This article was downloaded by: [Texas A&M University]

On: 29 November 2010

Access details: Access Details: [subscription number 915031382]
Publisher Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-
41 Mortimer Street, London W1T 3JH, UK

International Journal of Production Research

Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713696255

The cumulative capability 'sand cone' model revisited: a new perspective

for manufacturing strategy
Roger G. Schroedera; Rachna Shaha; David Xiaosong Pengb
a
Carlson School of Management, University of Minnesota, Minneapolis, MN 55455, USA b Department
of Information & Operations Management, Mays Business School at Texas A&M University, College
Station, TX 77843-4217, USA

First published on: 03 November 2010

To cite this Article Schroeder, Roger G. , Shah, Rachna and Xiaosong Peng, David(2010) 'The cumulative capability 'sand
cone' model revisited: a new perspective for manufacturing strategy', International Journal of Production Research,, First
published on: 03 November 2010 (iFirst)
To link to this Article: DOI: 10.1080/00207543.2010.509116
URL: http://dx.doi.org/10.1080/00207543.2010.509116

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.

Electronic copy available at: http://ssrn.com/abstract=1763440

 International Journal of Production Research
2010, 1–23, iFirst

The cumulative capability ‘sand cone’ model revisited: a new perspective

for manufacturing strategy
Roger G. Schroedera*, Rachna Shaha and David Xiaosong Pengb
a
Carlson School of Management, University of Minnesota, 3-150 CarlSMgmt Building,
321-19th Avenue South, Minneapolis, MN 55455, USA; bDepartment of Information
& Operations Management, Mays Business School at Texas A&M University, 320
Wehner Building, 4217 TAMU, College Station, TX 77843-4217, USA
(Received 26 January 2010; final version received 6 July 2010)
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

The cumulative capability or the ‘sand cone’ model (Ferdows and De Meyer 1990,
Lasting improvements in manufacturing performance: in search of a new theory.
Journal of Operations Management, 9(2), 168–184) has been central in the debate
on relations among dimensions of manufacturing performance. The central thesis
of this model is that manufacturing performance is cumulative and sequential,
with quality performance forming the foundation. An implicit assumption
underlying the model is that the indirect effects of quality leading to delivery, then
to flexibility and finally cost are stronger than direct effects among these same
performance dimensions. Despite its frequent use, the sand cone model has not
been empirically tested for direct and indirect effects or for the sequence of effects.
Most tests have used correlations or regression relationships that only establish
positive relationships among dimensions, not the specific sand cone sequence. We
conduct two new tests of the actual sequence of the sand cone theory using data
from 189 manufacturing plants. We do not find universal support for the sand
cone theory, since some plants in our data appear to be following the sand cone
sequence, while others are not. Our empirical tests support other limited evidence
in the literature that a contingency theory is needed rather than an outright
rejection or acceptance of the sand cone model.
Keywords: competitive capabilities; cumulative capability; manufacturing perfor-
mance; operations strategy; survey research

1. Introduction
Whether firms improve manufacturing performance in a cumulative manner or trade one
measure against another is not clear. The two polar end approaches to manufacturing
performance are popularly referred to as the cumulative capability and the trade-off
theory respectively. Existing literature shows mixed results in assessing the extent and
nature of trade-offs or cumulative capabilities in operations. While some papers have
supported cumulative capabilities, others have supported the trade-off theory (Corbett and
Van Wassenhove 1993, Noble 1995, Boyer and Lewis 2002). Still other papers have
concluded that trade-offs depend on factors such as distance from the efficient frontier,
country and industry (Clark 1996, Schmenner and Swink 1998, Flynn and Flynn 2004).

*Corresponding author. Email: [email protected]

ISSN 0020–7543 print/ISSN 1366–588X online

ß 2010 Taylor & Francis
DOI: 10.1080/00207543.2010.509116
http://www.informaworld.com
Electronic copy available at: http://ssrn.com/abstract=1763440
 2 R.G. Schroeder et al.

While many cumulative capability models have been proposed, the most popular is the
Ferdows and De Meyer’s (1990) sequence advocating that manufacturers acquire quality,
followed by delivery, flexibility and finally cost. This sequence, popularised as the sand cone
model, has been widely cited and examined in the operations management literature.
The purpose of our paper is to study if manufacturing performance measures are related in a
cumulative manner as the sand cone model implies. Further testing of the sand cone model is
needed because several past studies have shown conflicting results. Past studies use different
measures from the original Ferdows and De Meyer (1990) paper, or they use different
analytical approaches, which might explain the lack of agreement in results (Noble 1995,
Boyer and Lewis 2002, Corbett and Claridge 2002, Rosenzweig and Roth 2004).
Another point of contention is related to the type of data that should be used for
empirical examination. Because the sequence in the original sand cone model was stated in
terms of lasting improvement over time, it is logical to test the sand cone model with
longitudinal data, but, we contend that cross sectional data can be used to test the lasting
and cumulative effects of the sand cone if mediation or sequential effects are checked rather
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

than simply correlations or regression results. Using data collected at any one point in
time, if we can observe some sequences of the actual levels of the capabilities of quality,
delivery, flexibility and cost that are not consistent with the sand cone model, can provide a
basis for rejecting the universal nature of the model leading to further contingency theories.
We test the Ferdows and De Meyer model using cross sectional data from 189
manufacturing plants using path analysis and structural equation modelling, along with
sequence tests. Our results do not support the universal nature of the sand cone model.
Our findings point to the need to uncover explanations for those plants that are not
following the prescribed sequence. We propose some alternative theories that might
explain why organisations pursue different sequences or configurations than the sand cone
model prescribes.
The paper is organised as follows. First we present a literature review of the various
studies of the sand cone model and the conflicting results obtained, along with the
underlying theories and hypothesis that we will test. Then we describe the research design,
data collected for this study, and analysis. This is followed by a discussion of the results
and conclusions for the paper.

2. Theoretical foundation and literature review

The cumulative capability (sand cone) model originated largely in response to and as an
alternative to the trade-off model and its basic premise that ‘ . . . a production system
inevitably involves tradeoffs and compromises and so must be designed to perform a
limited task well, with that task defined by corporate objectives’ (Skinner 1969). The most
popular interpretation of the trade-off model posits that firms cannot achieve high levels
of performance for multiple competitive priorities simultaneously because an improvement
in one necessitates a decrease in another (Slack 1991). However, simultaneous improve-
ments on multiple dimensions achieved by world-class Japanese manufacturers during the
early 1980s led many to question the fundamental view underlying the trade-off model.
Following Nakane (1986), researchers have argued that firms are able to achieve
improvements on multiple fronts concurrently because the improvements reinforce each
other in a cumulative fashion (Corbett and Van Wassenhove 1993, Noble 1995, Wacker
1996, White 1996, Rosenzweig and Roth 2004). Additionally, some researchers have
 International Journal of Production Research 3

asserted that the improvements occur in a pre-specified sequence (Nakane 1986, Hall 1987,
Ferdows and De Meyer 1990, Größler and Grübner 2006).
Another school of thought has advocated reconciling the apparent differences between
the trade-off and the sand cone models through what is commonly referred to as the
integrative model (Clark 1996, Hayes and Pisano 1996, Schmenner and Swink 1998, Lapré
and Scudder 2004). According to the integrative model, trade-offs and simultaneous
improvements are a function of a firm’s location relative to its maximum potential
performance frontier. Schmenner and Swink (1998) conceptualise two performance
frontiers: an asset frontier, formed by a firm’s structural choices and an operating frontier,
formed by its infrastructural policies. Schmenner and Swink contend that when sufficient
slack is available and the firm lies below its operating and asset frontiers, it is capable of
making simultaneous improvements along multiple performance dimensions without
negatively impacting others. However, as firms become more efficient and remove slack
from their systems, it becomes increasingly more difficult to make improvements
simultaneously in multiple dimensions without negatively impacting other dimensions.
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

This suggests that firms have to choose between competing priorities when they are at the
frontier.
Even at the frontier, researchers argue that a firm can make improvements on multiple
dimensions simultaneously by shifting its operating frontier or changing its shape or by
making changes to the existing physical assets, thus in essence by shifting its asset frontier
or by changing the asset frontier’s shape (Clark 1996). Some manufacturers may seem to
improve along many dimensions simultaneously without experiencing any trade-offs,
because their operating frontier is far superior to their competitors at any particular point
in time (Schmenner and Swink 1998). In 2009 Toyota Motor Corporation became the
largest automobile manufacturer in the world by steadily improving multiple measures of
performance relative to its competitors. Thus, scholars studying the integrative model
agree that simultaneous improvements are possible for firms at the frontier also.
Over time, cumulative capability models have been examined extensively but the
evidence in their support is mixed at best (Nakane 1986, Hall 1987, Ferdows and De
Meyer 1990). Although there is consensus among researchers that quality improvement
forms the foundation of a cumulative capability model, a number of possible alternate
sequences have been specified over time and a vibrant debate continues about several
aspects of the sand cone model. We conducted an extensive review of the literature and
present the most relevant studies in Table 1. Our literature review points to four issues that
may have influenced the empirical results related to the sand cone model research. These
are: (1) conceptualisation of improvements; (2) actual sequence of improvements; (3)
measures used to operationalise improvements; and (4) methods used to examine the sand
cone model. Each of these issues is discussed briefly below.

2.1 Conceptualisation of improvements

In their seminal paper, Ferdows and De Meyer set out to explain how firms can make
improvements to their manufacturing performance. Their central thesis was that firms can
improve by building manufacturing capabilities cumulatively following a pre-specified
sequence. This distinction between manufacturing performance and manufacturing
capabilities seems to have been lost in the literature related to the sand cone model.
 Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

Table 1. Literature review to highlight conflicts in operationalising the sand cone model1.

Sample size Type of Conceptualisation Operational Research

(n) study Type of data Model tested of improvement measures methods Results

Ferdows and De 167 business Conceptual Longitudinal Cumulative Manufacturing Multi-item scales Associative ‘Sand cone model’ was
Meyer 1990 units model capability/perf. analysis first proposed in this
(QDeFC) study
Noble 1995 561 plants Empirical Cumulative Capabilities Multi-item scales Regression Preliminary support for
model analysis the cumulative model
(QDeDFCI) with variation across
countries.
Schmenner and NA Theory NA Cumulative Manufacturing NA NA ‘Integrative model’ was
Swink 1998 model capability in the outcome of this
(QDCF) QDCF study
Narasimhan and 215 mfg busi- Empirical Cross Cumulative Manufacturing goals Multi-item scales Correlation Nothing specific is con-
Jayaram 1998 ness units sectional model cluded about the
(QDeFC) relationship among
the manufacturing
goals
Boyer and Lewis 110 plants Empirical; no Cross Trade-off Competitive Multi-item scales Correlation 2 Some evidence that
2002 formal sectional priorities graphs trade-offs vary across
R.G. Schroeder et al.

hypothesis plant hierarchy. No

support for the sand
cone model.
Flynn and Flynn 165 plants Empirical 9 Cross sec- Trade-off Capabilities Single items No graphs Failed to find support
2004 hypotheses tional; 3 Correlation, for a sequential pro-
industries; Hierarchical gression; variation
5 countries regression across countries but
analysis no industry effects.
Rosenzweig and 81 business Empirical 7 Cross sec- Integrative Capabilities Single items Path analysis Evidence in support of
Roth 2004 units hypotheses tional; model sand cone model.
high tech
industry

Note: 1For a comprehensive review of the sand cone model, see Scudder (2001).
 International Journal of Production Research 5

In examining the sand cone model, researchers have conceptualised performance and
capabilities in many different ways contributing to the confusion.
Table 2 illustrates terms that researchers have used to discuss trade-offs and the sand
cone model. Most commonly used terms include manufacturing capability (Skinner 1969),
cumulative (manufacturing) capability (Nakane 1986, Ferdows and De Meyer 1990),
competitive priority (Noble 1995, Boyer and Lewis 2002), and competitive capability
(Rosenzweig and Roth 2004). While capabilities emphasise relative competitive strength
and competitive priorities denote a strategic emphasis on goals, another closely related set
of terms capture manufacturing performance including competence and its derivatives.
In this research, we propose to use the term ‘competitive performance’ over other
available alternatives because it embodies the outcome of a firm’s strategic manufacturing
emphasis relative to its main competitors. It seems that most other studies of the sand cone
model measure the outcomes of manufacturing (or manufacturing performance) and yet
they call them ‘capability’. To avoid confusion we use the term competitive performance
to refer to the outcomes of quality, delivery, flexibility and cost in the remainder of this
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

paper.

2.2 Sequence underlying the sand cone model

Ever since Nakane (1986) and Hall (1987) first proposed a sequence to cumulative
improvements, many possible alternate sequences have been suggested (Ferdows and De
Meyer 1990, Noble 1995). Nakane’s and Hall’s proposed sequence of quality ! (depend-
able) delivery ! cost (efficiency) ! flexibility is distinct from other sequences that
followed them because cost is an intermediate variable in their model where as in most
subsequent models, quality forms the base and cost is modelled as the ultimate apex of the
pyramid. Noble (1995) examined a variant of this model and positioned ‘product
innovation’ at the top of her cumulative capability model. In spite of many alternate
models, Ferdows and De Meyer’s sequence of cumulative capability, quality ! deliv-
ery ! flexibility ! cost, popularised as the sand cone model has gained broad acceptance.
For a comprehensive review of the various other sequences underlying the sand cone
model, see Scudder (2001).

2.3 Operationalising improvements

There is general consensus about the underlying dimensions used to measure improve-
ments. Most studies operationalise competitive performance along cost, quality, delivery
and flexibility dimensions. Previous research has shown that most of these dimensions are
multidimensional in nature and several different measures to operationalise improvements
exist in previous research. For instance, quality can be measured as conformance or
performance quality (Vickery et al. 1994, Ward et al. 1998). Similarly, delivery can be
measured as delivery speed or delivery reliability (White 1996, Vickery et al. 1997) and
flexibility as range and mobility flexibility (Koste and Malhotra 1999). To represent the
multidimensional character of performance measures, researchers can develop scales using
multivariate techniques (Boyer and Lewis 2002). In contrast, researchers have also used
single items to measure them (Rosenzweig and Roth 2004). Different approaches to
measuring performance are presented in Table 2.
 Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

6
Table 2. Literature review to highlight conflicting use of terms associated with capabilities, competence and performance.

Strategic management Operations management

Term Definition Term Definition

Capabilities The organisational characteristics that Manufacturing Strength in key manufacturing performance dimensions
enable an organisation to conceive, capabilities such as cost, quality, and time (Skinner 1969 1974,
choose, and implement strategies Hayes and Wheelwright 1984, White 1996).
(Barney 1991) Cumulative Capabilities Ability to simultaneously improve multiple dimensions
A set of business process strategically of manufacturing performance (Nakane 1986,
understood (Stalk et al. 1992). Ferdows and De Meyer 1990).
Activities that a firm can do better than its High performance in multiple manufacturing perfor-
competitors (Porter 1996). mance dimensions simultaneously (Flynn and Flynn
A combination of resources that creates 2004).
high-order competencies (Madhok 1997). Core mfg capabilities Fundamental proficiency in manufacturing (Swink and
Capabilities are a high-level routine, or Hegarty 1998).
collection of routines (Winter 2003). Competitive priorities Need for choosing among and achieving one or more
Ability to perform a coordinated task key manufacturing capabilities (Noble 1995, Ward
utilising organisational resources (Helfat et al. 1998, Boyer and Lewis 2002).
and Peteraf 2003). Competitive capabilities Actual, or realised competitive strengths relative to
The capacity to perform a task or activity primary competitors (Rosenzweig and Roth 2004).
R.G. Schroeder et al.

in an integrated manner (Hoskisson et al.

2004).
(Core) Ability to coordinate diverse production Competence A bundle of aptitudes, skills, and technologies that the
Competence skills and integrate multiple streams of firm performs better than its competitors, that is
technologies (Prahalad and Hamel 1990). difficult to imitate and provides an advantage in the
The local ability and knowledge which are market places (Coates and McDermott 2002).
fundamental to day-to-day problem Production competence A measure of the combined effects of a manufacturer’s
solving (Henderson and Cockburn 1994). strengths and weaknesses in certain key performance
A set of problem-defining and problem- areas (Cleveland et al. 1989).
solving insights that fosters the develop- Degree to which manufacturing performance supports
ment of idiosyncratic strategic growth the strategic priorities of the firm (Vickery et al.
alternatives. (Lei and Hitt 1996). 1994).
Distinctive activities enabled by integrated Manufacturing/opera- Often used interchangeably with manufacturing capa-
clusters of firm specific assets (Teece tional performance bilities to describe the strength in key manufacturing
et al. 1997). performance dimensions such as cost, quality, and
time (e.g., Ferdows and De Meyer 1990).
 International Journal of Production Research 7

2.4 Methods used to examine the sand cone model

Previous studies also differ from one another in the empirical methods used to examine the
sand cone model. While some studies provide simple descriptive statistics and bivariate
correlations as evidence to support or refute the sand cone model, others have used more
sophisticated techniques like path analysis. Both Ferdows and De Meyer (1990), and
Boyer and Lewis (2002) used simple descriptive statistics including bivariate correlations
to examine the model but reached opposing conclusions. While Ferdows and De Meyer
found support for the sand cone model, Boyer and Lewis concluded that trade-offs exist.
Noble (1995), and Flynn and Flynn (2004) both used regression analysis but found
contradictory results. Rosenzweig and Roth (2004) used path analysis and Größler and
Grübner (2006) used structural equation modelling, the most sophisticated statistical
techniques in the group and found evidence in support of the sand cone model.
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

3. Research hypothesis and tests

The purpose of our research is to empirically examine the sand cone theory following
Ferdows and De Meyer’s founding logic, which posits the quality ! delivery ! flexibil-
ity ! cost sequence of manufacturing competitive performance. Paraphrasing Ferdows
and De Meyer, we propose the following hypothesis for testing:
Hypothesis: Improvements in competitive performance are achieved by focusing first on
quality, then on quality and dependability, then quality, dependability and flexibility and
finally all three and cost.

3.1 Mediation tests

Past studies have used correlations or regression to test this hypothesis. While such tests
demonstrate association among the four dimensions they do not test the sequence.
For example, to establish the above hypothesis as true, positive correlations among the
four competitive performance dimensions are a necessary condition, but not sufficient.
Regression analysis could also show a similar positive relationship among the four dimen-
sions, but this does not insure a sequence of effects. The empirical implication of
the conceptual model suggests that the magnitude of a direct path connecting any two
non-adjacent performance dimensions should be less than the effect of following the
prescribed sequence. More specifically, the direct effect between non-adjacent performance
dimensions should be smaller than its indirect effect. For instance, the magnitude of the
direct path from quality to flexibility should be smaller than the magnitude of the path that
goes from quality to flexibility through delivery because the latter follows the suggested
sequence of performance improvement. Also, because quality and flexibility are
non-adjacent, the sand cone model mandates that the effect of quality on flexibility
should be mediated through delivery. Further, if the implications of the sand cone model
are correct then the direct effect of quality on flexibility should be significantly lower than
its indirect effect, mediated through delivery. Similarly, delivery and cost are not adjacent
in the sand cone model. Therefore, the direct effect of delivery on cost should be
substantially lower relative to its mediated effect through flexibility. Following this logic,
we test the sand cone model by comparing the direct and the mediated paths between every
 8 R.G. Schroeder et al.

two competitive performance dimensions using path analysis and structural equation
modelling (SEM).

3.2 Sequence tests

In addition to mediation tests, we use a second method which examines the frequencies of
sequences followed. Because Ferdows and De Meyer’s original conceptualisation of the
sand cone model was stated in terms of long lasting improvement, recently researchers
construed it to mean that the sand cone model must be tested with longitudinal data, but it
is our contention that cross sectional data can also be used to detect the long lasting effects
by examining the actual sequences that result at any point in time. We explain our logic
using Figure 1; it is a stylistic representation of the sand cone model and shows the
sequential progression of performance measures over time, where quality starts low, but
then continues to increase in a lasting way to form the base of the sand cone. Next delivery
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

starts to increase as prescribed by the sand cone while quality continues to increase and
so on. Thus, any horizontal line (such as the bold double dashed line, labelled ‘A’) shows
the manner in which a plant would progress along the four performance dimensions if it
were following the sand cone model and any vertical line (such as the single dashed lines
labelled T1, T2, T3, and T4) shows the levels of the four performance measures for a plant
at a single point in time. Thus for instance for any plant at time T1, one can see that quality
is higher than the other three performance measures while by time T4, the plant has
achieved higher performance on all four measures in the proper sequence. The points T1,
T2 and T3 show intermediate states in a plant’s progression. An absence of the sequential
pattern in the cross sectional snapshots implies that the sand cone model is not being
followed. For example, if a cross sectional slice indicates that a plant’s quality is low and
delivery is high, then we can conclude that it is not following the sand cone model
irrespective of the flexibility and cost level because increasing efforts and levels of quality
must be achieved at the base of the sand cone model before higher levels of delivery,
flexibility and cost advantage can be achieved. There are a number of other sequences that
are impossible under the sand cone model.
At the same time the sand cone is supported when an individual plant exhibits a
sequence consistent with the sand cone model. For instance consider QH, DH, FL, and CL
where QH indicates high quality, DH high delivery, FL low flexibility, and CL low cost
advantage (or relatively high unit cost). This sequence is consistent with the sand cone
model since it suggests that a plant has progressed through the first two dimensions of the
sequence, but not the last two. If we conceptualise that quality, delivery, flexibility and
cost advantage can take high or low levels there are 16 possible sequences that can be
observed. Our analysis will examine all 16 possible sequences and determine the frequency
of plants that follow each sequence, some being consistent with the sand cone theory and
others not consistent.
If we make the reasonable assumption that different plants begin following the sand
cone model at different points in time, then a cross sectional slice at any point in time
should exhibit various combinations of slices at points T1, T2, T3, and T4. While some
plants will undoubtedly be starting the sand cone, others may have progressed further
along, including all the way to achieving lower cost shown in T4. Examining cross sectional
data from a large sample of plants allows us to distinguish plants that are following the
sand cone from plants that are not following it. Finding a mixture of some plants following
 International Journal of Production Research 9

T1 T2 T3 T4

High
Performance level

Quality Delivery Flexibility Cost

A
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

Low

Time

Figure 1. Stylistic representation of the sand cone model.

the sand cone model and others not following it refutes the universal nature of the sand
cone model and points to the need for a contingency theory.

4. Research method
The analysis is intended to investigate the relationships among various manufacturing
performance dimensions. We first examine the correlation matrix of the variables to
establish the necessary condition that competitive performance dimensions are positively
related. Then we use path analysis and structural equation modelling to compare direct
and indirect effects. Finally, we test the existence of the Ferdows and De Meyer sequence
by examining all 16 possible sequences of the four dimensions for individual plants.

4.1 Data collection

The level of analysis is the manufacturing plant. While Ferdows and De Meyer’s (1990)
study was conducted at the business unit level, their choice has come under criticism
because a business unit often includes multiple manufacturing plants that may each
develop different capabilities (Szwejczewski et al. 1997, Boyer and Lewis 2002). In
contrast, we surveyed individual plants where manufacturing performance is actually
realised (Skinner 1969, 1974).
The sampling frame consists of medium to large size manufacturing plants (more than
250 employees) in machinery, electronics, and automotive supplier industry segments
 10 R.G. Schroeder et al.

(defined by four digit SIC codes) in six industrialised countries: Finland, Germany, Japan,
Korea, Sweden and the United States. The research team developed a set of 13 survey
instruments focusing on various aspects of a plant’s internal and external operations that
was administered to 21 different informants. While comprehensive data was collected in
2006, this paper uses data only from the plant manager who responded to the questions
on competitive performance dimensions.
The research team in each country contacted the plant manager of each randomly
selected plant to solicit the plant’s participation. The plants were selected from a
comprehensive database in the target industries that exceeded the minimum size of 250
employees. In return, each plant was promised a detailed profile of its own operations and
benchmarking information comparing it to the other plants in its industry segment. Upon
agreement to participate in the study, each plant manager appointed a survey coordinator
to work with the research team. The survey questionnaires and instructions for
administering the survey were mailed to the survey coordinator. After distributing the
questionnaire to randomly selected shop floor employees and named managers, the survey
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

coordinator collected and mailed the completed questionnaires to the research team. The
final sample includes 189 manufacturing plants, representing a response rate of 65%.
The response rate is based on the percentage of plant managers actually contacted that
agreed to return the questionnaires and subsequently returned them. Non-respondents
were compared to the original sample on size of plant and industry and no significant
differences were found.

4.2 Measures
The measurement instrument was developed from an extensive review of the relevant
literature and many of the scales were already tested for reliability and validity in the
literature. To insure content validity, in prior studies a panel of academic researchers and
operations managers reviewed each of the scales. The instrument was then pretested at
several manufacturing plants and revised as needed. To prevent loss of meaning during
translation, each survey questionnaire was translated into the local language of the
country where the plant was located and then translated back to English by a separate set
of researchers (Cua et al. 2001).
The current study uses 10 items related to competitive performance. The respondent
for each of the 10 items were the plant managers, who rated the plant’s performance
relative to competitors on a 5-point scale (Appendix). Because data was collected from six
countries and three industries, we conducted one-way ANOVA by country and by
industry to check if there were systematic differences in responses due to either of these
two variables (Flynn et al. 1990). ANOVA results do not indicate a significant difference in
any of the 10 items across countries or industries.
Ferdows and De Meyer (1990) used a single indicator for measuring each of the four
manufacturing performance dimensions in the sand cone model including quality, depend-
ability, flexibility, and cost efficiency. They operationalised quality as quality conformance
(i.e., producing according to the specifications), delivery as dependability (on-time
delivery), flexibility as the speed of new product introduction, and cost efficiency as unit
manufacturing cost.
However, using single indicators to operationalise manufacturing performance is
restrictive because it does not adequately capture the breadth of each manufacturing
 International Journal of Production Research 11

performance dimension (Ketokivi and Schroeder 2004a). Instead, a multi-indicator scale

should be developed for each performance dimension (Ketokivi and Schroeder 2004b).
Therefore, in addition to using single indicators to closely replicate the Ferdows and De
Meyer (1990) measurements, we also develop a multi-item scale for operationalising each
manufacturing performance dimension by selecting measures that are most commonly
used in previous research (Ward et al. 1998). We measure quality performance with
product performance quality and conformance to product specifications. The delivery
dependability dimension is measured by on time delivery performance and speed of
delivery. We assess flexibility as volume flexibility, mix flexibility and new product
introduction speed (Anand and Ward 2004). Then, we use three items to assess cost
performance: inventory turnover, cycle time, and unit cost of manufacturing.
We used a confirmatory factor analysis to examine the reliability and validity of the
measurement scales. Standardised factor loadings and composite reliability of each
measurement scale can be found in the Appendix. All standardised factor loadings are
greater than 0.50 and significant at 0.01. Likewise, the composite reliability of each scale
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

is greater than 0.65, suggesting satisfactory measurement reliability. We also computed

Cronbach’s alpha of each scale to check inter-item reliability within each scale and found
alpha greater than 0.65, indicating acceptable inter-item reliability. Finally, we conducted
pairwise chi-square (2 ) difference tests to ensure the discriminant validity of the
measurement scale. Again, the 2 difference tests indicate good discriminant validity
(Appendix).
Descriptive statistics and the Pearson correlation matrix are provided in Table 3. The
correlations are all positive and significant lending support to the lack of trade-offs as
observed in previous studies (Boyer and Lewis 2002) and providing a necessary condition
to support the hypothesis. Prior to conducting the analysis, we verified normality,
skewness and kurtosis of each variable and examined the assumptions of constant
variance. We did not find any of them to be a significant problem.

4.3 Path analysis

We use path analysis to investigate the direct and indirect effects in the single item models
that closely approximate the Ferdows and De Meyer (1990) measures and use structural
equation modelling (SEM) to examine the direct and indirect effects in the latent variable
models. The two methods are similar in that they both allow simultaneous assessment of
the direct and indirect effects (Shah and Goldstein 2006). Both path analysis and SEM can
use the correlation matrix or covariance matrix among the measurement items as the input
for conducting the analysis.
We specify four path models (Figure 2). Model 1(a) is labelled ‘purely mediated model’;
it includes only paths following the performance progression prescribed by the sand cone
model (1990). Models 1(b) and 1(c) are named ‘mediated model with one direct path’.
Model 1(b) includes all the paths in model 1(a) and the direct path from quality to
flexibility. Model 1(c) includes all the paths in model 1(a) and the direct path from delivery
to cost. Finally, model 1(d) includes all the paths in model 1(a) and the direct paths from
quality to flexibility and from delivery to cost.
To assess model fit, we provide commonly reported fit measures including root
mean square error of approximation (RMSEA), normed fit index (NFI), incremental fit
index (IFI), comparative fit index (CFI), and normed chi-square (2 /degree of freedom)
 Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

12

Table 3. Mean, standard deviation, and correlations among variables (n ¼ 189).

Mean S.D. Q1 Q2 D1 D2 F1 F2 F3 C1 C2

Product capability and performance Q1 3.90 0.75

Conformance to specifications Q2 3.84 0.71 0.49**
Fast delivery D1 3.89 0.84 0.29** 0.41**
On-time delivery performance D2 3.79 0.81 0.24** 0.28** 0.61**
Flexibility to change volume F1 3.34 0.90 0.33** 0.29** 0.25** 0.15*
Flexibility for product mix F2 3.90 0.74 0.16* 0.16* 0.38** 0.39** 0.31**
Speed of new product introduction F3 3.88 0.69 0.30** 0.13 0.26** 0.31** 0.30** 0.50**
Inventory turnover C1 3.39 0.89 0.27** 0.21** 0.33** 0.20** 0.29** 0.18* 0.25**
R.G. Schroeder et al.

Cycle time (raw materials to delivery) C2 3.45 0.75 0.33** 0.28** 0.37** 0.33** 0.41** 0.27** 0.25** 0.66**
Unit cost of manufacturing C3 3.15 0.90 0.27** 0.29** 0.38** 0.25** 0.31** 0.32** 0.21** 0.46** 0.42**

Note: *p 5 0.05; **p 5 0.01.

 International Journal of Production Research 13

b1 b2 b3
Quality Delivery Flexibility Cost

Model 1(a): purely mediated model.

Quality Quality

b1 b1
b4

Delivery Delivery β5
b2 b2
b3
Flexibility Cost b3
Flexibility Cost

Models 1(b) and 1(c): mediated model with one direct path.
Quality

b1
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

b4
Delivery b5

b2
β3
Flexibility Cost

Model 1(d): mediated model with two direct paths.

Figure 2. Conceptualisation of sand cone model tested with path analysis.

(Hair et al. 1995). If the sand cone model holds, then model 1(a) should have acceptable
model fit because it specifies the exact sequence of performance progression suggested by
the sand cone model. Additionally, Model 1(b) through Model 1(d) should not fit the data
substantially better than model 1(a) because according to the sand cone model, all the
direct paths linking any two non-adjacent performance dimensions should be relatively
small compared to the indirect paths linking the same pair of non-adjacent performance.
However, evaluating the four models using only the fit indices listed above is
inappropriate because the four models are nested (Shah and Goldstein 2006). For nested
models, the 2 difference test is highly recommended because it allows the researcher to
evaluate the significance of the change in 2 statistic with the corresponding change in the
degrees of freedom between any two nested models (Marsh and Hocevar 1985). Thus, we
use the 2 difference test to evaluate the marginal improvement over the sand cone model
in 1(a) achieved by adding one or more direct paths.
Fit statistics suggest that the data does not fit the sand cone model well. It can be seen
from Table 4 that in model 1(a) RMSEA ¼ 0.22, normed 2 ¼ 10.11, NFI ¼ 0.67,
IFI ¼ 0.69, and CFI ¼ 0.68; most are inferior to the Shah and Goldstein (2006)
recommended cut-off values (RMSEA for close fit 50.08, normed 2 4 2, NFI, IFI
and CFI 4 0.90). The fit statistics suggest that Model 1(a) fits poorly with the data,
indicating that the model may be misspecified and important paths may be missing from
the hypothesised specification. Adding one direct path from quality to flexibility in model
1(b)) or from delivery to cost in model 1(c) improves the model fit. The 2 difference test
between models 1(b) and 1(a) and models 1(c) and 1(a) is significant at 0.01. Finally, model
1(d) with two direct paths fits significantly better than all the other models and the 2
difference test is also significant at p 5 0.01. Only model 1(d) provides a reasonable fit
according to generally accepted fit measures (Shah and Goldstein 2006).
 Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

14

Table 4. Path analysis results associated with Figure 21.

Model 1(a) Model 1(b) Model 1(c) Model 1(d)

Q
Q
Q

D
D
F C

F C F C
Paths Q D F C

Q!D 0.41** (0.000) 0.41** (0.000) 0.41** (0.000) 0.41** (0.000)

Q!F – 0.22** (0.01) – 0.22** (0.01)
D!F 0.25** (0.000) 0.16* (0.01) 0.25** (0.000) 0.16* (0.01)
D!C – – 0.33** (0.000) 0.33** (0.000)
F!C 0.31** (0.000) 0.31** (0.000) 0.23** (0.001) 0.23** (0.001)
Fit statistics
R.G. Schroeder et al.

2 (df) 30.34(3) 23.18 (2) 9.92 (2) 2.19 (1)

Normed 2 10.11 11.59 4.96 2.19
RMSEA 0.22 0.24 0.15 0.08
NFI 0.67 0.75 0.90 0.98
IFI 0.69 0.76 0.91 0.99
CFI 0.68 0.75 0.91 0.99
Comparing direct and indirect effects
Q ! F direct – 0.22 – 0.22
Q ! F indirect 0.10 0.07 0.10 0.07
D ! C direct – – 0.33 0.33
D ! C indirect 0.08 0.06 0.18 0.04

Note:
1
The numbers in the top section of the table are path coefficients followed by (significance).
 International Journal of Production Research 15

In Table 4 an examination of the path coefficients connecting various performance

dimensions provides further evidence that the sand cone model does not fully capture the
relationships among the four performance dimensions. All paths are positive and
significant ( p 5 0.05). In models 1(b), 1(c) and 1(d), the indirect effects are smaller in
magnitude than the corresponding direct effects, in contradiction of our hypothesis.
In addition to the model fit improvement, we compare the magnitude of the indirect
and direct paths linking the same pair of non-adjacent performance statistically using
bootstrap analysis. The bootstrap analysis is conducted following a procedure recom-
mended by Shrout and Bolger (2002). The bootstrap analysis involves two steps and allows
us to assess the hypothesis by computing the 95% confidence interval around the mean
difference between the indirect and the direct effect. In the first step, a bootstrap sample of
N (189) size is created by randomly sampling from the original data set, with replacement.
The second step involves estimating the indirect effect and the direct effect using the
bootstrap sample created in step one and obtaining a difference score by subtracting the
direct effect from the indirect effect. A negative difference score suggests that the indirect
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

effect is smaller than the direct effect, leading us to reject the hypothesis. These two steps
are repeated 1000 times to obtain a sample from which the 95% confidence interval is
constructed. The sand cone model is disconfirmed when both the lower and upper
confidence limits are negative. The sand cone model is supported only when both the lower
and the upper confidence limits of the 95% confidence interval are positive.
Boot strap analysis results are also at odds with our hypothesis. We conducted two
bootstrap analyses: (1) compare the difference between the indirect effect of quality leading
to delivery and then flexibility (quality ! delivery ! flexibility) and the direct effect from
quality to flexibility; and (2) compare the difference between the indirect effect
delivery ! flexibility ! cost and the direct effect from delivery to cost. The first bootstrap
analysis indicates that the mean difference between the indirect and the direct effect is
0.22, and the 95% confidence interval is (0.52, 0.08). For bootstrap sample (2), the
mean difference is .24, and 95% confidence interval is (0.43, 0.04). In both cases the
indirect effect is not significantly larger than the direct effect, resulting in lack of support
for the sand cone model.

4.4 Structural equation modelling

To improve measurement reliability and validity, we specify four SEM Models (Figure 3).
The specification of each SEM model is identical to the corresponding path model: model
2(a) is the purely mediated model, model 2(b) and model 2(c) are mediated models with
one direct path, and model 2(d) is the mediated model with two direct paths. We use
AMOS 5 with maximum likelihood estimation method and raw data to run the SEM
models. To ensure that the model was identified, we included additional constraints to set
the scale for each of the latent variables which had fewer than three measurement items per
latent variable (Bollen 1989). The SEM models are assessed in a manner similar to the path
models using model fit statistics, 2 difference test and the direct and the indirect effects
comparison.
Each SEM model fits the data better than the corresponding path model (Table 5). The
goodness-of-fit improves as we move from model 2(a) towards model 2(d); however the
improvement is only marginal suggesting that adding direct paths incrementally does not
substantially improve model fit relative to model 1(a). The reason why SEM results differ
 16 R.G. Schroeder et al.

F2
Q1 Q2 F1 F3

b1 b2 b3
Quality Delivery Flexibility Cost

D1 D2 C1 C3
C2

Model 2(a): purely mediated model.

Q1 Q2 Q1 Q2

Quality Quality

b1
b1
b4
Delivery Delivery
b2 b5
b2
D1 D2 b3 D1 D2
Flexibility Cost b3
Flexibility Cost

F1 F2 F3 C1 C3
C2 F1 F2 F3 C1 C3
C2
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

Models 2(b) and 2(c): mediated model with one direct path.
Q1 Q2

Quality

b1 b4

Delivery
b2 b5
D1 D2
b3
Flexibility Cost

F1 F2 F3 C1 C3
C2
Model 2(d): mediated model with two direct paths.

Figure 3. Conceptualisation of sand cone model with latent variables tested with SEM.

from the path analysis results is that a larger number of parameters are estimated in a
latent variable model compared to its corresponding path model; thus, an additional path
does not affect overall model fit to the same degree in SEM as it does in the path model.
Because we are estimating a larger number of parameters in the SEM models, we
computed the power of our models based on the sample size of 189 plants and other
characteristics (Preacher and Coffman 2006). The results from the analysis show that the
power in each of our SEM models exceeds 80%. Because fit statistics alone do not provide
sufficient information about the relative magnitude of the direct and the indirect effects,
we examine the standardised path coefficients, 2 difference test and the boot strap
analysis results.
Similar to the path analysis results, the magnitude of the indirect path is smaller than
the direct path links but the difference between the indirect and the direct path is not as
large as the path analysis, especially for Q ! F. The 2 difference test of each of the
models from model 2(a) is significant at p 5 0.01. The bootstrap analysis results are also
similar to the ones obtained with path models. Specifically, the mean difference between
the indirect and the direct effect corresponding to quality ! delivery ! flexibility is 0.13,
and the 95% confidence interval around the mean difference is (0.34, 0.07). The mean
difference for delivery ! flexibility ! cost is 0.06, and the 95% confidence interval
is (0.26, 0.13). Thus the bootstrap samples provide strong evidence that the magnitude
 Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

Table 5. SEM results associated with Figure 3.

a
Model 2(a) Model 2(b) Model 2(c) Model 2(d)
Q
Q Q

D
D D
F C

F C F C
Paths Q D F C

b
Q!D 0.61** (0.000) 0.57** (0.000) 0.62** (0.000) 0.59** (0.000)
Q!F – 0.33* (0.01) – 0.26þ (0.10)
D!F 0.7** (0.000) 0.47** (0.001) 0.61** (0.000) 0.43** (0.01)
D!C – – 0.39** (0.01) 0.33** (0.01)
F!C 0.64** (0.000) 0.68** (0.000) 0.29* (0.01) 0.36** (0.005)
Fit statistics
2 (df) 86.11(32) 79.69(31) 79.19(31) 77.31(30)
Normed 2 2.69 2.57 2.55 2.58
RMSEA 0.10 0.09 0.09 0.09
NFI 0.85 0.86 0.87 0.87
IFI 0.90 0.91 0.91 0.91
CFI 0.90 0.91 0.91 0.92
International Journal of Production Research

Comparing direct and indirect effects

Q ! F direct – 0.33 – 0.26
Q ! F indirect 0.42 0.27 0.37 0.25
D ! C direct – – 0.39 0.33
D ! C indirect 0.45 0.32 0.18 0.16

Note:
þ
p 5 0.1; *p 5 0.05; **p 5 0.01;
a
measurement items are not shown in the path diagrams;
b
all path coefficients reported are standardised.
17
 18 R.G. Schroeder et al.

of the indirect effect is not greater than the magnitude of the corresponding direct effect.
The results from path analysis and SEM also do not provide support for the sand
cone model.

4.5 Sequence tests

We outlined above how enumerating the number of plants (frequency) that represent
various sequences in the data can be used to test the hypothesis. To represent high and low
levels of competitive performance, we divided each of the four performance dimensions
into high and low levels using the median. Then we constructed all possible sequences of
high and low levels that can be observed. We list the 16 sequences along with the number
of plants that fall under each sequence (Table 6). The last column in Table 6 indicates
whether the particular sequence is possible or not under the sand cone hypothesis.
For example, we see that the sequence QL, DL, FL and CL is possible for 21 plants
(bottom of Table 6). These plants could eventually follow the sand cone sequence but have
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

not started yet, since all four levels are low. (Note: CL represents low cost advantage that
is high unit cost.) We also count the first sequence in Table 6 QH, DH, FH, and CH as
possible, but we do not know the order in which all four high levels of competitive
advantage were achieved. On the other hand, the second sequence in Table 6 followed by
three plants in the sample QH, DL, FH, and CH does not conform to the sand cone
model, since delivery is low and should be high to enable high flexibility and eventually
high cost advantage. The third sequence in the table with 15 plants is also impossible to
observe QH, DH, FL and CH according to the sand cone model for similar reasons, since
flexibility is low in the sequence.
We should also note that the sequence QH, DH, FL, CL is a possible sand cone
progression that has only progressed through two stages with quality high and delivery
high, but nevertheless the sand cone sequence is still possible by these plants. At a future
point in time we would expect to see QH, DH, FH and CL if there is a continuation of the
progression. We also count the sequence QH, DH, FH and CL with 30 plants as a possible
sand cone sequence, since cost advantage could be high (representing low unit cost) in a
future cross sectional snapshot. When the total number of plants is counted for each of the
16 sequences, we arrive at the following frequencies:

Number of plants Frequency

NO–does not follow the sand cone 63 33%

Possibly following the sand cone 126 67%
TOTAL 189 100%

The table illustrates that 33% of the plants cannot possibly follow the sand cone
prescribed progression; that is, there are too many plants not following the sand cone
model casting doubt on the universal nature of the sand cone model. This analysis was
done with the four individual measures of competitive performance used by Ferdows and
De Meyer. Conducting a similar analysis with multiple measures shows that 47% of the
plants do not follow the sand cone model, resulting in even stronger rejection of the
universal hypothesis. However, we observe that the sand cone model is not always
rejected; it is followed by some plants, but not by others. Since there is high variance in
the sample, it is possible that contingencies such as different strategies or different
 International Journal of Production Research 19

Table 6. Number of plants exhibiting each particular sequence.

SEQUENCE EXHIBITED

Quality Delivery Flexibility Cost # of plants in Sequence progression according to

(Q) (D) (F) (C) the sample the sand cone model

H H H H 41 Possible
H L H H 3 No
H H L H 15 No
H L H L 9 No
H H H L 30 Possible
H L L H 4 No
H L L L 11 Possible
H H L L 23 Possible
L H H H 4 No
L L H H 1 No
L H L H 3 No
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

L L H L 4 No
L H H L 4 No
L L L H 3 No
L H L L 13 No
L L L L 21 Possible

Note:
In the table H represents high and L represent low on a median split for each competitive
performance dimension. The question asked was how does your plant compare to its competition in
your industry on a global basis (rated on a 5-point scale with 1 ¼ Poor, low end of industry,
3 ¼ Average, and 5 ¼ Superior). Thus, H (high) is desirable on each of four dimensions according to
the sand cone theory. High cost advantage represents low unit cost.

external environments might explain why some plants follow the sand cone model and
others do not.

5. Summary and conclusion

Ever since Ferdows and De Meyer published their seminal paper about the now famous
sand cone model in 1990, many studies have tested their model and/or offered competing
models. The results from these studies vary widely, ranging from full support, partial
support to no support of the sand cone model. In this study, we identify four factors that
may have resulted in inconsistent outcomes in the previous studies and attempt to address
each of them. We also point out that most past studies have found positive correlations or
positive regression results that are necessary conditions, but do not directly test the sand
cone sequence.
We conducted new tests of the sand cone model using path analysis and structural
equation modelling to test for mediation. Using this formulation, the direct effect between
any two non-adjacent competitive performance measures should be less than the indirect
effect. Neither of these analyses supported the sand cone model. Similarly, examining the
frequency of sand cone sequences for individual plants shows that 33% of plants in our
 20 R.G. Schroeder et al.

sample using individual measures, and 47% of plants using multiple measures, exhibit
particular sequences that contradict the model. We find high variance in the sequences
followed rather than a single sand cone sequence.
The implications of our research, we hope, will help shape the future research agenda
related to examining the sand cone model. The results clearly suggest that the sand cone
model is not a universal phenomenon. Rather, there may be contingencies guiding the
sequence that different plants follow.
Flynn and Flynn (2004) found substantial differences in the pattern of cumulative
capabilities between countries and some evidence of industry differences (also see Corbett
and Claridge 2002, Amoako-Gyampah and Meredith 2007). Using different methods we
lend support to their conclusions that contingency effects may be operating. Because a
large proportion of plants in our sample do not follow the prescribed sequence, it is
imperative for future research to pursue alternative theories and examine factors that
impact such plants’ decision to adopt a particular sequence.
One possible theory is strategic choice theory. Future research could, for example, test
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

the effect of operations strategy goals on the sequence of improvements. We should expect
different strategic goals to lead to different sequences. When plants are not at the efficient
frontier, there are many different ways of moving to the frontier depending on the
particular strategy chosen. This could explain why plants choose different performance
levels, configurations and sequences at any point in time. Also, environmental contingency
theory (e.g., industry, and country) could be advanced and tested, with some support
already from Flynn and Flynn (2004) along with Corbett and Claridge (2002), and
Amoako-Gyampah and Meredith (2007). The most bothersome thing about the sand cone
model is that it ignores strategy completely and does not recognise possible contingencies
or equifinality theories. New theories that explain why alternative sequences are observed
would be welcome.
We also argue that trade-off theory is still appropriate when operating at or near the
efficient frontier, particularly when the frontier is static. In this case, it would be interesting
to study which performance measures are subject to trade-offs and which ones are
mutually supportive.
We hope that this paper has helped illuminate the controversy surrounding the sand
cone model. It is only through further research using strategic choice theory, contingency
theory, or other explanations that progress can be made in gaining a greater understanding
of the relationships among competitive performance measures.

References

Amoako-Gyampah, K. and Meredith, J.R., 2007. Examining cumulative capabilities in a

developing economy. International Journal of Operations & Production Management, 27 (9),
928–950.
Anand, G. and Ward, P.T., 2004. Fit, flexibility and performance in manufacturing: coping with
dynamic environments. Production & Operations Management, 13 (4), 369–385.
Barney, J., 1991. Firm resources and sustained competitive advantage. Journal of Management,
17 (1), 99–120.
Bollen, K.A., 1989. Structural equations with latent variables. New York: John Wiley & Sons.
Boyer, K.K. and Lewis, M.W., 2002. Competitive priorities: investigating the need for trade-offs
in operations strategy. Production and Operations Management, 11 (1), 9–20.
 International Journal of Production Research 21

Clark, K., 1996. Competing through manufactuirng and the new manufacturing paradigm: is
manufacturing strategy passe. Production and Operations Management, 5 (1), 42–58.
Cleveland, G., Schroeder, R.G., and Anderson, J.C., 1989. A theory of production competence.
Decision Sciences, 20 (4), 655–668.
Coates, T.T. and McDermott, C.M., 2002. An exploratory analysis of new competencies: a resource
based view perspective. Journal of Operations Management, 20 (5), 435–450.
Corbett, C. and Van Wassenhove, L., 1993. Trade-offs? What trade-offs? Competence and
competitiveness in manufacturing strategy. California Management Review, 35 (4), 107–122.
Corbett, L.M. and Claridge, G.S., 2002. Key manufacturing capability elements and business
performance. International Journal of Production Research, 40 (1), 109–131.
Cua, K.O., McKone, K.E., and Schroeder, R.G., 2001. Relationships between implementation of
TQM, JIT, and TPM and manufacturing performance. Journal of Operations Management,
19 (6), 675–694.
Ferdows, K. and De Meyer, A., 1990. Lasting improvements in manufacturing performance: in
search of a new theory. Journal of Operations Management, 9 (2), 168–184.
Flynn, B.B., et al., 1990. Empirical research methods in operations management. Journal of
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

Operations Management, 9 (2), 250–284.

Flynn, B.B. and Flynn, E.J., 2004. An exploratory study of the nature of cumulative capabilities.
Journal of Operations Management, 22 (5), 439–457.
Größler, A. and Grübner, A., 2006. An empirical model of the relationships between manufacturing
capabilities. International Journal of Operations & Production Management, 26 (5), 458–485.
Hair, J.F., et al., 1995. Multivariate data analysis. Englewood Cliffs, NJ: Prentice-Hall.
Hall, R.W., 1987. Attaining manufacturing excellence. Homewood, IL: Dow Jones-Irwin.
Hayes, R.H. and Wheelwright, S.C., 1984. Restoring our competitive edge – competing through
manufacturing. New York: John Wiley & Sons.
Hayes, R.H. and Pisano, G.P., 1996. Manufacturing strategy: at the intersection of two paradigm
shifts. Production & Operations Management, 5 (1), 25–41.
Helfat, C.E. and Peteraf, M.A., 2003. The dynamic resource-based view: capabilities lifecycles.
Strategic Management Journal, 24 (10), 997–1010.
Henderson, R. and Cockburn, I., 1994. Measuring competence? Exploring firm effects in
pharmaceutical research. Strategic Management Journal, 15 (S1), 63–84. [Special issue:
Competitive organizational behavior: toward an organizationally-based theory of competitive
advantage].
Hoskisson, R.E., Hitt, M.A., and Ireland, R.D., 2004. Competing for advantage. Cincinnati, OH:
South-Western Educational.
Ketokivi, M. and Schroeder, R., 2004a. Manufacturing practices, strategic fit and performance
a routine-based view. International Journal of Operations & Production Management, 24 (2),
171–191.
Ketokivi, M.A. and Schroeder, R.G., 2004b. Perceptual measures of performance: fact or fiction?
Journal of Operations Management, 22 (3), 247–264.
Koste, L.L. and Malhotra, M.K., 1999. A theoretical framework for analyzing the dimensions of
manufacturing flexibility. Journal of Operations Management, 18 (1), 75–93.
Lapré, M.A. and Scudder, G.D., 2004. Performance improvement paths in the U.S. airline
industry: linking trade-offs to asset frontiers. Production and Operations Management,
13 (2), 123–134.
Lei, D. and Hitt, M.A., 1996. Dynamic core competences through meta-learning and strategic
context. Journal of Management, 22 (4), 549–569.
Madhok, A., 1997. Cost, value and foreign market entry mode: the transaction and the firm.
Strategic Management Journal, 18 (1), 39–61.
Marsh, H.W. and Hocevar, D., 1985. Application of confirmatory factor analysis to the study of
self-concept: first and high order factor models and their invariance across groups.
Psychological Bulletin, 97, 562–582.
 22 R.G. Schroeder et al.

Nakane, J., 1986. Manufacturing futures survey in Japan: a comparative survey 1983–1986. Research
report. System Science Institute, Waseda University, Tokyo, Japan.
Narasimhan, R. and Jayaram, J., 1998. Causal linkages in supply chain management: an exploratory
study of North American manufacturing firms. Decision Sciences, 29 (3), 579–605.
Noble, M.A., 1995. Manufacturing strategy: testing the cumulative model in a multiple country
context. Decision Science, 26 (5), 693–721.
Porter, M.E., 1996. What is strategy? Harvard Business Review, 74 (6), 61.
Prahalad, C.K. and Hamel, G., 1990. The core competence of the corporation. Harvard Business
Review, 68 (3), 79–91.
Preacher, K.J. and Coffman, D.L., 2006. Computing power and minimum sample size for RMSEA.
Available from: http://www.quantpsy.org.
Rosenzweig, E.D. and Roth, A.V., 2004. Towards a theory of competitive progression: evidence
from high-tech manufacturing. Production & Operations Management, 13 (4), 354–368.
Schmenner, R.W. and Swink, M.L., 1998. On theory in operations management. Journal of
Operations Management, 17 (1), 97–113.
Scudder, G.D., 2001. An analysis and update of the cumulative capabilities model. Working paper.
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

Owen Graduate School of Management, Vanderbilt University.

Shah, R. and Goldstein, S.M., 2006. Use of structural equation modeling in operations
management research: looking back and forward. Journal of Operations Management,
24 (2), 148–169.
Shrout, P.E. and Bolger, N., 2002. Mediation in experimental and nonexperimental studies: new
procedures and recommendations. Psychological Methods, 7 (4), 422–445.
Skinner, W., 1969. Manufacturing – missing link in corporate strategy. Harvard Business Review,
47 (3), 136–145.
Skinner, W., 1974. The focused factory. Harvard Business Review, 52 (3), 113–121.
Slack, N., 1991. The manufacturing advantage. London: Mercury Books.
Stalk, G., Evans, P., and Shulman, L.E., 1992. Competing on capabilities: the new rules of corporate
strategy. Harvard Business Review, 70 (2), 57–69.
Swink, M.H. and Hegarty, W.H., 1998. Core manufacturing capabilities and their links to product
differentiation. International Journal of Operations & Production Management, 18 (4), 374–386.
Szwejczewski, M., Mapes, J., and New, C., 1997. Delivery and trade-offs. International Journal of
Production Economics, 53 (3), 323–330.
Teece, D.J., Pisano, G., and Shuen, A., 1997. Dynamic capabilities and strategic management.
Strategic Management Journal, 18 (7), 509–533.
Vickery, S.K., Droge, C., and Markland, R.E., 1994. Strategic production competence: convergent,
discriminant, and predictive validity. Production & Operations Management, 3 (4), 308–318.
Vickery, S.K., Droge, C., and Markland, R.E., 1997. Dimensions of manufacturing strength in the
furniture industry. Journal of Operations Management, 15 (4), 317–330.
Wacker, J., 1996. A theoretical model of manufacturing lead times and their relationship to a
manufacturing goal hierarchy. Decision Sciences, 27 (3), 483–518.
Ward, P.T., et al., 1998. Competitive priorities in operations management. Decision Sciences, 29 (4),
1035–1046.
White, G.P., 1996. A meta-analysis model of manufacturing capabilities. Journal of Operations
Management, 14 (4), 315–331.
Winter, S.G., 2003. Understanding dynamic capabilities. Strategic Management Journal, 24 (10),
991–995.

Appendix: Measurement items

Please circle the number that indicates your opinion about how your plant compares to its
competition in your industry, on a global basis (rated on a 5-point scale with 1 ¼ Poor, low end of
industry, 3 ¼ Average, and 5 ¼ Superior)
 International Journal of Production Research 23

Quality a0.66, b0.65

c
Q1 Product capability and performance 0.68
Q2 Conformance to product specifications* 0.71
Delivery 0.75, 0.75
D1 Fast delivery 0.70
D2 On time delivery performance* 0.85
Flexibility 0.65, 0.67
F1 Flexibility to change volume 0.72
F2 Flexibility to change product mix 0.67
F3 Speed of new product introduction into the plant* 0.51
Cost 0.75, 0.77
C1 Inventory turnover 0.77
C2 Cycle time (from raw materials to delivery) 0.81
C3 Unit cost of manufacturing* 0.59
Chi-square (2 ) difference test
Downloaded By: [Texas A&M University] At: 20:52 29 November 2010

Pr(2 4 6.64) ¼ 0.01 with a degree of freedom 1

Quality Delivery Flexibility

d
Delivery 69
Flexibility 72.2 70.7
Cost 66.2 31.4 56.7
a
Cronbach’s alpha;
b
Composite reliability;
c
Standardised item factor loading;
d 2
 difference between two models: (1) a model where the covariance between a pair of scales is fixed
to one; (2) a second model where the covariance between the same two scales is free to assume any
value;
All items are used in the latent variable models (multi-indicator measures of manufacturing
performance);
*Items used for the path model (single indicator measures of manufacturing performance).