Content uploaded by Sokol Celo
Author content
All content in this area was uploaded by Sokol Celo on Feb 26, 2018
Content may be subject to copyright.
1
The Role of Internal and External Complexity in Global Factory
Performance: An NKC Application
Sokol Celo⁎, James Nebus, I. Kim Wang
Sawyer Business School, Su
ff
olk University, 120 T remont Street, Boston, MA 02108, USA
A B S T R A C T
The global factory literature suggests that MNCs can take advantage of global operations by
extensively offshoring and outsourcing activities. However, the added difficulty for the lead firm
to coordinate the resulting complex structure is often underestimated. Evidence could be found in
Boeing's 787 Dreamliner project, in which the external complexity disrupted MNC performance.
Motivated by the gap between theory and practice, this study focuses on systems of MNCs
connected with each other with supplier-client relationships and/or outsourcing. In particular it
investigates the interplay of the internal and external complexity in such systems and how their
balance affects the system performance. The study models the internal and external complexity
by using the NKC-simulation methodology and adjusting it to the specific MNC context. The NKC
methodology is widely used in organization theory to study complex systems. Simulations
comparing the performance of MNCs that use outsourcing to different degrees indicate that a
balanced level of internal and external complexity is beneficial in the context of global factory.
1.
Introduction
The concept of the “global factory” as a form of a firm's deployment of its international activities has been popularized in the IB
literature (Buckley, 2009, 2010, 2011; Buckley and Ghauri, 2004). The global factory is based on MNCs leveraging information and
communication technology to take advantage of location specific advantages in emerging economies and pockets of expertise around
the world. In short, the global factory firm is characterized by geographically dispersing its value chain activities, which are mostly
outsourced to supplier firms. The lead firm seeks to optimize its performance through efficiencies available from contractors in
different locations. In theory, it is hard to argue against the global factory concept.
However, in practice, there have been several examples of industry leading firms who have failed in implementing the global
factory concept. One notable example is Boeing's development of its 787 Dreamliner aircraft, the largest outsourcing event in its
history. Boeing contracted with
>
50 suppliers, 28 of them outside the US. After receiving subassemblies from multiple suppliers,
Boeing estimated that the final assembly task followed by a systems integration test would enable Boeing to turn out an aircraft in 2
or 3 days. However, it didn't turn out exactly as Boeing had planned. Delivery of the first plane was four years behind schedule due to
Boeing losing control of the process of farming out more design and production work than ever and not keeping close tabs on
suppliers.
The point of contrasting IB theory with practitioners' implementation experiences is to highlight an acknowledged, but often
overlooked, challenge of the global factory – the complexity of managerial coordination and control. Complexity has been defined as
1) the number and location of separated activities in a firm's value chain and 2) the interdependencies, connections, or interactions
among these activities. Our focus is on the second point. As such, our research complements the literature on outsourcing, make
versus buy decisions, and offshoring which address the first point. The decisions as to which value chain activities are outsourced or
where to offshore them are based on the economics and country specific advantages for these activities. These decisions are made to
maximize the MNC's performance for each value chain activity and together they determine the value chain's geographical footprint.
However, often these decisions are made without regard to the complexity and cost of interactions to manage the interdependencies
among all these separated value chain activities.
The theoretical relationship between a single firm's internal complexity and firm performance is well accepted in the literature as
an inverted U shaped curve (Ghemawat and Levinthal, 2008; McKelvey, 1999). That is, once complexity is increased beyond a certain
threshold, firm performance decreases as complexity increases. The level of complexity at which this performance inflection point
occurs may vary depending on type of organizational structure type, but the inverted U relationship holds regardless of the structure
type (Celo et al., 2015).
However, what is not clear is how complexity affects performance in the global factory organization, because the global factory is
not a single firm, but a “family” of firms, or the lead firm and its supplier firms. Theoretical arguments can be made for both sides -
that the global factory organization either increases complexity which decreases performance, or that it decreases complexity and
increases performance. Slicing the value chain into separately managed subunits, including suppliers and service providers, has the
beneficial effect of reducing the internal complexity of each subunit, which can result in increased performance due to easier co-
ordination within each subunit's internal activities. However, slicing the value chain into separately managed subunits also increases
the external complexity (Li et al., 2017) of the value chain by increasing the coordination necessary with all subunits that comprise
the value chain. The greater the value chain's external complexity the more separately made, but interdependent, decisions in each
2
subunit must be aligned towards a common goal to optimize performance. This alignment towards MNC optimum performance is
hindered by separately managed subunits, because each subunit has a goal to increase its own performance while at the same time
having limited insight as to how its decisions affect the overall global factory value chain performance.
As Buckley (2011: 272) points out: “The manager of the global factory has to ask two very straightforward questions of each
activity in the global network. Where should this activity be located? How should this activity be controlled? The first question of the
optimum location for each activity is of course complicated by managing the interrelationships between activities. The relocation of one piece of
the global network will have profound effects on many others” (italics are ours). It is the italicized part that is of central interest for us and
accordingly our research question is: How does the balance between internal and external complexity affect performance in a global
factory type of organization? That is, given a value chain distributed among a family of firms, as in the global factory organization,
how does the number of interactions among relative to those within operationally interdependent firms affect overall performance?
We are not concerned with the total amount of outsourcing, or the proportion of all value chain activities that are outsourced. Rather,
our focus is on the degree to which slicing the value chain results in a greater amount of external interdependent decisions among
separate subunits, and how the division between these interactions being internal versus external affects the performance of the
overall “family” of firm which comprise the global factory.
We address these research questions through a simulation model. Simulation is an especially powerful tool to model theory in
situations where field studies are impractical because large matched samples are unobtainable or the number of control variables
required to account for alternative explanations are too numerous (Davis et al., 2007; Lazer and Friedman, 2007; Venaik et al., 2004).
In our research questions, finding MNCs and global factory configurations whose value chains are identical but whose separation of
activities and geographic dispersion differs is impossible, and second, degree of internal and external complexity is but one of dozens
of variables that can affect MNC performance.
Our contributions are theoretical, practical, and valuable. On the theoretical side we quantify and explain the relationship be-
tween complexity and performance in the global factory structure. In doing so, we provide some insights into the gaps between global
factory theory prescriptions and practitioner implementation. On the practical side, we can help practitioners understand the limits of
outsourcing, but also show ways how to manage complexity by balancing the internal and external complexity. Finally, ours is a
valuable contribution as no previous literature has quantitatively addressed these research questions.
2.
Literature review and practitioner experiences
2.1.
Literature review - global factory, outsourcing, and offshoring
The concept of global factory as a new institutional form (Buckley, 2011) is used to describe the changing location and ownership
strategies of MNEs resulting in fine-slicing of the MNC's value chain and allocating these slices to specialized outside supplier firms in
optimum locations (Buckley and Ghauri, 2004). The advantages of the global factory include various types of efficiencies such as
lower production costs resulting from lower labor rates in emerging economies, better product design resulting from technology, and
specialized knowledge in clusters of expertise found only in certain geographic areas. Such increasingly complex strategies are
necessary in the face of technological changes, such as the rise of e-commerce and political changes that increase access to previously
closed economies (Buckley, 2011).
Because the supplier firms focus on specializing in only a small part of the value chain, they are likely to be more efficient and
perform better than the lead firm in this particular activity. Whereas this small part of the value chain is a strategic focus for the
supplier firm, it is a non-strategic activity of the lead firm. As a consequence, implementing best practices is the sole focus for the
supplier firm; however it is not the primary focus for the lead firm. Therefore, in theory the global factory form of organizing an
MNC's value chain is designed to maximize value by 1) combining the advantages of the lead firm with the advantages of supplier
firms to achieve better overall value chain efficiencies, 2) enabling the MNC to dynamically redeploy assets and reorganize activities
by changing supplier firms as the industry and geographic environment changes, and 3) achieving a better balance between global
integration and local responsiveness than is possible in a single hierarchical vertically integrated firm.
The logic behind the ‘fine-slicing’ of activities, which leads to global factory, is that every element of the firm's value chain can be
evaluated with regard to alternative ownership forms and locations. The former entails comparing the current activity with the
market alternative and – if profitable – externalize it, a strategy known as outsourcing. The latter deals with determining whether it's
profitable for the activities to be relocated overseas, known as offshoring.
Although both outsourcing and offshoring work in concert to determine the shape of the global factory, there are important
differences between them. Outsourcing is about firm restructuring with regard to ownership and hence concerned with the estab-
lishment of new boundaries of the firm and the decision of choosing from a variety of governance modes ranging from wholly owned
units via FDI through market relationships such as subcontracting, including joint ventures, to strategic alliances (Buckley and
Ghauri, 2004; Contractor et al., 2010; Schmeisser, 2013).
Offshoring, on the other side, is about (re)locating of activities across national borders (Schmeisser, 2013) or restructuring of the
firm along the geographical dimension (Contractor et al., 2010). To a certain extent, outsourcing and offshoring represent two
orthogonal decisions since outsourcing can take place in the home country and/or abroad, and offshored activities can be performed
internally by a MNE's foreign subsidiary or by an external foreign contractor (Contractor et al., 2010).
One of the most significant advantages of outsourcing is the access to knowledgeable resources (Quinn, 1999). As a matter of fact,
the majority (60% according to UNCTAD, 2013) of the trade in goods and services today “consists of trade in intermediate goods and
services that are incorporated at various stages in the production process of goods and services for final consumption” (UNCTAD,
2013: 19). It is common in the technology industries for firms to seek technical help from their suppliers to meet their technology
needs (Kapoor and Adner, 2012). Additionally, outsourcing allows a firm to form strategic networks to extend its access to partner
3
firms' knowledge base. Suppliers' commitments to the firm locks them in so that they are less likely to compete against the lead firm
(Quinn, 1999), or less likely to join the rival's camp. For example, Boeing engages Japanese firms Kawasaki, Mitsubishi, and Fuji in
building 787 Dreamliner in order to secure its share in the Japanese market. As a consequence, the Japanese Airlines only uses Boeing
airplanes.
Most of the outsourcing and offshoring research emphasizes their positive attributes. For instance, Bertrand (2011) finds that
offshore outsourcing increases export performance, primarily in the export markets where firms import intermediate goods. Firms
engaging in offshoring activities are found to have a higher rate of survival relative to purely domestic firms (Coucke and
Sleuwaegen, 2008) and improve their innovation performance (Nieto and Rodríguez, 2011). Di Gregorio et al. (2008) focus on SMEs
finding that offshore outsourcing enhances their international competitiveness.
Besides strategic benefits, outsourcing is advantageous for well-structured activities. A firm's gain from outsourcing efficiency
depends on its ability to adapt to changes in the environment (Weigelt and Sarkar, 2012). A firm's needed adaptation then correlates
with how a firm's activities are structured. The structure represents “the level of understanding of the K interactions among N
knowledge sets for a given problem” (Macher and Boerner, 2012: 1018). For well-structured activities in which the interaction
between knowledge sets is well understood, costs associated with coordinating and configuring activities will stay low (Walker and
Weber, 1984).
Other researchers have found that international outsourcing is not associated with increased firm performance (e.g., Mol et al.,
2004). In fact, almost three decades ago Kotabe (1989) had warned of the “hollowing out” of the offshore outsourcing client firm as a
result of the deterioration of its resources and competencies. With the firms using outsourcing not only as a source of cost savings but
also of competence acquisition, it becomes increasingly important to focus not just on short-term cost savings but also on long-run
consequences of outsourcing, which are much less understood (Kotabe et al., 2008).
Overall there is a certain ambiguity regarding the benefits of outsourcing and offshoring and the resulting global factory structure
for firm performance. As it is customary in such cases, the researchers have identified contingencies, such as the availability of
relevant host country information and cultural proximity to the firm's home country in the case of offshoring (Schmeisser, 2013), the
choice of strategic vs non-strategic activities to be outsourced (Mudambi and Tallman, 2010), and the degree to which competencies
between client and supplier are complementary vs overlapping (Kotabe et al., 2008). Also, outsourcing is preferred when an en-
vironment exhibits the extremes of being either stable or turbulent (Claussen et al., 2014). The costs associated with coordination and
cooperation are the key to outsourcing decision (Gulati et al., 2005). In a stable setting, outsourcing is preferred because a firm has
fewer changes to respond. The interactions between the firm and suppliers are low, so are the costs; in a turbulent setting where
changes are radical and frequent, outsourcing is still preferred because a firm can leverage outside resources to help it make a speedy
response to external changes (Claussen et al., 2014).
Our review above of the global factory, offshoring, and outsourcing literature found that the extant research rarely discusses the
role of complexity of management coordination and control and its effect on performance. When complexity of management co-
ordination and control is mentioned, it is descriptive, not quantitative. For example, Contractor et al. (2010: 1425) propose that
performance is maximized for some intermediate level of value chain disaggregation and dispersion since “the benefits of fine-slicing
and organizational and geographical dispersion come at a cost in terms of increased complexity and coordination.” Li et al. (2017)
employ complexity as a antecedent in an empirical study, but the study's dependent variable is equity vs. contract based governance,
not performance. Furthermore, when complexity is mentioned in the literature, no distinction is made between internal and external
complexity. It is partly due to this underdeveloped stage of theory building that, to our knowledge, no quantitative study has
addressed the relationship between complexity and global factory performance. Below we seek to make a contribution to fill this gap
by discussing theoretically the role of complexity for global factory performance – distinguishing between internal and external
4
Fig. 1. Boeing's global supply chain.
Source: The Boeing Company.
complexity – and conducting quantitative tests of the theory. But first we demonstrate how the difficulty of the lead firm's role as
coordinator and orchestrator in the complex global factory structure is often underestimated, as we will see by Boeing's im-
plementation experience described in the next section.
2.2.
Boeing experiences
Boeing's 787 Dreamliner serves an example of a global factory. Boeing was losing market share to Airbus in the late 1990s. Boeing
needed a new type of aircraft to compete with Airbus. In order to shorten the time-to-market as well as to tap into the technological
advantages of the suppliers, Boeing outsourced labor-intensive assembly, such as composite component manufacture, and formed a
network of suppliers with strong technological competence worldwide. Richard Aboulafia, an aviation analyst at the Teal Group in
Fairfax, Virginia, summarized Boeing's approach: “The idea was to get the risk off their books and get other people to do the heavy
lifting for them.”
In 2003 Boeing began contracting for specific parts of the 787. Boeing contracted with
>
50 suppliers, 28 of them outside the U.S.
As shown in Fig. 1, the Japanese firms Mitsubishi, Kawasaki, and Fuji provided the wing box, forward portions of the fuselage,
landing gear, wing fixed trailing edge, center wing box, and wheel well for the main landing gear (Slayton and Spinardi, 2016). The
Italian firm Alenia teamed with the U.S. Vought to build the horizontal stabilizer and portions of the fuselage (Mecham, 2003).
Another team of suppliers (e.g., Spirit AeroSystems, Global Aeronautica and Mitsubishi Heavy Industries) would put together the
parts in to subsystems before final assembly at Boeing. Spirit Aerosystems would deliver the forward portion of the fuselage to Everett
with the cockpit fully “stuffed” with electronics and controls (Tang et al., 2009). Boeing would then rapidly “snap together” each
aircraft from just seven parts—two wings, three fuselage sections, the horizontal stabilizer, and the vertical fin—and complete the
systems integration in just two to three days (Mecham, 2003).
Boeing's plan run aground as its supply chain management was incapable of working with suppliers (Madslien, 2010; Peterson,
2011; Slayton and Spinardi, 2016; Tang et al., 2009). Neither Boeing nor the suppliers had experience in handling a product of such a
massive scale (Slayton and Spinardi, 2016). For example, problems caused by supplier Alenia resulted in a temporary stop to flight
tests in 2012 (Norris, 2010; Slayton and Spinardi, 2016). Even after their technological capabilities sufficed, their organizational
5
capabilities did not. The complexity entailed by outsourcing was beyond what Boeing could handle. Unexpected need for co-
ordination resulting from outsourcing slowed the program (Slayton and Spinardi, 2016; Tang et al., 2009).
Boeing tried to regain the control of the schedule. CEO Mr. McNerney visited the largest suppliers - like Vought Aircraft Industries
in North Charleston, South Carolina; Mitsubishi Heavy Industries, Fuji Heavy Industries and Kawasaki Heavy Industries in Japan; and
Alenia Aeronautica in Italy - to get a better handle on the problems. He also visited Spirit AeroSystems of Wichita, Kansas, which
makes the front section of the fuselage at an old Boeing plant. Nevertheless, after four years of delay, Boeing delivered the first
airplane to All Nippon Airways in September 2011 (Slayton and Spinardi, 2016), and even then the 787 was grounded by much
publicized problems including its lithium batteries being a fire hazard.
Although Boeing's decision to global sourcing is in line with the prescription of the global factory literature, the actual con-
struction of Boeing 787 points out the limits. The initial scope that Boeing envisioned was too big to follow through. Problems in
managing a global supply chain contributed to a multi-year delay. Boeing was forced to reduce its scope of outsourcing. Boeing
bought the rear-fuselage plant from Vought in an effort to reel in its supply chain in 2008. Boeing also increased stakes in other major
suppliers to reduce inter-organization coordination. It can be concluded that a successful global outsourcing strategy depends on the
complexity of such a project. Boeing's competitors, Airbus, Bombardier, and Embraer were more conservative in outsourcing their
projects, subsequently saving them from the delays resulting from high complexity. For example, Airbus, although outsourcing its
manufacturing to India, expanded the project scope at a much slower pace. Bombardier's most notable outsourcing project was a non-
core customer support unit to India in 2005.
3.
Theory
–
complexity and global factory performance
In this section, we explain and address the control and coordination issues in the global factory building on resource dependence
theory, network theory, and complexity theory. The literature defines complexity, in general, in terms of the number of elements in a
system and the number of interactions among these elements (Anderson, 1999; Simon, 1969; Skyttner, 2006). Applying this defi-
nition to our context, complexity stems from the number of firm decisions and the interdependence among these decisions. In a
manufacturing value chain it is evident that decisions in the sales part of the business can be dependent on decisions regarding
inventory, components, materials, manufacturing, and sales forecasts in other parts of the business. In his pre-global factory writings,
Buckley acknowledges the problem of complexity in international business and discusses it in the parlance of system theorists who
distinguish between combinatorial complexity and organic complexity (Buckley and Casson, 2001).
Combinatorial complexity is created when multiple decisions whose choices together affect performance are made by different
decision makers. To appreciate the problem of combinatorial complexity, consider the simple case when each decision can be made in
only two possible choices, yes or no. If there are ten decisions which together affect performance, then there are 210, or 1024, unique
sets of decision choices. Given a firm which has made one set of these decision choices and currently has the corresponding per-
formance, selecting which of the other thousand sets of decision choices which will increase performance is difficult because of the
sheer number of possibilities and the uncertain performance outcomes. This exponential relationship between the number of deci-
sions and the possible sets of choices, which together determine performance, characterizes combinatorial complexity.
Another layer of complexity on top of combinatorial complexity is organic complexity (Buckley and Casson, 2001) which is due to
the interdependence among decisions in affecting performance. When one decision affects the effectiveness of other decisions, there
are feedback loops within the firm which make it difficult to determine cause and effect. Complexity theory tells us that performance
of a single business can be scaled up to a certain point as the number of decisions and their interdependence make it unmanageable.
After that point, performance decreases. We must take into account both combinatorial and organic complexity in addressing our
research question. Thus, we define internal complexity as both the number and interdependency of decisions within a single business
or subunit.
One way to scale performance of an MNC and make it more manageable is to slice the value chain into different activities which
are managed in different subunits or subsidiaries. This approach has the advantage of organizing the MNC into smaller, less complex
subsidiaries while at the same time locating them in countries to take advantage of country specific advantages. This disaggregation
and dispersion of the firm makes the internal complexity of each subsidiary less complex because fewer key decisions take place
within each subsidiary. Subsidiary management only makes decisions associated with a single value chain activity and focuses on
optimizing its performance of that one value chain activity. While disaggregating the value chain reduces subsidiary internal
complexity, it does not reduce the overall complexity of the MNC. The number of decisions and interdependence among them is the
same. Rather the tradeoff for reducing internal complexity is introducing interdependence of decisions among subsidiaries. These
interdependencies add more complexity and more costs in terms of management coordination and control efforts among subsidiaries
(Contractor et al., 2010). Therefore, an MNC must not slice its value chain too finely or it will add more complexity among the
subsidiaries than it reduces the subsidiaries' internal complexity. Therefore, too much complexity within each subsidiary adversely
affects performance, but also too much complexity among the subsidiaries adversely affects performance.
Within a single MNC, alignment of many decisions can be made more manageable because it is under control of a single man-
agement hierarchy. This is typically implemented when a business' overall performance goals are cascaded down by top management
as management objectives for lower levels in the reporting chain. Performance within a business will peak when the particular
decision choices made within the organization reinforce each other and align with its overall performance goals by achieving a
synergy of action. This is the best case scenario where all departments and subunits have “their noses pointed in the same direction”
as the Dutch idiom states. However, while some MNC organizational structures can help manage complexity better than others, all
MNC structures have limits on performance imposed by complexity despite the single management hierarchy and chain of control.
6
The literature supports the relationship between MNC complexity and firm performance as an inverted U shaped curve regardless of
the structure (Celo et al., 2015; Ghemawat and Levinthal, 2008; McKelvey, 1999).
Previous literature discusses that high complexity and high coordination costs decrease MNC performance (Ambos and
Birkinshaw, 2010; Bouquet and Birkinshaw, 2008; Wiersema and Bowen, 2011). However, this literature does not address the
uniqueness of the global factory structure because the global factory is not a single MNC, but a “family” of firms, or the lead firm and
its supplier and service provider firms. One of the tenets of resource dependence theory is that a firm's power comes from maintaining
control over important resources (Pfeffer and Salancik, 1978). Therefore, when a global factory lead firm establishes a buy-or-ally
relationship with a separate business to supply tangible or intangible resources, it loses some degree of control over its own destiny.
This is because the resource provider's primary goal is maximization of its own performance. While the resource provider is obligated
to meet the contractual terms of the relationship with the global factory lead firm, the decisions made within the supplier or service
provider to optimize its own performance may have the unintentional consequences of reducing overall global factory performance.
We define external complexity as the added complexity due to the interdependence of decisions among the various businesses that
comprise the global factory.
External complexity is more of a problem in the global factory organization because the different subunits in a global factory can
be independent, legally separate businesses with their own objectives. Even if the manager within the resource supplier is sensitive to
overall global factory performance, he or she may not be aware of the decisions being made concurrently, in other resource providers
or in the lead firm itself, on which his or her decision depends. Furthermore, even if the resource supplier is aware of what decisions
are made in other parts of the global factory, it may not understand the degree to which its decision interacts with these other
decisions to affect global factory performance. In light of these unique characteristics of the global factory, our research question
results from our impetus to better understand the tradeoff between internal and external complexity in this type of organization. Such
a tradeoff becomes even more crucial considering that often MNCs facing competitive pressures have no other option but to out-
source.
Network theory can help us represent the global factory complexity problem. The IB literature has a history of theorizing the MNC
as a network (Ghoshal and Bartlett, 1990; Gupta and Govindarajan, 2000; Nohria and Ghoshal, 1997). For the global factory, we can
represent each decision as a node in the network. The links can represent the interdependency of decisions, such that a link between
two nodes means the decisions are dependent. Network theory defines constructs such as network density which would indicate the
complexity of the entire network. Network theory also defines node degree as the number of connections a node, or decision, has to
other nodes (Wasserman and Faust, 1994). The average network node degree is an indication of the MNC's organic complexity.
While we can construct a network which represents the number and dependency of decisions in the global factory, network theory
has three limitations which prevent us from analyzing the network to answer our research question (McKelvey, 1999). First, we
would like to divide our above MNC network into subnetworks where the decisions made within each business of the global factory
are contained in one of the subnetworks. Then we would like to examine the density of each subnetwork, internal complexity, and
also calculate the density among subnetworks, or external complexity. However the network theory construct of density does not
enable us to calculate the density within multiple subnetworks and at the same time density of connections among subnetworks
(Wasserman and Faust, 1994). Second, we would like the links in our MNC network of decisions to be directional, such that arrows on
links going into a node represents decisions on which this decision is dependent in order to be effective in increasing performance. We
can draw networks with directional links to represent the dependency of decisions, but network theory's degree construct does not
typically take into account directionality. The third and most severe limitation of network view is that it is static, whereas, of course,
an MNC is not. If an MNC's set of decision choices do not meet the MNC performance objectives, an MNC will change its decisions in
order to improve performance. The MNC's approach to changing its decision may be more trial and error than analysis of its network
of decisions because of its limited insight and complexity. Nevertheless, the MNC will strive to incrementally move towards in-
creasing performance over time. In addition, an understanding of performance in the context of global factory requires also ac-
counting for the ‘moves’ made by other firms which might alter the value of solutions found by the focal firm and hence it's necessary
to track dynamically the interaction of all the parts.1 In light of these three limitations of network theory, we instead adopt Kauffman's
(1993) approach, in the following sections, which has none of these limitations.
4.
Research design - value chain example and simulation models
In this section we describe the research design we used to explore the questions of the effects of internal and external complexity
on the performance of the value chain for MNCs that comprise a global factory's architecture. At a high level, the research design is
straightforward. We have one MNC value chain whose complexity in terms of the number of value chain activities and the inter-
actions, or links, among these activities is the same for four different global factory configurations. In each global factory config-
uration the value chain activities are distributed across three different firms, MNCs, arranged in a global factory architecture. That is,
one MNC can be thought of as the lead firm and the other two as service providers or suppliers. The only difference among the four
configurations is the way in which the value chain activities are mapped onto the three MNCs that comprise a global factory. We
choose four mappings which differ only in the total number of external links, or links among the three MNCs within a global factory,
and the total number of internal links, or the links within the three MNCs of the global factory. The total number of links, external or
1
According to Fioretti (2 013) early social network analysis used to be a st ructure-to-actor description of social relations, while NKC and more in general ag ent-based
modeling are the opposite, i.e. bottom-up or actor-to-structure approach.
7
internal, is the same for all global factory configurations. We simulate the value chain's operation for each global factory config-
uration, and compare the global factory performance, which reflects the performance of all three MNCs within a global factory
configuration, or what we will call simply a configuration from here forward. The following paragraphs describe these configurations
in more detail.
Fig. 2 shows the value chain's 15 activities. Three are value chain support activities: Corporate finance, procurement, and a global
supply chain unit. The remaining twelve are primary value chain activities: two R&D units, four component manufacturing units,
three final assembly manufacturing units, marketing, and two regional sales offices. This base model MNC represents a completely
internalized value chain with no outsourcing. Fig. 2 shows all these value chain activities and also the linkages between them. In our
model linkages represent interdependence of decisions made by an MNC's units: a directed linkage from A to B represents a de-
pendency of the performance of unit B on a decision made in unit A. This is different from an organizational structure diagram where
linkages may represent lines of reporting. One simple example of our linkage is that performance of a unit performing final assembly
is dependent on another unit making subassemblies to ship to the final assembly unit at the time required. Obviously, there are a
whole range of reasons for these dependencies, such as budget allocation, procurement, R&D, manufacturing, and global supply chain
activities. All the linkages illustrated in Fig. 2 are double-headed arrows indicating mutual dependencies between units and these are
counted as two links making 146 links in total. However, in the general case, and often in our subsequent models, a link can also be
unidirectional.
In order to simulate the workings of the global factory, we developed four different configurations of three firms, shown in Fig. 3.
Considering that our main construct of interest is the performance of the entire system of global factory, we do not single out one
MNC as the focal firm and instead consider all three MNCs to have independent strategies. As Buckley (2011) argues, “(A)t the level of
the system, it makes sense to talk about the strategy of the global factory in the same way that it makes sense to talk about the strategy
of a unitary firm even though different managers, affiliates, divisions, business functions and units have separate strategies” (274;
italics in original).
What changes from one configuration to another is the distribution of the links between internal, or within firms, and external, or
between firms. For instance, in the first configuration, or (B, C, D), the MNC B outsources to C only one of R&D functions, MNCs C and
D each outsource the manufacturing of one component to B, and finally D outsources one of its R&D functions to C. Overall, in this
configuration 378 links are within and 60 between the MNCs. The numbers of internal and external links for each configuration are
shown in Fig. 3. Our choices are driven by the fact that value chain is no longer divided into large groupings such as R&D, Production,
or Marketing and fine-slicing can take place within each category (Contractor et al., 2010). Moving from the first to the last con-
figuration we increasingly let the MNCs specialize to reflect the resource endowments of their respective locations. For instance, in
the configuration (K, L, M) company L specializes in manufacturing of components, K in final assembly, and M in R&D. In this way we
distinguish between countries that have abundant cheap labor (such as L or K) and those with advanced technologies (such as M).
The four configurations shown in Fig. 3 can now be compared in terms of their overall performance using the NKC-simulation
methodology. Before presenting the results of the simulation, in the next section we shortly review simulation methods and especially
the NKC model and then describe in details how we have modified the base model to better describe the context of an MNC.
5.
An NKC simulation model for the MNC
5.1.
Simulation methods and the NKC-model
Simulation methods use computer software to model the operation of “real world” processes, systems, or events (Davis et al.,
2007; Law and Kelton, 1991). By offering a simplified and parsimonious representation that has some, but not all, of the char-
acteristics of that world (Lave and March, 1975), simulation methods are particularly useful to develop theory when the phenomena
of interest involve multiple and interacting processes. Simulations make more assumptions than typical empirical research, but
enable the researcher to control for all variables other than the antecedents being manipulated and follow the interactions of spe-
cifically configured firms over extended periods of times.
Within the family of simulation methods, the NKC-methodology has been very useful in addressing the performance implications
of dynamic processes in complex systems. Originating in the field of biology, it is used to model the evolution of biological systems
towards greater fitness (Kauffman, 1993; Kauffman and Johnsen, 1991) while the species move on the ‘fitness landscape’ (Wright,
1931, 1932). In the domain of social sciences, the biological notions of alleles and epistasis are replaced by decisions and inter-
dependence, respectively (Ganco and Hoetker, 2009). Levinthal (1997) first introduced NK methodology into the management and
organizational theory literature to examine complexity in an organizational context and McKelvey (1999) translated the NK-model
into a firm context by using value chain competencies as “parts” of firms. McKelvey (1999: 304) even states that “the assumptions for
firms are actually more straightforward than for organisms.” Many of these studies in this line of research examine the pattern of
interaction among different parts/units of organizations and how it affects different organizational outcomes. For instance, NK-
simulation has been used to explore the effects of modularization on the dynamics of innovation and performance in complex systems
(Ethiraj and Levinthal, 2004), the effect of the level of articulation of a strategy on performance (Ghemawat and Levinthal, 2008), the
relationship between the complexity of a successful business strategy and its ability to deter its imitation (Rivkin, 2000, 2001), and
patterned interactions in complex systems and their implications for organizational exploration. (Rivkin and Siggelkow, 2007).
NK-methodology enables insight into theory because it models a controlled system, free of limitations of empirical approaches,
which in face of combinatorial complexity is constrained by the expense and practical challenges of studying real-world systems
(Lazer and Friedman, 2007: 672). In McKelvey's (1999: 313) words, the use of this methodology allows us “… to go beyond the loose
8
(caption on next page)
9
Fig. 2. Base MNC -model.
Labeling convent ions:
•
All connectio ns between activities are b idirectional (accept ing inputs/direction, also giving status/feedback)
•
Numbers next to arrows indicate the number of connectio ns going into an acti vity (accepting inp uts)
•
No numbers nex t to an arrow indicates one input
•
Total number of connect ions in value chain is t he addition of all
“
arrows
”
in the diagram
Internal Links: 378; External Links: 60 Internal Links: 328; External Links: 110
Internal Links: 204; External Links: 234 Internal Links: 161; External Links: 277
Fig. 3. Con
fi
gurations with coupled
fi
rms and outsourcing.
insights of natural history case studies, to pursue questions about intricate complexities impossible to study in real world analyses.”
The NKC-model as an extension to the NK-model has also been introduced first in the field of biology (Kauffman, 1993, 1995;
Kauffman and Johnsen, 1991), but later applied to technological coevolution (Kauffman, 1995), management research (e.g., Baum,
1999), and coevolution of firm strategies and organizations (Ahouse et al., 1991; Stewart, 2001).
In the simplest form of organizational research N stands for the number of units in an organization. In each unit a decision is made
that can take one of two values. The payoff or the contribution to overall organization's fitness or performance of choosing one or the
other value in one particular unit, however, depends on decisions made in other K units (see McKelvey, 1999 for a thorough ex-
planation). The ‘position’ of an organization at any point in time consists of the decisions made in each unit and the overall per-
formance is calculated by considering the contributions of all units. The mapping of all possible organizational positions to the
performance values is often referred to as a ‘performance landscape’, a graphical depiction of the relation of inputs to the output
(Levinthal, 1997).
N and K jointly determine how ‘rugged’ the fitness landscape is. When there is little interaction (low K) among the parts, there is
one, or few optimal combinations or ‘hills’ and the landscape is ‘smooth’ (see Fig. 4a). On the other hand, as shown in Fig. 4b, firms
with a high degree of interaction among their subunits (a higher K value) result in more rugged, multi-peaked performance surface,
K
All FA
L
All R&D
All Comp.
All FA
M
All R&D
All Comp.
J
H
All FA
All Comp.
I
All Comp.
All FA
E
1x Comp.
1x FA
F
1x R&D
G
1x Comp., 1x FA
1x R&D
B
1x Comp.
C
1x R&D
D
1x Comp.
1x R&D
10
(a) Lower K, “Flatter” Landscape (b) Higher K, “Rugged” Landscape
Fig. 4. NK-landscapes.
since a change in one choice will influence many of the other subunits. The organization then moves on this landscape in search for
higher performance.
However, despite allowing to track the performance of an organization dynamically, a key feature of the NK performance
landscape is that it is fixed. By contrast, the NKC-model takes into account also the interactions between organizations or, in NKC-
parlance the fact that the landscapes of two or more organizations are coupled. A movement made by one organization on its own
landscape affects or ‘deforms’ the landscapes of the organizations with which it is coupled. In this way the NKC-model helps trace the
dynamics of a co-evolutionary system of organizations (Vidgen and Bull, 2011). Kauffman and Johnsen (1991) first introduced this
model to study the coevolution of different species, where adaptive moves by one species may change the fitness and the fitness
landscapes of the coevolutionary species. Famously they illustrated the model by describing the coevolution of frogs and flies: the
development of a sticky tongue by the frog alters the fitness of the fly, and given the frog's sticky tongue, the fly should develop
slippery feet (Kauffman and Johnsen, 1991). In another example that is more familiar to an IB-audience Ahouse et al. (1991) describe
Michelin's (a major European tire manufacturer) decision to enter the North American market and Goodyear's (a U.S. tire manu-
facturer) retaliation by lowering its tire prices in major European markets. According to Ahouse et al. (1991: 348) this example
“demonstrates the consequences of myopic behavior when strategic decisions are linked. Michelin tried to myopically improve its
market position, but at the same time “deformed” Goodyear's profitability landscape, which in turn led Goodyear to myopically cut
prices in Europe.”
The NK- and NKC-models have been already applied in many other fields outside biology. In the field of IB, the NK-methodology
has been recently introduced and used to investigate the effects of structural choices and level of internal complexity on MNC-
performance (Celo et al., 2015), but the NKC-model is as of now yet to be applied. In the rest of this section, we explain our
application of NKC methodology to model the effects of internal/external complexity and value chain configurations on MNC per-
formance.
5.2.
An NKC-model of MNCs
We start by showing the steps we go through to build our model illustrating with a very simple example of two MNCs, X and Y,
consisting each of three units (N
=
3) (see Fig. 5a): the headquarters (x1 and y1, respectively), an R&D department (x2 and y2), and a
foreign subsidiary (x3 and y3). The arrows between the subunits indicate interdependencies of unit decisions that affect overall firm
performance and not necessarily only authority, the hierarchy of reporting level, or communication or knowledge flows. While all
subunits report to the headquarters, each of them has the independence to make some business decisions regarding operations and
allocation of its resources, and overall firm performance is affected by the interdependence of these decisions. To illustrate, let's
assume first that each of the units in each of the MNCs will have to make a decision with two possible choices (0 or 1) and that the
outcome of the decision making will have implications for MNC performance. We also keep the types of decisions the same for both
MNCs. More specifically, the headquarters (x1 and y1) will have to choose between a strategy of local responsiveness (0) or global
integration (1). The R&D departments (x2 and y2) are considering whether to dedicate resources to introduce a portfolio of custo-
mized products to be sold in individual countries (0) or to process innovation (1), which would result in a generic product to be sold
in all countries where the MNC operates. Finally, both foreign subsidiaries (x3 and y3) are considering to grow by acquiring local
firms that offer similar products, which are more customized to country's specific preferences (0), or by aggressively marketing their
existing standardized product (1).
To understand the interdependencies among units, let's consider MNC X (the same can be said for Y) and start with the arrow
connecting HQs to the R&D-unit. It shows that the contribution to overall MNC-performance of having the R&D-unit choose one or
the other option depends also on the decision made at HQs. For instance, if the R&D-unit decides to develop a portfolio of products
11
(b)
(a)
Fig. 5. Example of two MNCs and the NKC-adjacency ma trix.
while HQs decides to allocate resources to pursue a global integration strategy (e.g., Dellestrand and Kappen (2011) discuss the
headquarters' resource allocation for innovation transfer projects), this would create inconsistencies within the organization and lead
to inferior performance. Likewise, the contribution to overall performance of choosing the growth strategy at the subsidiary depends
on the decision made at the HQ and R&D-unit. All these links are internal to each of the MNCs. However, there is one more link to
consider: the double-headed arrow connecting x3 and y3. What this means for x3 is that the contribution to performance of X of the
decision made at x3 depends not only internally on the decisions made at x1 and x2, but also externally at y3. This could be the case,
for instance, if x3 and y3 are located in the same host country and see each other as competitors (in a different context the units of two
different organizations might be collaborators).
The specific pattern of interdependence described above is reflected in the adjacency matrix or the NKC-matrix shown in Fig. 5b
and c. Fig. 5b gives the big picture with the internal links of firms X and Y located in the top-left and bottom-right corners, re-
spectively (labeled as KX and KY). The other two quadrants represent the effect of X on Y (labeled Cx → y), and of Y on X (labeled
Cy → x). Fig. 5c shows the full NKC-matrix for our example. We create a column and a row for each unit and in the intersection of a
row and a column we put a 1 if the row-unit depends on the column-unit and a 0 otherwise (we obviously have 1s in the main
diagonal of the matrix).
For instance, the second row in the NKC-adjacency matrix of Fig. 5c is (1, 1, 0, 0, 0, 0), meaning that x2 depends on x1 and itself,
but not on x3 or any of the Y-units. Finally, two columns have been added to the right of the adjacency matrix to that show the
number of other units on which the row unit depends internally (K) and externally (C), respectively.
Di
ff
erent from the original NK and NKC models, in which each subunit depends on exactly K other units internally and C units
externally (e.g., Caldart and Oliveira, 2010; Kau
ff
man and Johnsen, 1991; Vidgen and Bull, 2011), we propose a
fi
ner-grained
conceptualization of K and C in order to re
fl
ect the di
ff
erential roles of value chain activities. More speci
fi
cally, let k
j
be the number of
other units on which unit j depends internally and K the average across all units (K
=
(Σ
j
= 1
to N
k
j
) / N), i.e. we allow di
ff
erent units
to have di
ff
erent values of k. We follow the same logic to de
fi
ne c
j
and C for the external links. The full NKC matrix is a 2N
×
2N
adjacency matrix (e.g., Ethiraj and Levinthal, 2004; Ganco and Agarwal, 2009; Ghemawat and Levinthal, 2008; Rivkin and
Siggelkow, 2007). Equipped with the NKC-matrix we can generate the performance landscape as described in the next section.
5.3.
Generating the landscape of performance values
Returning to the illustration shown in Fig. 5a, we have altogether two MNCs with six units and a decision made in each unit that
takes on one of the two values (0 or 1) resulting in a total of 64 possible combinations (26) of 0s and 1s. Showing the entire table is not
only space consuming but also not essential to our example. Instead, in Fig. 6b and d we show parts of the table with all the
combinations for x1, x2, x3 (columns 1 to 3, respectively), and y3, the only unit of Y that has any influence on X (column 4).
More specifically, in Fig. 6b and d, we have all combinations for x1, x2, and x3. However, in Fig. 6b y3 is fixed at 0, and in Fig. 6d
y3 is fixed at 1. The next three columns in both tables (wx1–wx3) show the performance contributions of having a value of 0 or 1 in
x1–x3, respectively. During the simulation such values are generated randomly. A closer look at Fig. 6b and d reveals that there are
only two different values in wx1-column (0.6 and 0.55), reflecting the fact that x1 is completely autonomous in its decision-making,
i.e.
depends only on itself. The situation is different in column wx2: there are four different values (22), because the value changes any
time x1 or x2 itself change (e.g., compare the values in rows 1 and 3), but remains the same if the values in x1, and x2 remain
x
1
, y
1
: Headquarters
x
2, y2: R&D
x
3, y3: Subsidiaries
x1
y1
x2
x3
y2
y3
KX
C
X→Y
CY→X
K
Y
(c)
12
(a) (b)
Fig. 6. Example of coupled landscapes and NKC -matrix.
unchanged, regardless of the value of x3 and y4 (e.g., compare values in rows 3 and 4). Unlike the values for wx1 and wx2, which
remain the same when we move from Fig. 6b to d (since none of x1 and x3 depends on y3), there are 16 different values (24) in the
column wx3 (in both Fig. 6b and d), which means that the contribution of having x3 set at 0 or 1 depends also on the decisions made
outside the organization. Finally, the W-column shows the performance for the entire combination of the units of X as the average of
the three values in columns wx1, wx2, and wx3. While an exact 3D–visualization of the ‘landscape’ is not possible (with values in three
units and the resulting overall performance, the ‘surface’ would be 4-dimensional), in Fig. 6a and c we combine the values of x1 and x2
in the x-axis, of x3 in y-axis, and finally the performance in the z-axis.
In the general case, N represents the number of units or activities of an MNC value chain and each of the units will have to make a
decision of choosing between two options designated as 0 or
1.
2
In other words, each MNC's value chain's set of decisions is re-
presented as a vector (x1, x2,
…
,
xn), where each of the
x
j
has a value of 0 or 1 and hence the landscape is composed of 2N possible
decision choices. The overall performance value associated with the full vector F(x1, x2,
…
,
xN) is simply the average of the individual
contributions:
Σ
i = 1 to
Nf(xi
|
x
i1
, x
i2
,
…
,
xiKi, yi1, yi2,
…
,
yiCi) / N and each of the individual contributions is drawn randomly from the
uniform distribution from 0 to
1.
3
The landscape for MNC Y would be built in the same way we described above. While each of the
MNCs X and Y has their own performance landscapes, these landscapes are coupled in the sense that they are continuously reshaped
any time one of the relevant units of the other MNC makes a decision. In the next section we have a closer look at the movements of
the MNCs in search of better performance.
5.4.
NKC adaptive walk to search for maximum performance
Once the performance landscape is generated, the organization moves on the landscape in search of locations with better per-
formance, which we call adaptive walk. An MNC could in theory go through the exercise of generating such a landscape, which,
following our example in Figs. 5 and 6, would mean estimating each and every of the 48 elements of the matrices in Fig. 6b and d in
columns wx1–wx3 (the number of elements would have been higher if the C were higher), calculating the overall performance for
2
A more general case of more complicated decisions
–
i.e. decisions that take more t han two valu es
–
may be reduced to a sequence of bi nary choices; see McKe lvey
(1999).
3
The distribution from which payo
ff
values are randomly drawn cou ld a
ff
ect the outcome (for instance, a Gaussian distribution could be used), but Kau
ff
man points
to the fact that the sta tistical features of the resulting landscapes are
“
largely insen sitive to the choice made for the underlying distr ibution
”
(1993, pp. 44
–
45).
3
2
1
3
2
1
(c)
(d)
2
2
13
•
•
•
each row (column W), and finally finding the maximum value in that column. In practice, however, this process is very difficult or
even impossible for at least two reasons. First, as a manifestation of the organic complexity the estimation of the values becomes
increasingly difficult and inaccurate when the unit in question is influenced by many other units due to lack of data, large number of
variables involved, and their interactions. Second, in line with the combinatorial component of complexity, it is the sheer number of
elements to be estimated and since the number of possible combinations grows exponentially with the number of units, time re-
sources put constraints on finding the global optimum (Buckley and Casson, 2001; Frenken, 2006; Simon, 1969). Instead, the MNCs
try – as do the organizations in general – to make adjustments one step at a time, i.e. looking for better solutions in the area
surrounding their existing position consistent with the principle of satisficing (Simon, 1957) and local search (Cohen and Levinthal,
1990; March and Simon, 1958).
To illustrate, we turn to our example, but for the time being disregard the effects of Y on X by fixing the value of y3 to 0. By doing
this, the matrix in Fig. 6b represents a simple NK-landscape for the MNC X. Let us also assume that the initial position of X on the
landscape is the vector (0, 1, 1) with a performance level of 0.473. Now the MNC searches the neighborhood for a better position by
changing the value of one of its elements from 0 to 1 or from 1 to 0. If the value of x3 is changed from 1 to 0, the new position – (0, 1,
1)
– would have a performance of 0.447 and hence there is no reason for MNC to move. However, if x2 is changed from 1 to 0, the new
solution (0, 0, 1) with a performance of 0.617 is an improvement and the MNC moves there. If the landscapes of X and Y were not
coupled, firm X would continue its walk by changing in the next step x3 from 1 to 0 achieving the position (0, 0, 0) with a per-
formance of 0.683 (showed with a dashed circle in Fig. 6d). From there no further improvements would be possible; the MNC has
reached a peak: the headquarters has chosen a strategy of local responsiveness, the R&D department develops a portfolio of custo-
mized products, and the foreign subsidiary sells customized, country's specific products.
However, in an NKC system, each whole NK system is evolving with each other system to which it is coupled (Stewart, 2001). As a
result, before MNC X moves to position (0, 0, 0) it's the turn of MNC Y to make a move. Assuming that this move involves changing y3
from 0 to 1, the landscape of MNC X has changed from that of Fig. 6b to that of 6d. Also, the performance of the current position (0, 0,
2)
is now 0.64 and, more importantly, now a move to (0, 0, 0) is no longer justified, since its performance is lower. As a matter of fact,
none of the neighbors now constitutes an improvement and position 2 is the local optimum (the payoff might change again depending
on the next move of Y). While in this example the coupling of the landscapes is of a complementary nature and causes a sub-optimal
Table 1
Roadmap of NKC -modeling.
Step Theoretical modeling Alterations for m anagerial applications
Establish the internal structure: links
within
each
participating
fi
rm.
Establish the external str ucture for the
global factory
:
links
between
fi
rms representing outsourcing activities.
The
fi
rm can choose any numbe r of units (i.e. N).
The
fi
rm can customize the
level
of internal complexity (i.e. K)
as well as the
pattern
of internal depend ence to re
fl
ect its own
idiosyncrasies.
•
More than two de cisions can be made by each unit.
As above, the
fi
rm can customize the pattern of external
dependence and t he links going from the
fi
rm to the rest of
global factory or vice versa.
Map the internal and external structure to the NKC
adjacency matrix.
Generate performa nce landscapes. Uneve n distribution of weig hts among the units when
calculating performance to re
fl
ect di
ff
erences in importance.
Adding costs for each move a
fi
rm takes to better re
fl
ect the
dynamics a given indust ry.
Firms search for peaks (i.e. better solutions) on the
landscape which is simultaneo usly changed by the
moves of other
fi
rms.
Firms take turns in making their moves complet ing one
step at a time.
The order in which
fi
rms make their moves can be customized
to re
fl
ect the value chain.
When it's their turn,
fi
rms in the global fa ctory can be all owed
to make more than one optimization ste p.
•
•
•
•
•
•
•
•
•
•
14
solution to become optimal (at least locally), scenarios in which an optimal or good solution turns into an underperforming one are
also possible (coupling reflects competition). In fact, with high enough C and low K, the firms might keep walking on continuously
changing landscapes in what Kauffman (1993) calls ‘coupled dancing.’
Table 1 provides a roadmap of the entire simulation process. The ‘theoretical modeling’ column explains shortly the goal of each
step, the ‘step’ column visualizes in a stylized way each of the steps. Finally, in the last column or the ‘alterations for managerial
applications’ we provide some ideas how this model can be adapted to simulate real problems, a point we address in details in the
Discussion section.
5.5.
Applying the NKC model
The model introduced above can now be applied to the configurations described in Section 3. Each MNC has an N
=
15. Each unit
is equipped with fixed amount of resources, and their capability to adapt to the environment is equal. Also, each interaction has equal
strength. Further, MNCs' international experience is constant for all simulations. Finally, the local knowledge is equal for all sub-
sidiaries.
Following the same routine as illustrated above, each of the configurations can be translated into an NKC adjacency matrix. Again,
we have a 1 any time the column influences the row and a 0 otherwise (each unit depends on itself and hence there will be 1s in the
main diagonal in all NKC matrices).
In order to control for the overall amount of complexity (internal and external), the total number of 1s in the entire NKC-matrices
was kept constant. The changes made when we move from one configuration to another cause some of the internal links to become
external. So, moving from one configuration to the next the NK-matrices became less populated and the C-matrices more.
Accordingly, for configuration (B, C, D) the average, or configuration's K and C were 8.4 and 1.3, respectively. For (E, F, G) K and C
were 7.3 and 2.4, for (H, I, J) 4.5 and 5.2, and finally for (K, L, M) 3.6 and 6.1.
These configurations and the simulations described below allowed us to look at the performance implications of increasingly
distributed value chains without focusing on individual firms, but rather on systems of firms.
6.
Results
For each NKC-matrix the simulation software generates a performance landscape that takes into account the specific pattern of Ks
and Cs (in the same way as illustrated in Fig. 5). MNCs make their moves in rotation (for instance, B, C, D, B, C, D,
…
) after initial
positions are chosen randomly on the respective landscapes. We allow each MNC to perform up to 50 such steps to optimize its
performance to the local environment.
4
It may be that an MNC reaches a local optimum in
<
50 steps if a neighboring location does
not improve performance in the subsequent steps. We call this 50-step process
‘
a round of simulation
’
. The simulation software
records the MNC performance achieved after each simulation round and we average the three performances after 50 steps to calculate
the performance of the con
fi
guration.
The results of such a round can and will be affected by both the random assignment of values we use to generate the performance
landscape and the random choice of the initial position. In order to eliminate this effect, we repeat each round of simulation 1000
times. In other words, for each configuration of three MNCs, we generate 1000 different NKC-landscapes and every time start at a
randomly chosen initial position. By doing so, we obtain statistical significance below 0.01 (Ganco and Agarwal, 2009). The values
we present below are averages of the configuration performance from 1000 rounds of simulation. For instance, if in the next section
we report that the performance of a particular configuration after 50 steps is 0.64, we mean that 0.64 is the average of the per-
formance values recorded for the 1000 simulation rounds for the same configuration.
The results are shown in Table 2, in which besides the configurations and their performance values we include also the normalized
performance values along with the values of K, C, and the difference K-C. The configuration (H, I, J) with a K of 4.5 and a C of 5.2 has
the highest performance followed by (E, F, G), (K, L, M), and (B, C, D).
Further, in Table 3 we summarize the results of all the t-tests we run to compare each pair of configurations. The third column
shows the sign of the difference in performance between the first and second configuration being compared (in the first and second
column, respectively), the last column the significance of the corresponding two-sided t-test. With the exception of the (E, F, G)–(H, I,
J) comparison, which was marginally significant, all the comparisons were significant at 99% confidence level.
In Fig. 7, we report the normalized values of the configurations' performances (the far right column in Table 2) and plot them
against the difference K-C. In the same graph we also fit a second degree curve. The performance achieves its highest level for K-
C=
−
0.7 and drops on either side of this value suggesting an inverted U-shaped relationship.
The results provide evidence that balancing the level of internal complexity (the number of links within MNCs) with the external
one (the number of links between the MNCs) improves the performance of the entire configuration of the global factory. Such finding
is consistent with the view that favors responding to complexity via more complex strategies, structures, and decision processes,
called a complexity absorption strategy, over a complexity reduction strategy, which entails the pursuit of equilibrium via simplistic
internal arrangements (Ashmos et al., 2000; Boisot and Child, 1999; Walters and Bhuian, 2004).
It is important to repeat at this point that the overall number of links in any configuration (within and between MNCs) remains
constant. Also the nature of each link remains unchanged, that is whether the link is internal or external it preserves the logic and
4
Performance for all Ks converges aroun d step 50. We tried models with more steps, and results are consistent. Thus at here we report models running 50 steps.
15
Table 2
Summary of the results: performance.
Con
fi
guration
K
C
K–C
Performance
Normalized performance
BCD
8.4
1.3
7.1
0.6221751
0
EFG
7.3
2.4
4.9
0.6653361
0.9419509
HIJ
4.5
5.2
−
0.7
0.6679959
1
KLM
3.6
6.1
−
2.5
0.6605338
0.83714515
Table 3
Summary of the performa nce comparisons.
t
-Test signi
fi
cance
⁎⁎⁎
⁎⁎⁎
⁎⁎⁎
p
=
0.0506
⁎⁎⁎
⁎⁎⁎
Normalized Performance with Fitted Curve
1.4
1.2
1
0.8
0.6
0.4
0.2
0
-4 -2 0 2 4 6 8
K - C
Fig. 7. Con
fi
guration performance versus K
–
C (internal
–
external complexity).
sequence of the value chain activities. The only element that changes and to which the performance differentials we found can be
attributed is the distribution of the links between internal and external.
The results are best understood by considering two extreme scenarios. In the first scenario, the evolution and performance of the
global factory is largely driven by the (relatively high) internal complexity and the role of the external complexity is insignificant. The
firms try to optimize largely independently from each other and because of the high level of internal complexity they quickly reach
local suboptimal peaks and with little pressure coming from the external environment. Firms within this scenario are destined to
remain stuck on such peaks.
In the second scenario, firms within the global factory system face low internal and high external complexity. The advantage of
the low internal complexity is that it is less likely that the firms will get stuck on a suboptimal peak; the disadvantage is that
improvements while steady are very minimal. Under these circumstances, a firm would embark on a long journey of continuous
improvement on a relatively smooth landscape, if it wasn't for the external complexity that frequently changes its landscape. To be
sure the ‘new’ landscape would be smooth again, but a different one with the global optimum having changed location. Such high
dependence on the external environment leaves little room for long-term strategic actions by the firms.
The two types of complexity we modeled in our simulations play opposite roles with regard to the time needed for the entire
system to stabilize. Raising the level of internal complexity speeds up the process of finding an equilibrium, because it increases the
‘ruggedness’ of the landscape and hence the number of peaks. At the same time, however, the average height of the peaks decreases
(Kauffman, 1993). Increasing the level of external complexity slows down the process of finding an optimum (e.g., Baum, 1999),
since the landscape itself continuously changes albeit the number of peaks (not their locations) and their average height might not be
affected. By increasing the external complexity, the past actions of each firm become increasingly irrelevant.
In the next section we discuss our findings within the broader context of the performance implications of the interplay between
the internal and external complexity. We also translate the results into practical implications for the management of the global
factory.
Normalized
Performance
(a)
(b)
(a)
–
(b)
BCD
EFG
“
−
”
BCD
HIJ
“
−
”
BCD
KLM
“
−
”
EFG
HIJ
“
−
”
EFG
KLM
“
+
”
HIJ
KLM
“
+
”
⁎⁎⁎
Sign i
fi
cant
at
<
0.001.
16
7.
Discussion
One of the potential advantages in a successful global factory configuration is enabling the lead firm to apply its resources to
execute its unique value adding activities, while at the same time offloading other value chain activities to suppliers, subcontractors,
and service providers and extracting rents from their assets (Buckley, 2011). The main challenge of the lead firm in a global factory is
coordination and control of a network of operationally interdependent but legally separate firms. Often overlooked in this challenge
is the fact that the suppliers, subcontractors, and service providers are not captive to a single lead firm (Yamin, 2011). It is not in
these firms' interests to be dependent on a single customer. These firms typically have multiple lead firm customers and if they possess
a scarce resource or skill they may deliberately seek to increase their own performance by playing one lead firm against another. In
any case, it is clear that suppliers, subcontractors, and service providers may make decisions to optimize their own performance in a
way that does not optimize the performance of the lead firm.
Thus, the global factory architecture inherently includes two countervailing forces that affect its performance. The positive force
affecting performance is that the value chain can be divided, or sliced, and its activities allocated to separately managed firms, each of
which specializes in executing that activity. This separate, parallel managing of each activity reduces the internal complexity of each
firm in the global factory compared to the internal complexity one firm simultaneously managing all activities. The negative force
stems from the fact that these firms are interdependent in value chain performance but often independent in ownership. While
suppliers and service providers have contractual obligations to the lead firm, they may make decisions which seek to optimize their
own performance but result in suboptimal performance of the global factory due to the interdependence of these firms in executing
the value chain. While symbiotic relationships and mutualism are expected to occur between the firms in a global factory setting,
depending on the balance of power between customers and suppliers, there may well be exploitation as well (e.g., Vidgen and Bull,
2011). Also the complexity of the offshoring of activities, whether outsourced or not, makes it difficult for decision makers to consider
all important decision-making factors and increases the cost-estimation errors (Larsen et al., 2013).
These findings have important practical implications for MNCs in the context of global factory by suggesting a balance of internal
complexity and external complexity. This is important because IB researchers as well as practitioners need to know the boundaries,
limits, and contingencies of applying the global factory concept. Managers are incentivized to leave complexity to the external units
because doing so transfers financial and technical risk to their suppliers. If the incentives to outsourcing are embedded in the global
factory system, the IB literature should convey where this point of decreasing performance starts, otherwise practitioners will have to
learn the hard way, as did Boeing's CEO, W. James McNerney Jr., when he conceded that Boeing lost control of the process by farming
out more design and production work than ever. Similarly, Scott Carson, who led the commercial division said “I think there were
places where we went too far, …, clearly, we made some poor judgment calls.” The costs of fixing the poorly balanced complexity is
high as Boeing spent $580 million acquiring Vought, one of their suppliers, to bring back the production schedule. Additional
dividing and subcontracting of value chain activities beyond a point results in decreased performance, thus offsetting the benefits of a
global factory architecture. Because of lesson learned from Boeing, Airbus knows to allocate more resources for coordination between
engineering, manufacturing, and pilot production.
In a situation of high external complexity, it is tempting to simplify the complexity of incoming information and pressures by
economizing on the resources needed to respond and more in general choosing a complexity reduction strategy. However, a view of
organizations as complex adaptive systems (e.g., Brown and Eisenhardt, 1997) argues that performance may be enhanced when
internal organizational arrangements are designed to respond in a complexity absorption fashion via more complex strategies,
structures, and decision processes (Boisot and Child, 1999; Boisot and McKelvey, 2011; Brown and Eisenhardt, 1997). Our results are
consistent with this view and suggest that MNCs in a global factory setting should seek to balance the external level of complexity,
which is often not a parameter that can be adjusted, with the internal one. However, from a practitioners' standpoint a complexity
absorption strategy requires an understanding of the sources of external complexity as much as the estimation of its degree. For
instance, if the external complexity is driven by a combination of rapid innovation, volatility of demand, and efficiency, then in-
troducing internally multiple and ‘orthogonal’ performance criteria (e.g., Baum, 1999) based on innovation output, flexibility, and
cost reduction, will increase the level of internal complexity and also an MNC's ability to absorb the external complexity.
Our focus on complexity of the global factory and our emphasis on the need to balance internal and external complexity com-
plements rather than substitutes for the focus of other theoretical views and literature streams. For instance, we do not address
directly the country location decision which is a focus in the offshoring literature, albeit the pattern of specialization of different
MNCs within our configurations reflects country differentials in production factor endowments. We would argue that two global
factory configurations located in the same countries would still differ in performance if their internal versus external complexity
configurations are different within this same geographical footprint.
8.
Contributions and conclusion
Our contributions are threefold: academic contribution to global factory literature, practical contribution to MNCs of managing
outsourcing operations, and a methodological contribution of NKC-methodology to international business. Foremost is our con-
tribution to the global factory literature. To date, the global factory literature has been almost entirely theoretical (see Larsen et al.,
2013 for an exception), extolling its positive characteristics of enabling MNEs to be efficient and competitive by extracting value from
subcontractor assets and exploiting local country advantages (Buckley, 2009, 2011; Buckley and Ghauri, 2004). However, ours is the
first study which takes into account both these positive aspects as well as the negative aspects, of coordination and control, to
demonstrate their combined effects on performance. This is significant because all theory has boundaries and contingencies (Davis
17
et al., 2007) and ignoring them can result in good theory being applied under the wrong circumstances, with negative results. Our
study concludes by prescribing the boundaries within which the global factory architecture should be applied.
Second is this study's practical contribution to MNE's management of global value chains. This topic's practical significance stems
from the fact that more than $12T of trade, or about 60% of total world trade, is due to the fragmentation and international
dispersion of MNE production processes of goods and services (UNCTAD, 2013). Boeing's experience might leave many practitioners
with the conclusion that the disadvantages of outsourcing outweigh the advantages. However, all systems will fail when pushed
beyond their limits, so it important for practitioners to understand what the limits are of outsourcing value chain activities and find
the right amount of external relative to internal complexity.
In our methodology we model a typical, generic global value chain configuration (UNCTAD, 2013). However a global factory can
be arranged in a variety of ways depending on the value creation potential of each partner, their level of involvement and the
industry (Kedia and Lahiri, 2007). Besides being a methodology that helps build theory for academic purposes, NKC simulation can
be a powerful tool for company analysis of any type of global value chain configuration since several elements can be customized. In
other words, managers can make the model relevant to their particular firm by substituting the specific parameters for their own value chain
and global factory configuration. To be sure, they still have to make assumptions, disregard certain variables and events as insignif-
icant, and make simplifications to keep the complexity of the model at a manageable level. Table 1 illustrates the NKC the main
simulation steps that a practitioner would take in modeling a variety of value chains. As illustrated in our modeling exercise any
pattern of internal and external links can be modeled to represent more accurately a particular MNC and its environment. It is within
the discretion of the practitioner to decide which units are important enough to be included in the model and for any units A and B
that are selected, within the same firm or in different firms, whether a decision made in unit A has a strong enough influence on the
performance of unit B to justify having an arrow from A to B. If the practitioner feels that unit A influences unit B in more than one
way (for instance, A might decide a new way to calculate transfer prices and introduce a new expatriate compensation package), then
he/she might consider creating two units, A1 and A2 each having their own pattern of interaction with other units and between them.
Also, once a customized pattern of interaction is generated consisting of units and links, when calculating the overall performance
different weights can be attached to different units depending on their relative importance (for instance, the US-subsidiary of a Dutch
MNC might be weighed more heavily relative to the HQs than the Dutch subsidiary of a US MNC). Once a basic configuration is
created, managers and practitioners can test different scenarios of the what-if-type. For instance, they can test whether keeping HQs
strongly coupled with all the partners within the global factory is better than delegating part of this responsibility to lower-level units.
As in other models, customization might lead to a substantial increase in the number of units and interactions and potentially
create challenges computationally. However, while the size of the performance landscape increases exponentially on the number of
units, increasingly sophisticated computer codes are reducing the computational burden.5 Several computer codes are publicly
available (e.g., see Fioretti, 2013 for a summary of available platforms6) that practitioners can use to model concrete examples and
Malan and Engelbrecht (2013) provide a useful survey of techniques for characterizing fitness landscapes.
Using simulation directly is not the only way managers can make use of NK/NKC thinking. Practitioners can also use NK/NKC
methodology as a new lens to explain different business-related phenomena (e.g., Hall et al., 2012; Vidgen and Bull, 2011; Vidgen
and Wang, 2006) or as an inspiration to derive certain propositions, which can then be tested empirically (e.g., Alkemade et al., 2009;
Billinger et al., 2014; Fleming and Sorenson, 2001; Frenken, 2000).
Third, our application of the NKC simulation to IB theory is a methodological contribution to international business. NK simu-
lation methodology has been widely used in the literature on organization theory (Rivkin and Siggelkow, 2003), management
cognition (Gavetti and Levinthal, 2000), and strategy (Ganco and Agarwal, 2009; Ganco and Hoetker, 2009) and recently in IB (Celo
et al., 2015). The NKC methodology used herein takes this NK approach a step further by adding interactions outside a firm's
boundary to those internal to the firm's value chain. It is an especially powerful tool to model theory in situations where field studies
are impractical because large matched samples of cooperating MNCs are unobtainable and the number of control variables required
to account for alternative explanations of performance are too numerous (Davis et al., 2007; Lazer and Friedman, 2007; Venaik et al.,
2004).
In conclusion, our contributions are not only theoretical and practical but also unique in that they provide quantitative insights
into the divergence between global factory theory prescriptions and practitioner experiences. Our intention is not to criticize the
global factory literature but rather to expound upon the understanding of how and when to use it. The advantages of a global factory
architecture are well accepted, the disadvantages not so much. By taking into account both simultaneously, we strive to bridge the
gap between theory and practice.
18
References
Ahouse, J.J., Bruderer, E., Gelover-Santiago, A., Ko nno, N., Lazer, D., Veretnik, S., 1991. In: Nade l, L., Stein, D. (Eds.), Reality Kisses the Neck of Speculation: A Report
From the NKC Workgroup, pp. 331
–
353.
Alkemade, F., Frenken, K., Hekkert, M.P., Schwoon, M., 2009. A complex systems methodology to transition management. J. Evol. Econ. 19 (4), 527
–
543.
Ambos, T.C., Birkinshaw, J., 2010. Headquarters' attention and its e
ff
ect on subsidiary performance. Manag. Int. Rev. 50 (4), 449
–
469.
Anderson, P., 1999. Complexity theory and organization science. Organ. Sci. 10, 216
–
232.
5
For instance, the landscape is generated
‘
on-the-
fl
y
”
(Ganco and Hoetker, 2009) by populating the matrix (see Fig. 7) gradually only for the solutions that can be
visited at any gi ven time.
6
See also http://www.cs.unibo.it/~
fi
oretti/CODE/NK/.
19
Ashmos, D.P ., Duchon, D., McDa niel Jr., R.R., 2000. Organizational responses to complexity: the e
ff
ect on organi zational performance. J. Organ. Chang. Manag. 13 (6),
577
–
595.
Baum, J., 1999. Whole-part coevoluti onary competition in organizations. In: B aum, J., McKelvey, B. (Eds.), Variati ons in Organization Science: In Honour of Donald T.
Campbell. Sage, Thousand Oaks, CA, pp. 113
–
135.
Bertrand, O., 2011. What goes around comes around: e
ff
ects of o
ff
shore outsourcing on the export performance of
fi
rms. J. Int . Bus. Stud. 42 (2), 334
–
344.
Billinger, S., Stieglitz, N., Schumacher, T.R., 2014. Search on rugged landscapes: an experimental study. Organ. Sci. 25 (1), 93
–
108.
Boisot, M., Child, J., 1999. Organizations as adaptive systems in complex environments: the case of China. Organ. Sci. 10 (3), 237
–
252.
Boisot, M., McKelvey, B., 2011. Complexity and organization
—
environment relations: revis iting Ashby's law of requisite variety. In: The Sage Handbook of Complexity
and Management, pp. 279
–
298.
Bouquet, C., Birkinshaw, J., 2008. Weight versus voice: how foreign subsidiaries gain attention from corporate headquarters. Acad. Manag. J. 51 (3), 577
–
601.
Brown, S.L., Eisenhardt, K.M., 1997. The art of continuous change: linking complexity theory and time-paced evolution in relentlessly shifting organizations. Adm. Sci.
Q. 42 (1), 1
–
34.
Buckley, P., 2009. The impact of the global factory on economic development. J. World Bus. 44 (2), 131
–
143.
Buckley, P., 2010. The role of headquarters in the global factory. In: Ande rson, U., Holm, U. (Eds.), Managing in the Contempo rary Multinati onal
–
The Role of
Headquarters. Edward Elgar, Cheltenham, UK, pp. 60
–
84.
Buckley, P., 2011. International integration and co ordination in the global factory. Manag. Int. Rev. 22 (3), 227
–
245.
Buckley, P., Casson, M., 2001. Strategic complexity in international busi ness. In: Rugman, A.M., B rewer, T.L. (Eds.), The Oxford Handbook of Internati onal Business.
Oxford University Press, Oxford, pp. 88
–
126.
Buckley, P.J., Ghauri, P.N., 2004. Globalisation, economic geography and the strategy of multinational enterprises. J. Int. Bus. Stud. 35 (2), 81
–
98.
Caldart, A.A., Oliveira, F., 2010. Analysing industry pro
fi
tability: a
“
complexi ty as cause
”
perspective. Eur. Manag. J. 28 (2), 95
–
107.
Celo, S., Nebus, J., Wang, I.K., 2015. MNC structure, complexity, and performa nce: insights f rom NK method ology. J. Int. Manag. 21 (3) , 182
–
199.
Claussen, J., Kretschmer, T., Stieglitz, N., 2014. Vertical scope, turbulence, and the bene
fi
ts of commitme nt and
fl
exibility. Manag. Sci. 61 (4), 915
–
929.
Cohen, W.M., Levinthal, D.A., 1990. Absorptive capacity: a new perspective on learning and innovation. Adm. Sci. Q. 35 (1), 128
–
152.
Contract or, F.J., Kumar, V., Kundu, S.K., Pedersen, T., 2010. Reconcept ualizi ng the
fi
rm in a world of outsourcing and o
ff
shoring : the organizational and geographical
relocation of high-value company functions. J. Manag. Stud. 47 (8), 1417
–
1433.
Coucke, K., Sleuwaegen, L., 2008. O
ff
shoring as a survival strategy: evidence from manufacturing
fi
rms in Belgium. J. Int. Bus. Stud. 39 (8), 1261
–
1277.
Davis, J.P ., Eisenhardt, K.M., Bingham, C.B., 2007. Develo ping theory through simulation methods. Acad. Manag . Rev. 32 (2), 480
–
499.
Dellestrand, H., Kappe n, P., 2011. Headqu arters allocati on of resources to innovation transfer projects within the multinational enterprise. J. Int. Manag. 17 (4),
263
–
277.
Di Gregorio, D., Musteen, M., Thomas, D.E., 2008. O
ff
shore outsourcing as a source of international competitiveness for SMEs. J. Int. Bus. Stud. 40 (6), 969
–
988.
Ethiraj, E., L evinthal, D., 2004. Bounded r ationality and the search for organizational architecture: an evolutionary pe rspective on the design of organizations and thei r
evolvability. Adm. Sci. Q. 49 (3), 404
–
437.
Fioretti, G., 2013. Agent-based simulation models in organization science. Organ. Res. Methods 16 (2), 227
–
242.
Fleming, L., Sorenson, O., 2001. Technolo gy as a complex adaptive s ystem: evidence from patent data. Res. Policy 30 (7), 1019
–
1039.
Frenken, K., 2000. A complexity approach to innova tion networks. The case of t he aircraft industry (1909
–
1997). Res. Policy 29 (2), 257
–
272.
Frenken, K., 2006. A
fi
tness landscape a pproach to tech nological complexity, m odularity, and vertical disinteg ration. Struct. Chang. Econ. Dyn. 17 (3), 288
–
305.
Ganco, M., Agarwal, R., 2009. Perf ormance di
ff
erentials between diversifying entrants and entrepreneurial start-ups: a complexity approach. Acad. Manag. Rev. 34 (2),
228
–
252.
Ganco, M., Hoetke r, G., 2009. NK mode ling methodo logy in the strategy literature: bounded sea rch on a rugged landscape. Res. Method ol. Strateg. Manag. 5, 23 7
–
268.
Gavetti, G., Levinthal, D., 2000. Looking f orward and look ing backward: cognitive a nd experiential se arch. Adm. Sci. Q. 45 (1), 113
–
137.
Ghemawat, P., Levinthal, D., 2008. Choice interactions and business strategy. Manag. Sci. 54 (9), 1638
–
1651.
Ghoshal, S., Bartlett, C.A., 1990. The multinational corporation as an interorganizational network. Acad. Manag. Rev. 15 (4), 603
–
626.
Gulati, R., Lawrence, P.R., Puranam, P., 2005. Adaptation in vertical relationships. Strateg. Manag. J. 26 (5), 415
–
440.
Gupta, A.K., Govindarajan, V., 2000. Knowledge
fl
ows within multinational corporations. Strateg. Manag. J. 21 (4), 473
–
496.
Hall, J., Matos, S., Silvestre, B., 201 2. Understandin g why
fi
rms should invest in sustainable supply chains: a complexity approach. Int. J. P rod. Res. 50 (5), 1332
–
1348.
Kapoor, R., Adne r, R., 2012. What
fi
rms make vs. what they k now: how
fi
rms' production and knowledge bou ndaries a
ff
ect competitive advantage in the face of
technological change. Organ. Sci. 23 (5), 1227
–
1248.
Kau
ff
man, S.A., 1993. The Origin of Order: Sel f-organization and Selecti on in Evolution. Oxford University Press.
Kau
ff
man, S., 1995. At Home in the Universe. Oxfo rd University Press.
Kau
ff
man, S.A., Johnsen, S., 1991. Coevolution to the edge of chaos: couple d
fi
tness landscapes, poise d states, and coevolutionary avalanches. J. Theor. Biol. 149 (4),
467
–
505.
Kedia, B.L., Lahiri, S., 2007. International outsourcing of services: a partnership model. J. I nt. Manag. 13 (1), 22
–
37.
Kotabe, M., 1989.
“
Hollowing-out
”
of US multinationals and their global competitiveness: an intra
fi
rm perspective. J. Bus. Res. 19 (1), 1
–
15.
Kotabe, M., Mol, M.J., Ketkar, S., 2008. An evolution ary stage model of outsourcing and competence des truction: a triad comparison of the consumer electronics
industry. Manag. Int. Rev. 48 (1), 65
–
94.
Larsen, M.M., Manning, S., Pedersen, T., 2013. Uncovering the hidden cost s of o
ff
shoring: the inte rplay of complexity, organizational de sign, and experie nce. Strateg.
Manag. J. 34 (5), 533
–
552.
Lave, C., March, J.G., 1975. An Introduction to Models in the Soci al Sciences. Harpe r & Row, New York.
Law, A.M., Kelton, D.W., 19 91. Simulation Modeling and Analysis, 2nd ed. M cGraw-Hill, New York.
Lazer, D., Friedman, A., 2007. The network structure of exploration and exploitation. Adm. Sci. Q. 52 (4), 667
–
694.
Levinthal, D.A., 1997. Adaptation on rugged landscapes. Manag. Sci. 43 (7), 934
–
950.
Li, D., Eden, L., Josefy, M., 2017. Agent and task complexity in multilateral alliances: the safeguarding role of equity governance. J. Int. Manag. 23 ( 3), 227
–
241.
Macher, J.T., Boerner, C., 2012. Technological development at the b oundari es of t he
fi
rm: a kn owledge-based examination in drug development. Strateg. Manag. J. 33
(9), 1016
–
1036.
Madslie n, J., 20 10. Dreamliner's journey from crisis to
fl
ight. BBC newshttp://www.bbc.co.uk/news/business-10435529.
Malan, K.M., Engelbrecht, A.P., 2013. A survey of techniques for characterising
fi
tness landsca pes and some possible ways f orward. Inf. S ci. 241, 148
–
163.
March, J.G. , Simon, H.A. , 1958. Organizations. John Wiley & Sons, New York etc.
McKelvey, B., 1999. Avoiding complexity catastrophe in coe volutionary pockets: strategies for rugged landscapes. Organ. Sci. 10 (3), 294
–
321.
Mecham, M., 2003. Betting on supplie rs. The 7E7 is all about reducing costs, as Boeing's choice of how and where to build it may res hape the compa ny. In: Aviation
Week & Space Technology. 159(17). pp. 51
–
54.
Mol, M.J., Pauwels, P., Matt hyssens, P., Quintens, L., 2004. A technological contingen cy perspective on the de pth and scope of international outsourcing. J. I nt. Manag.
10 (2), 287
–
305.
Mudambi, S.M., Tallman, S., 2010. Make, buy or ally? Theor etical perspecti ves on knowledge process outsourcing through alliances. J. Manag. Stud. 47 (8),
1434
–
1456.
Nieto, M.J., Rodríguez, A., 2011. O
ff
shoring of R&D: looking abroad to improve inn ovation performance. J. Int. Bus. Stud. 42 (3), 345
–
361.
Nohria, N., Ghoshal, S., 1997. The Di
ff
erentiated Network: Organizing Multinational Corporations for Value Creation. Jossey-Bass Publishers.
Norris, G., 2010. Tail woes. In: Aviation Week & Space Tec hnology. 172(25). p p. 33.
Peterson, K., 2011. Special report
—
a wi ng and a p rayer: outs ourcing at Boe ing. Reutershttp://www.reuters.com/article/2011/01/20/us-boeing-dreamliner-
idUSTRE70J2UX20110120.
2
Pfe
ff
er, J., Salancik, G.R., 1978. The External Control of Organisations. (New Y ork).
Quinn, J.B., 1999. Strategic outsourcing: leveraging knowledge capabilities. MIT Sloan Manag. Rev. 40 (4), 9
–
21.
Rivkin, J.W., 2000. Imitation of comple x strategies. Manag. Sci. 46 (6), 824
–
844.
Rivkin, J .W., 2001. Reprodu cing knowledge: replicati on without imitati on at moderate complexit y. Organ. Sci. 12 (3), 274
–
293.
Rivkin, J.W., Siggelkow, N., 2003. Balancing search and stability: interdependencies among elements of orga nizational design. Manag. Sci. 49 (3), 290
–
311.
Rivkin, J.W., Siggelkow, N., 2007. Patterned interactions in complex systems: implications for exploration. Manag. Sci. 53 (7), 1068
–
1085.
Schmeisser, B., 2013. A systematic review of literature on o
ff
shoring of value chain activities. J. Int. Manag. 19 (4), 390
–
406.
Simon, H.A., 1957. Models of Man: Social and Rational. John Wiley & Sons, New York.
Simon, H.A., 1969. The Sciences of the Arti
fi
cial. MIT Press, Cambridge, MA & London.
Skyttner, L., 2006. Gene ral Systems Theory: Problems, Perspective, Practice. World Scienti
fi
c Publi shing Company, River Edge, NJ.
Slayton, R., Spina rdi, G., 2016. Radical i nnovation in scaling u p: Boeing's Dreamliner and the challenge of socio-tec hnical transitions. Technovation 47, 47
–
58.
Stewart, M., 2001. The Coevolving Organizatio n. Decomplexity Associates.
Tang, C.S., Zimmerman, J.D., Nelson, J. I., 2009. Managing ne w product deve lopment and supply chain risks: the Boeing 787 case . Supply Chain Forum 10 (2), 74
–
86.
UNCTAD, 2013. W orld Investment Re port 2013: Global Value Chains: Investme nt and Trade for Development. United Na tions, New York.
Venaik, S. , Midgley, D.F., Devinney, T.M., 2004. A new perspective on the integ ration-responsiveness press ures confront ing multinational
fi
rms. Manag. Int. Rev. 44
(1), 15
–
48.
Vidgen, R., Bull, L., 2011. App lications of Kau
ff
man's coevolutionary NKCS m odel to management and organization studies. In: The Sage Handbo ok of Complexity and
Management, pp. 201
–
218.
Vidgen, R., Wang, X., 2006. Fro m business process manageme nt to business process ecosystem. J. Inf. Technol. 21 (4), 262
–
271.
Walker, G., Weber, D., 1984. A transaction cost approach to make-or-buy decisions. Adm. Sci. Q. 29 (3), 373
–
391.
Walters, B.A., Bhuian, S.N., 2004. Complexity absorption and performance: a structural analysis of acute-care hospitals. J. Manag. 30 (1), 97
–
121.
Wasserman, S., Fa ust, K., 1994. S ocial Network Analysis: Methods and Applications. Cambridge University Press.
Weigelt, C., Sarkar, M.B., 2012. Pe rformance implications of outsourc ing for technological innovatio ns: managing the e
ffi
ciency and adaptability trade-o
ff
. Strateg.
Manag. J. 33 (2), 189
–
216.
Wiersema, M.F., Bowen, H.P., 2011. The relationship between internati onal diversi
fi
cation and
fi
rm performance: why it remains a puzzle. Glob. Strateg. J. 1 (1
–
2),
152
–
170.
Wright, S., 1931. Evolution in Me ndelian populations. Genetics 16, 97
–
159.
Wright, S., 1932. The role of m utation, inbreeding, cross-breeding a nd selection in evolution . In: Proceedings Xl I nternational C ongress of Genetics. 1. pp. 356
–
366.
Yamin, M. , 2011. A commentary on Peter Buckley's writing s on the global factory. Manag. Int . Rev. 51 (2), 285
–
293.
