The scientific theory-building process:

A primer using the case of TQM

Robert B. Handfield

Steven A Melnyk

The Department of Marketing and Supply Chain Management

The Eli Broad Graduate School of Management
Michigan State University,
East Lansing, MI, 48824-1122
 The scientific theory-building process:

A primer using the case of TQM

Abstract

As Operations Management (OM) researchers begin to undertake and publish

more empirical research, there is a need to understand the nature of the scientific
theory-building process implicit in this activity. This tutorial presents a process
map approach to this process. We begin by defining the nature of scientific
knowledge, and proceed through the stages of the theory building process, using
illustrations from OM research in Total Quality Management. The tutorial ends
with a discussion of the criteria for OM journal reviewers to consider in evaluating
theory-driven empirical research, and suggests a number of OM topic areas that
require greater theory development.

Keywords: Literature review, empirical research, theory building.

1.0 Introduction

Recently introduced broad-based business practices such as Lean Manufacturing, Total Quality
Management, Business Process Re-engineering, and Supply Chain Management have brought with them
increased functional integration, with managers from multiple areas working full-time on cross-functional
implementation teams. For researchers in Operations Management (OM), this means that we will need to
participate and share ideas with researchers working in areas such as organizational behavior, marketing,
and strategy. To do so, however, we will need to communicate using the language of theory. We must
know how to build, refine, test and evaluate theory. Theory and theory-building are critical to our
continued success, since “Nothing is so practical as a good theory” (Simon, 1987; Van de Ven, 1989).
Without theory, it is impossible to make meaningful sense of empirically-generated data, and it is
not possible to distinguish positive from negative results (Kerlinger, 1986, p. 23). Without theory,
empirical research merely becomes “data-dredging.” Furthermore, the theory-building process serves to
differentiate science from common sense (Reynolds, 1971). A major objective of any research effort is to
create knowledge. Knowledge is created primarily by building new theories, extending old theories and
discarding either those theories or those specific elements in current theories that are not able to
withstand the scrutiny of empirical research. Empirical research is, after all, the most severe test of all
theory and research. Whatever question we ask, whatever data we collect reflects the impact of either a
theory or framework (be it explicit or implicit). Whenever we analyze data, we are evaluating the
findings in light of these underlying theories or frameworks.
Given the increasing importance of theory, it is imperative that we have a clear and unambiguous
understanding of what theory is and the stages involved in the theory building process. Developing such
an understanding is the primary purpose of this paper. This primer borrows extensively from the
behavioral sciences, since the practice of theory driven empirical research has been relatively well
established and many of the issues now facing OM researchers have been previously addressed by
researchers there.
The process of transporting this existing body of knowledge to Operations Management is not an
easy task. First of all, Operations Management is a relatively new field with its own unique set of needs
and requirements. This is a field strongly linked to the “real world.” It is a field where little prior work
in theory-building exists. Until recently, much of the work in OM was directed towards problem solving
rather than theory building. Due to the nature of our field, most OM researchers intuitively think in
terms of processes. While several prior works identify the need for theory in OM (e.g. Swamidass, 1991;
Flynn, Sakakibara, Schroeder, Bates, and Flynn, 1990; McCutcheon and Meredith, 1993), there is no
published work which specifies the actual process used in carrying out a theory-based empirical study.

1
 2

Much of the existing body of knowledge pertaining to theory-building and testing has been organized
around concepts, definitions and problems in other fields such as marketing, strategy, sociology, and
organizational behavior. As a result, there is a critical need to restate this body of knowledge into a form
more consistent with the Operations Management frame of reference. This is the major objective of this
paper.
We provide a view of theory building and theory-driven empirical research that is strongly
process-oriented. This view of theory-building draws heavily from an initial model developed by
Wallace (1971). We begin with Wallace because he presents one of the few models in the theory-
building literature that is process based. However, it is important to note that the theory-building model
presented in this paper draws heavily on the thoughts and contributions from other researchers. As such,
it is an eclectic merger reflecting the contributions of many different writers from diverse areas. Finally,
given the application orientation of the Operations Management field, we illustrate the application and
power of this model by drawing on examples from Total Quality Management (TQM). We conclude
with guidelines for journal reviewers who evaluate and criticize empirical theory-building research.

2.0 OM as Scientific Knowledge

Underlying the notion of theory-driven empirical research is the view of operations management as
science. One of the major traits of a science is that it is concerned only with those phenomena that can be
publicly observed and tested. This is very relevant to Operations Management since we deal with a field
which is practically oriented. Practicing managers are one of the major consumers of the knowledge
created by OM researchers. These managers use this information to hopefully improve the performance
of their processes. Unless we can provide these “consumers” with knowledge pertaining to events which
are observed and tested, managers will quickly and ruthlessly discredit the resulting research.
An important point to note about OM research is that its basic aim is not to create theory, but to
create scientific knowledge. Most people want scientific knowledge to provide (Reynolds, 1971, p. 4):
*0 A method of organizing and categorizing “things,” (a typology)
*1 Predictions of future events
*2 Explanations of past events
*3 A sense of understanding about what causes events, and in some cases,
*4 The potential for control of events.

The creation of knowledge, while critical, is not sufficient for success. To be successful, the
research must be accepted and applied by other researchers and managers in the field. To gain such
acceptance, the research must improve understanding of the findings (Reynolds, 1971; Wallace, 1971)
and it must achieve one or more of the five above objectives of knowledge. Finally, it must pass the test
of the real world. An untested idea is simply one researcher’s view of the phenomenon – it is an
 3

educated opinion (nothing more). It is for this reason that empirical research is the cornerstone for
scientific progress, especially in a field such as Operations Management where research results may be
put to the test by managers on a regular basis.
A good example of how a great idea can later become accepted can be illustrated by the early
beginnings of TQM. In the 1920s, two Bell System and Western Electric employees, William Shewhart
(1931) and George Edwards, in the inspection engineering department, began noting certain
characteristics of problems result from defects in their products. Based on these observations, Edwards
came up with the idea that quality was not just a technical, but rather an organizational phenomenon.
This concept was considered novel at the time, but generally irrelevant even in the booming postwar
market (Stratton, 1996). Quality assurance was simply an idea. Its impact had yet to be extensively
tested in the real world; that task would fall to Deming, Juran and their disciples in postwar Japan. At
this point in history, however, few researchers and practitioners were aware of the importance of Quality
and Quality Assurance.
Clearly, one cannot specify how OM researchers should go about creating knowledge. However,
as we will show, theory is the vehicle that links data to knowledge. This is the process that we will focus
on in the next section.

3.0 The Scientific Theory-building Process

How are theories developed? Researchers have noted over the years that there exists no common
series of events that unfold in the scientific process. However, several leading philosophy of science
scholars have identified a number of common themes within the scientific process. The most common of
these was stated by Bergmann (1957: 31), and reiterated over the years by others (Popper 1961; Bohm,
1957; Kaplan, 1964; Stinchcombe, 1968; Blalock, 1969; and Greer, 1969): “The three pillars on which
science is built are observation, induction, and deduction.” This school of thought was later summarized
into a series of elements and first mapped by Wallace (1971) (see Figure 1). The map provides a useful
reference in identifying the different stages that must occur in the scientific process.

INSERT FIGURE 1 ABOUT HERE

Due to the cyclical nature of the process, there is really no unique starting point at which to begin
within this map. However, it makes sense to begin at the lower section, with “Observation.” Wallace
(1971: 17) summarized his mapping as follows:
 4

Individual observations are highly specific and essentially unique items of

information whose synthesis into the more general form denoted by empirical
generalizations is accomplished by measurement, sample summarization, and
parameter estimation. Empirical generalizations, in turn, are items of information
that can be synthesized into a theory via concept formation, proposition formation,
and proposition arrangement. A theory, the most general type of information, is
transformable into new hypotheses through the method of logical deduction. An
empirical hypothesis is an information item that becomes transformed into new
observations via interpretation of the hypothesis into observables, instrumentation,
scaling, and sampling. These new observations are transformable into new empirical
generalizations, (again, via measurement, sample summarization, and parameter
estimation), and the hypothesis that occasioned their construction may then be tested
for conformity to them. Such tests may result in a new informational outcome:
namely, a decision to accept or reject the truth of the tested hypothesis. Finally, it is
inferred that the latter gives confirmation, modification, or rejection of the theory.

Once again, note that there is no distinct pattern for the manner in which this process unfolds.
The speed of the events, the extent of formalization and rigor, the roles of different scientists, and the
actual occurrence of the events themselves will vary considerably in any given situation. However, the
model provides a useful way of conceptualizing the primary themes that take place. The model also
provides an initial template for OM researchers interested in theory-driven empirical research. Moving
through the different stages requires a series of trials. These trials are initially often ambiguous and
broadly staged, and may undergo several revisions before being explicitly formalized and carried out.
The left half of the model represents what is meant by the inductive construction of theory from
observations. The right half represents the deductive application of theory to observations. Similarly,
the top half of the model represents what is often referred to as theorizing, via the use of inductive and
deductive logic as method. The bottom half represents what is commonly known as doing empirical
research, with the aid of prescribed research methods. The transformational line up the middle
represents the closely related claims that tests of congruence between hypotheses and empirical
generalizations depend on the deductive as well as the inductive side of scientific work, and that the
decision to accept or reject hypotheses forms a bridge between constructing and applying theory, and
between theorizing and doing empirical research (Merton, 1957). With this model in mind, we can now
proceed to each quadrant of the model and illustrate the processes using the unfolding field of TQM as a
reference point to illustrate each process.

Step 1: Observation
Observation is a part of our daily lives, and is also the starting point for the scientific process. As Nagel
(1961: 79) points out:
Scientific thought takes its ultimate point of departure from problems suggested
 5

by observing things and events encountered in common experience; it aims to

understand these observable things by discovering some systematic order in
them; and its final test for the laws that serve as instruments of explanation and
prediction is their concordance with such observations.

Observation, however, is shaped by the observer’s prior experiences and background, including
prior scientific training, culture, and system of beliefs. Likewise, observations are interpreted through
scaling, among which certain specified relations are conventionally defined as legitimate. In this manner,
observations can be compared and manipulated. The assignment of a scale to an observation is by
definition a classificatory generalization. Summarizing a sample of individual observations into
“averages,” “rates,” and “scores” is by definition dependent on the sample. A biased sample will surely
affect the way that observations are interpreted, and will therefore also affect parameter estimation. The
transformation of observations into empirical generalizations is therefore affected by the choice of
measures, sample, and parameter estimation techniques employed.
This problem was noted by Kaplan (1964) in his paradox of sampling. This paradox states that
the sample is of no use if it is not truly representative of its population. However, it is only
representative when we know the characteristics of the population (in which case we have no need of a
sample!). This presents a dilemma, since samples are supposed to be a random representation of a
population. Although the paradox of sampling can never be completely resolved, OM researchers need
to carefully consider the attributes of their population in generalizing observations. Specifically,
researchers must consider the possible effects of industry, organization size, manufacturing processes,
and inter-organizational effects in setting boundary assumptions on their observations. Such precautions
taken early in the theory development process will result in greater rewards later in the theory testing
phase, and will enhance the power of the proposed relationships.
The underlying purpose and set of techniques associated with different types of observations are
summarized in the first two rows of Table 1, which can be better appreciated if the nature of the columns
is first understood. The first column, Purpose, describes the goals driving research at each stage; the
Research Questions column lays out some of the typical questions that a researcher might be interested in
answering at each stage of the process; the Research Structure column deals with the design of the study;
Data Collection Techniques presents some of the procedures that a researcher might draw on in collecting
material for analysis; Data Analysis Procedures summarizes some of the methods we might use to
summarize and study the results of the data collected. The techniques and procedures presented in the last
two columns are not intended to be exhaustive or comprehensive; rather they are intended to be
illustrative. Finally, we have also provided some illustrative examples of studies from the TQM literature
that are representative of each process stage in Table 2.
 6

INSERT TABLES 1 AND 2 ABOUT HERE

Step 2: Empirical Generalization

An empirical generalization is “an isolated proposition summarizing observed uniformities of
relationships between two or more variables” (Merton, 1957: 95). This is different from a “scientific
law”, which is “a statement of invariance derivable from a theory” (1957: 96). A theory, on the other
hand, can be defined as a statement of relationship between units observed or approximated in the
empirical world (Bacharach, 1989; Coehn, 1980; Dubin, 1969; Nagel, 1961). Moreover, empirical
generalizations do not have explanatory theories to explain them. The transformation from an
“idea”/“fantasy” into “understanding”/“law” is an important part of the inductive theory-building process,
but is somewhat difficult to describe precisely. There are a variety of different perspectives regarding
whether theories are induced from facts or from simple thought experiments (see, for example, Nagel,
1961, Popper, 1961, Watson, 1960).
Some of the techniques for transforming observations into empirical generalizations are
summarized in the first three rows of Table 1. “Discovery” creates awareness of a problem or an event
which must be examined or explained. In a sense, discovery uncovers those situations or events which
are mystifying, and leads to further inquiry. Another form of observation is “Description,” wherein one
tries to explain what is happening in those situations identified in the Discovery phase. Description is
often concerned with information gathering and identifying key issues. There are two major types of
descriptions: taxonomies and typologies. Taxonomies deal with a categorical analysis of the data (i.e.,
"What are the phenomena?") In contrast, a typology tries to describe what is the most important aspect
of the phenomena or activity under consideration. The goal in each case is to provide a thorough and
useful description of the event being studied. With this description completed, we can now proceed to
Mapping, where one attempts to identify the key variables and issues involved, without specifying the
actual structure of the problem. In essence, one is trying to generalize from a set of observations on a
very broad level. Specific problem structures are developed later during the relationship building phase,
(which occurs through concept formation in the theory-building process). Thus, discovery expands
boundaries; description provides a portrait of the new events or problems; mapping identifies the factors
that are important; and, relationship building provides structure.
Returning to our TQM example, we know that as early as 1941 Deming and Leon Geoffrey
(1941), while working for the Bureau of the Census, had shown that the introduction of quality control in
clerical operations saved the bureau several thousand dollars (1941). In 1946, the American Society for
Quality Control was formed. At this point, America was entering the post-war boom, and production
 7

issues tended to override concerns of quality control. As a result, quality was relegated second-class
status (Handfield, 1989). It was for this reason that Deming, along with Juran, traveled to Japan to teach
and implement methods of quality management. The Japanese recognized the importance of quality
control in rebuilding their industries and their economy.
By the late 1970’s, the American public became aware of the difference in quality between
Japanese and American-made products. In March 1979, Business Week published a major article
recognizing the need for American manufacturers to adopt a top-down quality management changeover.
The authors broke down the components of quality as defined by Juran (1974), Crosby (1979) and
Deming, and also noted how the Japanese had implemented these methods. The article identified the
importance of quality as a strategy: ". . . product quality can be a pivotal, strategic weapon - sometimes
even more crucial than price - in the worldwide battle for market share." (1979: 32). The article also laid
the full blame on top management, not the worker. It was also noted how the Japanese stressed defect
prevention as opposed to defect detection, which in turn relates to careful product design. These
discoveries began to lead to a major empirical generalization:

Empirical Generalization 1:
Japanese companies employ quality assurance techniques, produce high quality
products, and have penetrated a number of American markets. American companies do
not employ quality assurance techniques and are losing market share in several industries
to the Japanese.

This generalization began to suggest some of the key variables, and even hinted at a relationship,
but did not go the next step in answering “why”?

Step 3: Turning Empirical Generalizations into Theories

The creation of theories from empirical observations is a process of "disciplined imagination" (Weick,
1989), involving a series of thought trials establishing conditions and imaginary outcomes in hypothetical
situations. Once a problem has been identified, the researcher develops a set of conjectures in the form of
“If-Then statements.” In general, a greater number of diverse conjectures will produce better theories
than a few homogeneous ones. The conjectures are then chosen according to the researcher's selection
criteria, which should include judgments of whether the relationship is interesting, plausible, consistent,
or appropriate. The researcher must be careful at this stage to maintain consistency in criteria when
evaluating different conjectures. Examples of selection criteria include the following:

1) "That's Interesting" (Davis, 1971) - Is the relationship not obvious at first?

2) "That's Connected" (Crovitz, 1970) - Are the events related when others have assumed they are
 8

unrelated?
3) "That's Believable" (Polanyi, 1989) - Is the relationship convincing?
4) "That's Real" (Campbell, 1986, Whetten, 1989) - Is the relationship useful to managers?

The theory-building process at this stage should not be constrained by issues of testability,
validity, or problem solving. When theorizing is equated with problem-solving at this stage, the activity
becomes dominated by the question.
The process of taking an empirical generalization and developing it into a theory produces
concepts and propositions that specify relationships. Theories emerge when the terms and relationships
in empirical generalizations are made more abstract by introducing terms that refer to non-observable
constructs. The researcher may employ “heightened idealization,” by either dropping error terms that are
usually explicit in empirical generalization, or relegating them to an implicit status in the theory
(Wallace, 1971). In developing and building relationships between constructs, the researcher is seeking
to achieve two major objectives. The first is to generalize the nature of relationships between key
variables (often in the form: x  y). These relationships address the "how" component of theory.
Second, we also try to explain the reasons for these relationship. In other words, we provide the "why”
(Whetten, 1989). Theory must contain both how and why. It must also both predict and explain known
empirical generalizations.
Returning to the TQM example, Empirical Generalization 1 proposed a link between quality and
change in market share in the 1970’s and early 1980’s. Initially academics and practitioners conducted
case studies of Japanese manufacturers, thereby generating another set of empirical generalizations that
helped explain the relationship between quality and financial performance in greater detail. One of the
first was Hayes (1981). He described the typical Japanese factory as being clean and orderly, with small
inventories neatly piled up in boxes (as opposed to the large work-in-process inventories witnessed in
comparable American facilities). He noted that Japanese workers were extensively involved in
preventive maintenance of machines and equipment monitoring on a daily basis. Hayes also noted that
Japanese plants had defect rates of about 0.1%, as compared to around 5% in most U.S. plants. This was
achieved by "thinking quality " into the product through product design, worker training, quality circles,
and screening of materials and suppliers. From these observations, Hayes surmised that a philosophical
change in Japanese organizations had taken place due to Deming’s teachings.
Other researchers (e.g., Wheelwright, 1981; Tribus, 1984) in the 1980’s began to notice another
attribute of successful Japanese corporations: top management responsibility for quality. Association
between quality, flexibility, cost, dependability and overall corporate performance was also recognized.
The positive nature of these relationships led to the notion of simultaneity (where improving quality also
simultaneously reduces costs). These and other observations made by researchers during the 1980-1985
 9

period regarding quality management can be summarized in the following theoretical statement:

Proposition 1
Quality management practices driven by top management leads to fewer defects,
reduced costs, and higher customer satisfaction, which in turn leads to lower overhead
costs, higher market share, and improved financial performance.

Although this proposition partially helps explain why Japanese companies outperformed
American companies, it fails to explain the “how” behind the concept of TQM. Although American
executives believed quality was important, they were still unsure about the methods and procedures to
implement TQM within their companies. During this period, Deming returned to the U.S. from Japan
and began to work with American companies to achieve this objective.
The most famous set of prescriptions to emerge from Deming’s work were the Fourteen Points
and the Seven Deadly Sins (Deming, 1981; 1982; 1986). The emphasis of these points is essentially
about the attitudes that should exist and the nature of the relationships among people in successful
organizations (Stratton, 1996). From the fourteen points and their own observations, a number of
researchers began to develop concepts and theoretical relationships between them that specify the
relationship between TQM to financial performance in an increasingly abstract manner.
One method frequently used in OM to describe and explore an area when there is no a priori
theory is the case study (Eisenhardt, 1989; McCutcheon & Meredith, 1993). This type of theory-building
relies on direct observations with the objects or participants involved in the theory and its development
(Glaser & Strauss, 1967; Yin & Heald, 1975; Miles & Huberman, 1994). The potential output is a
description of events and outcomes that allow other researchers to understand the processes and
environment (with the focus often being on exemplary or revelatory cases (Yin & Heald, 1975)).
Several researchers (e.g., Sirota & Alper, 1993; Green, 1993; Handfield & Ghosh, 1994) have
used direct observation from case studies and developed theoretical statements about TQM. In these
studies, the focus of attention shifted from the product and process to the corporate culture. For example,
Sirota and Alper (1993) after interviewing employees at 30 companies, proposed that the greatest impact
of TQM occurred when companies switched from a “detection” to a “prevention” culture. In a set of
interviews conducted with 14 North American and European Fortune 500 quality managers, Handfield
and Ghosh (1994) found a relationship between changes in the corporate culture and performance. They
proposed the existence of a transformation process that consisted of a series of stages (Awareness,
Process Improvement, Process Ownership, and Quality Culture). Financial performance improved as the
firms progressed through the stages. Based on these and other studies, we might modify our prior
theoretical proposition as follows:
 10

Proposition 1a
Visionary leadership drives the integration of continuous improvement and
variation reduction processes into organizational culture, thereby enabling firms to
eliminate defects, reduce costs, and improve customer satisfaction, which in turn leads to
reduced overhead costs, higher market share, and improved financial performance.

Here again, we predict a relationship between observable/nonobservable constructs (continuous

improvement, variation reduction processes, organizational culture, visionary leadership, customer
satisfaction) and direct observables (defects, costs, market share, and financial performance.) At this
stage, as shown in Figure 2, the construction of theory from observation ends and the application of
theory to observations begins.

Step 4a: Hypothesis Generation

By this stage, we have in place all the elements of a theory. We have present the what, how, and why
demanded of a theory. Now, the researcher must begin to compare the theory to determine its relative
applicability to observations (Wallace, 1971). There are three types of comparisons that may be used to
determine the extent to which a given theory provides useful symbolic representations of actual and
possible observations. First, internal comparisons may be made, whereby some parts of the theory are
compared with other parts in order to test whether it is internally consistent and non-tautological.
Tautological theories cannot, by definition, be falsified. For instance, suppose we hypothesize that
“Reducing material defects leads to better quality.” If we find that quality is defined by the number of
material defects, then the relationship is tautological.
Secondly, the theory may be compared with other theories to test whether, all other things being
equal, it has a higher level of abstraction, is more parsimonious, or is more flexible. This may involve an
assessment of the "nomological net," which consists of the underlying groundwork of related theory
which exists in a particular field. The net is established through a literature review that establishes the
framework within which the new theory is embedded or framed (Cook & Campbell, 1979). The primary
purpose of this net is often to place one construct relative to other constructs. While the development of
a nomological net (with its boxes and arrows) serves to answer the “what” and “how” questions as posed
by Whetten (1989), the nomological net is not theory because the “why” question and the boundary
conditions are often not specified.
For instance, TQM has been examined in relation to the mechanistic, organismic, and cultural
models of organization which exist in the literature (Spencer, 1994). The author found that many of the
new ideas regarding TQM are associated with organismic concepts, whereas Deming’s work seems to
graft mechanistic and organismic concepts into a coherent whole. The cultural model also taps into the
 11

philosophical components of TQM and is useful for evaluating the deployment of the practice. This
assessment seems to provide reasonable support for Proposition 1a.
Finally, the theory may be compared with empirical facts by comparing its predictions or low-
level hypotheses with appropriate empirical generalizations to test the truth of the theory (Wallace, 1971).
For instance, we could partially compare our theoretical statement with the experiences of firms such as
Xerox, Ford, and Motorola, which have successfully employed TQM to respond to their various
problems (Greene, 1993). A very succinct empirical generalization regarding quality was made by the
CEO of a large US global multi-national company: "Quality equals survival" (Bonsignore, 1992). Such
generalizations also lend support to Proposition 1a.
The real test of a theory begins, however, when hypotheses are deduced from the theory. Once a
particular set of conjectures or propositions has been selected, the researcher must now put them into
empirically testable form. A key requirement of this form is that the researcher must be able to reject
them based on empirical data (Popper, 1961). Hypotheses act as the vehicle by which the researcher
discards old variables and relationships which have not been able to pass through the screen of
falsification and replaces them with new variables and relationships (which are again subject to
evaluation).
With this approach, all hypotheses are essentially tentative. They are always in the process of
either being developed or being refuted. Recalling that a theory is a "statement of relations among
concepts within a set of boundary assumptions and constraints", the hypothesis development involves an
explicit translation of concepts into measures. Because many of the concepts used in OM have no
specific metrics, OM researchers often find that a system of constructs and variables will have to be
created before testing can proceed.
Constructs are approximated units, which by their very nature, cannot be observed directly.
Constructs (otherwise referred to as latent variables) are created for the explicit purpose of testing
relationships between associated concepts. Variables (or manifest variables), on the other hand, are
observed units which are operationalized empirically by measurement. "The raison d’être of a variable is
to provide an operational referent for a phenomenon described on a more abstract level" (Bacharach,
1989: 502). In testing a theory, the researcher is testing a statement of a predicted relationship between
units observed or approximated in the real world. Thus, constructs are related to each other by
propositions, while variables are related by hypotheses. The whole system is bounded by the researcher's
assumptions.
Since OM is a relatively new field, researchers may need to borrow at times measures from other
more developed fields (such as marketing and organizational behavior) while in other cases developing
new detailed measures of their own. If newly developed multiple measures are needed to accurately
 12

assess a construct, the variables used to define the constructs in the proposed relationship must be
coherent, and must pass a series of tests which address their measurement properties. These include tests
of content and face validity, convergent, discriminant, construct, and external validity, reliability, and
statistical conclusion validity (Flynn et. al, 1990). These factors will in turn depend on the choice of
instrumentation (case studies/interviews vs. surveys), scaling (nominal, ordinal, or ratio), and sampling
(defining the population and the relative power of the test).
Returning to our TQM theory, we find that Proposition 1a consists of a number of rather broad
constructs. Several researchers have undertaken the task of developing a set of constructs and measures
related to the concepts proposed in our theoretical statement. Powell (1995) developed a number of
constructs and measures related to continuous improvement (quality training, process improvement,
executive commitment) and organizational culture (openness, employee empowerment, and closer
relationships with suppliers and customers). Others such as Flynn et. al (1994), Black and Porter (1996),
and Ahire et. al (1996) have also developed measures for constructs commonly associated with TQM by
the Baldrige Award, including top management commitment, customer focus, supplier quality
management, employee involvement, employee empowerment, employee training, process improvement,
teamwork, SPC usage, and others. Suppose that among the different types of process variation
techniques, we limit our analysis to the implementation of process capability in manufacturing. This
involves the cross-functional cooperation between design and manufacturing personnel to ensure that
product specifications are wider than process specifications, in order to ensure that natural variations in
the process do not result in product defects. Let us now suppose that we are also interested in the effect
of this form of TQM practice on new product success. We can then specify the following hypothesis:

Hypothesis 1:
Companies that employ process capability studies in new product development have
products that are positively associated with improved market performance.

This hypothesis embodies certain important traits. First, causality is proposed, which assumes a
strict time precedence – the introduction of process capability in new product development must precede
product introduction and change in market performance. Second, the theory specifies a relationship
between two constructs: the use of process capability studies and market performance. Third, on a more
detailed level, we have now begun to operationalize our constructs through bounding assumptions, and
have limited the level of abstraction first from TQM in general to variation reduction practices, and
finally to the use of process capability studies. At this point, we can identify measurable or manifest
variables associated with each of the constructs within the hypothesis. For instance, we could measure
the use of process capability through direct measures (e.g. C p or Cpk to measure process capability,
 13

percent market share for market performance), or indirect measures (multiple Likert scales). We would
then need to set up conditions for testing the effect of one variable on another, and apply the appropriate
set of tests to determine whether the occurrence was not just mere chance.
In testing our hypothesis, we may have several options available (as shown in the “Theory
Validation” portion of Table 1). We might begin by conducting a set of single or multiple case studies
using structured interviews that allow “pattern matching” (Yin, 1989), where the pattern of actual values
of product market share versus their comparable C pk ratings in the product development process are
compared. If the constructs and measures are relatively well-developed, we might conduct a survey of
automotive part manufacturers that elicits multiple responses from design and development engineers,
assessing the extent to which process capability studies are carried out on a routine basis. We may need
to develop indirect measures for new constructs, such as routine product development team meetings,
information system implementation, training of design engineers in process capability methods, etc. In
turn, this could lead us to develop a set of statistical tests (e.g., regression or structural equations)
specifying the impact of these elements on new product market success, determined by the percent of
satisfied customers called at random, number of warranty calls, number of complaints, ratings in Ward’s
Auto World, etc. When such a study has been carried out, and the statistical results summarized, we are
ready for the next step.

Step 4b: Hypothesis Testing

When we have finished with the summary of the empirical data, we return to “observations”
(Figure 1). At this point a set of findings has been constructed to correspond logically to a given
theoretically-deduced hypothesis. We are now interested in internal validity or the extent to which the
variables identified by the hypotheses are actually linked or related in the way described by the
hypothesis (Hedrick, Bickman & Rog, 1993, p. 39). The hypothesis is highly testable if and when it can
be shown to be false by any of a large number of logically possible empirical findings and when only one
or a few such findings can confirm it. In other words, the researcher is concerned with whether x indeed
affects y in the way that we predicted based on the initial theory and hypotheses developed.
The decision to accept or reject a hypothesis is not always straightforward. Popper (1961: 109-
110) suggests that the test procedure is like a trial by jury. All of the elements in the theory-building
process are put “on trial,” including the originating theory, its prior support, the steps of deducing
hypotheses, and the interpretation, scaling, instrumentation, and sampling steps involved.
 14

Step 5: Logical Deduction

Next, we close the gap between theory and the empirical results. Logical deduction requires that
we return to our original research question, and ask ourselves if the results make sense or at least
contribute to the theory from on a more specific level. In general, there are three possible outcomes at
this point: (1) “end confirmation to” the theory by not disconfirming it; (2) “modify” the theory by
disconfirming it, but not at a crucial point; or (3) “overthrow” the theory by disconfirming it at a crucial
point in its logical structure, in its history of competition with rival theories (Wallace, 1971).
Irrespective of the outcome, the theory is affected to some extent.
Suppose that we have generated reasonable empirical support for our hypothesis. What does this
tell us about our theory outlined in Proposition 1a? When support for the hypothesis is found, the
researcher may proceed to theory extension/refinement. This set of activities associated with the theory-
building process focuses on external validity, or the “extent to which it is possible to generalize from the
data and context of the research study to broader populations and settings (especially those specified in
the statement of the original problem/issue)” (Hedrick et al., 1993, p. 40). As shown in the last section of
Table 1, theory extension/refinement involves applying the theory and the hypotheses in different
environments to assess the extent to which the results and outcomes indicated by the hypotheses are still
realized. If we had tested our hypothesis using a sample of domestic automotive parts manufacturers,
then one might argue that the same hypotheses on process capability studies (and by association, process
variation practices) be tested both in other industrial settings (e.g. electronics manufacturers, machine
tool manufacturers, hard disk manufacturers, etc.) and other countries (e.g. Japan, Germany, Brazil and
Malaysia). Generally speaking, OM focuses at the level of individuals, groups, processes and plants. The
greater the range of settings in which a theory can be successfully applied, the more general the theory
and the more powerful it is.
The strongest statement possible from our “imaginary study” is that there is statistical evidence
to support the fact that process capability studies can improve performance. Because of the path taken in
developing this hypothesis, there is also support for the theory that variation reduction practices can
reduce product defects, and by extension, improve financial performance. The researcher will now
probably seek to publish this result in an academic journal.

4.0 Evaluating Empirical Theory -Driven Research

As an increasing number of OM researchers submit empirically based articles, reviewers are faced by the
challenge of evaluating these manuscripts. In judging the contribution of a theory
building/testing/extension study, reviewers should pay attention to four major criteria (as identified by
 15

researchers such as Blalock (1970), Wallace (1971), Reynolds (1969), Whetten (1989) and Simon
(1967)): (1) “not wrong,” (2) falsifiability, (3) utility, and (4) parsimony.

4.1 “Not Wrong”

The criterion of “not wrong” is a test applied to the overall approach of the paper and the procedures used
by the researchers. This is a simple test in which we, as reviewers, examine the paper to ensure that the
research carried out and described has been executed correctly. This test begins by looking at whether
the research methodology used within the study is appropriate given the nature of the research problem
stated. If the research problem is essentially exploratory in nature, then using a statistical procedure such
as linear regression (which is more appropriate for evaluating well developed hypotheses) should raise
concerns
The “not wrong” criterion also focuses on whether the constructs defined by the researcher are
consistent with the manner in which they are implemented. For example, if the researcher uses a single
indicator to measure or implement a multi-trait construct, then this should raise a “red flag” in the mind
of the reviewer. It is not appropriate to measure a complex construct such as quality with a simple, single
indicator such as “Number of defects per million parts.”
Next, the “not wrong” criterion forces the reviewer to assess whether the researcher has used the
research methodology correctly. This assessment embraces a number of different issues. On the lowest
level, it forces the reviewer to determine if the research project or the data set violated any of the major
assumptions on which the procedures being used are based. It also forces the reviewer to determine if the
data is reported correctly. In some cases, this may mean that the reviewer must ensure that they accept
the “correctness” of such indicators as degrees of freedom or the p-statistic or the standard errors. This
criterion also requires that the researcher provide sufficient data so that the reviewer can judge
independently the “correctness” of the results. For example, if the researcher were to analyze a data set
using Structural Equation Modeling (SEM) (see Bollen, 1989), then it would be useful for the reviewer to
have either the variance/covariance matrix or the correlation matrix. At the highest level, we must
determine if the researcher is using the most powerful or suitable tool or whether the researcher is
involved in “fashion statistics.” That is, in our field as in others, there emerge new statistical techniques
that suddenly attract a great deal of attraction. Suddenly, it seems, nearly every paper uses this
technique, even though it may not be appropriate.
Finally, this criterion forces the reviewer to determine whether the question being posed is “post
hoc” (after the fact) or “ad hoc” (before the fact). That is, the reviewer must assess the extent to which
the theory and questions are driving the resulting analysis, or whether the data and its subsequent analysis
 16

are driving the theory and question. The latter approach, indicative of statistically driven data fitting, is
highly inappropriate within any theory driven empirical research framework.

4.2 Falsifiability

Falsifiabilty requires that the proposed theory be coherent enough to be refuted, and that it specify a
relationship among concepts within a set of boundary assumptions and constraints (Bacharach, 1989).
Anytime we propose a relationship between two concepts (often specified by arrows between "boxes" in
a diagram) we are faced with the task of demonstrating causality. In the case of OM, most researchers
are interested in identifying what the effect of various managerial decisions are on system or firm
performance. While the notion of cause and effect is fairly simple to understand, demonstrating causality
for these types of situations is by no means an easy undertaking.
Causality represents the "holy grail" since researchers have always striven to achieve it yet have
acknowledged that it can never be irrefutably "proven". The role of causality in science has its origins in
the work of Hume (see Cook and Campbell, 1979) who stressed three conditions for inferring that a
cause X has an effect Y: 1) both the proposed cause and effect must be concurrent and contiguous, 2)
there is temporal precedence, (the cause always occurs before the effect in real time), and 3) there must
be demonstrated conjunction between the two. Humes' approach challenged the early positivist views of
causality (that correlation implied causation), and later essentialist views (that causation referred to
variables which were necessary and sufficient for the effect to occur).
Mill's work (1843) has had the greatest influence on current paradigms of causation, and posits
three conditions for establishing causation:
 Cause precedes effect in time (temporal precedence)
 Cause and effect have to be related (simultaneity)
 Other explanations have to be eliminated (exclusivity).
Of these three requirements, the third is often the most difficult to achieve in an OM setting, for
it implies control over alternative explanations. While sources of variation and simultaneity can be
established in simulation and mathematical modeling, these factors are not so easily controlled for in
empirical research designs. Most OM studies take place in field settings which are subject to a great
number of possible sources of variation. An empirical OM researcher must strive to develop criteria
within the research design which provide some evidence that the conditions for causality are being met,
by ruling out those variables that are possible "causes" of the effects under study.
Blalock (1970) notes two primary problems that empirical researchers encounter:

One is in the area of measurement, and in particular the problem of inferring indirectly
 17

what is going on behind the scenes (e.g., in people’s minds) on the basis of measured
indicators of the variables in which we are really interested. The second area where
precise guidelines are difficult to lay down is one that involves the linking of
descriptive facts (whether quantitative or not) with our causal interpretations or theories
as to the mechanisms that have produced these facts.

Statistical methods are important in proving that there is sufficient evident that a relationship
between theoretical constructs exists. However they are not sufficient by themselves. In many cases, the
reviewer must look beyond the tables and numbers (Blalock, 1970). The reviewer should judge the
validity of the proposed relationship based on the soundness of the logic used in measuring the
constructs, and should also look for supporting evidence of the relationship, even if it is anecdotal. One
area which should receive special attention is that of the “quality” of the measures being used. Theory-
driven research is very sensitive to this aspect. Unlike studies where the goal is prediction, the intent is
that of explaining. To insure that this objective is met, the researcher must recognize the threats to
validity posed by the potential presence of missing variables, common methods, errors in variables and
other similar problems and must have controlled for them within the study.
Authors can strengthen their argument by employing data “triangulation.” That is, they should
be able to supplement their statistical results with case studies, quotes, or even personal insights that may
help to portray the results in a vivid way and provide additional insights. Research that draws on different
kinds of data that converge to support a relationship certainly provides a stronger case for inferring
causality. For instance, perhaps the OM researcher can interview members of a process improvement
team, or even check with the plant’s customers. The combination of methodologies to study the same
phenomenon borrows from the navigation and military strategy that use multiple reference points to
locate an object’s exact point (Smith, 1975; Jick, 1983).
One problem that OM researchers often encounter is that managers want to be represented in the
best possible light. Triangulation can be an important tool to combat this problem. For instance, a
manager may circle the “Strongly Agree” response on a Likert scale survey question asking whether “We
use TQM in our plant.” However, if the researcher conducts interviews in the plant and finds that few
workers agree with the philosophy of TQM and even fewer understand or use the tools of Statistical
Process Control, then considerable doubt is cast on the validity of the survey question.
Finally, researchers should look for multiple methods to verify statements deduced from prior
empirical studies. What may be appropriate in one setting may not be appropriate in another.

4.3 Utility

The third attribute is more commonly referred to usefulness. Using the basic questions and practical
styles of a journalist, Whetten (1989) suggests that the essential ingredients of a value-added theoretical
 18

contribution are explicit treatments of Who? What? Where? When? Why? and How? These criteria are
not enough however. A theory that is complete, comprehensive and exhaustive in its analysis but which
addresses an issue seldom if ever encountered in the field is not useful. In OM, research funding often
comes from private industries that are impatient in realizing the practical significance and utility of
abstract theories. Our research should be applied and the outputs potentially applicable to the OM
environment. Therefore, useful theories should have the following traits:

 The theory must deal with a problem of "real importance."

 The theory must point to relationships or uncover potentially important variables overlooked in prior
studies.
 The theory must direct the researcher to issues and problems not previously examined (but which are
still of interest).
 The theory must explain or provide insight into behavior or events occurring in other areas.
 The theory must be operationalized.
 The theory and its output must be interesting.

Of these traits, the last one requires further explanation. First advocated by Davis (1971),
interesting theories were ones which caused the readers to ‘sit up and take notice” (Davis, 1971, p. 310).
To be interesting, theories had to present an attack on an assumption which was taken for granted by the
readers. Interesting theories present one of two types of arguments (Davis, 1971, p. 311):

 What is seen as non-X is really X, or,

 What is accepted as X is actually non-X.

The assumptions attacked by interesting theories cannot be ones strongly held by the readers.
Papers presenting such arguments are examples of That’s Absurd and are often summarily dismissed by
the readers, not to mention reviewers and discussants! For example, if we were to read a paper arguing
that there is no linkage between corporate strategy, corporate performance and manufacturing
capabilities, our initial reaction would be to dismiss the paper out-of-hand without reading it any further.
Why? Because we see economic performance as being strongly influenced by manufacturing
capabilities. This view has been shaped by a long line of research going back to Adam Smith!
Second, interesting theories must consider both the theoretical and the practical dimensions
(Simon, 1967). They must be seen as being of real practical significance to the audience. This
significance might lie in directing research into new directions; it could indicate new research
methodologies. If the practical consequences of a theory are not immediately apparent, the theory will be
rejected. Theories lacking such practical significance are examples of the category of Who Cares?
Third, interesting theories must challenge. Theories which merely confirm views, assumptions
or frameworks already accepted by the audience are not interesting. Such theories represent the That’s
 19

Obvious category. As can be seen from this discussion, to generate theories that are interesting, the
writers must identify and understand their audience. What may be obvious to one audience may be
absurd to another and interesting to a third.
It should be noted here that the discipline of OM is one in which practical knowledge,
accumulated from years of experience, has surpassed scientific knowledge built upon theories that have
withstood many empirical attempts at falsification. To establish OM as a scientific field, it may therefore
be necessary in some cases to relax the stringent condition of “That’s Obvious”. What is obvious in OM
is often based on anecdotal evidence with few formally constructed statements of conceptual
relationships. Therefore, in the early stages of our field’s emergence as a scientific discipline, perhaps it
is all right, if not imperative, that such “obvious” theories be generated and empirically examined.

4.4 Parsimony

In addition to “not wrong,” causality and utility, a fourth trait of good theory is parsimony (Reynolds,
1971). A good theory should be rich enough to capture the fewest yet most important variables and
interactions required to explain the events or outcomes of interest. Why is parsimony so important?
Because the power of any theory is inversely proportional to the number of variables and relationships
that it contains. The theory should be free of redundancy, and if it could do as well or better without a
given element of form or content, that element is an unnecessary complexity and should be discarded
(Wallace, 1971). As Popper noted, “Simple statements . . are to be prized more highly than less simple
ones because they tell us more; because their empirical content is greater; because they are better
testable” (1961: 142).
Researchers should be aware that the need to be parsimonious introduces its own set of
challenges. By excluding certain dimensions to focus on other more important dimensions, the
researcher runs the risk of potentially overlooking or omitting important factors in the development of a
theory. Important extensions to current theories are often uncovered by researchers who examine those
factors which are either omitted or treated in a very superficial or simplified manner. In deciding whether
a theory contains extraneous elements, the reviewer should consider whether it adds to our overall
understanding of the phenomenon, or whether it can be simplified to its essential elements. The 80/20
rule may again apply in such cases!

5.0 Conclusion

Theory development is a dynamic process with theories essentially being “work-in-process.” At any point
in time, some segments of the theory are being tested for internal or external validity and other segments
being discovered, described or mapped. Each stage in this process is driven by different types of research
 20

questions and has different objectives. As a result, there is a need to apply different sets of research
methodologies as one undertakes various activities. What may be highly appropriate for one stage of
theory building may be inappropriate for another. These relationships between stage in theory
development and research methodologies are summarized in Table 1.
Although the process of theory-building is not always strongly sequential, it often begins with
discovery and ultimately culminates with theory validation, extension, and refinement. However,
researchers can enter at any stage in the theory building process. The stage at which a researcher enters is
often influenced by their academic research training and skills, and the degree to which he or she feels
comfortable in dealing with the methodologies typically employed at each stage. Working at the early
stages of theory building requires that the OM researcher be out in the field and in close contact with the
environments being studied. In later stages (e.g., theory validation), a large portion of our knowledge
and frameworks now come from previously completed research (published and private). In these stages,
the researcher can choose to maintain distance from the situation being studied by drawing data from
large scale mailings of survey questionnaires, or employing computer simulation models.

INSERT TABLE 3 ABOUT HERE

We are fortunate to find ourselves in a comparatively young field with many areas ripe for theory
development. Some of these areas have been classified according to their stage of theory development in
Table 3. This table is not intended to be an exhaustive list of topics in OM, but is intended to provoke
discussion. Areas such as inventory theory, job shop scheduling, and manufacturing planning are
comparatively well-developed. Much of the current research here involves extending and refining
existing theories. Other concepts such as Time-based competition, Total Quality Management, Lean
Manufacturing, and Cross-functional Teaming have been around for a number of years, but are still in the
Mapping and Relationship Building stage from a theoretical context. While the concepts are fairly well-
defined, there remains considerable work to be done in establishing the critical implementation success
factors within organizations that lead to improved performance. Finally, a number of emerging areas in
operations are still in the embryonic stage, including Environmentally-Conscious Manufacturing, Supply
Chain Management, and the Virtual Organization.
This article proposes a road map for OM researchers to follow in developing theory-driven
research, and has also outlined a number of key attributes for evaluating this research. From this point,
we need to assess the critical areas for theory development through a number of different streams of
activity. First, theory-building needs to be emphasized in our research methods seminars for doctoral
students. Second, we should encourage cross-fertilization of our doctoral students in other fields to
broaden our theoretical foundations. Finally, professional development seminars at conferences should be
 21

developed in order to “re-tool” academics interested in theory building. It is hoped that OM researchers
can employ the material presented to guide them in building better and more consistent theories, and
progress towards a better understanding of the radical change taking place in the field.
 22

References

Ahire, S.L., D.Y. Golhar, and M.A. Waller, 1996. “Development and validation of TQM
implementation constructs”, Decision Sciences Journal, vol. 27, no.1, pp. 23-56.

Anderson, J. C., M. Rungtusanatham, and R.G. Schroeder, R.G, 1994. “A theory of quality
management underlying the Deming management method”, Academy of Management
Review, vol. 19, no. 3, pp. 472-509.

Bacharach, S.B., 1989. "Organizational theories: Some criteria for evaluation”, Academy of
Management Review, vol. 14, no. 4, pp. 496-515.

Bergmann, G., 1957. Philosophy of Science, The University of Wisconsin Press , Madison, WI.

Black, S.A., and Porter, L.J., 1996. “Identification of the critical factors of TQM”, Decision
Sciences Journal, vol. 27, no.1, Winter, pp. 1-22.

Blalock, H.M. Jr., 1969. Theory Construction, Prentice Hall, Englewood Cliffs, NJ.

Blalock, H.M. Jr., 1970. Introduction to Social Research, , Prentice Hall, Englewood Cliffs, NJ.

Bohm, D., 1957. Causality and Chance in Modern Physics, Routledge and Kegan Paul, London.

Bonsignore, M. R., (CEO, Honeywell Inc.), 1992. “Quality implementation in a global

environment”, Keynote address at the American Society of Heating, Refrigeration and
Air Conditioning Engineers, Baltimore, Maryland, June 27.

Business Week, 1979. "American manufacturers strive for quality - Japanese style”, March 12,
pp. 32B-32W.

Business Week, 1994. “Quality: how to make it pay”, August 8, pp. 54-59.

Campbell, D.T., 1986. "Science's social system of Validity-enhancing Collective Belief change
and the Problems of the Social Sciences." In D.W. Fiske & R.A. Shweder (Eds).
Metatheory in Social Science: Pluralisms and Subjectivities , University of Chicago
Press, Chicago, pp. 108-135.

Coehn, B., 1980. Developing Sociological Knowledge: Theory and Method, Prentice Hall,
Englewood Cliffs, NJ.

Cook T., and D. Campbell, 1979. Quasi-Experimentation: Design and Analysis Issues for Field
Settings, Rand McNally, Chicago, IL.

Crosby, P. B.. 1979. Quality is Free, McGraw-Hill, New York.

Crovitz, H.F., 1970. Galton's Walk, Harper, New York.

 23

Davis, M., 1971. “That's Interesting! Towards a Phenomenology of Sociology and a Sociology
of Phenomenology”, Philosophy of the Social Sciences, vol. 1, pp. 309-344.

Deming, W.D., and L. Geoffrey, 1941. "On sample inspection in the processing of census
returns", Journal of American Statistical Association, vol. 36, pp. 351-360.

Deming, W.E., 1981. “Improvement of quality and productivity through action by

management”, National Productivity Review, vol. 1, pp. 12-22.

Deming, W.E., 1982. Quality, Productivity and Competitive Position, MIT Center for Advanced
Engineering Study, Cambridge, MA.

Deming, W.E., 1986, Out of the Crisis, MIT Center for Advanced Engineering Study, Cambridge, MA.

Dubin, R. 1969. Theory Building, Free Press, New York.

Eisenhardt, K.M. 1989. "Building theories from case study research”, Academy of Management
Review, vol. 14, no. 4, pp. 532-550.

Flynn,.B. B., S. Sakakibara, R. G. Schroeder, K. Bates, and E. J. Flynn, 1990. "Empirical

research methods in operations management", Journal of Operations Management, vol. 9,
no. 2, pp. 250-284.

Flynn, B.B., R.G., Schroeder, and S.A. Sakakibara, 1994 . “A framework for quality
management research and an associated measurement instrument”, Journal of Operations
Management, vol. 11, pp. 339-366.

Glaser, B.G., and Strauss, A.L., 1967. The Discovery of Grounded Theory, Aldine Publishing
Co., Chicago, IL.

Greene, R.T., 1993. Global Quality: A Synthesis of the World's Best Management Models, , ASQC
Press, Milwaukee, WI.

Greer, S., 1969. The Logic of Social Inquiry, Aldine Publishing Co., Chicago, IL.

Handfield, R., 1989. "Quality management in Japan versus the United States: an overview",
Production and Inventory Management, Second Quarter, pp. 79-85.

Handfield, R., and S. Ghosh, 1994, “Creating a total quality culture through organizational
change: a case analysis”, Journal of International Marketing, vol. 2 , pp. 15-30.

Hayes, R., 1981. "Why Japanese factories work", Harvard Business Review, July-Aug, pp. 57-
66.

Hedrick, T. E., L. Bickman, and D.J. Rog, 1993. Applied Research Design: A Practical Guide.
Sage Publications, Newbury Park, CA.

Hendricks, K., and V. Singhal, 1998. “Quality awards and the market value of the firm: An
 24

empirical investigation,” forthcoming in Management Science.

Hempel, C.G., 1965. Aspects of Scientific Explanation, Free Press, New York.

Hempel, C.G., 1952. “Methods of concept formation in science,” in International Encyclopedia

of Unified Science, , University of Chicago Press, Chicago, IL.

Jick, T.D., 1983. “Mixing qualitative and quantitative methods: triangulation in action”,
reprinted in J. Van Maanen (Ed.), Qualitative Methodology Sage, Newbury Park, CA.

Juran, J.M., 1974. Quality Control Handbook, 3rd ed., McGraw-Hill Company, San Francisco,
CA.

Kaplan, A., 1964. The conduct of inquiry. Chandler, San Franciso, CA.

Kerlinger, F. N., 1986. Foundations of Behavioral Research (3rd Edition), Harcourt Brace
Jovanovich College Publishers, Fort Worth, TX.

McCutcheon, D.M., and Meredith, J.R., 1993. “Conducting case study research in operations
management”, Journal of Operations Management, vol. 11, pp. 239-256.

Merton, R., 1957. Social Theory and Social Structure The Free Press, Glencoe, IL.

Miles, M.B., and Huberman, A.M., 1994. Qualitative Data Analysis: A Sourcebook of New
Methods, : Sage Publications, Newbury Park, CA.

Mill, J.S., 1843. A System of Logic, Ratiocinative, and Inductive, 2 vols. Parker, London.

Misterek, S., 1995. The Performance of Cross-Functional Quality Improvement Project Teams.
Ph.D. dissertation, University of Minnesota, Minneapolis, MN.

Nagel, E., 1961. The Structure of Science: Problems in the Logic of Scientific Exploration,
Harcourt, Brace and World New York, NY.

Osigweh, C. A.B., 1989. "Concept fallibility in organizational science", Academy of

Management Review, vol. 14, no. 4, pp. 579-594.

Platt, J.R., 1964. “Strong Inference”, Science, vol. 146, p. 347-353.

Polanyi, L., 1989. Telling the American Story, MIT Press, Cambridge, MA.

Popper, K.R. 1961. The Logic of Scientific Discovery, Science Editions, New York.

Powell, T., 1995. “Total quality management as competitive advantage: A review and
empirical study”, Strategic Management Journal, vol. 16, pp. 15-37.

Reynolds, P.D., 1971. A Primer in Theory Construction, Bobbs-Merrill Company,

Inc.,.Indianapolis, IN
 25

Shewhart, W.A., 1931. Economic Control of Quality of Manufactured Product, D. Van Nostrand
Company, Inc., New York, NY.

Simon, H.A., 1967. “The business school: A problem in organizational design”, Journal of
Management Studies, vol. 4, pp. 1-16.

Sirota, L. and Alper, A., 1993. “Employee survey research”, in: A. Hiam (Ed.), Does Quality
Work? A Review of Relevant Studies, The Conference Board ,New York.

Sitkin, S. B., K. M. Sutcliffe, and R.G. Schroeder, 1994. “Distinguishing control from learning
in total quality management: a contingency perspective”, Academy of Management
Review, vol. 19, no. 3, pp. 537-564.

Smith, H.W., 1975. Strategies of Social Research: The Methodological Imagination, Prentice
Hall, Englewood Cliffs, NJ.

Spencer, B., 1994. “Models of organization and total quality management: a comparison and
critical evaluation”, Academy of Management Review, vol. 19, no. 3, pp. 446-471.

Stinchcombe, A.L., 1968, Constructing Social Theories, Harcourt, Brace and World, New York.

Stratton, B., 1996. “Not the best years of their lives”, Quality Progress, May, pp. 24-30.

Swamidass, Paul M., 1991. “New frontiers in operations management research”, Academy of
Management Review, vol. 16, no. 4, October, pp. 793-814.

Tribus, M, 1984. "Prize-winning Japanese firms' quality management programs pass

inspection", Management Review, February, pp. 31-32, 37.

Wall St. Journal, 1985. "Cause of quality control problems might be managers, not workers",
April 10, p. 1.

Wallace, W. 1971. The Logic of Science in Sociology, Aldine Atherton, Chicago, IL,.

Watson, W.H., 1938. “On methods of representation,” from On Understanding Physics,

University Press, Cambridge (England). Reprinted in Philosophy of Science, edited by
Arthur Danto and Sidney Morgenbesser, 1960, Cleveland, World Publishing Company,
pp. 226-244.

Weick, K.E., 1989. "Theory construction as disciplined imagination”, Academy of Management

Review, vol. 14, no. 4, October, pp. 515-531.

Wheelwright, S. C., 1981. "Japan - where operations really are strategic", Harvard Business
Review, July-Aug, pp. 68 - 80.

Whetten, D.A., 1989. "What constitutes a theoretical contribution?", Academy of Management

Review, vol. 14, no. 4, October, pp. 490-495.
 26

Van De Ven, A.H. 1989. "Nothing is quite so practical as a good theory," Academy of
Management Review, vol. 14, no. 4, October, pp. 486-489.

Yin, R.K., and Heald, K.A., 1975. “Using the case survey method to analyze policy studies”,
Administrative Science Quarterly, vol. 20, pp. 371-381.
 27

Figure 1 - The Principal Information Components, Methodological Controls, and

Information Transformations of the Scientific Process (Wallace, 1971, p. 18)

Theories
Step 3

Concept
Conceptand
and
Proposition
Proposition Logical
Logical
Logical
Logical
Formation,
Formation,and
and Deduction
Deduction
Deduction
Deduction
Arrangement
Arrangement Step
Step55

Decisions to
Empirical Accept or Reject
Generalizations Hypotheses Hypotheses
Step 2 Step 4a

Tests
Testsof
of
Hypotheses
Hypotheses Interpretation,
Interpretation,
Measurement,
Measurement, Step
Sample Step4b
4b Instrumentation,
Instrumentation,
SampleSummari-
Summari- Scaling,
zation, Scaling,and
and
zation,Parameter
Parameter Sampling
Estimation Sampling
Estimation

Observations
Step 1

Note: Informational components are shown in rectangles; methodological controls are

shown in ovals; information transformations are shown by arrows.
 28

Table 1
Match Research Strategy with Theory Building Activities

Purpose Research Research Examples of Examples of

Question Structure Data Data Analysis
Collection Procedures
Techniques

1a. Discovery What is going on In-depth case Observation Insight

Uncover areas for here? studies Interviews Categorization
research & theory Is there something Unfocused, Documents Expert Opinion
development interesting enough longitudinal field Elite Interviewing Descriptions
to justify research? study

1b.Description What is there? In-depth case Observation Insight

Explore territory What are the key studies Interviews (group Categorization
issues? Unfocused, or individual) Expert Opinion
What is longitudinal field Documents Descriptions
happening? study Elite Interviewing Content Analysis
Critical Incident
Technique

2. Mapping What are the key Few focused case Observation Verbal Protocol
Identify/describe key variables? studies In-depth interviews Analysis
variables What are the In-depth field Diaries Cognitive Mapping
draw maps of the salient/critical studies Survey Repertory grid
territory themes, patterns, Multi-site case questionnaires technique
categories? studies History Effects Matrix
Best-in-class case Unobtrusive Content Analysis
studies measures

3. Relationship What are the Few focused case Observation Verbal Protocol
Building patterns or studies In-depth interviews Analysis
Improve maps by linkages between In-depth field Diaries Cognitive Mapping
identifying the variables? studies Survey Repertory grid
linkages between Can an order in Multi-site case questionnaires technique
variables the relationships studies History Effects Matrix
Identify the “why” be identified? Best-in-class case Unobtrusive Content Analysis
underlying these Why should these studies measures Factor Analysis
relationships relationships exist? Multidimensional
Scaling
Correlation
analysis
Nonparametric
statistics
 29

4. Theory Are the theories Experiment Structured Triangulation

Validation we have generated Quasi-experiment Interviews Analysis of
Test the theories able to survive the Large scale Documents variance
developed in the test of empirical sample of Open and closed- Regression
previous stages data? population ended Analysis
Predict future Did we get the questionnaires Path Analysis
outcomes behavior that was Lab experiments Survival Analysis
predicted by the Field experiments Multiple
theory or did we Quasi-experiments comparison
observe another Surveys procedures
unanticipated Nonparametric
behavior? statistics

5. Theory How widely Experiment Structured Triangulation

Extension/ applicable/generali Quasi-experiment Interviews Analysis of
zable are the Large scale Documents variance
Refinement theories that we sample of Open and closed- Regression
To expand the map
have developed? population ended Analysis
of the theory
To better structure Where do these questionnaires Path Analysis
the theories in light theories apply? Lab experiments Survival Analysis
of the observed Where don’t these Field experiments Multiple
results theories apply? Quasi-experiments comparison
Surveys procedures
Documentation Nonparametric
Archival Research statistics
Meta Analysis
 30

Table 2 - Recasting Table 1 from a Process and TQM Perspective

PROCESS STEP PURPOSE ILLUSTRATIVE TQM

REFERENCES
Discovery Juran, 1974
Step 1a Observation Deming, 1981; 1982; 1986
Shewhart, 1931
Description Business Week, 1979
Step 1b Observation Crosby, 1979

Step 2 Empirical Mapping Hayes, 1981

Generalizations Wheelwright, 1981
Tribus, 1984
Step 3 Theories Relationship Building Green, 1993
Handfield & Ghosh, 1994
Sirota & Alper, 1993

Step 4 a, 4b Hypothesis Theory Validation Ahire, Goldhar & Waller,

Testing 1996
Black & Porter, 1996
Flynn, Sakakibara, Schroeder,
1994
Hendricks & Singhal, 1997
Misterek, 1995
Powell, 1995

Step 5 Logical Deduction Theory Extension/Refinement Anderson, Rungtusanatham &

Schroeder, 1994
Sitkin, Sutcliffe, & Schroeder,
1994
Spencer, 1994
 31

Table 3 - Stages of Research Areas in Operations Management

Theory Discovery / Mapping / Theory Validation /

Building Description Relationship-Building Extension /
Stage Refinement

Research Area • Computer Integrated • Order Entry and Release • Inventory Theory
Manufacturing Systems • Manufacturing Planning
•Environmentally-friendly • Just-in-Time and Control
Manufacturing / Design • Total Quality Mgmt. • Jobshop Scheduling
• Supply Chain Mgmt. • Time-based Competition • Lotsizing
• Global Manufacturing • Lean Manufacturing • Statistical Process
• "Extended Enterprises” • World Class Mfg. Control
• Learning Organizations • Concurrent Engineering • Project Management
•Inter-Organizational • Manufacturing Strategy • Focused Factories
Information Systems • Mass Customization • Production Competence
• Cross-functional Teaming