13 (1), 19–33.

Wu, Y., Zhang, D.Z., 2007. Demand fluctuation and chaotic behaviour byinteraction between customers and suppliers. International Journal ofProduction Economics 107 (1), 250–259.

Yang, J., Wang, J.J., Wong, C.W.Y., Lai, K.H., 2008. Relational stability and allianceperformance in supply chain. Omega 36 (4), 505–508.

Yin, R.K., 1994. Case Study Research: Design and Methods, second ed Sage,Thousand Oaks, CA.