“SIS

 

20‘0 LIBRARY
Michigan State
U liversity

 

 

 

This is to certify that the
dissertation entitled

MODULARITY AND COMPLEXITY: AN EXAMINATION OF
THE EFFECTS OF PRODUCT STRUCTURE ON THE
INTRICACY OF PRODUCTION SYSTEMS

presented by

Delvon B. Parker

has been accepted towards fulﬁllment
of the requirements for the

PhD. degree in Business Administration

 

 

 

Major Professor’s Signature U

Ol’ll-IO

 

Date

MSU is an Aﬂimnative Action/Equal Opportunity Employer

 

o-:-_~9--l-O-I-l-I-IALI-l-A- .—

n.---"--o-n-o-n-I-|--a-O-O-I-I-A-I-u-o-c-o—c-0-c-n-lgLu-Vl-t-Lw‘ﬂ-I.

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5108 K:IProj/Acc&Pres/ClRC/DateDue.indd

 

 

MODULARITY AND COMPLEXITY: AN EXAMINATION OF THE EFFECTS OF
PRODUCT STRUCTURE ON THE INTRICACY OF PRODUCTION SYSTEMS
By

Delvon B. Parker

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHIOLOSOPHY
Business Administration

2010

ABSTRACT

MODULARITY AND COMPLEXITY: AN EXAMINATION OF THE EFFECTS OF
PRODUCT STRUCTURE ON THE INTRICACY OF PRODUCTION SYSTEMS

By
Delvon B. Parker

Applications of product modularity have increased sharply over the last decade,
as more ﬁrms are adopting product modularity as their principal product design
strategy due to its potential to increase the speed of product development in the
face of dynamic competitive environments. Product modulan'ty has also been
advocated an effective means of managing complexity. While this may be the
case for the structure of products themselves, the effects of implementing of a
modular product design strategy on the organizational systems that support the
manufacture of modular products are not as well known. This study represents
an attempt at assessing this impact, as it investigates the effects of product

modularity on the complexity of production systems.

This study applies a systems theoretic framework and Seemingly Unrelated
Regression (SUR) to empirical analysis of the effects of product modularity on
production system complexity. The results suggest the product modularity is
positively associated with the complexity of internal manufacturing processes and
supply networks in situations of high outsourcing and high environmental

uncertainty. The study also provides empirically validated perceptual measures

of product modularity, manufacturing process complexity, organizational

complexity, and supply base complexity.

TABLE OF CONTENTS

List of Tables ............................................................................................. vi
List of Figures ............................................................................................ viii
Chapter 1 - Introduction ............................................................................ 1
Chapter 2 — Literature Review ................................................................... 5
2.1.0 Modularity Research in OSM ................................................. 5
2.2.0 The Relationship Between Modularity and Complexity .......... 8
2.3.0 Composition of the Modularity and Complexity Constructs 12
2.4.0 Chapter Summary .................................................................. 19
Chapter 3 — Theory, Research Model and Hypotheses ............................. 20
3.1.0 Theoretical Basis .................................................................... 20
3.2.0 Research Models ................................................................... 22
3.4.0 Chapter Summary .................................................................. 34
Chapter 4 — Research Design .................................................................... 35
4.1.0 Questionnaire Design and Pretesting ..................................... 35
4.2.0 Target Population and Sampling Frame ................................. 38
4.3.0 Survey Administration ............................................................ 41
4.4.0 Response Statistics and Sample Characteristics ................... 43
4.5.0 Test for Non-Response Bias .................................................. 43
4.6.0 Sample Characteristics .......................................................... 44
4.7.0 Management of Missing Data ................................................. 47
4.8.0 Chapter Summary .................................................................. 47
Chapter 5 — Measurement ......................................................................... 49
5.1.0 Product Modularity ................................................................. 49
5.2.0 Complexity ............................................................................. 54
5.3.0 Chapter Summary .................................................................. 66
Chapter 6 - Hypothesis Testing ................................................................. 67
6.1.0 Pairing of Responses ............................................................. 67
6.2.0 Seemingly Unrelated Regression ........................................... 68
6.3.0 Preliminary Analysis ............................................................... 70
6.4.0 Results ................................................................................... 71
6.5.0 Chapter Summary .................................................................. 82

Chapter 7 — Discussion .............................................................................. 84

7.1.0 Measurement Outcomes ........................................................ 84
7.2.0 Results and Conclusions of Hypothesis Testing .................... 89
7.3.0 Contributions to Business Theory and Practice ...................... 93
7.4.0 Limitations and Directions for Future Study ............................ 95
Appendix A - Research Questionnaire ...................................................... 100
Appendix B — Sample Invitation and Follow-up Letters .............................. 104
Appendix C - Sample Breakdown by Various Categories ......................... 107
Appendix D - Technical CFA ..................................................................... 108
Appendix E - Managerial CFA ................................................................... 109
Appendix F - SEM Analysis of Exploratory Model ..................................... 111
Bibliography ............................................................................................... 1 13

LIST OF TABLES

Table 2.1 Summary of Product Modularity Literature in OSM ................... 11
Table 2.2 Dimensions of Key Constructs .................................................. 19
Table 4.1 Perceptual Scales of Modularity and Complexity in OSM ......... 36
Table 4.2 Response by Category .............................................................. 43
Table 4.3 Sample Characteristics ............................................................. 45
Table 5.1 CFA-O1 ...................................................................................... 50
Table 5.2 CFA-02 ...................................................................................... 53
Table 5.3 CPA-03 ...................................................................................... 57
Table 5.4 CPA-04 ...................................................................................... 59
Table 5.5 CPA-05 ...................................................................................... 62
Table 5.6 CFA-06 ...................................................................................... 65
Table 6.1 Descriptive Statistics for Model Variables ................................. 70
Table 6.2 SUR Exploratory Model ............................................................. 72
Table 6.3 SUR Contingency Model - Closed System ............................... 74
Table 6.4 SUR Contingency Model - Open System ................................. 76
Table 6.5 SUR Moderation Model - Low Uncertainty ............................... 79
Table 6.6 SUR Moderation Model — High Uncertainty .............................. 80
Table 6.7 Summary of Hypothesis Testing ............................................... 83
Table A.1 Research Questionnaire — Technical Section ........................... 100

vi

Table A2 Research Questionnaire - Managerial Section ........................ 101

vii

LIST OF FIGURES

Figure 3.1 Overarching Research Framework .......................................... 21
Figure 3.2 Exploratory Model .................................................................... 27
Figure 3.3 Contingency Model .................................................................. 31
Figure 3.4 Moderation Model .................................................................... 33
Figure 5.1 Manufacturing Process Complexity MIMIC Model ................... 56
Figure 5.2 Organizational Complexity MIMIC Model ................................. 61
Figure 5.3 Supply Base Complexity MIMIC Model .................................... 64

viii

1.0.0 Introduction

Business environments in the 21 st century are becoming increasingly
challenging due to global competition, rapid advancements in technology,
growing pressure to reduce product development cycle times, and the
emergence of more sophisticated and demanding customers. The concept of
modularity has been advanced as one effective strategy to offset some of the
increasing complexity faced by businesses in this era of frequent and rapid

change.

Modularity refers to the hierarchic structure of a system whereby it is composed
of smaller subsystems that can be designed independently, yet function together
as a whole (Baldwin and Clark, 1997). Modularity has been credited with
enabling the innovation and commercialization surge experienced by the
computer industry over the last three decades (Baldwin and Clark, 2000), and its
rapid diffusion into other industries such as automotive, consumer electronics,
home appliances, and even services has begun to receive the increased
attention of top managers, analysts and researchers (Andrews, 2000; Worren,
Moore and Cardona, 2002; Salz, 2006; Hagel and Brown, 2008). Although
modularity is not a new concept (it is over 40 years old), the recent increase in its
deployment raises new questions regarding the broader implications of its
application. This study is focused on understanding the effects of product
modularization beyond its inﬂuence on the product itself, to its impact on the

organizational systems that support the manufacture of modular products.

Complexity refers to the presence of many, varied parts, patterns or elements,
connected in a way that makes an object difﬁcult to understand fully (Baldwin and
Clark, 2000). Extant literature suggests that a relationship exists between
modularity and complexity at the product level (Simon, 1962; Ulrich, 1995;
Salvador et al., 2002) and also at the organizational level (Orton and Weick,
1990; Galunic and Eisenhardt, 2001; Schilling and Steensma, 2001), but few
studies have examined the relationship between modularity and complexity
across organizational levels. The research presented here begins to address

this void by investigating the following research questions:

1) Does employing modularity at the product level impact the level of
complexity of supporting production systems?
2) What factors may impact how modularity affects production system

complexity?

A production system is herein deﬁned as a ﬁrm’s internal manufacturing process,

its organizational structure, and its supply base.

Consistent with the heritage of rigorous academic research on modularity and
complexity, a systems perspective has been adopted for this study. Systems
Theory draws on the notion that both man-made and naturally—occurring systems

have common characteristics (Boulding, 1956). This allows for the application of

insights drawn from the study of natural systems to the examination of other
organizational phenomena. In this study, elements of Systems Theory are
applied to the understanding of product modularity’s inﬂuence on production

system complexity.

The research presented here makes a number of signiﬁcant contributions to the
Operations and Supply Chain Management body of knowledge and to business
practice. First, to support the examination of the phenomena of interest, this
research provides empirically validated perceptual measures of product
modularity, manufacturing process complexity, organizational complexity, and
supply base complexity. Perceptual measures of modularity and complexity
have been called for in prior studies of these concepts (Choi and Krause, 2006;
Salvador, 2007). Additionally, the results provide insight regarding the
applicability of systems theoretic perspectives in predicting structural
characteristics of manufacturing systems. Finally, the results provide insight to
managers regarding the system consequences of the choice to modularize

products.

This study is organized as follows: Chapter 2 provides a review of the literature
pertinent to this investigation, followed by a discussion of the theoretical
framework applied to this analysis, and the presentation of hypotheses in
Chapter 3. Chapter 4 provides detail regarding the research design and data

analysis is covered in Chapter 5 (Measurement Analysis) and Chapter 6

(Hypothesis Testing). The study concludes with a discussion of the
contributions and limitations of this research, as well as recommendations for

future study in Chapter 7.

2.0.0 Literature Review

The literature reviewed in this Chapter covers three topics: 1) modularity research
in Operations and Supply Chain Management; 2) the relationship between
modularity and complexity; and 3) the composition of the modularity and

complexity constructs.

2.1.0 Modularity Research in Operations and Supply Chain Management
While much of the literature on modularity and modular systems resides in the
engineering domain, there is a small, but solid base of research on the topic that
has a high degree of relevance to Operations and Supply Chain Management
(OSM). This section segments these studies by method of analysis (i.e.,
conceptual, empirical, analytical and simulation), and discusses their

contributions to the OSM body of knowledge.

2.1.1 Conceptual Foundation

Starr (1965) introduced the concept of Modular Production to the academy. His
work suggested that the production process be broken up into two distinct
process types: transformational processes and assembly processes.
Transformational processes deal with the conversion of raw materials into
product components, while assembly processes deal with the connection of
product components to produce a ﬁnished product. He suggested that
segmenting the process in this way allowed for greater emphasis to be be placed

on producing components that could be combined into a variety of ﬁnished

product conﬁgurations. This introduction to modular production represented a
signiﬁcant departure from the typical mass manufacturing logic that dominated

the era.

The interest of the OSM research community in modularity was reignited nearly
three decades later as studies by Langlois and Robertson (1992) and Ulrich
(1995) suggested that product modularity could facilitate part standardization,
which enables a potential reduction in the complexity of purchasing, inventory
management, quality control and ﬁeld service. Interest in the concept continued
to gain momentum as Fine (1998) and Baldwin and Clark (2000) added more
reﬁned theoretical perspectives through their discussion of the interrelated nature
of product structures and organizational task structures. They suggest that the
concurrent design of products, manufacturing processes and supply networks is
crucial in order to effectively produce and extract maximum value from modular

products.

2.1.2 Empirical Analysis

Empirical studies on modularity in OSM have primarily focused on the effects of
product modularity on ﬁrm performance. Worren et al. (2002) found an indirect
relationship between product modularity and ﬁrm performance (mediated by
model variety) in their study of the US and UK appliance industries, while Lin
(2003) failed to detect a signiﬁcant relationship between product modularity and

ﬁrm performance in his analysis of computer and electronics manufacturers in

Taiwan. Signiﬁcant relationships between product modularity and delivery,
ﬂexibility, and customer service were, however, detected by Antonio et al. (2007)
in their analysis of manufacturers of electronics, toys, and plastics in Hong Kong.
These mixed ﬁndings suggest that there is much more to be learned about the
intermediary mechanisms that link modular product architecture to ﬁrm
performance. Other empiIical studies on modularity in OSM have sought to
establish modularity’s ability to facilitate a mass customization strategy (Duray et
al., 2000; Tu et al., 2004), and found strong evidence to illustrate its performance

in this application.

2.1.3 Analytical and Simulation Modeling

Analytical and simulation studies examining modularity within OSM contexts
have been somewhat more scarce. Ernst and Kamrad (2000) develop an
analytical model to illustrate the value of coupling product modularity and
logistics postponement. The model is used as a basis for a static simulation of
three cases which demonstrate the costs beneﬁts of pursuing the two strategies
simultaneously. Fine et al. (2005) use a goal programming approach to
investigate whether complementarity exists between product structures and
supply chain structures. Results suggest that modular product - modular supply
chain combinations, as well as integral product — integral supply chain
combinations outperform mixed structure combinations (i.e., modular product —

integral supply chain, and integral product — modular supply chain combinations).

Although the studies reviewed here broadly establish that product modularity
may have an impact on the structure and performance of production systems,
explicit tests of modularity effects on structural and operational Characteristics of

manufacturing systems have yet to be performed.

2.2.0 The Relationship between Modularity and Complexity

Throughout the operations and organizational management literature on
modularity, there exists the assumption that the modularity-based strategies
reduce complexity (Ethiraj & Levinthal, 2004). Baldwin and Clark (2000), in their
retrospective of the evolution of the computer industry, suggest that
modularization reduces the structural complexity of products in the design phase
by minimizing the number of interdependencies among product components.
Langlois (2002) implies that lower levels of structural complexity in modular
products results from the reduction in the number of components from which they
are comprised. This rationalization of parts is credited to the consolidation of
lower level components into pre-assembled subsystems. Pil and Cohen (2006),
on the other hand, suggest that it is the perceived complexity of a product that is
actually lowered through the use of modular design, in that its readily
decomposable structure makes it easier to ascertain how the product's

conﬁguration supports and enables its functionality.

Product modularity is also posited to reduce managerial complexity. Sanchez

(1995) and Sanchez and Mahoney (1996) explain that a great deal of the

coordination required for product development is embedded in the structure of a
modular product through the use of standardized interfaces. These interface
standards set forth speciﬁc criteria required for the integrating of a component
into the ﬁnished product, and therefore reduce the need for coordination between
component designers during product development. Ulrich (1995) suggests that
the elimination of these task interdependencies allows for a more systematic
approach to the management of product development projects. He further
suggests that the reuse of common components, enabled by employing a
modular product structure, simpliﬁes the supply-side tasks of purchasing and
quality control. This sentiment is echoed from the demand side by Feitzinger and
Lee (1997) who suggest that product modularity, coupled with a postponement
strategy, simpliﬁes product customization, inventory management and market

mediation.

Perhaps the most pervasive view of modularity’s effect on complexity considers
the adaptable nature of modular products, and how this adaptability enables a
rapid product response from ﬁrms competing in dynamic business environments.
Langlois and Robertson (1992) and Garud and Kamuraswamy (1995) explain
that in dynamic environments, the cognitive complexity facing managers
increases under the constant pressure to innovate and adapt. They suggest that
the modularization of products enables components to be more easily
outsourced, thereby reducing the scope of internal development and production

activities. The resultant external development capability allows for potentially

faster and broader component innovation, through competition in the supply
market (Fine, 2000), with marginal or no increase in internal operation

complexity.

Management scholars laud the value of organizational modularity in coping with
environmental complexity. In their study of 10 divisions of a Fortune 100 high-
technology ﬁrm, Galunic and Eisenhardt (2001) show that organizational
modularity facilitates the reorganization of ﬁrm resources to quickly respond to
emerging opportunities, as well as to stave off competitive threats and shore up
performance slippage in active operations. Also, Schilling and Steensma (2001)
demonstrate in an empirical analysis of data on 330 manufacturing industries that
modular forms of organization are a common response to high levels of
heterogeneity and technological change in a variety of competitive and complex

business environments.

More recently, however, researchers are beginning to challenge the assumption
that modularity-based strategies reduce complexity. Salvador (2007) points out
that complexity arises as a result of a system’s ability to take on many
conﬁgurations, and Staudenmayer et al. (2005) discovered in their case studies
of seven ﬁrms from various industries that employing modularity in products
increased the complexity of product development activities considerably. These
observations suggest that some ambiguity may exist regarding modularity’s

effect on complexity. The purpose of this study is to provide further insight into

10

the relationship between modularity and complexity by performing careful

measurement of these constructs, and executing tests of the relationship

between product modularity and production system complexity in various

competitive contexts.

Table 2.1 below offers a summary of the literature reviewed here, as well as

some other key modularity research themes in Operations and Supply Chain

Management.

Table 2.1 Summary of product modularity literature relevant to Operations and Supply

Chain Management

 

Key Themes

References

 

Modularity in products is a key enabler of
mass customization

Starr (1965); Duray et al. (2000); Tu et al.
(2004)

 

Product modularity has a positive effect on ﬁrm
performance

Worren et al. (2002); Lin (2003); Antonio et al.
(2007)

 

Product modularity accelerates innovation

Henderson and Clark (1990); Langlois and
Robertson (2002); Sanchez and Mahoney
(1996); Baldwin and Clark (1997, 2000); Fine
(2000)

 

Product modularity facilitates the design and
coordination of supply chains

Fine (2000); Fixson (2005); R0 at al. (2007)

 

Product modularity reduces product complexity

Baldwin and Clark (1997, 2000); Langlois
(2002); Pil and Cohen (2006)

 

 

Product modularity reduces management
complexity in new product development

Sanchez (1995); Ulrich (1995); Baldwin and
Clark (2000)

 

 

11

 

Table 2.1 (cont.)

 

 

Product modularity may increase complexity in Salvador et al. (2002); Staudenmayer et al.
operations and supply management (2005)

 

 

2.3.0 Composition of the Modularity and Complexity Constructs

2.3.1 Modularity

Modularity refers to the hierarchic structure of a system whereby it is composed
of smaller subsystems that can be designed independently, yet function together
as a whole (Baldwin and Clark, 1997). Each dimension of the modularity

construct will be introduced and discussed in the following sections.

Functional Binding

Functional Binding refers to the compartmentalization of the designed functions
of a system. If a system exhibits high functional binding, each designed function
is conﬁned to a minimal number of components, with one-to-one correspondence
between designed functions and system components representing an ideal
condition. Designed functions refer to those functions which are intentionally
included in the system, and can be isolated, controlled and modiﬁed by the
designers of the system.1 The term functional binding is credited to Stevens,
Myers and Constantine (1974), but was reintroduced to management literature

by Salvador (2007).

 

1 This Classiﬁcation excludes emergent functions (i.e., anomalies, glitches, etc.)
which are unintended and occur as a result of unplanned interaction of designed
functions.

12

 

Functional binding underlies modularity in that its presence suggests a system
that is partitioned in a way that reduces the interdependence of components and
allows for their independent design. Although different terms are used to
describe the concept (e.g., architecture; function to component allocation;
functional mapping; module partitioning), the underlying meaning remains
consistent, and it is acknowledged as a critical dimension of modularity by a
number of seminal works in the area (Henderson & Clark, 1990; Ulrich, 1995;

Baldwin & Clark, 1997; Fixson, 2005).

Several attempts to measure modularity using perceptual scales have included
items that tap the functional binding dimension. Duray et al. (2000) do so by
inquiring about the interchangeability of product features and options, while
Worren et al. (2002) and Antonio et al. (2007) ask questions regarding the ability
of designers to make Changes in key product components without changing
others. Also, Duray et al. (2000) and Tu et al. (2004) assess functional binding
by including items addressing the ability to directly add features and options

through the addition of a module.

Interface Standardization
Interface standardization is the presence of common, previously agreed upon
conduits for interaction between the complementary components of a system.

Some characteristics that are typical of standardized interfaces are: 1) they tend

13

to remain unchanged over extended periods of time; 2) their speciﬁcations are
widely known among stakeholders of the system components to be connected;
and 3) they enable the open interchange of complementary system components.
Interface standardization has also been widely recognized by scholars as
undergirding the modularity construct (Garud & Kumaraswamy, 1995; Ulrich,

1995; Sanchez & Mahoney, 1996; Baldwin & Clark, 1997).

Of the studies that used perceptual measures of modularity, only two have
included items that tap the interface standardization dimension. Lin (2003)
simply asks respondents to indicate the degree to which their product’s interfaces
are standardized, whilst Worren et al. (2002) attempt to assess the level of
interface standardization present in organizations by inquiring about the extent to

which business processes are standardized across departments.

Decomposability

Decomposability describes the ease with which a system can be separated into
its constituent parts. It also implies that the system can be readily reconstructed
from the same, similar, or supplementary components without necessarily
suffering signiﬁcant performance deterioration. Decomposability is the aspect of
modularity most responsible for its ﬂexibility, as it makes component swapping (a

key beneﬁt of modularity) practical.

14

Of the three dimensions of modularity, decomposability is the one most
frequently represented in its perceptual measurement. Items tapping this
dimension ask directly about the separability of products (Worren et al., 2002;
Antonio et al., 2007) or the ability to recombine or reassemble products (Lin,
2003; Tu et al., 2004). Tu et al. (2004) also address the decomposability of

manufacturing processes by posing questions about their adjustability.

2.3.2 Complexity

Complexity is the presence of many, varied parts, patterns or elements,
connected in a way that makes an object difﬁcult to understand fully (Baldwin and
Clark, 2000). In the sections below, each of the proposed complexity dimensions

are deﬁned and discussed.

Number of Components

There is wide consensus among scholars of many disciplines that the number of
components of a system contributes to its complexity (Simon, 1962; Kauffman,
1995; Choi & Krause, 2006). The inclusion of this dimension in the complexity
construct can be grounded via the concept of bounded rationality - the notion
that there are constraints to the information processing capabilities of decision
makers, and that their behavior in response to receiving additional information is
markedly less rational (Simon, 1957). As the number of components of a system
increases, the amount of information on hand also increases, approaching or

surpassing the information processing limits of the decision maker. This inhibits

15

the decision makers ability to accurately perceive the system, thereby making it

appear more complex.

Many studies in Operations Management (OM) that have employed perceptual
measures of complexity have captured the number of components dimension. At
the product level, Sawhney and Piper (2002) asked respondents to report the
number of layers in printed circuit boards as one component of a three-item
complexity measure, and Devaraj, Hollingworth and Schroeder (2004) capture
the number of product lines in respondents’ manufacturing plants as part of their
complexity measure. At the market and supply chain levels, Flynn and Flynn
(1999) include the number of customers served in their measure of
manufacturing environment complexity and Vachon and Klassen (2002) used an
estimate of supply network size as a component of their measure of supply chain

complexity.

Heterogeneity

Heterogeneity refers to the presence of characteristic differences in the units that
make up a system. For physical systems, this could refer to differences in the
geometry, materials of construction, or functions of units; and for non-physical
systems, heterogeneity could refer to differences in objectives, procedures or
protocols of system components. Like the number of components, heterogeneity
is also broadly accepted as a dimension of complexity (Thompson, 1967; Child,

1972; Holland, 1995) and can be conceptually linked through the concept of

16

bounded rationality. As diversity in system components increases, the types of
information to be processed by the decision maker also increase. This expanded
scope of information present in the system may exceed the processing
capabilities of the decision maker, making the system more difﬁcult to perceive

accurately and therefore more complex.

The representation of heterogeneity in perceptual measures of complexity in
Operations Management (OM) studies has been relatively low, appearing in only
two identiﬁed studies. Flynn and Flynn (1999) used averages of multi-item
scales measuring labor diversity, output diversity, manufacturing diversity and
process diversity in their manufacturing environment complexity construct and
Vachon and Klassen (2002) used a measure of product variety to capture

heterogeneity in their supply chain complexity construct.

Interdependence

Interdependence refers to the level of connectedness among the components of
a system, as indicated by the number of linkages among components and/or the
number of interactions that occur among components. The interdependence
dimension is central to the complexity construct because it is the element that
most reﬂects a system’s potential to exhibit unwieldy behavior. In highly
interdependent systems, a change to one part of the system tends to propagate
throughout the system. This propagation is capable of producing unanticipated

effects (i.e., glitches) and non-linear response, both of which can obscure cause

17

and effect relationships (Stacey, 1995). The resultant causal ambiguity is a key
component of system complexity. This phenomenon has been borne out in a
number of physical experiments and simulations that illustrate the magnitude of
increased interdependence on system complexity (Holland, 1995; Kauffman,

1995).

Interdependence is the least represented of all dimensions of complexity as it
pertains to perceptual measurement in OM applications. The literature search
produced one study that related interdependence to complexity, albeit indirectly.
Lewis (2004) included both complexity and interdependence in a construct
labeled “task demands,” a comparative measure of the additional effort required
to complete one task relative to another. Furthermore, none of the studies
discussed above used all three proposed complexity dimensions in their
measurement. Most scales used one dimension (i.e., number, heterogeneity, or
interdependence) in conjunction with items that measure the level of difﬁculty
associated with an activity or items that directly ask about the complexity of an

object.

Table 2.2 below offers a summary of the proposed dimensions of modularity and

complexity, and references literature that supports the association of each

dimension with its respective construct.

18

Table 2.2: Dimensions of Key Constructs

 

 

 

 

 

 

 

Dimensions of Supporting Literature Dimensions of Supporting Literature
Modularity Complexity
Functional oSalvador (2007) Number of oChoi & Krause (2006)
Binding oFixson (2005) Elements oNovak & Eppinger (2004)
oUlrich (1995) oFlynn & Flynn (1999)
oThom pson (1967)
Interface oSalvador (2007) Heterogeneity oChoi/Krause (2006)
Standardization oBaldwin & Clark (1997) oNovak/Eppinger (2004)
oSanchez & Mahoney oFlynn/Flynn ( 1999)
(1996) oThompson (1967)
oGarud &
Kumaraswamy (1995)
Decomposability oSaIvador (2007) Interdependence oChoi/Krause (2006)
oSChilling (2000) oNovak/Eppinger (2004)
oSimon (1 967)

 

 

2.4.0 Chapter Summary

In this chapter, a review of the literature on modularity research in Operations

and Supply Chain Management was performed, as well as an analysis of the

literature concerning the relationship between modularity and complexity.

Finally, the dimensions of the modularity and complexity were researched, and a

discussion of prior attempts at perceptual measurement of the modularity and

complexity constructs was offered. The next chapter introduces the study’s

theoretical framework and presents the hypotheses to be tested.

19

 

3.0.0 Theory, Research Models and Hypotheses
This chapter provides the theoretical basis for the investigation, introduces each
of three research models to be analyzed, and presents a set of twelve

hypotheses to be tested.

3.1.0 Theoretical Basis

To guide the examination of the effects of product modularity on production
system complexity, a systems theoretic perspective has been adopted. Systems
Theory is relevant for analysis of this phenomenon for a number of reasons.
First, it provides a grounded framework for the examination of multi-Ievel
phenomena — situations where the effect of a Change or stimulus within a
particular locale has implications beyond its deﬁned bounds (Short, Ketchen Jr.,
Palmer, & Hult, 2000). This comes to bear in the current study, as it examines
the effect of structural choices at the product level on the structural
characteristics of the supporting production system. Second, by virtue of its
interdisciplinary nature, Systems Theory allows for importation of concepts and
explanatory mechanisms describing system behavior from a vast repository of
scientiﬁc learning (Boulding, 1956). In the present study, Characteristics of open
and closed systems are applied to the study of complexity in manufacturing
systems. Third, Systems Theory has a history of use in aiding the understanding
of organizational behavior (Thompson, 1967), and recent studies of modularity
and complexity have adopted the systems perspective. Schilling (2000) used

General Systems Theory as a framework to develop a model to explain the

20

motivations for interﬁrrn product modularity, and Baldwin and Clark’s (2000)
study of the transformation of the computer industry (ignited by the use of
modular product architecture) has Complex Adaptive Systems as its theoretical

foundation.

The basic proposition advanced by Systems Theory, as it is operationalized for

this study, is summarized in Figure 3.1 below:

Fi_gure 3.1 Overarching Research Framework

 

Production System Structure and
Operational Characteristics

 

Internal
Manufacturing
Process

 

 

 

 

 

Product Structure Organization

 

 

 

 

 

 

 

 

Supply Base

 

 

 

 

 

 

 

 

 

21

3.2.0 Research Models

This investigation is divided into three separate research models: an exploratory
model, a contingency model, and a moderation model. The exploratory model
seeks to answer the question of whether or not product modularity is associated
with complexity in production systems, while the contingency model is directed
toward understanding what factors determine whether product modularity is
associated with either increased or decreased production system complexity.
Finally, the moderation model examines whether or not the strength of the
proposed relationship between product modularity and production system
complexity is affected by the level of uncertainty in the manufacturing

environment.

3.2.1 Exploratory Model

When modularity is employed at the product level, there are likely to be
consequences for a ﬁrm's intemal manufacturing processes. Some researchers
suggest that manufacturing processes used in the creation of modular products
are simpler. Schilling (2000) argues that product modularity facilitates the
decoupling of manufacturing processes, allowing processing steps to be
performed independently on separate subsystems — a simpler approach in
comparison to the highly coordinated processing requirements for integrated
products. Salvador et al. (2002) suggest that it is the reduction in the number of

required operations owed to the receipt of completed sub-assemblies that

22

simpliﬁes the manufacturing process, while Fixson (2005) argues that the use of
plug-and-play type component interfaces reduces the need for a variety of
approaches to integrating components into ﬁnished products. Other scholars
suggest that production of modular products could make manufacturing
processes more complex. Baldwin and Clark (1997) note that product modularity
can destabilize the manufacturing process, as frequent modiﬁcations are likely
required to produce the larger range of product variants enabled by employing a
modular product architecture, and Nyaga et al. (2007) indicate that the conﬁgure-
to-order (CTO) manufacturing processes associated with modular products often
require increased manufacturing capacity to effectively address the demand for a

broader range of product variants.

While the literature offers somewhat conﬂicting perspectives on the relationship,
it strongly suggests that product modularity inﬂuences the structural and
operational Characteristics of manufacturing processes. As such, the ﬁrst
hypothesis proposes that product modularity impacts the complexity of

manufacturing processes.

H1: Product modularity inﬂuences the complexity of internal

manufacturing processes.

Adopting a modular product architecture may also have an impact on the way an

organization is structured. Henderson and Clark (1990) suggest that the

23

structure of product creating organizations tends to mirror the structure of their
products. If organizations do indeed reﬂect the structures of their products, then
within the context of product modularity, organizations would be expected to
exhibit a more decentralized structure, with authority to set objectives and policy
distributed among the divisions of a ﬁrm. In this situation, it could be argued that
the organization itself would be more complex, due to the presence of many
different policies and business processes across the organization. Alternatively,
Baldwin and Clark (2000) noted IBM’s need to add a powerful central authority
during the design and production of their ﬁrst modular computer system in order
to monitor progress and ensure compliance with established system standards.
Langlois (2002) also highlights the critical need for centralized coordination when
employing modularity in products to ensure that modules are integrated in
manner which extracts maximum value from system production. In this case, the
presence of common policies and standard practices that are consistent with

centralized organizations would suggest simpler, more uniform organizations.

Although there is not unequivocal support in the literature for either perspective
on product modularity’s effect on organizational designs, it suggests that
modularity dictates some organizational response in order to effectively manage
the production and commercialization of products. Hypothesis two examines
whether or not product modularity affects the complexity of manufacturing

organizations.

24

H2: Product modularity inﬂuences the complexity of organizational design

and coordination.

A ﬁrm's supply base is also likely to be affected by the adoption of a modular
product architecture. Modularity allows for the incorporation of a variety of
complementary product components through the use of standardized interfaces,
and while theoretically their use should reduce the number of interdependencies
among firms collaborating to produce modular products (Sanchez and Mahoney,
1996), empirical ﬁndings suggest the contrary. The following was observed in
recent case studies of ﬁrms employing modular product architectures: “While an
environment of interﬁnn (product) modularity should in theory eliminate
interdependencies among ﬁrms. . .interdependencies continually emerged
throughout the product development process, despite attempts to limit them. . .the
quantity and variegated nature of external relationships made their management
exceedingly difﬁcult. . .The sheer complexity was daunting, given both the size of
the external network as well as the number of ties per external collaborator.”
(Staudenmayer, Tripsas, & Tucci, 2005, p. 303) This ﬁnding suggests that

product modularity increases supply base complexity.

Alternatively, product modularity may also reduce the complexity of a ﬁrm’s
supply base. Consider the situation where a firm initially produces a product
component internally. All the individual parts that make up that component must

be procured, and in many cases separately. Now consider the situation where a

25

ﬁrm procures the same component as a modular sub-assembly from a ﬁrst-tier
supplier. In this situation, the focal ﬁrm has reduced the size of their supply base
by as much as n-1 (where n equals number of components in the sub-
assembly)2, and effectively shifted the coordination effort of parts procurement
and perhaps even component development to their supplier. Case studies
performed by Salvador et al. (2002) provide examples of this scenario from the

home appliance and heavy truck industries.

Again, the literature does not provide unequivocal insight regarding product
modularity's inﬂuence on the supply base, however, it does suggest that the
supply base is affected by the deployment of modularity at the product level.
Accordingly, hypothesis three (H3) proposes a relationship between product

modularity and supply base complexity.

H3: Product modularity inﬂuences the complexity of a ﬁrm ’8 supply base.

Hypotheses one (H1) through three (H3) are summarized in Figure 3.2 below.

 

2 . .
Assumes Single sourcmg

26

Figure 3.2: Exploratory Model

 

     

Mfg. Process
Complexity

     

 

H1
Product H2 Organizational
Modularity Complexity
H3

     

Supply Base
Complexity

 

 

 

3.2.2 Contingency Model - Open versus Closed Systems

An aspect of Systems Theory that is relevant to the study of complexity is the
concept of entropy. Entropy is a measure of disorder in a system, and it
increases as the number of conﬁgurations that a system can occupy increases
(Kauffman, 1995). The number of conﬁgurations a system can occupy is
proportional to the number of system elements, the diversity of elements present
in the system (i.e., its heterogeneity), and the number of connections that are
possible among the elements of a system (i.e., its level of interdependence).

From the discussion of the dimensions of complexity in Chapter 2, it can be

27

observed that entropy and complexity are closely related concepts. In fact, prior
studies have even used measures of entropy as indicators of manufacturing
system complexity (Deshmukh, Talavage, & Barash, 1998). This study proposes
that entropic behavior in physical systems can provide critical insights regarding

complexity behavior in production systems.

Katz and Kahn (1978) distinguish between two types of systems based primarily
on their entropic behavior, Closed systems and open systems. Closed systems
are purported to respond to Changes introduced into the system by increasing in
entropy, while open systems respond by either maintaining their current level or
decreasing in entropy. For organizations, “closure" and “openness" can refer to
their span of control (Thompson, 1967), or in an extended enterprise context,
their span of inﬂuence. When viewed in this way, closure resulting in increased
entropy (complexity) and openness leading to decrease in the level of entropy
(complexity) can be reconciled. As a case in point, when manufacturing
organizations grow, the scope of operations that sustain them necessarily grows
— new products are introduced into the portfolio; new technologies are added to
manufacturing processes; new markets emerge that require different marketing
strategies, legal compliance requirements, and so on. To the extent that
organizations strive to maintain these emergent activities ﬁrmly under their span
of control or inﬂuence, operations become more complex. This occurs because

organizations are boundedly rational (Williamson, 1975). On the other hand, if

28

organizations relinquish their authority over certain aspects of operation,

complexity may be limited.

Viewing production systems associated with modular product architectures as
open and Closed systems may provide insight into why conﬂicting views exist in
the literature regarding modularity’s inﬂuence. In the case of a closed system, a
ﬁrm employing product modularity may act as both a “module producer” and a
“system integrator,” retaining responsibility for design and production of system
components, as well as system architecture design, system integration, and
system conﬁguration. Based on the argument presented above, a ﬁrm in this
situation is likely to experience higher levels of complexity. Alternatively, a ﬁrm
resembling an open system would function primarily as “system integrator,”
leaving the design and production of system modules to external suppliers. This
decrease in the scope of operations would likely result in reduced complexity.
Hypotheses four through six represent this contingency view of product

modularity’s inﬂuence on production system complexity.

H4a: In closed systems, product modularity is positively associated with

the complexity of internal manufacturing processes.

H5a: In closed systems, product modularity is positively associated with

the complexity of organizational design and coordination.

29

H6a: In Closed systems, product modularity is positively associated with

the complexity of ﬁrms’ supply bases.

H4b: In open systems, product modularity is negatively associated with

the complexity of internal manufacturing processes.

H5b: In open systems, product modularity is negatively associated with

the complexity of organizational design and coordination.

H6b: In open systems, product modularity is negatively associated with

the complexity of ﬁrms’ supply bases.

Hypotheses 4a through 6b are summarized in Figure 3.3 below.

30

Figure 3.3: Contingency Model

 

     

  

 

 
   

     

 

     

    

 

  

Closed System
Mfg. Process
Complexity
H4a
(+l
Product H5a Organizational
Modularity (4.) Complexity
H6a
(+l
Supply Base
Complexity
Open System
Mfg. Process
Complexity
Mb
(-I
Product H5b Organizational
Modularity (.) Complexity
H6b

I-I

    

Supply Base
Complexity

  

 

 

 

31

3.3.3 Moderation Model — Effects of Environmental Uncertainty

Product Modularity is often Cited as an effective strategy for offsetting the effects
of environmental uncertainty (Langlois and Robertson, 1992; Garud and
Kamuraswamy, 1995). The loosely coupled structure of a modular product
architecture allows a product to be partitioned in such a way that changes to a
segment of a product (i.e., a module) can be executed without signiﬁcantly
altering the design or operational characteristics of the product’s other modules.
These module-level Changes, in most cases, can be made much more rapidly
than the changes required to modify a product with an integral structure (Baldwin
and Clark, 2000). This piecewise innovation capability enables ﬁrms to quickly
respond to customer demand for new product features and incorporate new

technologies.

Since this rapid response is likely to produce the greatest beneﬁt in situations of
high uncertainty, it is proposed that applications of product modularity will be
higher in these situations, therefore causing its effect on production system
characteristics to be more pronounced relative to its effect in low uncertainty

environments.

H7: The effect of product modularity on internal manufacturing process

complexity is stronger under conditions of high environmental uncertainty,

as compared to low environmental uncertainty.

32

 

H8: The effect of product modularity on organizational complexity is
stronger under conditions of high environmental uncertainty, as compared

to low environmental uncertainty.
H9: The effect of product modularity on supply base complexity is
stronger under conditions of high environmental uncertainty, as compared

to low environmental uncertainty.

Hypotheses seven (H7) through nine (H9) are summarized in Figure 3.4.

 

Pwre 3.4: Moderation Model

   
   
   

Environmental
Uncertainty

     

Mfg. Process
Complexity

H7

  

Product
Modularity

 
   

Organizational
Complexity

 

H9

     

Supply Base
Complexity

 

 

 

33

3.4.0 Chapter Summary

This chapter introduced and applied Systems Theory to the analysis of Product
Modularity’s relationship with Production System Complexity. Three research
models were presented: an exploratory model, a contingency model, and a
moderation model. For each model, hypotheses were developed based on
deductive reasoning, literature support, and predictions from theory. The next

chapter discusses the research design and implementation.

34

4.0.0 Research Design

In order to address the research questions presented in Chapter 1, an empirical
study employing perceptual measures of the study’s key variables was chosen
as the most appropriate approach for the following reasons: First, the study’s
intent to span a large cross-section of manufacturing industries requires the use
of a set of generic measures that are applicable across all contexts. Perceptual
measures are well suited for this application because they can be designed
speciﬁcally to extract the required information, while controlling for context-
speciﬁc detail. Second, the study is concerned with capturing the outcomes of
inter-related decisions of individuals at multiple levels of organizations. The
number of decisions involved, their interaction, as well as the random component
present in decision-making makes empirical analysis relatively efﬁcient, as
compared to purely analytical methods. Third, the intent of the study is to test
and extend theory, which requires an assessment of the proposed relationships

as they exist in actual production environments.

This chapter details the methods employed to collect the data required to
address the study's primary research questions. Many of the procedures
employed here were adapted from the recommendations of Dillman, Smyth and

Christian (2009).

4.1.0 Questionnaire Design and Pretesting
The items appearing on the questionnaire were compiled using a search, select,

modify and generate procedure. The search stage of the procedure involved a

35

broad-based search of the terms modularity, modular production, and complexity
on the academic search engines ProQuest and Google Scholar. The search was
narrowed to empirical studies by including the term empirical as well as several
terms associated with statistical analysis methods (i.e., Regression, Structural
Equation Modeling, Factor Analysis, etc.) into the search criteria. Studies
generated by the search were then evaluated for their relevance to Operations
and Supply Chain Management applications by assessing source publication
titles and article content. Cross-referencing of the selected studies was
conducted via bibliographical analysis and the cited reference search feature on
lSl’s Web of Knowledge. This was performed in order to ensure inclusion of
source studies from which the original items were generated, and subsequent
works where measurement items were reused or reﬁned. Recently published
literature reviews (Fixson, 2007; Salvador, 2007; Roth et al., 2008) were also
consulted as a ﬁnal check against potential omissions. The ﬁnal set included
fourteen studies — ﬁve containing perceptual modularity scales, and nine with
perceptual complexity scales. Table 4.1 below lists the sources for candidate

measurement items evaluated for inclusion in the study.

Table 4.1: Perceptual Scales of Modularity and Complexity in OSM Research

 

 

 

 

 

 

Scale Name Source Study
Modularity through fabrication Duray et al. (2000)
Modularity through standardization Duray et al. (2000)
Modular products Worren et al. (2002)
Modular structure Worren et al. (2002)

 

 

36

 

Table 4.1 (cont)

 

Modular processes

Worrent et al. (2002)

 

Product modularity

Lin (2003)

 

Product modularity

Tu et al. (2004)

 

Process modularity

Tu et al. (2004)

 

Product modularity

Antonio et al. (2007)

 

Process structure complexity

Gupta and Lonial (1998)

 

Product line complexity

Gupta and Lonial (1998)

 

Manufacturing environment complexity

Flynn and Flynn (1999)

 

Process/product com plicatedness

Vachon and Klassen (2002)

 

Management system complicatedness

Vachon and Klassen (2002)

 

Product complexity

Sawhney and Piper (2002)

 

Project complexity

Sarin and McDermott (2003)

 

Process structure complexity

Deveraj et al. (2004)

 

Product line complexity

Deveraj et al. (2004)

 

Complexity

Lewis (2004)

 

Project complexity risk

Wallace et al. (2004)

 

 

Information complexity

 

Prasad et al. (2005)

 

Each item appearing in the identiﬁed perceptual scales of modularity and

complexity was then evaluated for potential inclusion in the study. An item was

selected if: (1) it exhibited a high degree of ﬁt with the phenomenon under study;

and (2) it demonstrated consistency with the theoretically derived dimensions of

modularity and complexity. Items failing to meet both criteria were dropped.

Items meeting both of these conditions were retained. Those meeting either of

the two conditions were modiﬁed. New items were generated for cases where

existing items did not provide sufﬁcient coverage of a construct’s theoretical

 

domain or cases where items were not relevant to the types of modularity and

complexity under study.

Items measuring Environmental Uncertainty were adapted from Calantone et al.
(2002). These items represent an abbreviated version of the widely adopted
Miller and Droge (1986) Environmental Uncertainty scale. A single objective item
measuring the percentage of components of a ﬁrm or SBU’s primary products
that are outsourced was included to serve as a proxy for system openness. Also,
objective items capturing: 1) the number of internal manufacturing processes; 2)
the number of divisions; 3) the number of organizational levels; 4) the number of
worldwide locations; and 5) the number of suppliers utilized, were designed for

use in this study.

An initial version of the questionnaire was pretested by a panel of four
manufacturing professionals and three professors of Operations and Supply
Chain Management. Reﬁnement of the questionnaire was performed on the
basis of recommendations this advisory panel. A copy of the questionnaire is

available in Appendix A.

4.2.0 Target Population and Sampling Frame
The primary target for study was US. based producers of technically complex
products. “Technically complex” refers to those products which are composed of

a high number of components that interact in a non-simple way (Simon, 1962),

38

 

and have a high proportion of moving parts (Novak and Eppinger, 2001 ). This
group was targeted because modularity is primarily advocated as a means to
manage the complexity of products, hence the presence of modularity in this
population is likely. However, to ensure that sufﬁcient variability was present in
the sample for employing variance partitioning data analysis techniques, a
secondary population — manufacturers of products which cannot be Classiﬁed as
technically complex, was included. The unit of analysis is a manufacturing unit,
which in some cases was represented by an entire ﬁrm, and in others by a

Strategic Business Unit (SBU).

The sampling frame for the study was generated by compiling contact

information for prospective participants in three broad categories:

Categom 1 — Producers of technically complex products and select Tier-1
suppliers

Category 2 — Individuals afﬁliated with Michigan State University who are
employed by ﬁrms that produce technically complex products

Category 3 - Contacts drawn from databases of professional societies.

For category one, six focal ﬁrms were targeted for participation: 1) a commercial
aircraft manufacturer; 2) an aircraft engine manufacturer; 3) a household
appliance manufacturer; 4) an automobile manufacturer; 5) a manufacturer of

heavy commercial, construction and agricultural equipment; and 6) a

39

manufacturer of diverse commercial, construction and industrial equipment.
Firms 2, 3 and 5 above agreed to participate in the study, and provided contact
information for select Tier-1 suppliers. This effort resulted in a total of 516

contacts.

Contacts in category two were complied by analyzing databases of afﬁliates of
the Department of Marketing and Supply Chain Management at Michigan State
University. For each entry, a subjective judgment was made regarding contact
inclusion based on Characteristics of their ﬁrm's primary products, as identiﬁed by
information available via ﬁrms' websites. This effort resulted in a total of 274

contacts.

For category three, electronic mailing lists were rented from two professional
societies, the Society of Manufacturing Engineers (SME), and the Institute of
Industrial Engineers (ME). The selection criteria applied for the creation of each
list were as follows: (1) NAICS industry classiﬁcations - 334 (Computer and
Electronic Product Manufacturing) and 336 (Transportation Equipment
Manufacturing); (2) Job function - Corporate Executive, Manufacturing
Management, Manufacturing Engineering, Product Design and Development; (3)

Location - United States. This effort produced a total of 1,426 contacts.

In total, 2,216 contacts were compiled. While many of the contacts obtained

were producers/assemblers of technically complex products, many of the Tier-1

40

suppliers from category 1 were not. As mentioned above, these ﬁrms were
included in the sampling frame for the purpose of assuring a sufﬁcient level of
variance in the data for the application of analysis techniques that involve

partitioning of sample variance.

4.3.0 Survey Administration

The survey was administered in two parts with the intent of collecting data at two
organizational levels. Part one addressed technical aspects of the enterprise —
product and process characteristics, while part two covered managerial and
administrative aspects of the enterprise - organizational and supply chain
characteristics. Invitations sent to prospective participants asked individuals to
complete the section that was most relevant to their current job function, and to
supply a referral for an individual within their organization who was qualiﬁed to
answer the remaining section. However, it was suggested that if the responding

individual was qualiﬁed to answer both sections, to proceed in that manner.

The survey was administered online. The questionnaire was translated into
HTML format via templates available through the survey hosting website.
lnvitees were required to enter a password in order to enter the survey.
Questions were randomized within categories in order to reduce the potential for
common method bias. Responses were saved upon completion of the ﬁnal
question, or at the end of a speciﬁed timeframe (48 hours) for which respondents

could exit and then re-enter the survey at the same question.

41

For category one and two contacts (n=790), e-mail invitations sent directly to
prospective participants were the primary mode of contact; however for a portion
of the sample (n=269) invitation letters were sent via US. mail, and for a smaller
portion (n=39) invitations were sent via e-mail to Tier-1 suppliers directly from a
focal ﬁrm participant. Category one supplier contacts received invitation letters
which contained logos and authorized signatures of the ﬁrm that provided the
contact information. In all cases, individuals were directed to the survey website
to complete the questionnaire. A total of two reminder notices were sent to
invitees based on whether or not a response had been recorded between
mailings. Mailings were spaced at a maximum of ﬁve business days apart, and
a minimum of three days apart. All reminder notices were sent via e-mail.

Samples of invitation and reminder notices are available in Appendix B.

For category three, contact information for invitees was not disclosed. All
communication to invitees was conducted through a third party service provider
hired by the professional society supplying the list. All invitations and reminders
for this group were sent via email. As was the case for category one and two
contacts, a total of two reminder notices were sent, with a minimum of three days
and a maximum of ﬁve days between mailings. However, in this case invitees
received reminder notices whether or not their response was recorded. This

action was caused by the inability to identify individuals who had already

42

responded due to non-disclosure of contact information for recipients by SME

and HE.

4.4.0 Response Statistics and Sample Characteristics

A total of 96 invitations were returned as undeliverable, bringing the total number
of individuals initially contacted to 2,120. The number of individuals submitting
usable responses totaled 281 for an effective response rate of 13.3 percent (281
/ 2,120). In addition, 30 referrals were received, nine whom consented to
participate and produced usable responses, bringing the total response rate up to
13.5% (290/ 2,150). See Table 4.2 below for a summary of response by contact

category.

Table 4.2: Response by Category

 

 

 

 

 

Category 1 Category 2 Category 3 Total
No. Contacted 482 227 1441 2150
No. of Responses 112 16 162 290
% 23.2% 7.0% 11.2% 13.5%

 

 

 

 

 

4.5.0 Test for Non-Response Bias

An assessment of non-response bias was made using a similar method to that

which is recommended by Armstrong and Overton (1977). However, instead of

using late respondents as a proxy group for non-respondents, this study uses

individuals who provided partial responses (i.e., those who provided responses to

less than 10% of the questionnaire items) to approximate the characteristics of

43

 

non-respondents. In 51 partial response cases, individuals provided information
regarding the number of employees in their ﬁrm or business unit, and in 36 cases
partial responses contained information regarding total annual sales. T-tests
were performed on mean values of these demographic items for partial
responses versus complete responses (i.e, observations that contained
responses to more than 90% of the questionnaire items). No evidence of a
signiﬁcant difference between the mean number of employees for complete
versus partial response groups was found at a 10% signiﬁcance level (p=0.313),
nor was a signiﬁcant difference in mean annual sales detected between the two
groups (p=0.108). This result suggests that non-response bias may not pose a
problem for making inferences about the population based on the ﬁndings

produced from analysis of the sample collected for this study.

4.6.0 Sample Characteristics

As discussed in Section 4.3.0, the survey was administered in two parts. As a
result, two separate samples were obtained, a sample containing responses to
items of a technical nature (n=169) and a sample containing responses to
managerial and administrative items (n=188). In addition, a third sample
containing responses for organizations that provided responses to both
managerial and technical sections was constructed from the data collected
(n=94). Sixty-seven responses (67) in this paired sample resulted from
individuals who answered both the technical and managerial sections of the

survey, accounting for the difference between total reported responses received

44

(n=290) and the combined responses to the managerial and technical sections of

the questionnaire (n= 357). Further detail regarding the pairing of responses is

provided in Section 6.1.0. Table 4.3 provides a summary of the key

characteristics of each sample.

Table 4.3: Sample Characteristics

 

 

Managerial
Sample
(n=188)

 

Number of Employees:
Annual Sales:

Industry:

Median=365

Median=$61 Million (US)

 

Frequency (Percentage)
Fabricated Metal Product Manufacturing 31 (16.5%)
Electrical Equip., Appliance and Component Mfg. 24 (12.8%)
Aerospace 23 (12.2%)
Transportation Equipment Manufacturing 23 (12.2%)
Primary Metal Manufacturing 15 (8.0%)
Machinery Manufacturing 14 (7.4%)
Miscellaneous Manufacturing 14 (7.4%)
Computer and Electronic Product Manufacturing 9 (4.8%)
Plastic and Rubber Products Manufacturing 7 (3.7%)
Other 28 (14.9%)
Respondent Title:
uenc Percenta e

President I CEO 17 (9.0%)

Vice President / COO 20 (10.6%)
Director 22 (1 1 .7%)
Supply Chain Lead (or equivalent) 17 (9.0%)
Manager 69 (36.7%)
Engineer 27 (14.4%)
Other 16 (8.5%)

 

45

 

Table 4.3 (cont)

 

 

 

Technical Number of Employees: not asked

Sample

(n=169) Annual Sales: not asked
Industry:

F rguency (Percentage)
Aerospace 27 (16.0%)
Fabricated Metal Product Manufacturing 26 (15.4%)
Electrical Equip., Appliance and Component Mtg. 23 (13.6%)
Transportation Equipment Manufacturing 20 (11.8%)
Miscellaneous Manufacturing 11 (6.5%)
Computer and Electronic Product Manufacturing 10 (5.9%)
Primary Metal Manufacturing 7 (4.1%)
Machinery Manufacturing 6 (3.6%)
Plastic and Rubber Products Manufacturing 5 (3.0%)
Other 34 (20.1 %)
Respondent Title:
Freguency (Pergntage)

President / CEO 4 (2.4%)
Vice President / COO 10 (5.9%)
Director 9 (5.3%)
Manager 42 (24.9%)
Engineer 77 (45.6%)
Other 27 (16.0%)

Paired

Sample Number of Employees: Median=363

(n=94)

 

Annual Sales:
Industry:

Fabricated Metal Product Manufacturing
Aerospace

Electrical Equip., Appliance and Component Mfg.

Transportation Equipment Manufacturing
Primary Metal Manufacturing

Computer and Electronic Product Manufacturing
Machinery Manufacturing

Miscellaneous Manufacturing

Plastic and Rubber Products Manufacturing
Other

Median=$80 Million (US)

Frguency (Percentage)
18 (19.1%)

14 (14.9%)
11 (11.7%)
10 (10.6%)
8 (8.5%)
7 (7.4%)
6 (6.4%)
6 (6.4%)
4 (4.3%)
10 (10.6%)

Respondent Title (n=121 due to dual respondents for 27 cases):

President I CEO

Vice President / COO

Director

Supply Chain Lead (or equivalent)
Manager

Engineer

Other

Fryueney (Percentage)

6 (5.0%)

13 (10.7%)

10 (8.3%)
7 (5.8%)

42 (34.7%)

32 (26.4%)

11 (9.0%)

 

46

 

4.7.0 Management of Missing Data

Because missing data can create issues for the application of the statistical
methods adopted for analysis, a data imputation procedure was implemented.
The Expectation-Maximization (EM) method of imputation was selected due to its
ability to produce reliable estimates of missing data for situations where
responses are missing at a relatively low rate (Hair et al., 2006). This is the case
for the data analyzed in this study, as less than 10% of responses are missing on
over 90% of the items selected for study from the retained observations. EM’s
use, however, is only advisable in situations where the data can be considered
Missing Completely at Random (MCAR). Little’s MCAR test was used to assess
the viability of the EM imputation procedure. In all three samples to be analyzed
(Managerial, Technical and Paired), the data was considered to be MCAR, as
evidenced by failure to reject the null hypothesis of the data being MCAR at the
10% signiﬁcance level. As such, EM was used to impute values for missing

responses in each sample.

4.8.0 Chapter Summary

This chapter discussed the methods used to obtain the data to be analyzed in
order to address the research questions posed for study. An empirical analysis
employing perceptual measures of key variables and constructs was selected as
an appropriate and efﬁcient means of carrying out the study. A questionnaire
was developed and deployed to collect data primarily from ﬁrms that produce

technically complex products. Three samples were generated from the data

47

collected: 1) a sample containing responses to questions of a technical nature;
2) a sample containing responses to questions that are managerial in nature; and
3) a sample containing observations from organizations that responded to both
technical and managerial sections of the questionnaire. Preliminary analysis of
the data revealed no evidence of non-response bias, and no discemable pattern
to missing data. Values were imputed for missing data in order to produce
complete datasets to be used for the assessment of measurement properties and

the testing of hypotheses.

48

5.0.0 Measurement

It has been argued, both in Chapter 2 of this study and in recent scholarly works
(Salvador, 2007; Choi & Krause, 2006), that Modularity and Complexity are multi-
dimensional constructs, meaning that the latent concepts of interest are
comprised of more than one distinct, measurable element. This study assesses
the viability of employing a multi-dimensional measurement approach by ﬁrst
establishing the psychometric properties of each proposed dimension of the
study constructs, and subsequently testing statistical models that assess the

level of association of each dimension with the underlying theoretical construct.

The psychometric properties of each dimension are assessed according to the
recommendations for establishing construct validity via Conﬁn'natory Factor
Analysis (CFA) proposed by O’Leary-Kelly and Vokurka (1998), and tests for
association of each dimension with its underlying construct are performed using
the Multiple-Indicator and Multiple-Causes (MIMIC) model of Joreskog and
Goldbeger (1975). A MIMIC model positions sets of observed values as both
causes and reﬂective indicators of a latent construct in order to identify the model

and estimate the parameters of interest.

5.1.0 Product Modularity
Theoretically, the Product Modularity construct is composed of three dimensions:
Functional Binding, Interface Standardization, and Decomposability, as deﬁned in

Chapter 2. An initial CFA containing multi-item measures for these three

49

dimensions, as well as for Process Heterogeneity and Process Interdependence
(dimensions of Process Complexity) was performed. The data ﬁt the model well
according to the cutoff criteria for covariance structure modeling suggested by Hu
and Bentler (1999) for samples with less than 250 observations, with CFI=0.963
and SRMR=0.052. However, the dimensions of Product Modularity failed to
meet the requirement of Discriminant Validity. Functional Binding exhibited an
estimated correlation near 1.00 with Decomposability, and a correlation of 0.99
with Interface Standardization. These results suggest that Function Binding as a
dimension cannot be distinguished from Decomposability or Interface
Standardization from a statistical standpoint. This condition also invalidates the
model, due to the presence of zero and negative error variances, which violate
key assumptions of covariance structure modeling. Results for the initial CFA

(CFA-01) are summarized in Table 5.1.

Table 5.1: CFA-01 — Dimensions of Product Modularity and Process Complexity

 

 

 

 

 

 

 

 

Scale (Reliability) / Question Std.
LoadinL

Functional Binding (0.720)

PD-FBI For our products, we can make changes in components without 0625"“r
changi_ng other components.

PD-F 82 For our products, functions can be directly added or deleted by 0.682"
adding or removing components.

PD-FB3 Our product design is such that a given product function is Dropped
performed by a single product component.

PD-FB4 Our products have interchangeable features and options. 0.735"

 

 

 

50

 

Table 5.1 (cont)

 

Interface Standardization (0.775)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PD-IS1 Our product components have standardized interfaces. 0.519"

PD-ISZ The interfaces of our product components are designed to accept a 0.887"
variety of components.

PD-IS3 Our product components are able to accept a wide range of 0.824“
complements (such as modules or peripherals).

PD-IS4 Each component interface is unique. (reversed item) Dropped
Decomposability (0.700)

PD-DC1 Our products can be decomposed into separate modules. 0.556"

PD-DCZ Our products are designed to be easily disassembled and 0.535"
reassembled.

PD-DC3 Our products are designed to be easily reconﬁgured. 0.853“
Process Interdependence (0.836)

PC-lNT1 Our production processes are broken down into sub-processes that 0.653“
can operate independently. (reversed item)

PC-INT2 Sub-processes can be added to or removed from our production 0.881 **
processes without changigg other sub-processes. (reversed item)

PC-INT3 Our production processes are designed so that sub-processes can 0.852"
be added or removed without signiﬁcant disruptions/changes to
other sub-processes. (reversed item)
Process Heterogeneity (r = 0.387")

PC-HTR1 The products we make have similar processing and assembly 0.534"
requirements. (reversed item)

PC-HTR2 Only minor equipment modiﬁcations are required to produce different 0.718"
products. (reversed item)

PC-HTR3 We are frequently forced to make changes in our manufacturing Dropped
processes to satisfy differing customer requirements.
Fit Statistics: **p<0.05

 

Chi-Square=103.5, df=67; CFI=0.963; GFI=O.921;
SRMR=0.053; RMSEA=0.057

 

 

Results from the initial CFA suggest that although the dimensions of Product

Modularity are distinguishable from a theoretical standpoint, they may actually

form a single dimension statistically. Thus, an alternate measurement approach,

which positions all eleven original items reﬂective items of a single modularity

construct, is tested in CFA-02. Although all items loaded signiﬁcantly on the

modularity construct, six items were retained in order to improve model

51

 

parsimony and ﬁt. Only items with a factor loading of 0.700 or higher were

_. retained, although careful consideration was made to ensure that the all three
theoretical dimensions of modularity were represented in its measurement. This
measurement model, which also contains two other constructs measured with
data from the technical sample (Process Heterogeneity and Process
Interdependence), exhibited a good ﬁt to the data according to the Hu and
Bentler (1999) criteria, with CFI=0.953 and SRMR=0.048. In addition, the
requirements of Unidimensionality are met through the signiﬁcant loadings of all
items on their respective constructs and the acceptable ﬁt of the overall model,
and Convergent and Discriminant Validity are satisﬁed through the detection of

signiﬁcant differences in the chi-square statistic for the null model versus the

measurement model (xzdm,o=849.5, df=14, pz0.00), and the constrained model

(covariances of constructs constrained to 1.0) and the measurement model

(xzdiﬁ,1=254.0, df=3, p=0.00). The reliability of each construct/dimension was

assessed via Cronbach’s Alpha (Cronbach, 1951 ). Reliability for Product
Modularity and Process Interdependence are 0.907 and 0.836 respectively, well
above the minimum acceptable limit of 0.600 for new scales (Nunnally, 1978).
However, calculation of Cronbach’s Alpha for Process Heterogeneity was
precluded due to the retention of only two measurement items. Pearson
correlation was calculated for the Process Heterogeneity construct as a
surrogate measure of reliability (r = 0.387). Results for CFA-02 are summarized

in Table 5.2.

52

Table 5.2 CFA-02 -Product Modularity and Dimensions of Process Complexity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Scale (Reliability) / Question Std.
LoadirLg

Product Modularity (0.907)

PD-MOD1 Our products are designed to enable the swapping of components. 0.720"

PD-F B1 For our products, we can make changes in components without Dropped
changlig other components.

PD-F 32 For our products, functions can be directly added or deleted by 0.706"
adding or removing components.

PD-F B3 Our product design is such that a given product function is Dropped
mrformed by a single product component.

PD-FB4 Our products have interchangeable features and options. 0.768"

PD~IS1 Our product components have standardized interfaces. Dropped

PD-ISZ The interfaces of our product components are designed to accept a 0.846”
variety of components.

PD-IS3 Our product components are able to accept a wide range of 0.818“
complements (such as modules or peripherals).

PD-IS4 Each component interface is unique. (reversed item) Dropped

PD-DC1 Our products can be decomposed into separate modules. Dropped

PD-DCZ Our products are designed to be easily disassembled and Dropped
reassembled.

PD-DC3 Our products are designed to be easily reconfigured. 0.768"
Process Interdependence (0.836)

PC-INT1 Our production processes are broken down into sub-processes that 0.653"
can operate independently. (reversed item)

PC—INT2 Sub-processes can be added to or removed from our production 0.881“
processes without changﬂg other sub-processes. (reversed item)

PC-INT3 Our production processes are designed so that sub-processes can 0.852’"Ir
be added or removed without signiﬁcant disruptions/changes to
other sub-processes. (reversed item)
Process Heterogeneity (r = 0.387“)

PC-HTR1 The products we make have similar processing and assembly 0.518"
requirements. (reversed item)

PC-HTR2 Only minor equipment modiﬁcations are required to produce diﬁerent 0.741 **
products. (reversed item)

PC-HTR3 We are frequently forced to make changes in our manufacturing Dropped
processes to satisfy differing customer requirements.
Fit Statistics: **p<0.05

 

Chl—Square=82.1, df=41; CFI=0.953; GFI=O.916;
SRMR=0.048; RMSEA=0.078

 

 

53

 

Due to the inability to distinguish among the dimensions of the modularity
construct, and because a valid one dimensional measure of modularity has been
established, analysis of the associations of each dimension to the underlying

construct via MIMIC is precluded.

5.2.0 Complexity

Complexity, as discussed in Chapter 2, is theoretically composed of three
dimensions: Number of Elements, Heterogeneity, and Interdependence. Each
of the three types of complexity: Manufacturing Process Complexity,
Organizational Complexity, and Supply Base Complexity, are examined to
assess the composition of the construct in terms of the theoretical dimensions of

complexity, and to establish valid and reliable measures of each construct.

5.2.1 Process Complexity

As shown in CFA-02 (Table 5.2), the two dimensions of Process Complexity that
are captured with multi-item measures, Process Heterogeneity and Process
Interdependence, are distinct and exhibit reasonable psychometric properties.
The remaining dimension, Number of Elements, was captured with a single
objective response item asking respondents to report the number of lntemal

manufacturing processes used to manufacture their primary products.

A MIMIC model was ﬁt to the data to examine the association of the Number of

lntemal Manufacturing Processes, Process Heterogeneity, and Process

54

Interdependence to the Manufacturing Process Complexity Construct (Figure
5.1). Three items purported to measure the Process Complexity construct
directly were included to identify the model (PC-CMP1, PC-CMP2, PC-CMP3).
The model did not exhibit a very strong ﬁt to the data, with CFI=0.944 and
SRMR=0.089. In addition, only the Number of lntemal Manufacturing Processes

was signiﬁcantly related to the Process Complexity construct.

The results of the MIMIC model analysis suggest that Manufacturing Process
Complexity cannot be reliably measured as a composite of its theoretically
derived dimensions as they have been measured for this study. As such, the
examination proceeded to establishing the measurement properties of the direct
Manufacturing Process Complexity measurement items. CFA-03 (Table 5.3)
includes the Manufacturing Complexity Construct as measured by these items.
This model displays a good ﬁt to the data, with CFI=0.962 and SRMR=0.051.
Conditions of Unidimensionality, Convergent Validity and Discriminant Validity
are met with signiﬁcant factor loadings of all included items, and signiﬁcant chi-

square differences statistics between the null model and the measurement model

(xzdiﬁ,o=828.6, df=10, p=0.00), as well as between the constrained model and

the measurement model (xzdiﬁ,1=287.9, df=1, p==0.00). The Manufacturing

Process Complexity construct also exhibits good reliability, with an alpha value of

0.893.

55

Figure 5.1 Manufacturing Process Complexity MIMIC Model

 

 

 

 

PC-CMP1 PC-CMPZ PC-CMP

 

 

 

 

 

 

 

 

 

0.878"

 

   
 

Mfg. Process
Complexity

-0.042

 

 

   
   

  

 

 

 

 

No. of Mfg. Process Process
Processes Interdependence Heterogeneity
Fit Statistics:
Chi-Square=58.617, df=22; CFI=0.944; GFI=O.924;
SRMR=0.089; RMSEA=0.100

Notes: **p<0.05

 

 

56

 

Table 5.3 CFA-03 —Product Modularity and Process Complexity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Scale (Reliability) / Question Std.
Loading

Product Modularity (0.907)

PD-MOD1 Our products are designed to enable the swapping of components. 0.719"

PD-FB1 For our products, we can make changes in components without Dropped
changing other components.

PD-F 82 For our products, functions can be directly added or deleted by 0.708"
adding or removing components.

PD-F B3 Our product design is such that a given product function is Dropped
performed by a single product component.

PD-F B4 Our products have interchangeable features and options. 0.762"

PD-IS1 Our product components have standardized interfaces. Dropped

PD—ISZ The interfaces of our product components are designed to accept a 0.850“
variety of components.

PD-IS3 Our product components are able to accept a wide range of 0.823”
complements (such as modules or peripherals).

PD-IS4 Each component interface is unique. (reversed item) Dropped

PD-DC1 Our products can be decomposed into separate modules. Dropped

PD-DC2 Our products are designed to be easily disassembled and Dropped
reassembled.

PD-DC3 Our products are designed to be easily reconﬁgured. 0.761 **
Manufacturing Process Complexity (0.893)

PC-CMP1 Our production system is very sophisticated in terms of its operating 0.808"
characteristics.

PC-CMP2 Our production system is very complex. 0.883"

PC-CMP3 Our production system is very sophisticated in terms of 0.884"
structure/layout.
Fit Statistics: **p<0.05

 

Chi-Square=58.7, df=26; CFI=0.962; GFI=O.927;
SRMR=0.051; RMSEA=0.087

 

 

57

 

5.2.2 Organizational Complexity

The individual dimensions of the Organizational Complexity construct
(Organizational Heterogeneity and Organizational Interdependence) were
analyzed along with dimensions of Supply Base Complexity (Supply Base
Heterogeneity and Supply Base Interdependence) and Environmental
Uncertainty in CFA-04 (Table 5.4). This measurement model exhibited a good ﬁt
to the data, with CFI=0.956 and SRMR=0.049. All factor loadings are signiﬁcant,
indicating satisfactory Unidimensionality, and Convergent and Discriminant
Validity conditions are satisﬁed, as illustrated by signiﬁcant chi-square difference

statistics for the comparison of the null model and the measurement model

(xzdiﬁ,o=908.0, df=26, p=0.00), and the constrained model and measurement

model (xzdm,1=374.6, df=10, p=0.00). Scale reliabilities are also acceptable, with

alphas of 0.807, 0.720, 0.802, and 0.753 for Organizational Interdependence,
Organizational Heterogeneity, Environmental Uncertainty and Supply Base
Interdependence respectively. Calculation of Cronbach’s Alpha for Supply Base
Heterogeneity was precluded because the construct was measured with only two
items. Pearson correlation was calculated for the Supply Base Heterogeneity

construct as a surrogate measure of reliability (r = 0.362).

58

Table 5.4 CFA-04 — Environmental Uncertainty and Dimensions of Organizational
Complexity and Supply Base Complexity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Scale (Reliability) / Question Std.
Load igl

Organizational Heterogeneity (0.807)

OG-HTR1 Sub-units in our ﬁrm/SBU are subject to the same policies (ex.: 0.742”
quality standards, compensation, absenteeism, etc. ). (reversed
item)

OG-HTR2 Sub-units in our ﬁrm/SBU employ similar or compatible business 0.778"
practices. (reversed iteny

OG-HTR3 Sub-units in our firm/SBU have similar or compatible information 0.683“
systems/technologies. (ﬁversed item)

OG-HTR4 Sub-units in our firm/SBU use common mechanisms for 0.670"
coordinating work. (reversed item)

OG-HTR5 The organizational structures of sub-units in our ﬁrm/SBU are Dropped
similar. (reversed item)
Organizational Interdependence (0.720)

OG-INT1 Project teams are self-directed. (reversed item) 0.601 **

OG-INT2 Project teams operate autonomously. (reversed item) Dropped

OG-INT3 Project teams operate independently from other organizational Dropped
entities. (reversed item)

OG-INT4 Project teams are free to develop links with other teams and/or 0.845"
internal units in our ﬁrm/SBU. (reversed item)

OG-INT5 Project teams are free to develop links with external units such as 0.616"
suppliers, customers, etc. (reversed itemL
Environmental Uncertainty (0.802)

EN-UNC1 Sales forecasts for our primary products are likely to be: (very Dropped
inaccurate. . .very accurate)

EN-UNC2 Predicting the actions of our competitors is: (very easy. . .very 0.635“
difﬁcult)

EN-UNC3 Predicting consumer tastes in our market(s) is: (very easy. . .very 0.797”
difﬁcult)

EN-UNC4 Monitoring market trends for our primary products is: (very 0.758"
easy. . . very difficult)

EN-UNCS Predicting demand for our primary product line is: (very 0.664"
easy. . . very diﬂicult)
Supply Base Interdependence (0.753)

SB-INT1 We have frequent face-to-face communication with key suppliers. 0.611"

SB-INT2 For a given supplier, there are multiple points of contact between Dropped
our oLganization and the supplier's.

SB-INT3 We have long-term relationships with our suppliers. 0.794”

SB~INT4 We have strong ties to our suppliers. 0.759"r

 

 

 

59

 

Table 5.4 (cont)

 

Supply Base Heterogeneity (r = 0.362")

 

 

 

 

SB-HTR1 Our suppliers are diverse in terms of size, location, and products 0.632“
supplied.

SB-HTR2 Effective management of procurement requires taking different 0.573"
approaches with different suppliers.
Fit Statistics: **p<0.05

 

Chi-Square=146.4,df=94; CFI=0.944; GFI=O.915;
SRMR=0.049; RMSEA=0.055

 

 

A second MIMIC model was ﬁtted to examine the association of Organizational

Complexity with its theoretical dimensions (Figure 5.2). The Number of Elements

dimension for Organizational Complexity is represented by the sum of three

objective response items: (1) the number of organizational divisions; (2) the

number of organizational levels; (3) the number of worldwide locations. Three

direct measures of Organizational Complexity (OG-CMP1, OG-CMP2, OG-

CMP3) were included to identify the model. The model exhibited a poor ﬁt to the

data, with CFI=0.892 and SRMR=0.075. However, both the Number of

Organizational Elements and Organizational Heterogeneity were signiﬁcantly

related to the Organizational Complexity construct in this model.

60

 

Figure 5.2 Organizational Complexity MIMIC Model

 

 

 

 

OG-CMP1 OG—CMPZ OGCMP3
I

 

 

 

 

 

 

 

 

 

     
 

0.61 1"

 

Organizational
Complexity

0.427“

 

 

     

 

 

 

 

No. Of DIVISIO'W’ Organizational Organizational
+N°' °f 0'9' Le“ Interdependence Heterogeneity
+No. of Locations

Flt Statistics:
Chi-Square=111.751, df=39; CFI=0.892; GFI=O.910;
SRMR=0.075; RMSEA=0.101

Notes: "p<0.05

 

 

 

As was the case with Manufacturing Process Complexity, Organizational
Complexity was not measured reliably as a composite of its theoretical
dimensions as they have been measured for this study. Consequently, the
examination of the psychometric properties of the direct Organizational
Complexity measures followed. CFA-05 (Table 5.5) substitutes the
Organizational Complexity construct as measured by the direct items for the
individual dimensions of Organizational Complexity in CFA-04. The model
exhibited a fair ﬁt to the data, with a CFI=0.915 and SRMR=0.073.

Unidimensionality, Convergent Validity and Discriminant Validity were all satisﬁed

61

(xzdiff,o=607.8, df=18, pzooo; xzdm,1=382.4, df=6, p=0.00). Reliability for the

Organizational Complexity is acceptable, with an estimated alpha of 0.763.

Table 5.5 CFA-05 -Environmental Uncertainty, Organizational Complexity and
Dimensions of Supply Base Complexity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Scale (Reliability) / Question Std.
Loading

Organizational Complexity (0.763)

OG-CMP1 Our ﬁrm/SBU has a very complex organizational structure. 0.842“

OG-CMP2 Our ﬁrm/SBU is organized in a straight-forward, uncomplicated 0.754“
way. (reversed item)

OG—CM P3 Coordinating work between sub-units in our ﬁrm/SBU is difficult. 0.606“
Environmental Uncertainty (0.802)

EN-UNC1 Sales forecasts for our primary products are likely to be: (very Dropped
inaccurate. . .very accurate)

EN-UNCZ Predicting the actions of our competitors is: (very easy. . . very 0.633"
difﬁcult)

EN-UNC3 Predicting consumer tastes in our market(s) is: (very easy. . .very 0.797"
difﬁcult)

EN-UNC4 Monitoring market trends for our primary products is: (very 0.759"r
easy. . .very difficult)

EN-UNC5 Predicting demand for our primary product line is: (very 0.663“
easy. . .very difficult)
Supply Base Interdependence (0.753)

SB-INT1 We have frequent face-to-face communication with key suppliers. 0.589“

SB-INT2 For a given supplier, there are multiple points of contact between Dropped
our organization and the supplier's.

SB-INT3 We have long-term relationships with our suppliers. 0.824”

SB-INT4 We have strong ties to our suppliers. 0.745"
Supply Base Heterogeneity (r = 0.362“)

SB-HTR1 Our suppliers are diverse in terms of size, location, and products 0.692"
supplied.

SB-HTR2 Effective management of procurement requires taking different 0.52 3**
approaches with different suppliers.
Fit Statistics: **p<0.05

 

Chi-Square=87.0,df=48; CFI=0.938; GFI=O.930;
SRMR=0.058; RMSEA=0.066

 

 

62

 

 

5.2.3 Supply Base Complexity

The dimensions of Supply Base Complexity (Supply Base Heterogeneity and
Supply Base Interdependence) were included in both CFA-04 and CFA-05 and
exhibited reasonable measurement properties (see tables 5.4 and 5.5). The
Number of Elements dimension for Supply Base Complexity was measured with
a single objective response item asking respondents to report the approximate

number of suppliers used by their ﬁrm or SBU.

A third MIMIC model examining the association of Supply Base Complexity to its
proposed dimensions was ﬁt to the data (Figure 5.3). As in prior MIMIC
analyses, three direct measures of Supply Base Complexity were included to
identify the model. The model ﬁt was poor, with CFI=0.899 and SRMR=0.070,
however, Supply Base Heterogeneity and Supply Base Interdependence were
signiﬁcantly related to the construct, although the path coefﬁcient from Supply

Base Interdependence was negative.

63

Figure 5.3 Supply Base Complexity MIMIC Model

 

 

 

 

SB-CMP1 SB-CMP2 SB-CMP3
I

 

 

 

 

 

 

 

 

 

    
       
 

 

Supply Base
Complexity

41.637"

0.638“

 

 
 
    

 

  

 

 

 

 

. Supply Base Supply Base
No. Of Suppliers Interdependence Heterogeneity
Fit Statistics:
Chi-Square=57.422, df=22; CFI=0.899; GFI=O.936;
SRMR=0.070; RMSEA=0.094

Notes: **p<0.05

 

 

 

Similar to the cases of Organizational Complexity and Manufacturing Process
Complexity, Supply Base Complexity was not reliably measured as a composite
of its theoretical dimensions as they have been measured for this study. Hence,
measurement with items that tap Supply Base Complexity directly (SB-CMP1,

SB-CMP2, SB-CMP3) was evaluated for validity and reliability. The

64

measurement model (CFA-06, Table 5.6) exhibited a fair ﬁt to the data, with

CFI=0.943 and SRMR=0.061. Constructs measured in this model met the

criteria of Unidimensionality, Convergent Validity and Discriminant Validity

(xzdiﬁ,o=748.8, df=20, pz0.00; xzdiﬁ,1=365.0, df=6, p20.00), and the alpha

reliability estimate for Supply Base Complexity was acceptable at 0.672.

Table 5.6 CFA-06 -- Environmental Uncertainty, Organizational Complexity and Supply
Base Complexity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Scale (Reliability) / Question Std.
Loadigq

Organizational Complexity (0.763)

OG-CMP1 Our ﬁrm/SBU has a very complex organizational structure. 0.842”

OG-CMPZ Our ﬁrm/SBU is organized in a straight-forward, uncomplicated 0.754“
way. (reversed item)

OG-CM P3 Coordinating work between sub-units in our ﬁrm/SBU is difﬁcult. 0.606"
Environmental Uncertainty (0.802)

EN-UNC1 Sales forecasts for our primary products are likely to be: (very Dropped
accurate. . .very inaccurate)

EN-UNCZ Predicting the actions of our competitors is: (very easy. . .very 0.633“
difficult)

EN-UNC3 Predicting consumer tastes in our market(s) is: (very easy. . .very 0.797"
difﬁcult)

EN-UNC4 Monitoring market trends for our primary products is: (very 0.759“
easy. . .very difﬁcult)

EN-UNCS Predicting demand for our primary product line is: (very 0.663"
easy. . .very difﬁcult)
Supply Base Complexity (0.672)

SB-CMP1 Coordinating procurement and manufacturing operations is 0.602“
problematic for our ﬁrm/SBU.

SB-CMP2 The structure of our supply chain is difﬁcult to describe or illustrate. 0.799“

SB-CMP3 Our supply chain is a very complex network. 0.538"
Fit Statistics: **p<0.05

 

Chi-Square=124.5, df=71; CFI=0.932; GFI=O.914;
SRMR=0.059; RMSEA=0.063

 

 

65

 

5.3.0 Chapter Summary

In this chapter, an assessment was performed of the viability of multi-dimensional
measurement approaches for the constructs under study: (1) Product
Modularity, (2) Manufacturing Process Complexity, (3) Organizational
Complexity, and (4) Supply Base Complexity. The results indicate that only the
Product Modularity construct could be measured validly and reliably employing
all theoretically derived dimensions, albeit the theoretical dimensions of the

construct were all reﬂected in a single dimension statistically.

Each theoretically derived dimension of complexity showed a signiﬁcant
association with complexity on at least one organizational level (i.e.,
manufacturing process, organization, or supply base), but in no case were all
three dimensions signiﬁcantly linked to a complexity construct. Instead, reliable
and valid measures of each type of complexity were obtained by tapping the
construct directly, as opposed to measuring its constituent dimensions. The next
chapter discusses the procedures used to test the hypotheses presented in

Chapter 3.

66

6.0.0 Hypothesis Testing

In this chapter, hypotheses are tested which relate variables measured in the
technical section of the survey to variables measured in the managerial section
of the survey. Hence, only those cases for which both sections of the
questionnaire received responses are observed (n=94). The lower sample size
precludes the straight-fonrvard application of common multivariate methods for
analyzing relationships among latent variables (such as Structural Equation
Modeling - SEM). Therefore analysis will proceed employing summated scales
of the measures that were validated using the larger samples in Chapter 5, and
Seemingly Unrelated Regression (SUR) — an econometric method of analyzing
systems of equations that is less restrictive than SEM in terms of required
sample size. The use of summated scales is an acceptable alternative means of
representing latent constructs in multivariate analysis, provided that they meet
the requirements of validity and reliability (Hair Jr., Black, Babin, Anderson, &

Tatham, 2006), as demonstrated in Chapter 5.

6.1.0 Pairing of Responses

Of the 94 cases analyzed, 27 cases are paired responses from different
individuals within the same organization. The remaining 67 responses were
cases where a single individual responded to both the technical and managerial
sections of the questionnaire. The average Intraclass Correlation Coefﬁcient
(ICC) was used to assess the level of agreement in responses provided by

different individuals from the same organization. The computed value for the

67

average ICC on four performance items for which different individuals within the
same organization responded was 0.58, just below the minimum threshold for
agreement of 0.60 recommended by Boyer and Verrna (2000) for Operations
Management and Strategy research. Although the minimum threshold for
agreement was not exceeded, the decision was made to proceed with the pairing
of matched responses because: (1) they constitute a relatively low proportion of
the sample (29%); and (2) the measures of interest in this study were not asked

of both respondents.

6.2.0 Seemingly Unrelated Regression (SUR)

Seemingly Unrelated Regression (Zellner, 1962) is an econometric technique of
efﬁciently estimating parameters of sets of equations that appear to be
independent (i.e., they do not exhibit simultaneity - having variables that are
exogenous with respect to one equation in the set, but endogenous in another),
but may exhibit error correlation across equations. The SUR model uses a
Feasible Generalized Least Squares (FGLS) approach to correct for error
correlation across equations and produces more efﬁcient parameter estimates

than OLS (Kennedy, 2003).

Because SUR ﬁrst performs Ordinary Least Squares (OLS) regression on the
individual equations, and subsequently uses the residuals to produce estimates
of variances and covariances, it does not require as large a sample to produce

consistent parameter estimates as SEM. It only requires that the number of

68

observations used to estimate each equation be larger than the number of
equations. However to produce reliable estimates, it is recommended that the
number of observations be “quite large” compared to the number of equations
(Kennedy, 2003). Like OLS regression, the conditions of the Gauss—Markov
Theorem (excluding the requirement of spherical disturbances, which the SUR
model corrects for) must be met to produce the best linear unbiased estimates.
In addition, because the SUR model estimates a system of equations, order and

rank conditions must also be met in order for the model to be identiﬁed.

Seemingly Unrelated Regression uses the same goodness of ﬁt measures as

OLS for overall model ﬁt: the F-test for lack of ﬁt and the Coefﬁcient of

Deterrnlnation, R2. The F-test evaluates the null hypothesis that all regression

coefﬁcients in the model are jointly equal to zero, while the Coefﬁcient of
Determination offers an estimate of the proportion of variance in the dependent
variable that is explained by the independent variables. SUR estimates an F-
statistic and Coefﬁcient of Determination for each equation In the model.
Evaluation of the statistical signiﬁcance of an individual regression coefﬁcient is

performed via the t-statistic, just as in OLS regression.

The SUR model is an appropriate technique for hypothesis testing in this study
because each observation in the set of equations to be estimated represents a
production system, in which the constituent parts operate together in the creation

of products. Therefore it can be assumed that the equations to be estimated are

69

not independent; hence their error terms may be correlated. In addition, the
lower sample size requirement makes SUR an attractive choice for hypothesis
testing here, especially when considering the need to parse the sample in order
to test for the hypothesized contingency and moderating effects of System

Openness and Environmental Uncertainty.

6.3.0 Preliminary Analysis

Prior to hypothesis testing, preliminary diagnostics were run on the data.
Bivariate scatterplots of the dependent variables versus the independent
variables were produced in order to detect outlying observations. This procedure
identiﬁed six (6) outliers. Upon careful review of the characteristics of these
observations, they were excluded from further analysis. Descriptive statistics for

the study variables are included in Table 6.1.

Table 6.1: Descriptive Statistics for Model Constructs

 

 

 

 

 

 

 

 

 

Construct Mean 3:23:3‘
I’é‘l‘é‘iﬁé,“‘7‘3§3lﬁl’¥caes) 22.00 0.07
31023225733332? 12-93 4.02
323323225233)“: 9.50 0.77
23173535353033'251333 70.04 3.07

 

70

Table 6.1 (cont)

 

 

 

 

 

 

 

guittzﬁrgggolnt scale) 3.16 1 .81
Z"I§371m?'§.?1#22372'j’y 17.22 3.39
III-3357 223579795232?“ 4-13 1.95
”"72? £11273 0337': Ions 2.83 2.06
”urgtr’neryfiaiziﬂicerZ) 3.32 1.59

 

 

 

6.4.0 Results

The SUR Model analyzed for hypothesis testing is deﬁned as follows:

(1) COMPLEXITYprocess = I30 f B1'MODULAR'TYproduct + BZ’NUMBERmfgprocesses + 8
(2) COItIIPLEXI'I'Yorg = Bo + B1'MODULARITYproduct + B2'NUMBERorgdivisions + 8

(3) COMPLEXITYsuppbease = Bo 1' B1'MODULARITYproduct + BZ'NUMBERsuppIiers + 8

The isubscript representing individual observations has been dropped from each

variable and error term for simpliﬁcation.

The COMPLEXITY terms on the left-hand side of each equation above represent
the dependent variables: Manufacturing Process Complexity, Organizational

Complexity, and Supply Base Complexity. The MODULARITY terms or the right-
hand side of equation above represent the primary independent variable, Product

Modularity. The NUMBER variables in the model above (Number of

71

Manufacturing Processes, Number of Organizational Divisions, and Number of

Suppliers) function as instrumental variables, aiding in the identiﬁcation of the

model and controlling for potential simultaneity.

6.4.1 Exploratory Model

Hypotheses one (H1) through three (H3) were tested by ﬁtting the SUR model

above to the full sample, absent the outliers purged in preliminary analysis

(n=88). Results are summarized in Table 6.2 below. Equations (1) and (3)

exhibited a good ﬁt to the data, with signiﬁcant F-statistics (p < 0.05) and R2

values of 0.15 and 0.13 respectively. Equation (2) did not exhibit a good ﬁt to the

data, with a nonsigniﬁcant F-statistic (p=0.262) and an R2 value of 0.04.

Table 6.2: SUR Exploratory Model — Full Sample (n=88)

 

Goodness of Fit Measures

 

 

 

 

 

 

 

 

 

 

 

 

Equation # Dependent Variable R2 F-statistic p-value
(1) Process Complexity 0.1519 7.11 0.001
(2) Organizational Complexity 0.0442 1.35 0.262
(3) Supply Base Complexity 0.1301 4.75 0.009
Parameter Estimates
Coefﬁcient Standard Error t-statistic p-value
Equation (1)
Process Complexity
Constant 8.597 1.254 6.86 0.000
Product Modularity 0.080 0.041 1.93 0.055
No of Mfg. Processes 0.61 5 0.202 3.04 0.003

 

72

 

Table 6.2 (cont)

 

 

 

 

 

 

 

 

 

 

Coefﬁcient Standard Error t-statistic p-value
Equation (2)
Org. Complexity

Constant 8.139 1.041 7.82 0.000
Product Modularity 0.038 0.042 0.90 0.366
No. of Divisions 0.200 0.1 76 1 .14 0.257

Equation (3)

Supply Base Complx.

Constant 6.655 1.151 5.78 0.000
Product Modularity 0.091 0.036 2.51 0.013
No. of Suppliers 0.402 0.195 2.06 0.041

 

 

Breusch-Pagan Test of Independence: Chi-square=20.615, df=3 p=0.000

 

Hypothesis one (H1) states that Product Modularity inﬂuences the complexity of

lntemal manufacturing processes. This hypothesis was not supported, as the t-

statistic on the regression coefﬁcient for Product Modularity in equation (1) was

not found to be signiﬁcant at the 5% alpha level (p=0.055). Hypothesis two (H2)

states that Product Modularity inﬂuences the complexity of organizational design

and coordination. This hypothesis was not supported, as the t-statistic on the

Product Modularity regression coefﬁcient in equation (2) was not found to be

signiﬁcant at an alpha level of 0.05 (p=0.366). Hypothesis three (H3) states that

Product Modularity inﬂuences the complexity of a ﬁrm’s supply base. This

hypothesis was supported, as the t-statistic on the regression coefﬁcient for

Product Modularity in equation (3) was found to be signiﬁcant at the 0.05 alpha

level (p=0.013).

73

 

6.4.2 Contingency Model

Hypotheses 4a through 6a were tested by ﬁtting the SUR model to a partial
sample. The sample of 88 observations was split into two groups based on
values of the proxy for system openness —.the reported percentage of the
components of the ﬁrm or SBU’s primary products that are outsourced. Values
falling below the sample mean of 3.16 (7-point Likert scale with anchors: 1=less
than 10% outsourced and 7=more than 85% outsourced) were considered the
low-outsourcing or “closed system” group. Fifty-four (54) observations were
included in this group. Results of the SUR model estimation for the “closed
system” group are summarized in Table 6.3 below. Equation (1) exhibited a

good ﬁt to the data for the “closed system” group, with a signiﬁcant F-statistic

(p=0.027) and an R2 value of 0.12. Equations (2) and (3) did not exhibit good ﬁts

to the data, with nonsigniﬁcant F-statistics (p > 0.05), and low R2 values, 0.02

and 0.04 respectively.

Table 6.3: SUR Contingency Model - Closed System Sample (n=54)

 

Goodness of Fit Measures

 

 

 

 

 

Equation # Dependent Variable R2 F-statistic p-value
(1 ) Process Complexity 0.1 199 3.71 0.027
(2) Organizational Complexity 0.01 53 0.23 0.798
(3) Supply Base Complexity 0.042 0.74 0.477

 

74

 

Table 6.3 (cont)

 

 

Parameter Estimates

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Coefﬁcient Standard Error t-statistic p-value
Equation (1)
Process Complexity
Constant 9.067 1 .671 5.43 0.000
Product Modularity 0.027 0.056 0.49 0.622
No of Mfg. Processes 0.748 0.282 2.65 0.009
Equation (2)
OQComplexity
Constant 9.382 1.510 6.21 0.000
Product Modularity -0.001 0.062 —0.02 0.983
No. of Divisions 0.169 0.256 0.66 0.51 1
Equation (3)
Supply Base Complx.
Constant 8.190 1 .460 5.61 0.000
Product Modularity 0.055 0.049 1.13 0.262
No. of Suppliers 0.138 0.250 0.55 0.581

 

 

Breusch-Pagan Test of Independence: Chi-square=18.666, df=3 p=0.000

 

Hypothesis 4a states that in Closed systems, Product Modularity is positively

associated with the complexity of internal manufacturing systems. This

hypothesis was not supported, as the t-statistic on the Product Modularity

coefﬁcient for equation (1) was not signiﬁcant at the 0.05 alpha level (p=0.622).

Hypothesis 5a states that in closed systems, Product Modularity is positively

associated with the complexity of organizational design and coordination. This

hypothesis was not supported, as the t-statistic on the regression coefﬁcient for

 

Product Modularity in equation (2) was not signiﬁcant at the 0.05 alpha level
(p=0.983). Hypothesis 6a states that in closed systems, Product Modularity is
positively associated with the complexity of a ﬁrm or SBU’s supply base. This
hypothesis was not supported, as the t-statistic on the Product Modularity
coefﬁcient in equation (3) was not found to be signiﬁcant at the 0.05 alpha level

(p=0.262).

Hypotheses 4b through 6b were tested by ﬁtting the SUR model to the high
outsourcing or “open system” group. This group (n=34) is composed of
observations whose reported value on the proxy for system openness was above
the sample mean of 3.16. Results of the SUR model estimation for the “open

system” group are reported in Table 6.4 below. Equation (1) through (3)

exhibited good ﬁts to the data, with signiﬁcant F-statistics (p < 0.05), and R2

values of 0.22, 0.21 and 0.28 respectively.

Table 6.4: SUR Contingency Model - Open Sﬁtem Sample (n=34)

 

Goodness of Fit Measures

 

2

 

 

 

 

Equation # Dependent Variable R F-statistic p-value
(1) Mfg. Process Complexity 0.2191 4.38 0.015
(2) Organizational Complexity 0.2088 3.91 0.023
(3) Supply Base Complexity 0.2814 5.47 0.006

 

76

 

Table 6.4 (cont)

 

Parameter Estimates

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Coefﬁcient Standard Error t-statistic p-value
Equation (1 )
Process Complexity
Constant 8.662 1 .905 4.55 0.000
Product Modularity 0.1 51 0.060 2.51 0.014
No of Mfg. Processes 0.343 0.303 1.13 0.260
Equation (2)
052- Complexity
Constant 6.137 1.261 4.87 0.000
Product Modularity 0.075 0.049 1.55 0.124
No. of Divisions 0.417 0.21 7 1 .92 0.058
Equation (3)
Suppl) Base Complx.
Constant 4.506 1.845 2.44 0.016
Product Modularity 0.140 0.053 2.62 0.010
No. of Suppliers 0.744 0.310 2.40 0.018

 

 

Breusch-Pagan Test of Independence: Chi-square=2.385, df=3 p=0.497

 

Hypothesis 4b states that in open systems, Product Modularity is negatively
related to the complexity of internal manufacturing systems. Although the t-
statistic is signiﬁcant (p=0.014) at the 0.05 alpha level for the Product Modularity
coefﬁcient in equation (1 ), the hypothesis is not supported because the direction
of the relationship is counter to what theory predicts. Hypothesis 5b states that in
open systems, Product Modularity is negatively related to the complexity of
organizational design and coordination. This hypothesis is not supported, as

evidenced by a nonsigniﬁcant t-statistic (p=0.124) on the Product Modularity

77

 

coefﬁcient in equation (2). Hypothesis 6b states that in open systems, Product
Modularity is negatively related to the complexity of supply bases. Although the
t-statistic on the Product Modularity coefﬁcient in equation (3) is signiﬁcant at the
0.05 alpha level (p=0.01), the hypothesis is not supported because the direction

of the relationship is counter to what theory predicts.

6.4.3 Moderation Model

The ﬁnal set of hypotheses, H7 through H9, considers the effects of
Environmental Uncertainty on the relationship between Product Modularity and
Complexity at the three observed organizational levels. These hypotheses
conjecture that Environmental Uncertainty has a moderating effect on the
relationship. In other words, it is argued that the presence of Environmental
Uncertainty affects the strength of the relationship between the study variables.
In order to test these hypotheses, the sample was again split into two (2) groups,
a high uncertainty group (n=39) and a low uncertainty group (n=49), with the
mean value of the summated 4-item Environmental Uncertainty scale (mean =
17.22) serving as the separator. The SUR model was ﬁt to the data for both
groups separately, and results of each estimation were compared in order to test
hypotheses seven (H7) through nine (H9). Results of the SUR model estimation
for the high and low uncertainty groups are summarized in Tables 6.5 and 6.6

below.

78

Table 6.5: SUR Moderation Model - Low Environmental Uncertainty Sample (n=49)

 

Goodness of Fit Measures

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Equation # Dependent Variable R2 F-statistic p-value
(1) Mfg. Process Complexity 0.1434 3.18 0.045
(2) Organizational Complexity 0.0673 0.96 0.386
(3) Supply Base Complexity 0.1459 2.91 0.058
Parameter Estimates
Coefﬁcient Standard Error t-statistic p-value
Equation (1)
Process Complexity
Constant 10.18 1.729 5.88 0.000
Product Modularity 0.030 0.048 0.61 0.541
No of Mfg. Processes 0.602 0.241 2.49 0.014
Equation (2)
0L9 Complexity
Constant 8.212 1.437 5.71 0.000
Product Modularity 0.01 7 0.053 0.32 0.751
No. of Divisions 0.304 0.237 1 .28 0.202
Equation (3)
Supply Base Complx.
Constant 6.312 1.806 3.50 0.001
Product Modularity 0.073 0.050 1.46 0.146
No. of Suppliers 0.612 0.273 2.24 0.026

 

 

Breusch-Pagan Test of Independence: Chi-square=12.346, df=3 p=0.006

 

79

 

Table 6.6: SUR Model — High Environmental Uncertainty Sample (n=39L

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Goodness of Fit Measures
Equation # Dependent Variable R2 F-statistic p—value
(1) Mfg. Process Complexity 0.1527 3.47 0.035
(2) Organizational Complexity 0.0464 0.92 0.403
(3) Supply Base Complexity 0.1352 3.06 0.051
Parameter Estimates
Coefﬁcient Standard Error t-statistic p-value
Equation (1)
Process ComplexitL
Constant 7.554 1.850 4.08 0.000
Product Modularity 0.154 0.075 2.06 0.042
No of Mfg. Processes 0.369 0.369 1 .00 0.319
Equation (2)
Org Complexity
Constant 7.748 1.520 5.10 0.000
Product Modularity 0.078 0.069 1 .1 3 0.260
No. of Divisions 0.097 0.260 0.37 0.709
Equation (3)
Supply Base Complx.
Constant 6.819 1.515 4.50 0.000
Product Modularity 0.137 0.055 2.47 0.015
No. of Suppliers 0.026 0.304 0.08 0.933

 

 

Breusch-Pagan Test of Independence: Chi-square=8.541, df=3 p=0.036

 

In both groups, equation (1) exhibited a good ﬁt to the data, with signiﬁcant F-

statistics (p < 0.05) and R2 values of 0.14 and 0.15. Equation (3) exhibited a

marginal ﬁt to the data in the low and high uncertainty groups with F-statistics of

 

2.91 (p=0.058) and 3.06 (p=0.051) and R2 values 0.15 and 0.14 respectively.

Equation (2) exhibited a poor ﬁt to the data in both groups with nonsigniﬁcant F-

statistics (p >> 0.05) and R2 values below 0.07.

Hypothesis seven (H7) states that the effect of Product Modularity on lntemal
Manufacturing Process Complexity is stronger under conditions of high
environmental uncertainty, relative to low environmental uncertainty. This
hypothesis is supported, as evidenced by two ﬁndings: (1) the coefﬁcient on
Product Modularity is not signiﬁcant for the low uncertainty group at an alpha
level of 0.05, but is positive and signiﬁcant for the high group; (2) the point
estimate for the Product Modularity coefﬁcient for the high uncertainty group
(0.154) is not contained within the 95% conﬁdence interval of the Product
Modularity coefﬁcient for the low uncertainty group (0066 to 0.125). Both of
these ﬁndings suggest that the relationship between Product Modularity and
lntemal Manufacturing Process Complexity is stronger under conditions of high

environmental uncertainty.

Hypothesis eight (H8) states that the effect of Product Modularity on the
complexity of organizational design and coordination is stronger under conditions
of high environmental uncertainty, relative to low environmental uncertainty. This
hypothesis was not supported. The coefﬁcient on Product Modularity was

nonsigniﬁcant in both the low and high uncertainty groups.

81

Hypothesis nine (H9) states that the effect of Product Modularity on Supply Base
Complexity is stronger under conditions of high environmental uncertainty,
relative to low environmental uncertainty. This hypothesis received partial
support. In the low uncertainty group, the coefﬁcient on Product Modularity was
not signiﬁcantly different from zero at the 0.05 alpha level, whereas in the high
uncertainty group, the coefﬁcient was positive and signiﬁcantly different from
zero. This is an indication that the relationship between Product Modularity and
Supply Base Complexity is stronger for the high environmental uncertainty group.
However, the point estimate for the Product Modularity coefﬁcient for the high
uncertainty group (0.137) falls within the 95% conﬁdence interval of the Product
Modularity coefﬁcient for the low uncertainty group (-0.026 to 0.171). This result

indicates that the equality of the two coefﬁcients cannot be ruled out statistically.

6.5.0 Chapter Summary

In this chapter, discussions of preliminary analysis procedures and the method of
analysis for hypothesis testing, Seemingly Unrelated Regression (SUR), were
provided. Tests of twelve hypotheses were performed and a number of key
ﬁndings were observed. First, Product Modularity has a positive and statistically
signiﬁcant inﬂuence on Supply Base Complexity. This result was present in the
entire sample (n=88), as well as for the “open system" group (n=34), and the high
environmental uncertainty group (n=39). Product Modularity was also found to
have a positive and statistically signiﬁcant effect on lntemal Manufacturing

Process Complexity. This result was evident in the “open system” group (n=34)

82

and the high environmental uncertainty group (n=39). Product Modularity,
however, was not found to have a statistically signiﬁcant effect on the complexity
of organizational design and coordination for any context examined in this study.
A summary of the hypotheses tested and their results are provided in Table 6.7

below.

Table 6.7: Summary of Hypothesis Testing

 

 

 

Statement of Hypothesis Result

H1 Product modularity inﬂuences the complexity of Not supported
internal manufacturing processes.

H2 Product modularity inﬂuences the complexity of Not supported

organizational design and coordination.

 

H3 Product modularity inﬂuences the complexity of a Supported
ﬁrm '3 supply base.

 

H4a In closed systems, product modularity is positively Not supported
associated with the complexity of internal
manufacturing processes.

 

H5a In closed systems, product modularity is positively Not supported
associated with the complexity of organizational
design and coordination.

 

H6a In closed systems, product modularity is positively Not supported
associated with the complexity of ﬁrms’ supply
bases.

 

H4b In open systems, product modularity is negatively Not supported
associated with the complexity of internal
manufacturing processes.

 

H5b In open systems, product modularity is negatively Not supported
associated with the complexity of organizational
design and coordination.

 

H6b In open systems, product modularity is negatively Not supported
associated with the complexity of ﬁrms’ supply
bases.

 

H7 The effect of product modularity on internal Supported
manufacturing process complexity is stronger under
conditions of high environmental uncertainty.

 

H8 The effect of product modularity on organizational Not supported
complexity is stronger under conditions of high
environmental uncertainty.

 

 

H9 The effect of product modularity on supply base Partially supported
complexity is stronger under conditions of high
environmental uncertainty.

 

 

 

83

 

7.0.0 Discussion

The primary research questions posed in this study investigated whether
employing modularity at the product level has an impact on the complexity of
supporting production systems, and if so, under what conditions does this result
hold. The investigation of these questions involved breaking the production
system down into three components: 1) the lntemal manufacturing process; 2)
the organization (ﬁrm or SBU); and 3) the supply base. For each of these
organizational levels, complexity was operationalized and then measured using
perceptual scales. Operationalization and measurement of product modularity
was conducted in similar fashion. These constructs were then used in the testing
of twelve (12) hypotheses that address the primary research questions from a
systems perspective. The results of measurement analysis and hypothesis
testing are discussed in this Chapter, as well as the limitations of the study and

directions for future research.

7.1.0 Measurement Outcomes

Multi-dimensional measurement was attempted for each of the study constructs:
Product Modularity, Manufacturing Process Complexity, Organizational
Complexity, and Supply Base Complexity. For each of concept, constituent
dimensions were derived from the extant literature on modularity and complexity.
For modularity, three dimensions were proposed: 1) functional binding, 2)

interface standardization, and 3) decomposability. For complexity, three

84

dimensions were also studied: 1) number of elements, 2) heterogeneity of

elements, and 3) interdependence of elements.

Each of the dimensions of the study’s key constructs was evaluated for the
ﬁtness of its measurement properties. All constructs exhibited sound

measurement properties according to the Nunnally (1978) criteria.

The results of measurement analysis suggest that manufacturing process
complexity is driven primarily by the number of internal manufacturing processes
employed to produce an organization’s primary products. When considering a
typology of manufacturing processes (i.e., project, batch, job shop, continuous,
and assembly-line), for each type, the addition of a new process element
requires an extra level of coordination in order to produce a ﬁnished product.
This coordination may not be trivial, as in many cases it requires the
establishment of new exchange protocols, the consideration of processing times,
and the introduction of new quality Checkpoints. It can be reasoned that this
increased coordination effort stemming from an increase in the number of

processes results in higher perceived complexity.

The interdependence and heterogeneity of the manufacturing process were not
found to be signiﬁcantly associated with its complexity. Heterogeneity may have
little meaning in this context because processes are typically grouped by similar

or complementary operations; hence by their very nature they are different from

85

other processes. How different they are may have no effect on coordination
required to produce the ﬁnished good. Regarding interdependence, the
processes required to produce a ﬁnished good are necessarily dependent,
however the level of dependence may depend on the type of process employed,
with continuous and line ﬂow processes tending toward higher dependence, and
project, batch and job shop processes tending toward lower dependence.
Because process interdependence may affect the level of complexity in different
ways based on the type of process employed, its speciﬁc effect could be masked

in the aggregate sample analyzed in this study.

From an organizational standpoint, both the number of organizational elements
and the heterogeneity of organizational elements were found to signiﬁcantly
contribute to organizational complexity, while the interdependence of
organizational elements showed no signiﬁcant association with complexity.
These results suggest that the larger and more dispersed an organization, and
the more it employs dissimilar business systems, the more complex it is — a fairly
intuitive ﬁnding. Less intuitive, however, is the absence of a relationship between
organizational complexity and organizational interdependence. It may be that
while interdependence increases the structural complexity of an organization,
coordination is simpliﬁed due to better information ﬂow. Consider the case of a
matrix structure, a highly interdependent organizational design: the vertical
(project team) and horizontal (functional group) assignment of human resources

creates a more intricate organizational structure, but the bilateral communication

86

that results could simplify coordination due to higher potential for shared
knowledge. These offsetting effects could be responsible for the absence of a
signiﬁcant relationship between organizational interdependence and

organizational complexity in this study.

Supply base complexity was shown to be inﬂuenced by both supply base
heterogeneity and supply base interdependence; however, the relationship with
supply base interdependence was negative. With regard to heterogeneity, the
necessity of managing relationships with suppliers in different ways due to
differences in size, location, and industry may serve to increase the amount of
effort required for effective coordination. This is especially salient in situations
where these differences create imbalance in the power distribution between
buyer and suppliers, causing posturing and poor information ﬂow to create a
more adversarial and complex procurement environment (Kraljic, 1983, Benton
and Maloni, 2005). The negative relationship between supply base
interdependence and supply base complexity suggests that increased
interdependence with suppliers results in decreased supply base complexity.
This is likely owed to a recent increase in supplier integration initiatives, and
ﬁrms’ consolidation of supply requirements with partner organizations. This
approach not only simpliﬁes supply management through reduction in supplier
count, but also the increased transparency and information ﬂow allows for

improved coordination and cost performance (Lee et al. 1997).

87

Surprisingly, the number of suppliers had no signiﬁcant inﬂuence on supply base
complexity. It could be that supply management organizations largely employ
standard systems for managing suppliers, and increases In the number of
suppliers pose no additional coordination burden, provided that the

characteristics of transaction are not unique.

Overall, it was shown that the theoretically derived dimensions of Number,
Heterogeneity, and Interdependence all have an inﬂuence on complexity.
However, both the presence and the magnitude of these inﬂuences vary by level
of analysis. This ﬁnding has important implications from both theoretical and
practical perspectives. Complexity theory implies that the simultaneous
presence of these three dimensions gives rise to complexity; however, it has
been demonstrated here that the composition of the complexity construct is
dependent upon the type of complexity being evaluated. In fact, in no case
observed were all three dimensions signiﬁcantly related to the construct. As
such, it should be noted that direct perceptual measures of complexity are
preferable to indirect measures due to the contingent nature of this construct.
Measurement of complexity as a composite of its constituent dimensions may
result in the inclusion of items that serve only to increase measurement error. In
addition, direct perceptual measurement offers a more parsimonious approach,

as fewer items are required to achieve sound measurement properties.

88

Product modularity, on the other hand, was shown to be reliably measured as a
composite of its theoretically derived dimensions, although its dimensions were
not found to be independent of one another. Each dimension exhibited very
strong correlation with its counterparts, which is an indication of the unity of the
modularity construct. These results supports recent literature that suggest that
modularity is a multi-dimensional concept (Schilling, 2000; Salvador, 2007), but
also illustrate that its measurement need not establish each dimension as a
distinct measurable element. Rather measurement may proceed by merely

ensuring that each dimension is adequately represented in the modularity scale.

7.2.0 Results and Conclusions of Hypothesis Testing

Product modularity was found to be positively associated with the complexity of
lntemal manufacturing processes under conditions of high outsourcing and high
environmental uncertainty, and the relationship was nearly signiﬁcant (p=0.055)
in the general sample. A quick tour through a modern automotive assembly plant
Is all that is required to understand this ﬁnding. In this setting, it is not
uncommon for assemblers to receive cockpits, seating systems, headliners, and
even drivetrains as pre-assembled modules. The assembly of these modules
into a ﬁnished vehicle is a symphony of highly coordinated deliveries, automated
part storage and retrieval, and integrated component assembly, inspection, and
certiﬁcation. The conﬁguration of these plants represents a modern spin on the
Henry Ford assembly line, where standardized components have been replaced

by modules with standardized interfaces. The tremendous number of product

89

conﬁgurations made possible through the receipt of different ﬂavors of each
module requires that the assembly process itself has enough sophistication (or
complexity) to maintain high volume throughput, while simultaneously providing
the level of customization demanded by consumers. The necessary level of
process sophistication increases with the number of assemblies received from
external suppliers and the variability in consumer demand for different product

conﬁgurations.

Product modularity was also found to be positively associated with the complexity
of the supply base in general, as well as under conditions of high outsourcing
and high environmental uncertainty. This result is best illustrated through the
example of the Boeing 787 Dreamliner, a highly modularized and extensively
outsourced aircraft design. Problems with the integration of modules and
coordination difﬁculties with suppliers for component design and delivery plagued
this product from the start, and resulted in a launch delay of over two years. As a
result, Boeing has vowed to reign in much of the design for the next generation of
this product in efforts to reduce the complexity in its supply chain (Sanders,
2009). This result was also typical of software, telecommunications and
consumer electronics ﬁrms evaluated in a study by Staudenmeyer et al. (2005),
where modularization and concurrent design of products across partnering ﬁrms
resulted in increases in supply base complexity, as well as difﬁculty in managing

the design and release of new products.

90

Product modularity was not, however, found to have an association with the
complexity of organizational design and coordination for any context examined.
Through conversations with focal ﬁrm participants, it became clear that the
organizational transformation that follows product modularization quite often
represents a ﬁnal phase of transition. It was observed that only after a ﬁrm had
experience in designing and launching modular products, and had decided to
adopt modularization as its primary product design strategy, that it began to
realize how its organization could best be restructured to support this strategy.
Because modular design is still relatively new to many of the industries under
study here, it is possible that many of the ﬁrms observed have yet to reach this

critical stage of development.

As evidenced by these findings, the open and closed systems framework may
not accurately characterize the relationship between modularity and complexity in
production systems. It was hypothesized that in an open system setting that
product modularity would trigger a decrease in complexity, and the converse was
hypothesized for closed systems. It was discovered, however, that in all cases
where product modularity was signiﬁcantly related to production system
complexity, the result was an increase in complexity. One reason for this result
may be that the proxy used to represent system openness (degree of
outsourcing) was not effective in segmenting the sample according to the true
characteristics of open and closed systems. However, the degree of outsourcing

is a good indicator of the extent of control an organization has over its product

91

design, which is consistent with the arguments presented to support open and
closed systems hypotheses in section 3.2.2. Another reason could be that the
extent of use of modularity in closed or low-outsourcing systems observed in the
study was systematically low, hence lower levels of modularity application and
decreased variability preclude the ability to detect a relationship with complexity.
However, a comparison of the means and standard deviations of modularity
scale values between low and high outsourcing groups indicates that both the
extent of modularity use and its corresponding variability are not statistically
different across groups at a 5% signiﬁcance level. It is likely then that another
theoretical perspective offers a better explanation of the phenomenon under

study than open and closed systems.

As an alternative explanation, a Law of Conservation of Complexity is proposed.
Similar to the Law of Conservation of Energy which states that the energy in an
isolated system cannot be destroyed, this proposition suggests that complexity in
a production system is not eliminated — like energy, it merely Changes form. As
such, the application of modular product architecture may reduce complexity at
the product level by lowering part count and minimizing interdependencies
(Baldwin and Clark, 2000), while increasing complexity in other parts of the
production system. The ﬁndings herein support this notion of a complexity shift,
as all signiﬁcant relationships observed indicate an increase in other forms of
complexity (i.e., manufacturing process complexity and supply base complexity)

associated with product modularity.

92

7.3.0 Contributions to Business Theory and Practice

This study contributes to the Operations and Supply Chain Management (OSM)
body of knowledge and business practice in a number of ways. First, it provides
empirically validated perceptual measures of Product Modularity, Manufacturing
Process Complexity, Organizational Complexity and Supply Chain Complexity.
Measures of these concepts have been called for by Salvador (2007) and Choi
and Krause (2006). Product modularity was measured using a six-item scale,
with items representing each of its three theoretically derived dimensions:
functional binding, interface standardization, and decomposability. Complexity
measures were each composed of three items that tap the construct directly by
asking questions about structural intricacy and difﬁculties in coordination. These
scales can be utilized, as well as further reﬁned and validated, in future studies

observing modularity and complexity in an OSM context.

Next, it was found that complexity is composed of a different mix of its
theoretically derived dimensions, contingent upon the type of complexity under
study. Organizational complexity was found to be driven by the number and
heterogeneity of organizational elements; supply base complexity was shown to
be positively inﬂuenced by supplier heterogeneity and negatively inﬂuenced by
the interdependence between organizations and their suppliers; and
manufacturing process complexity was found to be a function of only the number

of manufacturing processes employed in the creation of ﬁrms’ primary products.

93

This ﬁnding is important in that it highlights the areas which most impact
complexity at different organizational levels, thereby allowing for more focused

efforts to be applied toward its reduction.

The study examined the impact of a product-level characteristic on the
Characteristics of its supporting production system. This was done by employing
a multi-level research design, whereby information was collected about both the
technical and managerial aspects of an organization. Very few studies in
Operations and Supply Management have employed this type of design, even
though it is crucial to understanding the broader impacts of strategically-

motivated decisions.

The ﬁndings demonstrate that the Open / Closed Systems framework does not
accurately characterize the relationship between modularity and complexity in
production systems. It was hypothesized that in an open system setting that
product modularity would trigger a decrease in complexity, and the converse was
hypothesized for closed systems. It was discovered, however, that in all cases
where product modularity was signiﬁcantly related to production system
complexity, the result was an increase in complexity. A new framework, the Law
of Conservation of Complexity, was proposed as an alternative explanation for
the behavior exhibited. Perhaps this framework can be applied to further

examination of complexity behavior in other organizational systems.

94

Finally, the study provides managers with deeper insight regarding the
consequences of pursuing modularity as a design strategy. Results indicate that
product modularity is associated with increases in manufacturing process
complexity and supply base complexity. This is a key ﬁnding, as most studies on
the subject focus on the performance beneﬁts of employing modularity, such as
improved ﬂexibility (Worren et. al, 2002) and faster product development cycles
(Fine, 2000), without discussing the potential challenges that may arise as a
result of its application. The ﬁndings indicate that there is a tradeoff to pursuing
a modular product design strategy. Although modularity may serve to simplify
product designs, enabling a quicker turn around on product updates, it may
create stress in other parts of the manufacturing and product delivery system. In
cases where complexity is more cost-effectively managed on the support side of
manufacturing, this tradeoff may be worthwhile, as in the case of highly complex
products. Conversely, this may not be the case for products with simpler product
designs, as the cost of managing changes at the product level may be low in
comparison to the costs of managing complexity in manufacturing processes and
supply chains. This information should help managers make more informed
decisions regarding the choice to pursue product modularity, and better prepare

them for the likely impact on their production support systems.

7.4.0 Limitations and Directions for Future Study
The use of single respondents could have resulted in common method bias.

While this study did not rely solely on single respondents, the percentage of

95

paried observations from two individuals within an organization (29%) was fairly
low in comparison to observations received from a single respondent (71%).
This approach, in conjunction with the randomization of questions during survey
implementation and the varying of the sentence structure of questions, both
recommended by Podsakoff et al. (2003), was employed to reduce the potential
for common method bias. As a test to determine if the use of common methods
skewed the results of the study, the parameter estimates of the SUR model ﬁt to
the data for single respondents were compared against those produced by the
data from dual respondents. The was no statistical difference in the regression
coefﬁcients for the main study variables between the two groups, indicating that
common method bias does not pose a serious threat to the validity of the study.
However, future studies of this nature should strive to employ the multiple-

respondent approach to a greater extent.

The level of agreement between respondents from the same organization, as
evaluated using the Intraclass Correlation Coefﬁcient (ICC), was slightly lower
than recommended. Boyer and Venna (2000) suggest 0.60 as a minimum value
for the average ICC across questions asked of multiple respondents from a
single organization in operations research. The average ICC across four
questions asked of managers and technical personnel in this study was 0.58.
The was likely due to the fact that respondents from the same organization were
recruited for participation based on differences in their skill sets and knowledge

bases. It is unlikely that individuals in the technical sphere of an organization

96

share the same perspectives of the business as those in management, therefore
greater differences were likely to be detected here than in situations where
respondents of similar function are recruited to respond to the same questions.
Future studies of this type should place greater emphasis on identifying those
questions which have a high likelihood of being perceived similarly by individuals
operating at different levels of the organization in order to provide a stronger test

of agreement.

Although the sample sizes for both the technical (n=169) and managerial (n=188)
samples were comparable to most large scale studies performed in Operations
and Supply Management, the number of merged observations (n=94) was
somewhat lower than desired. As a result, the study could not employ the purely
multivariate analysis methods for testing hypotheses that are typical of this kind
of research. However, the methods selected were carefully chosen to minimize

the potential shortcomings of small sample analysis.

In addition, the study employed two-item measures for manufacturing process
heterogeneity and supply base heterogeneity. Studies intending to test
hypotheses including these concepts should employ more items for
measurement to ensure that these constructs exhibit sound psychometric

properties.

97

This study raises many new questions concerning both modularity and
complexity, as they relate production systems, which were not addressed with
analyses performed for this study. The ﬁndings of measurement analysis
revealed that a different mix of the theoretically derived dimensions of complexity
were signiﬁcantly associated with various types of production system complexity.
It was reasoned that different types of processes and organizational designs
could be responsible for the differences in these ﬁndings. Future studies could
examine this phenomenon by explicitly accounting for these differences in
process and organizational composition. Also, interdependence with suppliers
was found to be negatively related to supply base complexity. The diffusion of
the principles of Supply Chain Management, and in particular the increasing
deployment of Supplier Integration as a strategic initiative, was offered as a
possible explanation for this result. Follow up studies could perform an explicit

test of the relationship between supplier integration and supply base complexity.

Finally, the role of time was introduced as a possible explanation for the failure to
detect a relationship between product modularity and organizational complexity.
Future studies should employ a longitudinal design in order to effectively
examine this relationship. Longitudinal analyses would also be well suited for
testing the Law of Conservation of Complexity proposition. This approach would
allow researchers to observe the mechanism by which prOduct modularity

facilitates the transformation of production support systems.

98

APPENDICES

99

APPENDIX A. Research Questionnaire

Table A.1 - Technical Section

 

Please indicate the extent to which you agree or disagree with the following
questions regarding the characteristics of your primary products.

(Please answer anywhere from 1=Strongly Disagree to 7=Strongly Agree}

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

PD-FB1 For our products, we can make changes in components without changing other
components.

PD-FBZ For our products, functions can be directly added or deleted by adding or
removing components.

PD-F83 Our product design is such that a given product function is performed by a single
product component.

PD—FB4 Our products have interchangeable features and options.

PD-IS1 Our product components have standardized interfaces.

PD-ISZ The interfaces of our product components are designed to accept a variety of
components.

PD-IS3 Our product components are able to accept a wide range of complements (such
as modules or peripherals).

PD-IS4 Each component interface is unique. (reversed item)

PD-DC1 Our products can be decomposed into separate modules.

PD-DC2 Our products are designed to be easily disassembled and reassembled.

PD-DC3 Our products are designed to be easily reconﬁgured.

Please indicate the extent to which you agree or disagree with the following
questions regarding the characteristics of your production processes.
{Please answer anywhere from 1=Strongly Disagree to 7=Strongly Agree}

PC-INT1 Our production processes are broken down into sub-processes that can operate
independently. (reversed item)

PC-INT2 Sub-processes can be added to or removed from our production processes
without changing other sub-processes. (reversed item)

PC-INT3 Our production processes are designed so that sub-processes can be added or
removed without signiﬁcant disruptions/changes to other sub-processes. (reversed
item)

PC-HTR1 The products we make have similar processing and assembly requirements.
(reversed item)

PC-HTR2 Only minor equipment modiﬁcations are required to produce different products.
(reversed item)

PC-HTR3 We are frequently forced to make changes in our manufacturing processes to

 

satisfy differing customer requirements.

 

100

 

Table A.1 (cont)

 

 

 

 

PD-MOD1 Our products are designed to enable the swapping of components.

PC-CMP1 Our production system is very sophisticated in terms of its operating
characteristics.

PC-CM P2 Our production system is very complex.

PC-CM P3 Our production system is very sophisticated in terms of structure/layout.

 

 

Please answer the following question regarding the nature of your operation
by selecting from the options provided below.

 

 

 

 

Number of internal manufacturing processes:

(1) —1

(2) —2or3
(3) —4or5
(4) —6or7
(5) —80r9
(6) —10or11

(7) — 12 ormore

 

Table A2 - Managerial Section

 

 

Please indicate the extent to which you agree or disagree with the
following questions regarding the characteristics of your firm or SBU.

{Please answer anywhere from 1=Strongly Disagree to 7=Strongly Agree}

 

 

 

 

 

 

 

 

 

 

OG-HTR1 Sub—units in our ﬁrm/SBU are subject to the same policies (ex.: quality
standards, compensation, absenteeism, etc. ). (reversed item)

OG-HTR2 Sub-units in our ﬁrm/SBU employ similar or compatible business practices.
(reversed item)

OG-HTR3 Sub-units in our ﬁrm/SBU have similar or compatible information
systems/technologies. (reversed item)

OG-HTR4 Sub-units in our firm/SBU use common mechanisms for coordinating work.
(reversed item)

OG-HT R5 The organizational structures of sub-units in our ﬁrm/SBU are similar. (reversed
item)

OG-INT1 Project teams are self—directed. (reversed item)

OG-INT2 Project teams operate autonomously. (reversed item)

OG-INT3 Project teams operate independently from other organizational entities.
(reversed item)

OG-INT4 Project teams are free to develop links with other teams and/or internal units in

 

our ﬁrm/SBU. (reversed item)

 

101

 

 

Table A2 (cont.)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

OG-INT5 Project teams are free to develop links with external units such as suppliers,
customers, etc. (reversed item)

OG-CMP1 Our ﬁrm/SBU has a very complex organizational structure.

OG-CMP2 Our ﬁrm/SBU is organized in a straight-forward, uncomplicated way. (reversed
item)

OG-CMP3 Coordinating work between sub-units in our ﬁrm/SBU is difficult.
Please indicate the extent to which you agree or disagree with the
followlng questions regarding the characteristics of your supply chain.
{Please answer anywhere from 1=Strongly Disagree to 7=Strongly Agree}

SB-INT1 We have frequent face-to-face communication with key suppliers.

SB-INT2 For a given supplier, there are multiple points of contact between our
organization and the supplier's.

SB-INT3 We have long-term relationships with our suppliers.

SB-INT4 We have strong ties to our suppliers.

SB—HTR1 Our suppliers are diverse in terms of size, location, and products supplied.

SB—HTR2 Effective management of procurement requires taking different approaches with
different suppliers.

SB—CMP1 Coordinating procurement and manufacturing operations is problematic for our
ﬁrm/SBU.

SB—CMP2 The structure of our supply chain is difficult to describe or illustrate.

SB-CMP3 Our supply chain is a very complex network.
Please complete the following statements regarding the market for your
primary product line by selecting from the responses provided below.

EN-UNC1 Sales forecasts for our primary products are likely to be: (very inaccurate. . . very
accurate)

EN-UNC2 Predicting the actions of our competitors is: (very easy. . .very difficult)

EN-UNC3 Predicting consumer tastes in our market(s) is: (very easy. . .very difﬁcult)

EN-UNC4 Monitoring market trends for our primary products is: (very easy. . .very difﬁcult)

EN-UNCS Predicting demand for our primary product line is: (very easy. . . very difficult)

 

 

102

 

Table A2 (cont)

 

 

Please provide the following demographic data for your firm or SBU by
selecting from the options provided below.

 

Number of worldwide locations managed by your ﬁrm or SBU:

(1) — 1

(2) — 2 to 10

(3) - 11 to 20

(4) — 21 to 30

(5) - 31 to 40

(6) — 41 to 50

( 7) - more than 50

 

Number of organizational levels in your ﬁrm or SBU - from the lowest level
employee to the top management position:

(1) —1

(2) -2or3

(3) -4or5

(4) —6or7

(5) —80r9
(6)-100r11
(7) -12 ormore

 

Number of operating divisions:

(1) —1

(2) -2or3

(3) —4or5

(4) —60r7

(5) —80r9

(6) -10or11
(7) —12ormore

 

Number of suppliers used by your ﬁrm or SBU:

(1) -Iess than 10
(2) — 10 to 50

(3) - 51 to 100

(4) - 101 to 500

(5) - 501 to 1000

(6) — 1001 to 5000

( 7) — more than 5000

 

 

Please provide the following demographic data for your ﬁrm or SBU:

Approximate values are acceptable

 

Number of employees

 

 

 

Total sales

 

103

 

 

APPENDIX B. Sample Invitation and Follow-up Letters

Exhibit 3.1 — Invitation Letter

From: Delvon B. Parker [msubcrt@bus.msu.edu]
Sent: Tuesday, April 14, 2009
Subject: Modularity Research with Michigan State University

Dear <recipients full name>,

I am writing you today to ask for your participation in a study on the concept of Modularity that I
am conducting with the Department of Supply Chain Management at Michigan State University's
Eli Broad College of Business. I am polling manufacturing professionals, such as yourself, in the
electronics and transportation industries to get an idea of how extensive modularity is applied in
US-based ﬁrms.

Participation in the study is completely voluntary and involves only the completion of a 25-minute
online survey. Your insights are critical in helping us to evaluate modularity’s ability to impact key
performance measures, reduce operational complexity, and accelerate innovation. All participants
will receive a summary report of the study's ﬁndings.

The survey, which is split into two parts (technical and managerial), asks a series of very general
questions about your products, the manufacturing processes that produce them, your
organization and its supply base. Most of the questions are agree/disagree type questions that
address characteristics of your products, processes and organization in a very indirect way.

In order to access the survey, please follow the link for the section that is most relevant to your
current role, and enter the password when prompted. The links for each section, and the
password (which works for either section) are provided below:

 

 

Survey Link: Managerial Section
Survey Link: Technica_l Section
Password: modres2009

Be assured that information collected for this study will be kept conﬁdential to the extent allowable
by local, state and federal law. Completed surveys will be stored securely and will be available
only to members of the research team. No reference will be made in any oral or written report
that would link you individually, or your company, to the study.

Thanks in advance for your interest and participation in the study. If you have any questions or
concerns, you can reach me by phone during normal business hours at (517) 432-6455, or at any

time by e-mail at msubcrt@bus.msu.edu.

Thanks again,

Delvon B. Parker

Doctoral Candidate, Operations Management
Eli Broad College of Business

Michigan State University

104

Exhibit 8.2 - Follow-up Letter #1

From: Delvon B. Parker [msubcrt@bus.msu.edu]
Sent: Tuesday, April 21, 2009
Subject: Follow up request for participation in modularity research with MSU

Dear <recipients full name>,

I hope you received my prior invitation to participate in research on Modularity with Michigan
State University. Interest in this concept has grown steadily over the past decade, and many
ﬁrms have begun adopting modularity as their principal product design strategy. However, the
impact of modularity on operational performance is still somewhat unclear. Your insights will help
us gain a better understanding of how exactly modularity affects various aspects of performance
(such as costs, quality, innovation, etc.). Even if you are not currently employing modularity, your
responses are extremely valuable for comparison purposes. Results will be made available to all
study participants.

Over 120 ﬁrms in electronics, transportation and closely related industries have already
participated, and I hope that you will strongly consider participation as well. The two-part survey
(technical and managerial) contains very general questions about your products and
manufacturing processes, your organization and its supply base. Nearly all of the questions are
of the agree/disagree type, and the survey takes only 25 minutes to complete. You may even
elect to participate anonymously.

In order to access the survey, please follow the link for the section that is most relevant to your
current role, and enter the password when prompted. The links for each section, and the
password (which works for either section) are provided below:

Survey Link: Managerial Section
Survey Link: Technical Section
Password: modre32009

Please do not hesitate to contact me if you have any questions regarding any aspect of the study.
I can be reached by phone Monday through Friday, 8:00 am. to 5:00 pm. EST at (517) 432-
6455, or at any time by e-mail at msubcrt@bus.msu.edu.

Thank you very much for further considering participation,

Delvon B. Parker

Doctoral Candidate, Operations Management
Eli Broad College of Business

Michigan State University

Phone: (517) 432—6455

105

Exhibit 8.3 — Follow-up Letter #2

From: Delvon B. Parker [msubcrt@bus.msu.edu]
Sent: Monday, April 27, 2009
Subject: Thanks for your patience and participation

Dear <recipients full name>,

Thank you very much for your patience in receiving my prior e-mails. Michigan State University
relies strongly on industry professionals like you to provide a much needed real-world perspective
to our research and educational efforts, which is why I have contacted you so frequently. Without
your support, the relevance of the work we do with industry, and in the classroom, suffers
tremendously.

If you have already participated in the modularity research project by completing our online
survey, we extend you our most sincere thanks. You will receive a report on the major ﬁndings of
the research as soon as our analysis is complete.

If you have not yet had the opportunity to complete the survey, I would like to take this ﬁnal
opportunity to ask for your participation in this very important research. The survey will close on
Friday, May 1, 2009 at 5:00 pm. (EST). Twenty-ﬁve minutes of your time will go a long way
toward helping us learn more about modularity applications and their true value to the
manufacturing sector.

To access the survey, please follow the link for the section that is most relevant to your current
role, and enter the password when prompted.

Survey Link: Managerial Section
Survey Link: Technical Section
Password: modres2009

Thanks again for your interest, your patience, and your participation. Your response will help
shape education on modularity, as well as its future application.

Sincerely,

Delvon B. Parker

Doctoral Candidate, Operations Management
Eli Broad College of Business

Michigan State University

Phone: (517) 432-6455

E-mail: msubcrt@bus.msu.edu

106

APPENDIX C. Sample Breakdown by Various Categories

 

By Contact Information Source

Institute of Industrial Engineers (IIE)

Heavy Equipment Manufacturer and Tier 1 Suppliers

Society of Manufacturing Engineers (SME)

Household Appliance Manufacturer and Tier 1 Suppliers

Department of Marketing and Supply Chain Management (MSU) Mailing List

Aircraft Engine Manufacturer and Tier 1 Suppliers

Total Responses

No. in
sample
(% of total)
97 (33%)
72 (25%)
65 (22%)
29 (10%)
16 (6%)
11 (4%)
290

 

By Industry (Industry Classiﬁcation Code)

Transportation Equipment Manufacturing (NAICS: 336)
Fabricated Metal Product Manufacturing (NAICS: 332)
Electrical Equipment, Appliance, and Component Manufacturing (NAICS: 335)
Miscellaneous Manufacturing (NAICS: 339)

Primary Metal Manufacturing (NAICS: 331)

Machinery Manufacturing (NAICS: 333)

Computer and Electronic Product Manufacturing (NAICS: 334)
Plastics and Rubber Products Manufacturing (NAICS: 326)
Printing and Related Support Activities (NAICS: 323)

Food Manufacturing (NAICS: 311)

Nonmetallic Mineral Product Manufacturing (NAICS: 327)
Wood Product Manufacturing (NAICS: 321)

Petroleum and Coal Products Manufacturing (NAICS: 324)
Furniture and Related Product Manufacturing (NAICS: 337)
Textile Product Mills (NAICS: 314)

Leather and Allied Product Manufacturing (NAICS: 316)
Chemical Manufacturing (NAICS: 325)

Other (Non-manufacturing industries)

Total Responses

83 (28.6%)
39 (13.4%)
38 (13.1%)
38 (13.1%)
18 (6.2%)
17 (5.9%)
16 (5.5%)
9 (3.1%)

5 (1.7%)
4(1.4%)

3 (1.0%)

2 (0.7%)

2 (0.7%)

2 (0.7%)

1 (0.3%)

1 (0.3%)

1 (0.3%)
11 (3.8%)
290

 

By Questionnaire Type

Managerial
Technical

Total Questionnaires

188 (52.7%)
169 (47.3%)
357*

 

By PairedINon-Palred Cases
Paired (Both Managerial and Technical Responses per SBU)
Non-Paired (Either Managerial or Technical Response Only per SBU)

Total Cases

94 (35.7%)
169 (64.3%)
263"

 

 

*Sixty-seven ( 67) individuals completed both the Managerial and Technical questionnaire.

290+67=357

“Twenty-seven (27) cases had two responses per SBU, one technical and one managerial.

290-27=263

 

107

 

APPENDIX D. Technical CFA - All Product and Process-Related Constructs

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Scale (Reliability) l Question Std.
Loading

Product Modularity (0.907)

PD-MOD1 Our products are designed to enable the swapping of components. 0.720"

PD-FB2 For our products, functions can be directly added or deleted by 0.707“
adding or removing components.

PD-F B4 Our products have interchangeable features and options. 0.767“

PD-ISZ The interfaces of our product components are designed to accept a 0.846"
variety of components.

PD-IS3 Our product components are able to accept a wide range of 0.820"
complements (such as modules or peripherals).

PD-DC3 Our products are designed to be easily reconfigured. 0.765“
Process Interdependence (0.836)

PC-INT1 Our production processes are broken down into sub-processes that 0.660"
can operate independently. (reversed item)

PC-INT2 Sub-processes can be added to or removed from our production 0.875"
processes without chang_ir_rg other sub-processes. (reversed item)

PC-INT3 Our production processes are designed so that sub-processes can 0.855"
be added or removed without signiﬁcant disruptions/changes to
other sub-processes. (reversed item)
Process Heterogeneity (2-item correlation=0.387)

PC-HTRl The products we make have similar processing and assembly 0.533"
requirements. (reversed item)

PC-HTR2 Only minor equipment modiﬁcations are required to produce different 0.720“
products. (reversed item)
Process Complexity (0.893)

PC-CMP1 Our production system is very sophisticated in terms of its operating 0.809"
characteristics.

PC-CMP2 Our production system is very complex. 0.882“

PC-CMP3 Our production system is very sophisticated in terms of 0.884”
structure/layout.
Fit Statistics: **p<0.05

 

Chi-Square=164.3, df=71; CFI=0.924; GFI=0.874;
SRMR=0.0731; RMSEA=0.089

 

 

 

108

 

APPENDIX E. Managerial CFA — All Organization and Supply Base-Related

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Constructs

Scale (Reliability) / Question Std.
Loading

Organizational Heterogeneity (0.807)

OG-HTR1 Sub-units in our ﬁrm/SBU are subject to the same policies (ex.: 0.739"
quality standards, compensation, absenteeism, etc. ). (reversed
item)

OG-HTR2 Sub—units in our firm/SBU employ similar or compatible business 0.779"
practicesdreversed item)

OG-HTR3 Sub-units in our firm/SBU have similar or compatible information 0.682"
systems/technologies. (reversed item)

OG-HTR4 Sub-units in our firm/SBU use common mechanisms for 0.677”
coordinating work. @versed item)
Organizational Interdependence (0.720)

OG-INT1 Project teams are self-directed. (reversed item) 0.598"

OG-INT4 Project teams are #66 to develop links with other teams and/or 0.829"
internal units in our ﬁrm/SBU. (reversed item)

OG-INT5 Project teams are free to develop links with external units such as 0.635"
suppliers, customers, etc. (reversed item)
Organizational Complexity (0.763)

OG-CMP1 Our ﬁrm/SBU has a very complex organizational structure. 0.803“

OG-CM P2 Our ﬁrm/SBU is organized in a straight-forward, uncomplicated 0.763“
way. (reversed item)

OG-CMP3 Coordinating work between sub-units in our firm/SBU is difficult. 0.62 7**
Environmental Uncertainty (0.802)

EN-UNC2 Predicting the actions of our competitors is: (very easy. . .very 0.631 **
difﬁcult)

EN-UNCS Predicting consumer tastes in our market(s) is: (very easy. . . very 0.801 **
difﬁcult)

EN-UNC4 Monitoring market trends for our primary products is: (very 0.758“
easy. . .very difficult)

EN-UNCS Predicting demand for our primary product line is: (very 0.660"
easy. . .very difﬁcult)
Supply Base Interdependence (0.753)

SB-INT1 We have frequent face-to-face communication with key suppliers. 0.608“

SB-INT3 We have long-term relationships with our suppliers. 0.799“

SB-INT4 We have strong ties to our suppliers. 0.756“

 

 

 

 

109

Appendix E (cont)

 

 

Supply Base Heterogeneity (2-item correlation=0.362)

 

 

 

 

 

 

 

 

SB—HTR1 Our suppliers are diverse in terms of size, location, and products 0.628“
supplied.

SB-HTR2 Effective management of procurement requires taking different 0.577"
approaches with different suppliers.
Supply Base Complexity (0.672)

SB-CMP1 Coordinating procurement and manufacturing operations is 0.636“
problematic for our ﬁrm/SBU.

SB-CMP2 The structure of our supply chain is difﬁcult to describe or illustrate. 0.677“

SB-CMP3 Our supply chain is a very complex network. 0.620”
Fit Statistics: **p<0.05

Chi-Square=365.3,df=188; CFI=0.876; GFI=0.854;
SRMR=0.067; RMSEA=0.071

 

 

 

 

 

110

 

 

APPENDIX F. SEM Analysis of Exploratory Model

 

   
   
    
   
      
  
 
   
   

  

Process
Complexity

0.235 (p=0.053)

     

   

Product
Modularity

0.160 (p=0-171) Organizational

Complexity

0.313" (p=0.022)

Supply Base
Complexity

Fit Statistics:
Chi-Square=194.9, df=87; CFI=0.835; GFI=0.788;
SRMR=O.132; RMSEA=0.119

Notes: **p<0. 05

 

 

 

111

BIBLIOGRAPHY

112

Andrews, F. (2000, January 26). Dell, It Turns Out, Has a Better Idea Than Ford.
The New York Times, p. 12.

Antonio, K. W., Yam, R. C., 80 Tang, E. (2007). The Impact of Product Modularity
on Competitive Capabilities and Performance: An Empirical Study.
International Journal of Production Economics, 105, 1-20.

Armstrong, J. S., 80 Overton, T. S. (1977). Estimating Nonresponse Bias in Mail
Surveys. Journal of Marketing Research, 14, 396-402.

Baldwin, C. Y. 80 Clark, K. B. (2000). Design Rules. The Power of Modularity
(Vol. 1). Cambridge, MA: The MIT Press.

Baldwin, C. Y., 80 Clark, K. B. (1997). Managing in an Age of Modularity. Harvard
Business Review, 75(5), 84-93.

Benton, W. C., & Maloni, M. (2005). The Inﬂuence of Power Driven Buyer/Seller
Relationships on Supply Chain Satisfaction. Journal of Operations
Management, 23(1), 1-22.

Boulding, K. E. (1956). General Systems Theory - The Skeleton of Science.
Management Science, 2(3), 197-208.

Boyer, K. K., & Venna, R. (2000). Multiple Raters in Survey-Based Operations
Management Research: A Review and Tutorial. Production and
Operations Management, 9(2), 128-140.

Calantone, R., Droge, C., & Vickery, S. (2002). Investigating the Manufacturing-
Marketing Interface in New Product Development: Does Context Affect the
Strength of Relationships? Journal of Operations Management, 20(3),
273-287.

Child, J. (1972). Organization Structure, Environment and Performance: The
Role of Strategic Choice. Sociology, 6(1), 1-22.

Choi, T. Y., & Krause, D. R. (2006). The Supply Base and its Complexity:
Implications for Transaction Costs, Risks, Responsiveness, and
Innovation. Journal of Operations Management, 24, 637-652.

Cronbach, L. (1951). Coefﬁcient Alpha and the lntemal Structure of Tests.
Psychometrika, 16(3), 297-334.

113

Deshmukh, A. V., Talavage, J. J., & Barash, M. M. (1998). Complexity in
Manufacturing Systems, Part 1: Analysis of Static Complexity. llE
Transactions, 30, 645-655.

Devaraj, S., Hollingworth, D. G., & Schroeder, R. G. (2004). Generic
Manufacturing Strategies and Plant Performance. Journal of Operations
Management, 22, 313-333.

Dillman, D. A., Smyth, J. D., & Christian, L. M. (2009). lntemet, Mail, and Mixed-
Mode Surveys: The Tailored Design Method. Hoboken, NJ: John Wiley &
Sons.

Duray, R., Ward, P. T., Milligan, G. W., & Berry, W. L. (2000). Approaches to
Mass Customization; Conﬁgurations and Empirical Validation. Journal of
Operations Management, 18, 605-625.

Ernst, R., & Kamrad, B. (2000). Evaluation of Supply Chain Structures. European
Journal of Operations Research, 124, 495-510.

Ethiraj, S. K., & Levinthal, D. (2004). Modularity and Innovation in Complex
Systems. Management Science, 50 (2), 159-173.

Feitzinger, E., & Lee, H. L. (1997). Mass Customization at Hewlett Packard: The
Power of Postponement. Harvard Business Review, 75 (1 ), 116-121.

F ine, C. H. (1998). Clockspeed: Winning Industry Control in the Age of
Temporary Advantage. Reading, MA: Perseus Books.

Fine, C. H. (2000). CIockspeed-Based Strategies for Supply Chain Design.
Production and Operations Management, 9(3), 213-221.

Fine, C. H., Golany, B., & Naseraldin, H. (2005). Modeling Tradeoffs in Three-
Dimensional Concurrent Engineering: A Goal Programming Approach.
Journal of Operations Management, 23, 389-403.

Fixson, S. K. (2007). Modularity and Commonality Research: Past Developments
and Future Opportunities. Concurrent Engineering, 15(2), 85-111.

114

Fixson, S. K. (2005). Product Architecture Assessment: A Tool to Link Product,
Process, and Supply Chain Decisions. Journal of Operations
Management, 23, 345-369.

Flynn, B. B., & Flynn, J. E. (1999). Inforrnation-Processing Alternatives for
Coping with Manufacturing Environment Complexity. Decision Sciences,
30(4), 1021-1052.

Galunic, D. C., & Eisenhardt, K. M. (2001). Architectural Innovation and Modular
Corporate Forms. The Academy of Management Journal, 44(6), 1229-
1249.

Garud, R., & Kumaraswamy, A. (1995). Technological and Organizational
Designs for Realizing Economies of Substitution. Strategic Management
Journal, 16, 93-109.

Gupta, Y. P., & Lonial, S. C. (1998). Exploring Linkages Between Manufacturing
Strategy. Business Strategy, and Organizational Strategy. Production and
Operations Management, 7(3), 243-264.

Hagel, J., & Brown, J. S. (2008, February 28). Learning from Tata's Nano; The
Breakthrough Innovations of the $2,500 Nano Car Carry a lot of Important
Lessons for Western Executives. Business Week Online .

Hair Jr., J., Black, W., Babin, B., Anderson, R., & Tatham, R. (2006). Multivariate
Data Analysis (6th ed.). USA: Pearson/Prentice Hall.

Henderson, R. M., & Clark, K. B. (1990). Architectural Innovation: The
Reconﬁguration of Existing Product Technologies and the Failure of
Established Firms. Administrative Science Quarterly, 35(1), 9-30.

Holland, J. H. (1995). Hidden Order: How Adaptation Builds Complexity. New
York: Basic Books.

Hu, L., & Bentler, P. (1999). Cutoff Criteria for Fit Indexes in Covariance
Structure Analysis: Conventional Criteria Versus New Alternatives.
Structural Equation Modeling, 6(1), 1-55.

Joreskog, K. G., & Goldberger, A. S. (1975). Estimation of a Model with Multiple
Indicators and Multiple Causes of a Single Latent Variable. Journal of the
American Statistical Association, 70 (351 ), 631 -639.

115

Katz, D., & Kahn, R. L. (1978). The Social Psychology of Organizations (2nd
Edition ed.). New York: John Wiley & Sons, Inc.

Kauffman, S. (1995). At Home in the Universe. New York: Oxford University
Press.

Kennedy, P. (2003). A Guide to Econometrics. Cambridge, MA: The MIT Press.

Kraljic, P. (1983). Purchasing Must Become Supply Management. Harvard
Business Review, 61 (5), 109-117.

Langlois, R. N. (2002). Moudularity in Technology and Organization. Journal of
Economic Behavior & Organization, 49, 19-37.

Langlois, R. N., & Robertson, P. L. (1992). Networks and Innovation in a Modular
System. Research Policy, 21, 297-313.

Lee, H. L., Padmanabhan, V., & Whang, S. (1997). Information Distortion in a
Supply Chain. Management Science, 43(4), 546-558.

Lewis, K. (2004). Knowledge and Performance in Knowledge-Worker Teams: A
Longitudinal Study of Transactive Memory Systems. Management
Science, 50 (11), 1519-1533.

Lin, B. W. (2003). Cooperating for Supply Chain Effectiveness: Manufacturing
Strategy for Chinese OEMs. International Journal of Manufacturing
Technology and Management, 5 (3), 232-245.

Miller, D., & Droge, C. (1986). Psychological and Traditional Determinants of
Structure. Administrative Science Quarterly, 31 (4), 539-560.

Novak, S., & Eppinger, S. D. (2001). Sourcing By Design: Product Complexity
and the Supply Chain. Management Science, 47(1), 189-204.

Nunnally, J. (1978). Psychometric Theory (2nd ed.). New York: McGraw-Hill.

Nyaga, G. N., Closs, D. J., Rodrigues, A., & Calantone, R. J. (2007). The Impact
of Demand Uncertainty and Conﬁguration Capacity on Customer Service
Performance in a Conﬁgure-to-Order Envimoment. Journal of Business
Logistics, 28 (2), 83-104.

116

O'Leary-Kelly, S., & Vokurka, R. (1998). The Empirical Assessment of Construct
Validity. Journal of Operations Management, 16 (4), 387-405.

Orton, J. D., & Weick, K. E. (1990). Loosely Coupled Systems: A
Reconceptualization. The Academy of Management Review, 15 (2), 203-
223.

Pil, F. K., & Cohen, S. K. (2006). Modularity: Implications for Imitation,
Innovation, and Sustained Advantage. Academy of Management Review,
31 (4), 995-1011.

Podsakoff, P. M., MacKenzie, S. B., Lee, J. Y., & Podsakoff, N. P. (2003).
Common Method Biases in Behavior Research: A Critical Review of the
Literature and Recommended Remedies. Journal of Applied Psychology,
88(5), 879-903.

Prasad, 8., Tata, J., & Madan, M. (2005). Build to Order Supply Chains in
Developed and Developing Countries. Journal of Operations
Management, 23(5), 551-568.

Ro, Y. K., Liker, J. K., & Fixson, S. K. (2007). Modularity as a Strategy for Supply
Chain Coordination: The Case of US. Auto. IEEE Transactions on
Engineering Management, 54 (1 ), 172-189.

Roth, A. V., Schroeder, R. 6., Huang, X., & Kristal, M. (2008). Handbook of
Metrics for Research in Operations Management. Thousand Oaks, CA:
Sage Publications, Inc.

Salvador, F. (2007). Toward a Product System Modularity Construct: Literature
Review and Reconceptualization. IEEE Transactions on Engineering
Management, 54(2), 219-240.

Salvador, F., Forza, C., & Rungtusanatham, M. (2002). Modularity, Product
Variety, Production Volume, and Component Sourcing: Theorizing Beyond
Generic Prescriptions. Journal of Operations Management, 20(5), 549-
575.

Salz, P. A. (2006, June 29). A Modular Approach in Business Simpliﬁes lntemal
Processes, Boosts Innovation and Brings Sustained Competitive
Advantage. The Wall Street Journal, p. A10.

117

Sanchez, R. (1995). Strategic Flexibility in Product Competition. Strategic
Management Journal, 16, 1 35-1 59.

Sanchez, R., & Mahoney, J. T. (1996). Modularity, Flexibility, and Knowledge
Management in Product and Organizational Design. Strategic
Management Journal, 1 7, 63-76.

Sanders, P. (2009, July 2). Boeing Tightens Its Grip on Dreamliner Production -
Company is in Talks to Buy Fuselage Factory From Supplier; Supply
Chain Woes Have Dogged 787 Program. The Wall Street Journal, p. B1.

Sarin, S., & McDermott, C. (2003). The Effect of Team Leader Characteristics on
Learning, Knowledge Application, and Performance of Cross-Functional
New Product Development Teams. Decision Sciences, 34(4), 707-739.

Sawhney, R., & Piper, C. (2002). Value Creation Through Enriched Marketing-
Operations Interfaces: An Empirical Study in the Printed Circuit Board
Industry. Journal of Operations Management, 6(4), 259-272.

Schilling, M. A. (2000). Toward a General Modular Systems Theory and Its
Application to Interﬁrm Product Modularity. The Academy of Management
Review, 25 (2), 312-334.

Schilling, M. A., & Steensma, K. H. (2001). The Use of Modular Organizational
Forms: An Industry Level Analysis. The Academy of Management Journal,
44 (6), 1 149.

Short, J. C., Ketchen Jr., D. J., Palmer, T. B., & HuIt, G. T. (2000). Firm, Strategic
Group, and Indsutry Inﬂuences on Performance. Strategic Management
Journal, 28, 147-167.

Simon, H. A. (1957). Models of Man, Social and Rational. New York: John Wiley
and Sons, Inc.

Simon, H. A. (1962). The Architecture of Complexity. Proceedings of the
American Philosophical Society, 106, pp. 467-482.

Stacey, R. D. (1995). The Science of Complexity: An Alternative Perspective for
Strategic Change Processes. Strategic Management Journal, 16(6), 477-
495.

118

Starr, M. K. (1965). Modular Production - A New Concept. Harvard Business
Review, 43 (6), 131 -1 42.

Staudenmayer, N., Tripsas, M., & Tucci, C. L. (2005). lnterﬁrm Modularity and Its
Implications for Product Development. Journal of Product Innovation
Management, 22, 303-321.

Stevens, W., Myers, 6., & Constantine, L. (1974). Structured Design. IBM
Systems Journal, 13 (2), 115-139.

Thompson, J. D. (1967). Organizations in Action. New York: McGraw-Hill.

Tu, Q., Vonderembse, M. A., Ragu-Nathan, T., & Ragu-Nathn, B. (2004).
Measuring Modularity-Based Manufacturing Practices and Their Impact on
Mass Customization Capability: A Customer Drive Perspective. Decision
Sciences, 35 (2), 147-167.

Ulrich, K. (1995). The Role of Product Architecture in the Manufacturing Firm.
Research Policy, 24(3), 419-440.

Vachon, S., & D, K. R. (2002). An Exploratory Investigation of the Effects of
Supply Chain Complexity on Delivery Performance. IEEE Transactions on
Engineering Management, 49(3), 218-230.

Wallace, L., Keil, M., & Rai, A. (2004). How Software Project Risk Affects Project
Performance: An Investigation of the Dimensions of Risk and an
Exploratory Model. Decision Sciences, 35(2), 289-321.

Williamson, 0. E. (1975). Markets and Hierarchies: Analysis and Antib’ust
Implications. New York: The Free Press.

Worren, N., Moore, K., & Cardona, P. (2002). Modularity, Strategic Flexibility,
and Firm Performance: A Study of the Home Appliance Industry. Strategic
Management Journal, 23, 1 1 23-1 140.

Zellner, A. (1962). An Efﬁcient Method of Estimating Seemingly Unrelated

Regressions and Tests for Aggregation Bias. Journal of the American
Statistical Association, 57 (298), 348-368.

119

   

"'IIIIIIIIIILIIIJIIIIIIIIIIQIIIIIIIIIIIII“