Journal of Operations Management 19 (2001) 39–58

The impact of total productive maintenance practices on

manufacturing performance
Kathleen E. McKone a,∗ , Roger G. Schroeder b , Kristy O. Cua b
a Babson College, Babson Park, MA 02457, USA
b Department of Operations and Management Science, Carlson School of Management,
University of Minnesota, 321 19th Avenue South, Minneapolis, MN 55455, USA
Received 2 March 1999; accepted 26 October 1999

Abstract
In this paper we investigate the relationship between Total Productive Maintenance (TPM) and manufacturing performance
(MP) through Structural Equation Modeling (SEM). We find that TPM has a positive and significant relationship with low cost
(as measured by higher inventory turns), high levels of quality (as measured by higher levels of conformance to specifications),
and strong delivery performance (as measured by higher percentage of on-time deliveries and by faster speeds of delivery).
We also find that the relationship between TPM and MP can be explained by both direct and indirect relationships. In
particular, there is a significant and positive indirect relationship between TPM and MP through Just-In-Time (JIT) practices.
© 2001 Elsevier Science B.V. All rights reserved.
Keywords: Maintenance and reliability; Empirical research; Statistical analysis

1. Introduction MP. However, TPM can be thought of as integral

to a World Class Manufacturing Strategy that also
The purpose of this paper is to present an empiri- involves JIT, TQM, and EI. In particular, Schon-
cal analysis of Total Productive Maintenance (TPM). berger (1986) argues that JIT, TQM, EI, and TPM
While Just-In-Time (JIT), Total Quality Management are critical components of World Class Manufactur-
(TQM) and Employee Involvement (EI) have been ing. Therefore, it is hypothesized that companies that
recognized as strong contributors to manufacturing implement TPM will not only be able to enhance
performance (MP) both in the practitioner literature their maintenance practices but also improve their
(Schonberger, 1986, Miller and Schenk, 1997) and MP.
the academic literature (Cleveland et al., 1989; Flynn This paper focuses on the relationship between TPM
et al., 1995; Jarrell and Easton, 1997; Sakakibara et al., and MP. We propose a conceptual framework to ex-
1997), there has been limited recognition (Maier et al., amine the nature of this relationship. Since TPM, JIT,
1998) of the role that maintenance plays in improving and TQM are critical to a world class manufacturing
strategy, we believe that it is necessary to consider JIT
∗ Corresponding author. Tel.: +1-781-239-4245; and TQM when assessing TPM. Therefore, our frame-
fax: +1-781-239-5272. work considers both direct and indirect relationships
E-mail address: kmckone@babson.edu (K.E. McKone). (through JIT and TQM) between TPM and MP. After

0272-6963/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 2 7 2 - 6 9 6 3 ( 0 0 ) 0 0 0 3 0 - 9
40 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

proposing our framework, we then test it using survey TPM is designed to maximize equipment effec-
data collected from 117 plants across three industries tiveness (improving overall efficiency) by estab-
and four countries. lishing a comprehensive productive-maintenance
The remainder of this paper is organized as follows. system covering the entire life of the equipment,
In Section 2 of the paper, we define our model and our spanning all equipment-related fields (planning,
hypotheses. In Section 3, we describe our data. In Sec- use, maintenance, etc.) and, with the partici-
tion 4, we discuss the measurement of our model vari- pation of all employees from top management
ables. In Section 5, we present our analysis approach. down to shop-floor workers, to promote produc-
Then, in Section 6, we present and discuss the results tive maintenance through motivation management
from our study. Finally, we present our conclusions. or voluntary small-group activities. (Tsuchiya,
1992, p. 4)
TPM provides a comprehensive company-wide ap-
2. Framework definition proach to maintenance management, which can be
divided into long-term and short-term elements. In
the long-term, efforts focus on new equipment design
In this section, we define the components of our
and elimination of sources of lost equipment time and
framework (shown in Fig. 1) relating TPM and MP.
typically require the involvement of many areas of the
After discussing the components of the framework,
organization. In this paper, we focus on the short-term
we present the theory that supports this framework
maintenance efforts that are normally found at the
and discuss the hypothesized relationships that will be
plant level of the organization. In the short-term,
analyzed in this paper.
TPM activities include an autonomous mainte-
nance program for the production department and a
2.1. TPM elements planned maintenance program for the maintenance
department.
Seiichi Nakajima, vice-chairman of the Japanese In- Throughout this paper, we measure TPM as in
stitute of Plant Engineers (JIPE), the predecessor of McKone et al. (1999). We consider seven elements
the Japan Institute of Plant Maintenance (JIPM), pro- of TPM in the paper: four elements of autonomous
moted TPM throughout Japan and has become known maintenance — housekeeping on the production
as the father of TPM. In 1971, TPM was defined by line, cross-training of operators to perform main-
JIPE as follows: tenance tasks, teams of production and mainte-
nance personnel, and operator involvement in the
maintenance delivery system; and three elements
of planned maintenance — disciplined planning of
maintenance tasks, information tracking of equip-
ment and process condition and plans, and schedule
compliance to the maintenance plan. These seven el-
ements will be discussed in more detail in Section 4,
when we discuss the measurement of our framework
variables.

2.2. MP dimensions

There are many different ways of measuring MP.

However, the most predominant approach in the lit-
erature is to use cost, quality, delivery, and flexibility
as the four basic dimensions of MP. In some studies,
these dimensions have been expanded to include sev-
Fig. 1. Framework. eral additional measures (Hayes et al., 1988; Miller
 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 41

and Roth, 1994). We consider the four basic dimen- are seeing 50% reductions in breakdown labor rates,
sions because the plant is most concerned with these 70% reductions in lost production, 50–90% reductions
measures. In our study, we have two components of in setups, 25–40% increases in capacity, 50% increases
cost–cost as a percentage of sales and inventory turns, in labor productivity, and 60% reductions in costs per
two components of delivery–percentage of on-time de- maintenance unit (Koelsch, 1993). Many companies,
liveries and speed of delivery, and one component each such as Steelcase (Koelsch, 1993), Tennessee East-
for quality and flexibility. man (Garwood, 1990), Nissan (Suzuki, 1992), Nip-
Use of the four basic dimensions to measure MP pondenso (Teresko, 1992), and Michigan Automotive
can be traced back to Skinner (1969) who launched Compressor (MACI, 1995) have told similar success
the current interest in manufacturing strategy and stories. All claim that TPM had a significant impact
MP measurement with his now classic article. Skin- on their maintenance effectiveness and their MP.
ner has been followed by many others who have The academic literature also supports the idea that
also advocated the four basic dimensions, includ- TPM, which enhances the technology base of the
ing Schroeder (1993) and Ward et al. (1995). These plant, can lead to improved MP. Adler and Shenhar
authors have sometimes referred to the four di- (1990) indicate that companies that develop their tech-
mensions as competitive priorities or manufactur- nological base are able to capitalize on technology’s
ing capabilities, but we refer to them here as MP ability to make a positive contribution to perfor-
dimensions. mance. TPM can improve the technological base of
a company by enhancing equipment technology and
2.3. JIT and TQM elements improving the skill of employees (improving two of
Adler and Shenhar’s dimensions of technology — the
technology and organizational assets). Therefore, by
In this paper, we consider comprehensive measures
improving the technology of the plant, TPM should
of the level of JIT and TQM implementation. We
help improve MP.
capture multiple aspects of JIT development: ven-
Furthermore, TPM helps to improve the organiza-
dor relations, customer relations, and several aspects
tion’s capabilities by enhancing the problem-solving
of JIT production — the management of materials,
skills of individuals and enabling learning across
scheduling of resources where and when needed, and
various functional areas. Tyer (1991) and Tyer and
setup reduction (Sakakibara et al., 1993, 1997). We
Hauptman (1992) found that successful change in
also consider several aspects of TQM development:
technology depends on the deployment of organiza-
supplier management, customer involvement, the in-
tional structures that enable individuals to work across
ternal system for quality, and top management lead-
functional boundaries to identify problems, develop
ership for quality (Flynn et al., 1994, 1996). These
solutions, and execute plans. Similarly, Hayes and
are indicators of the level of implementation of JIT
Wheelwright (1984) suggest that companies need to
and TQM.
build the skills of their workforce and develop worker
participation in order to compete through World Class
2.4. Hypotheses Manufacturing. TPM changes the structure of the or-
ganization to break down traditional barriers between
2.4.1. TPM positively influences MP maintenance and production, foster improvement by
We first hypothesize that TPM implementation has looking at multiple perspectives for equipment op-
a positive influence on MP. This hypothesis is based eration and maintenance, increase technical skills of
on the experiences of numerous companies as well as production personnel, include maintenance in daily
the theory discussed in the technology and strategy production tasks as well as long-term maintenance
literature. plans, and allow for information sharing among differ-
The benefits from implementing TPM have been ent functional areas. Therefore, TPM should develop
well documented at numerous plants. Constance Dyer, the capability of the organization to identify and re-
Director of Research and TPM Product Development, solve production problems and subsequently improve
Productivity Inc., says that companies that adopt TPM MP.
42 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

Our first hypothesis considers the relationship R1 without considering the other relevant organizational
in Fig. 1 and is referred to as characteristics.
TPM, when part of a world class manufacturing
H1. TPM has a positive and direct relationship with
strategy that incorporates JIT and TQM, should lead to
MP.
improved MP. The importance of the relationship be-
2.4.2. TPM indirectly affects MP through JIT and tween JIT and TPM is clear. JIT’s emphasis on waste
TQM reduction creates an environment where inventory
JIT, TQM, EI, and TPM programs have often is reduced, production processes are interdependent,
been referred to as components of “World-Class and the plant operation is susceptible to breakdowns
Manufacturing” (Schonberger, 1986, 1990; Stein- of any process. TPM provides dependable equip-
bacher and Steinbacher, 1993). The relationship be- ment, reduces the number of production disturbances,
tween JIT, TQM, and MP has been supported in and increases plant capacity by providing effective
academic research (Flynn et al., 1995, 1996). McK- equipment maintenance. A study by Sakakibara et al.
one et al. (1999) showed that the implementation (1997) showed that there was not a significant rela-
level of TPM was closely linked to the implemen- tionship between the use of JIT practices and MP;
tation level of JIT, TQM, and EI. Companies with however, the combination of JIT management and
higher implementation levels of JIT, TQM, and EI infrastructure practices were related to MP. Similarly,
also had higher implementation levels of TPM. A we believe that TPM practices indirectly influence
more general study by Tunälv (1992) showed empir- MP by supporting JIT practices.
ically that business units with a manufacturing strat- The relationship between TPM and TQM is also im-
egy placed significantly more emphasis on product- portant. TQM aims to reduce variation in the product
and process-related programs (such as JIT, quality and eliminate defects. A strong maintenance program
management practices, and preventive maintenance) is needed to provide reliable equipment maintenance
than those without a strategy. These same business and reduce equipment process variation. Flynn et al.
units were also more successful in their financial (1995) found that quality practices focusing solely on
performance. quality improvement might not be a sufficient means
In this study, we have not included EI as a sepa- for a plant to attain and sustain its competitive posi-
rate component in our framework since it pervades tion. It is likely that the use of TPM to improve equip-
all the other World Class Manufacturing compo- ment performance and increase the skills of workers
nents — TPM, JIT, and TQM — and, therefore, is could be an additional factor in supporting TQM and
implicitly included in our framework. Moreover, we explaining competitive advantage. Therefore, we be-
cannot comprehensively capture the implementation lieve that TPM indirectly improves MP by supporting
of EI as a separate component of World Class Man- TQM efforts.
ufacturing due to limitations of the database that we In this paper, we consider the indirect effect of TPM
are using. Therefore, the following discussion con- on MP through JIT and TQM. Barley (1990) indicates
siders the relationships between JIT, TQM, TPM, that technologies change organizational and occupa-
and MP. tional structures by transforming patterns of action
Hayes and Wheelwright (1984) emphasized the and interaction and that roles and social networks are
need to match the facilities and technology choice held to mediate technology effects. Similarly, organi-
with business manufacturing programs and people. zational practices, such as JIT and TQM, may support
A manufacturing program is successful only when TPM (a program focused on improving the technol-
it improves MP and is aligned with the business ogy base) and its effect on MP.
strategy. Similarly, the system frameworks of various Our second set of hypotheses considers the indirect
authors (Gerwin, 1976; Galbraith, 1977; Van de Ven relationships between TPM and MP through JIT (R2
and Ferry, 1980) all hypothesized that consistency and R4 in Fig. 1) and through TQM (R3 and R5 in
among organizational design characteristics leads Fig. 1). Our hypotheses for the indirect relationships
to higher performance. These studies suggest that between TPM and MP (given that the direct relation-
we should not consider the impact of TPM on MP ship between TPM and MP is considered) are:
 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 43

H2a. JIT accounts for a significant portion of the 12 production workers. This battery of questionnaires
positive relationship between TPM and MP. allowed for multiple respondents for each question,
thereby providing greater reliability of the data. In
H2b. TQM accounts for a significant portion of the addition, it allowed respondents to address their par-
positive relationship between TPM and MP. ticular area of expertise. For example, certain people
responded to the TPM questions and others responded
It is important to mention that our framework, as to the MP questions. Also, we used two types of
presented in Fig. 1, is one possible set of relationships. questions: objective and perceptual. The objective
Clearly, the framework explores both the direct and questions were answered by one respondent in each
indirect links between TPM and MP. While other re- plant and addressed topics which can be measured on
lationships are possible (e.g., TQM influences JIT or an objective basis such as: “what percentage of the
TPM), this paper concentrates on TPM and its rela- maintenance in the plant is performed by the work-
tionship with MP. ers rather than by a separate maintenance crew?”.
The perceptual questions were arranged in multi-item
scales to ensure accurate representation of the con-
3. Description of data structs of interest. Each scale consisted of several
questions pertaining to the same construct; the an-
The data used for empirical analysis of the frame- swers to the questions were averaged to arrive at a
work were collected as part of the World Class Man- scale score. By using different types of measures and
ufacturing (WCM) Study (Flynn et al., 1994) being various respondents, we eliminated potential prob-
conducted by a team of researchers at several universi- lems with common method or common respondent
ties in the US, Europe, and Asia. The WCM database bias.
used for our research was assembled in 1997 from In the next section, the constructs of interest con-
three different regions of the world and three different cerning TPM, JIT, TQM, and MP are described. These
industries using a common set of questionnaires. The constructs are measured by a combination of percep-
database addresses TPM, JIT, and TQM and includes tual scales and objective measures from the WCM
117 different manufacturing plants. database.
The WCM database contains data from plants in the
US, Italy, Germany, and Japan. These four countries
partially represent the three major regions of the in- 4. Measurement of variables
dustrialized world: North America, Europe, and Asia.
In each country, plants were selected from three indus- As shown in Fig. 1, we selected seven TPM mea-
tries: electronics, machinery, and automobile indus- sures, one measure each for JIT and TQM, and six MP
tries. A stratified design was used to randomly select measures from the WCM database which are briefly
an approximately equal number of plants in each coun- discussed in this section. In our database 41 cases had
try and each industry. For this study, we did not in- a single missing value, 7 cases had two missing val-
vestigate cross-country or cross-industry differences. ues, and 3 cases had three missing values out of 15
We utilized the worldwide dataset in order to test our measures. Table 1 provides summary statistics of the
hypotheses with a wide variety of plants. 15 raw observed measures.
The selected plants were contacted by a member of For our analysis, we transformed the 15 measures
the WCM research team to participate in the study. using optimal Box–Cox transformations to satisfy
Two-thirds of the plants contacted decided to join the normality. Then we standardized the measures by
study. This relatively high response rate was assured industry since we are not interested in cross-industry
by contacting the plants personally and by promising differences. Cross-country standardization was not
that they would receive a plant profile for comparison performed since plants compete globally. Where nec-
with other plants. essary we replaced missing values with the mean
The data were collected in each plant using ques- measurement value for the industry. All measures
tionnaires that were completed by 11 managers and were adjusted so that a high value reflects a high
44 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

Table 1
Summary statistics of observed measuresa
Measure N Mean Standard deviation

1. HOUSEKP 117 3.7659 0.4943

2. XTRAIN 117 3.6559 0.4499
3. TEAMS 117 3.6771 0.4933
4. OPINVOLV 110 37.7000 33.0264
5. INFOTRAC 117 3.3493 0.5546
6. PLANNING 117 2.9541 0.4905
7. SKEDCOMP 95 72.2211 30.6147
8. JIT 117 3.2712 0.3083
9. TQM 117 3.4859 0.3472
10. LOWCOSTb 115 0.7144 0.1822
11. INVTURN 111 8.8998 10.5478
12. CONFQLTYb 102 5.3240 6.7818
13. ONTIMEDV 116 88.2586 12.7730
14. FASTDVb 110 53.2384 66.0944
15. FLEXIBLEb 113 2.2566 1.0158
a The summary statistics were calculated using the responses to the items in the survey that have not been statistically adjusted.
b Indicates that a low value of the measure reflects good performance.

level of program implementation or good perfor- tenance efforts; (2) this is a plant-level study and
mance. A correlation matrix and the variances of cannot assess the organization-wide maintenance ef-
the 15 statistically adjusted measures are shown in forts; and (3) this is not a longitudinal study and can-
Table 2. not evaluate the long-term efforts well. See Appendix
We also performed Box’s M test to determine A for details of the questions used for our analy-
if the combination of data from three industries sis. Rather than simply measuring the existence of a
and four countries is suitable for structural equa- TPM program, our questions assess the level of TPM
tion modeling (SEM). The tests performed on the implementation.
statistically adjusted data provide no evidence to The autonomous maintenance variables in-
conclude that there is significant difference among clude three perceptual measures for housekeeping,
the covariance matrices of the measures across the cross-training, and teams, and an objective measure
three industries and the four countries. Based on this for operator involvement. For housekeeping, we uti-
analysis, we concluded that the use of our statisti- lized a five-question scale from the WCM database.
cally adjusted dataset was sufficient for our analysis These questions relate closely to the 5-S approach,
approach. a system for industrial housekeeping practices that
is discussed in books by Nakajima (1988), Suzuki
4.1. Measurement of TPM (1992), Shirose (1992), and Tajiri and Gotoh (1992).
To assess the level of cross-training, we used five
For this study, we have selected questions from questions that relate to the amount of cross-training
the WCM database that fit well with our literature that is provided and utilized within the plant. Our
review on TPM and concentrate on the daily main- measure evaluates the skills of operators and specifies
tenance efforts that could be normally found at the whether or not an organization has established an en-
plant-level of the organization. These short-term TPM vironment where cross-training is possible. Similarly,
efforts include both autonomous and planned main- for the autonomous maintenance team measure, we
tenance activities. We have chosen to concentrate on measured the general level of team involvement within
short-term daily efforts for three reasons: (1) typi- the plant. We utilized a five-question team scale that
cally TPM efforts begin with these in-plant main- assesses the environment that is established for pro-
K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 45
46 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

duction and maintenance teams. Finally, for operator manufacturing cost, we measured the manufactur-
involvement, we used an objective measure of the ing cost of goods sold as a percentage of sales. We
percentage of operators who are directly involved in measured inventory costs as the inventory turnover
the maintenance delivery process. This measure pro- ratio. A high turnover ratio indicates a low cost po-
vides another indicator of the implementation level of sition. Both of these ratios are dimensionless and
autonomous maintenance. are not subject to currency differences between
While both the operators and maintenance per- countries.
sonnel are involved in the planning and execution Quality from a manufacturing point of view is
of maintenance within a TPM program, the main- measured as the percentage of good products that are
tenance personnel are ultimately held accountable produced (conformance to specifications) or if the
for long term maintenance planning and the state of quality is good enough defects can be measured in
readiness of the equipment. With the data that was ppm (parts per million). Of course, quality can also
available, we considered three measures of planned be measured by customer satisfaction, which is in-
maintenance: two perceptual measures for disciplined fluenced not only by the lack of defects but also by
planning and information tracking, and an objective product designs and after-sales service. Since product
measure for schedule compliance. A disciplined plan- design and service are cross-functional responsibili-
ning approach typically dedicates time for scheduled ties, we do not include customer satisfaction here as
maintenance activities, assigns tasks to specific peo- a MP measurement and only consider conformance
ple and inspects for good quality workmanship. We quality.
considered four questions that address the planning of Delivery performance includes two different mea-
the maintenance department. An information system sures: the percentage of orders delivered on time (or
that tracks past and current equipment performance filled from stock) and the manufacturing lead-time
is also important to a successful maintenance depart- from when an order is placed until it is delivered.
ment. We assessed the information tracking systems These measures are indicative of a plant’s ability to
that are relevant to equipment performance through deliver quickly and as promised.
five questions. Finally, compliance to a planned main- Finally, flexibility can be measured in a number of
tenance schedule is a measure of the successful appli- different ways. We have chosen to use one measure:
cation of the maintenance tools and execution of the the length of time that it takes to change the master
plans. We used a self-reported schedule compliance production schedule. Most plants have a frozen pro-
measure as another indicator of planned maintenance duction horizon inside of which they do not take addi-
implementation. tional orders or make changes to existing orders. This
production horizon measures a plant’s capability to
4.2. Measurement of MP make changes and, of course, a shorter horizon offers
more flexibility.
In this study, we are measuring MP at the plant
level. Since the plant does not control sales or costs 4.3. Measurement of JIT and TQM
outside the plant, overall financial measures of plant
performance are not appropriate. Rather, the basic di- Our goal for measuring JIT and TQM was to mea-
mensions of plant performance which are controlled sure the general level of program implementation
by the plant are used: cost, quality, delivery, and flexi- rather than to simply consider the existence of the
bility (Skinner, 1969; Schroeder, 1993; Ward et al., specific program. Since the focus of this study is the
1995). We discuss our measurement of each of the relationship between TPM and MP, we measured JIT
four dimensions of plant performance in this section. and TQM at an aggregate level, using one manifest
Appendix B has the details of our survey questions variable for each, combining several aspects of pro-
on MP. gram implementation. We did not include the years
Cost is interpreted to mean not only the tradi- of implementation efforts since we were more con-
tional accounting cost of manufacturing, but also cerned with the level of JIT or TQM implementation
the economic costs associated with inventories. For than the time since initial adoption. See Appendix
 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 47

C for details of the survey questions used for our for each scale ranged from 0.77 to 0.90. Since all α
measures. scores were considerably higher than the 0.70 accept-
To measure the implementation of JIT, we cons- able level advocated by Nunnally (1978), all scales
idered various JIT practices and developed a linear exhibit a high degree of reliability.
combination of five scales used in Sakakibara et al.
(1993, 1997). Our measurement captures JIT deliv-
ery by suppliers, JIT link with customers, pull sys- 5. Method of analysis
tem support, repetitive nature of the master production
schedule, and setup reduction efforts within the plant. Our analysis focused on understanding the nature
This is a comprehensive measurement of JIT involv- of the relationship between TPM and MP. Through
ing five different scales that measure different aspects SEM using AMOS 3.61 (Arbuckle, 1997), we tested
of JIT. our specified framework (refer to Fig. 1 for a diagram
To measure the implementation level of TQM, we of the model and the testable paths). We evaluated
considered customer involvement, rewards for quality, our measurement model and considered the relation-
supplier quality management, and top management ship between observed TPM measures and a latent
leadership for quality. Previous studies have found TPM construct, and between plant performance mea-
that these aspects of TQM adequately represent a sures and a latent MP construct. We also tested our
broad-based view of the construct (Flynn et al., 1994, specified hypotheses between the latent TPM and
1996). MP constructs and the observed JIT and TQM mea-
sures. The results of the SEM analysis allowed us
4.4. Validity and reliability of scale measures to describe the correlation between variables, to un-
derstand which TPM variables best explain the TPM
Our research used data from the WCM Study and construct, to understand the nature (direct and indi-
many of the constructs have been used and tested in rect) of the relationship between TPM and MP, and
previous studies (Sakakibara et al., 1993, 1997; Flynn to understand which MP measures best explain the
et al., 1994, 1996; McKone et al., 1999). In addi- MP construct (the variables that are highly influenced
tion, the items used for each construct fit well with by TPM).
the concepts of TPM, JIT, and TQM discussed in the We took a hierarchical (staged) approach to test-
framework and existing literature and therefore have ing hypothetical models that describe the relationship
a high degree of content validity. Although it may between both observed and unobserved measures for
be difficult to completely separate the concepts of TPM, JIT, TQM, and MP. This staged approach, sim-
TPM, JIT, and TQM, the measures used for these con- ilar to hierarchical regression, allows us to determine
structs are not identical. The discriminant validity of if the addition of new set of relationships adds signi-
the factors for TPM, JIT, and TQM were tested us- ficantly to our explanation of the variation in the data.
ing confirmatory factor analysis approach (Bagozzi, Therefore, we can test H2a and H2b by evaluating the
1980; Burnkrant and Page, 1982). In the tests, mod- difference in model fit when the indirect relationships
els of separate but correlated factors were compared are added to the model.
to models in which the pair of factors was hypothe- At each stage of the model testing process, we
sized to have a unity correlation or to be unidimen- verified that the assumptions required for SEM were
sional. The difference between the two models was met. The critical ratio of Mardia’s (1970, 1974, 1985)
evaluated using a change in χ 2 test with one degree coefficients of separate and joint multivariate kurtosis
of freedom. In the comparisons of the TPM–JIT mod- and skewness did not indicate significant differences
els and the TPM–TQM models, the χ 2 difference val- from zero (Bollen, 1989). Tests using Mahalanobis
ues were significant, indicating that TPM is indeed distance showed no evidence of the existence of out-
a separate scale and is only correlated with JIT and liers. χ 2 plots of the squared Mahalanobis distances
TQM. did not exhibit any systematic curvature. Thus, there
We used Cronbach’s coefficient α to evaluate the is no evidence to conclude that the data does not
reliability of the scales at the plant-level. The α scores satisfy multivariate normality.
48 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

For each of our models, maximum likelihood arate a good-fitting model from a poor-fitting model
estimation was used, the procedure converged for and can determine if one model provides a signifi-
estimation, the model was identified, and all residual cantly better fit than another model. To test our hy-
variances were positive. At each step of the model potheses, we need to determine if the addition of a
testing process, we compared the fit of the mod- new relationship to our model helps to improve the
els. For model evaluation, we used several standard explanation of the variation in the data. Therefore,
model evaluation criteria. (1) The Degrees of Free- comparing the difference in LRT statistics (dLRT)
dom (DF) represents the difference between the with the associated dDF is important to testing our
number of independent statistics and the model pa- hypotheses.
rameters fitted. (2) The Likelihood Ratio Test (LRT)
statistic is minimized and is usually interpreted as
a χ 2 variate. (3) The Likelihood Ratio Test to De- 6. Results and discussion
grees of Freedom (LRT/DF) Ratio is a relative χ 2
measure for model fit. A value of less than 5 for Through SEM, we tested our hypothesized relation-
this ratio indicates acceptable fit (Wheaton et al., ships between TPM and MP. Each stage of our ana-
1977; Marsh and Hocevar, 1985). (4) The Goodness lysis resulted in a new model, the results of which are
of Fit Index (GFI) rescales the fit of the observa- shown in Table 3. Models 2, 3, and 4 are a nested se-
tions and the expectations between 0 and 1, where quence of models that allow us to provide information
1 represents a perfect fit. (5) Bollen’s (1989) Incre- about distinct aspects of the structural equation model
mental Fit Index (IFI) basically represents the point embedded within the sequence. By using the dLRT
at which the model being evaluated falls on a scale statistics, we are able to isolate where fit and lack of
running from the null model (where all correlations fit arise in the model in the nested sequence and can
are zero) to a perfect fit, where a perfect fit would test hypotheses H2a and H2b. In this section, we will
equal 1. This index is adjusted for the DF of the review our analysis results and discuss the meaning of
model. (6) The Root Mean Square Residual (RMR) the results. First, we discuss our test for hypothesis H1
index represents the average size of the residual and then proceed to discuss tests for both hypotheses
correlations. (7) The Root Mean Square Error of H2a and H2b.
Approximation (RMSEA) is a measure of the pop- The first step in our analysis was to test hypothe-
ulation discrepancy that is adjusted for the DF for sis H1, the direct relationship between TPM and MP.
testing the model. A value of 0.08 or less for RMSEA We considered a model without JIT and TQM mea-
would indicate a reasonable error of approximation sures (Model 1 in Table 3) and found that TPM has
(Browne and Cudeck, 1993). a positive and significant relationship with MP. The
There is one common problem encountered in test- model showed that a 0.80 coefficient explained the
ing all model hypotheses in SEM. The LRT can be relationship between the latent TPM and MP con-
interpreted as a χ 2 variate for testing the null hypothe- structs. The fit of the model without JIT and TQM
ses of zero residual correlations; however, the χ 2 vari- was good, with LRT/DF=1.73, GFI=0.88, IFI=0.85,
ate is sensitive to sample size (Cochran, 1952; Bentler RMR=0.07, and RMSEA=0.08. 1 As a result, we
and Bonnet, 1980). For example, an insignificant χ 2 cannot reject hypothesis H1 that TPM has a positive
value does not always indicate a poor model fit and relationship with MP.
does not suggest that a model is not meaningful (Hay- Our results show that the TPM construct primarily
duk, 1996). Instead, we need to look at other model consists of six measures: three autonomous mainte-
fit statistics and also compare the difference in fit be- nance measures–housekeeping (item loading=0.48),
tween models. By comparing the difference in LRT cross-training (0.62), and teams (0.75); and three
statistics (dLRT) with the associated difference in de-
grees of freedom (dDF), we can test whether a model 1 We have considered values of the model fit criteria that are
improves the fit over another nested model (Ander- slightly below the cutoff values recommended by some authors.
son and Gerbing, 1988; Mulaik et al., 1989; McArdle Since we are primarily interested in the differences in fit between
and Prescott, 1992). In this way, we are able to sep- models, we considered this model as an acceptable starting point.
 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 49

Table 3
Results of SEM
The parameter estimates are standardized maximum likelihood estimates
Parameter estimates Models
Model 1: TPM related Model 2: TPM related Model 3: TPM related Model 4:a TPM related
to MP without JIT and to MP and unrelated to to MP and related to to MP with indirect effect
TQM in the model JIT and TQM JIT and TQM through JIT and TQM
Factor loadings
TPM→housekp 0.48 0.48 0.45 0.45
TPM→xtrain 0.62 0.62 0.63 0.64
TPM→teams 0.75 0.75 0.74 0.74
TPM→opinvolv 0.07* 0.07* 0.09* 0.09*
TPM→planning 0.79 0.79 0.78 0.78
TPM→infotrac planning 0.79 0.79 0.81 0.81
TPM→skedcomp 0.27 0.27 0.27 0.27
MP→lowcost −0.11* −0.11* −0.12* −0.03*
MP→invturn 0.46 0.46 0.46 0.43
MP→confqlty 0.45 0.45 0.44 0.44
MP→ontimedv 0.60 0.60 0.61 0.57
MP→fastdv 0.38 0.38 0.38 0.44
MP→flexible 0.12* 0.12* 0.10* 0.18*

Path coefficients
TPM→MP (R1) 0.80 0.80 0.80 0.81
TPM→JIT (R2) 0.00b 0.66 0.64
TPM→TQM (R3) 0.00b 0.87 0.89
JIT→MP (R4) 0.00b 0.00b 0.46
TQM→MP (R5) 0.00b 0.00b −0.35*

Evaluation criteria
DF 64 91 89 87
LRT 110.44 336.09 150.30 137.19
LRT/DF 1.73 3.59 1.69 1.58
GFI 0.88 0.76 0.86 0.87
IFI 0.85 0.52 0.88 0.90
RMR 0.07 0.19 0.07 0.07
RMSEA 0.08 0.15 0.08 0.07
a Indicates that the model is the best fitting model.
b Indicates a fixed parameter.
∗ Indicates a free parameter that is not significantly different from zero (t<2.0).

planned maintenance measures–information track- on different scales than the other TPM measures.
ing (0.79), disciplined planning (0.79), and schedule These objective measures also have more miss-
compliance (0.27). One of the autonomous main- ing values. It is possible that these measures have
tenance measures — operator involvement — does measurement error or do not accurately represent
not have a significant relationship with our TPM the nature or level of TPM implementation at the
construct. In addition, schedule compliance, while plant.
significant, is poorly explained by our TPM con- Even though two of the TPM measures have low
struct (squared multiple correlation of 0.07). Both item loading, they were not excluded from subsequent
of these measures are objective measures, measured analyses because item loading of all measures remain
50 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

stable whether or not they were excluded from the and Pisano (1996) and Menda and Dilts (1997) ad-
model. Furthermore, the exclusion of the low load- vocate compatibility of manufacturing dimensions
ing measures from the models led to poorer overall and the absence of trade-offs. While maintenance
model fit. Our MP construct consists of four statisti- has traditionally been seen as a means of con-
cally significant measures of performance. High MP trolling cost, our results show that TPM simulta-
is consistent with low cost (as measured by higher neously impacts components of cost, quality, and
inventory turns [item loading=0.46]), high quality delivery.
levels (as measured by higher levels of conformance Next, we tested hypotheses H2a and H2b. To test
to specifications [0.45]), and strong delivery perfor- these hypotheses, we explored a series of nested mod-
mance (as measured by a higher percentage of on-time els that included our TPM and MP constructs as well
deliveries [0.60] and by faster speeds of delivery as our JIT and TQM measures. The results of these
[0.38]). Therefore, our TPM construct influences our models are shown in Table 3. First, we tested the di-
MP construct and is associated with low cost inven- rect relationship between TPM and MP, when JIT and
tory positions, high internal quality, and responsive TQM are unrelated to TPM and MP (Model 2). For
delivery. this model, all relationships in Fig. 1, except R1, are
However, our MP construct, and therefore our TPM set equal to zero. The fit of Model 2 is significantly
construct, has no significant relationship with low cost better than a model where the relationship between
(as measured by manufacturing cost as a percentage of TPM and MP is also set to zero (dLRT=39.13 on
sales) or flexibility (as measured by the time horizon dDF=1). However, overall this model has a rela-
of the fixed production schedule). The non-significant tively poor fit (LRT/DF=3.59, GFI=0.76, IFI=0.52,
relationship with low cost may be explained by the RMR=0.19, and RMSEA=0.15). The poor fit is
following: (1) maintenance is a small portion of to- due to the fact that we consider there to be no re-
tal costs, and therefore, a change in maintenance costs lationship between JIT and TQM and any other la-
has a non-significant impact on our cost measure; (2) tent or manifest variable in the model. It is likely
manufacturing costs are calculated in different ways that JIT, TQM, TPM, and MP are related in some
depending upon the company; therefore, it is diffi- manner.
cult to compare the results between companies; or (3) Next, we considered several models where the
TPM allows for effective use of the budgeted mainte- relationships between TPM and JIT and TPM and
nance expenses and is able to improve inventory turns, TQM are tested. Model 3 permits TPM to influence
quality, and delivery while maintaining stable produc- MP as well as the level of JIT and TQM imple-
tion costs. The non-significant relationship with flex- mentation, allowing R1, R2, and R3 from Fig. 1 to
ibility could be a result of our measure of flexibility. be non-zero while R4 and R5 are set equal to zero.
It is difficult to change the planning horizon without When the fit of Model 3 is compared to the fit of
process, equipment, and planning system modifica- Model 2, there is a significant improvement in fit
tions. Another possible explanation is that the trans- (dLRT=185.79 on dDF=2). This suggests that when
formation of our flexibility measure, while helpful a plant has multiple manufacturing practices, they
in satisfying condition for normality, may not rep- cannot be considered to be independent; rather, they
resent the non-linear relationship between TPM and can be considered to be mutually supportive of each
flexibility. other. In this case, TPM has a significant and positive
Nevertheless, it is interesting to notice the influence on both JIT and TQM implementation, in-
multi-dimensionality of our MP construct. TPM dicating a reliable association of TPM with JIT and
does not impact only one dimension of MP but im- TQM and supporting our hypothesized relationships.
pacts several dimensions of performance. The idea This result is also consistent with McKone et al.
of compatibility of manufacturing dimensions had (1999) where higher levels of TPM implementation
been discussed in recent literature. Ferdows and De were associated with higher levels of JIT and TQM
Meyer (1990) proposed a “sandcone” model which implementation.
represents a sequential approach to achieving com- Model 4 enables us to test the set of hypothe-
patibility among the four dimensions. Also, Hayes ses H2a and H2b, the indirect relationship of TPM
 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 51

with MP (R1, R2, R3, R4, and R5 are all al- TPM has both direct and indirect relationships with
lowed to be non-zero). This model has a good fit MP. By comparing Model 3 to Model 4 (dLRT=13.11
(LRT/DF=1.58, GFI=0.87, IFI=0.90, RMR=0.07, on dDF=2), we see that while Model 3 provides a
and RMSEA=0.07) and also improves the fit over pre- good fit, Model 4 provides a better fit to our data.
vious models. When compared to Model 2, where only Model 4, not only considers the direct relationships
direct relationships between TPM and MP are con- between TPM and JIT, TQM and MP but also the indi-
sidered, Model 4 significantly improves the fit of the rect relationships between TPM and MP (through JIT
model (dLRT=198.90 on dDF=4). This suggests that and TQM).

Fig. 2. Best fitting model — Model 4. A value along an arrow is a standardized factor loading or path coefficient. A value above an
endogenous variable indicates the squared multiple correlation (SMC) between that variable and the variables (other than residual variables)
directly affecting it. The “*” indicates a free parameter that is not significantly different from zero (t<2.0).
52 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

Our final model, Model 4, is the best fitting model infrastructure practices were related to MP. Our re-
and is pictured in Fig. 2. It is important to notice that sults suggest that JIT practices need to be supported
the relationship between TQM and MP, R5 in Fig. 1, by TPM efforts and that together JIT and TPM can
is non-significant. This leads us to reject hypothesis improve MP.
H2b. There are three possible explanations for this re-
sult. (1) Our definition of TPM included some mea-
sures that could be included in TQM. In fact, the 7. Conclusions
variance of our TQM measure is mostly explained
(0.78 is the squared multiple correlation, shown in The results of the analyses indicate that TPM, as
Fig. 2) by our TPM construct. This suggests that TPM measured for this paper, has a strong positive impact
and TQM are interrelated. (2) TQM represents an in- on multiple dimensions of MP. While TPM directly
tegrated theory of management philosophy (Powell, impacts MP, there is also a strong indirect relationship
1995) rather than a clearly defined technology or set of between TPM and MP through JIT. Our results are
techniques. It is feasible that TQM could invoke a goal important for two reasons. (1) Maintenance programs
of improving quality without dictating a well-defined have long been used as a means to control manufac-
routine for accomplishing it (Westphal et al., 1997). turing costs. Our results show that TPM does more
Campbell (1994, p. 7) mentions that when TQM ac- than control costs, it can improve dimensions of cost,
quires institutional status, quality practices may be quality, and delivery. TPM can be a strong contrib-
evaluated by a “logic of social appropriateness” rather utor to the strength of the organization and has the
than a “logic of instrumentality”. (3) Loose coupling ability to improve MP. (2) World Class Manufactur-
may occur between TQM practices designed for cus- ing programs, such as JIT, TQM, and TPM, should
tomer demands and the activities on the plant floor not be evaluated in isolation. They are closely related
designed for plant performance. TQM is found to and in combination can help foster better MP. Future
have a stronger impact on customer satisfaction than research should further consider the relationships be-
plant performance (Choi and Eboch, 1998). Perhaps tween these practices and their combined impact on
TQM, as measured in this study, considers the so- performance.
cially accepted aspects of the program rather than the We plan to continue our research in this area to fur-
instrumental aspects of the program that would di- ther explain the relationship between manufacturing
rectly improve MP. This helps explain why TQM, as practices and MP. In particular, we plan to identify the
measured in this paper, does not contribute to MP common infrastructural and unique practices of TQM,
and why TPM does provide such a large explana- JIT, and TPM, and test their interrelationships and im-
tion of MP. Our TPM construct has a clearly defined pact on MP. Also, we would like to investigate the
set of methods for improving performance while our nature of the relationships in different contextual situ-
TQM measure considers only general management ations (for example, cross-country and cross-industry
approaches. differences), combining the work from McKone et al.
While TQM does not provide a significant expla- (1999) and this paper. In addition, we would like to
nation of the positive relationship between TPM and consider the life cycle of the practices and evaluate
MP, the relationship between TPM and MP through the impact of the development time on MP. Hopefully,
JIT is significant and positive (R2 and R4 in Fig. 1). this type of research will support and encourage suc-
Therefore, we cannot reject hypothesis H2a. Our re- cessful implementation of TQM, JIT, and TPM.
sults support our theory that TPM helps improve the Based on this research, the authors recommend that
equipment performance which in turn supports JIT’s practitioners pay closer attention to their maintenance
efforts to reduce inventory, shorten lead-times, and management practices and their impact not only on
eliminate other wastes. The impact of TPM should not costs but also on quality and delivery performance.
be considered in isolation but must be considered with Our future research will provide additional details
respect to the other manufacturing practices. These re- about specific practices that lead to improved perfor-
sults augment those of Sakakibara et al. (1997), who mance in various environmental and organizational
showed that a combination of JIT management and situations.
 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 53

Appendix A. Measurement of TPM implementation

Concept Factor Measure

Autonomous House-keeping, Our plant emphasizes putting all tools and fixtures in their place.
maintenance α=0.85632 We take pride in keeping our plant neat and clean.
Our plant is kept clean at all times.
I often have trouble finding the tools I need.3
Our plant is disorganized and dirty.3
Cross-training, Employees receive training to perform multiple tasks.
α=0.80052 Employees at this plant learn how to perform a variety of
tasks/jobs.
The longer an employee has been at this plant, the more tasks or
jobs they learn to perform.
Employees are cross-trained at this plant so that they can fill in
for others if necessary.
At this plant, employees only learn how to do one job/task.3
Teams, α=0.88122 During problem solving sessions, we make an effort to get all team
members’ opinions and ideas before making a decision.
Our plant forms teams to solve problems.
In the past 3 years, many problems have been solved through small
group sessions.
Problem solving teams have helped improve manufacturing pro-
cesses at this plant.
Employee teams are encouraged to try to solve their problems as
much as possible.
Operator involvement What percent of the maintenance on the machines involved in the
production of this product is performed by the workers, rather than
by a separate maintenance crew? 4
Planned Disciplined planning, We dedicate a portion of every day solely to maintenance.
Maintenance α=0.76662 We emphasize good maintenance as a strategy for achieving qual-
ity and schedule compliance.
We have a separate shift, or part of a shift, reserved each day for
maintenance activities.
Our maintenance department focuses on assisting machine opera-
tors perform their own preventive maintenance.
Information tracking, Charts plotting the frequency of machine breakdowns are posted
α=0.81322 on the shop floor.
Information on productivity is readily available to employees.
A large percent of the equipment or processes on the shop floor
are currently under statistical quality control.
We use charts to determine whether our manufacturing processes
are in control.
We monitor our processes using statistical process control.

2α refers to Cronbach’s alpha, used to measure the reliability of the scale.

3 Indicates that the variable is reversed scored.
4 Response is in terms of percentage. All other responses are in the scale score format with 1 being strongly disagree and 5 being

strongly agree.
54 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

Schedule compliance What percent of the time is the maintenance schedule (for equip-
ment used to produce this product) followed?4

Appendix B. Measurement of MP

Concept Meaning Measure

Low cost (LOWCOST)5 Manufacturing cost of Manufacturing costs
goods sold as a percentage
of sales.
Sales value of production
Inventory turnover (INVTURN) Manufacturing cost of Manufacturing costs
goods sold as a percentage
of average inventory.
Value of average annual finished
goods inventory
Value of average annual work-in-
process inventory
Value of average annual raw materials
inventory
Quality (CONFQLTY)5 Conformance to specifications. What is the percentage of internal scrap
and rework?
On-time delivery (ONTIMEDV) Ability to deliver as promised. What percentage of the orders are
shipped on time?
Fast delivery (FASTDV)5 Ability to deliver quickly. What is the average lead-time from the
receipt of an order until it is shipped
(in days)?
Flexibility (FLEXIBLE)5 Flexibility to change mas- What is the time horizon for the fixed
ter production schedule. production schedule?
(1) 1 day, (2) 1 week, (3) 1 month, (4)
3 months or more

Appendix C. Measurement of JIT and TQM Implementation

Concept Factor Measure

JIT, α=0.90456 JIT delivery by suppliers Our suppliers deliver to us on a JIT basis.
We receive daily shipments from most suppliers.
Our suppliers are certified, or qualified, for quality.
We have long-term arrangements with our suppliers.
Our suppliers deliver to us on short notice.
We can depend upon on-time delivery from
our suppliers.
Our suppliers are linked with us by a pull system.
JIT link with customers Our customers receive JIT deliveries from us.

5 Indicates that the variable is adjusted so that a high value reflects good performance.
6α refers to Cronbach’s alpha, used to measure the reliability of the scale.
 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 55

Most of our customers receive frequent shipments from

us.
We are expected to supply on short notice to our cus-
tomers.
We always deliver on time to our customers.
We can adapt our production schedule to sudden pro-
duction stoppages by our customers.
Our customers have a pull type link with us.
Pull system support We use a back-flushing system, where components are
subtracted from inventory every time a product is made.
We have laid out the shop floor so that process and
machines are in close proximity to each other.
Direct Labor is authorized to stop production for
quality problems.
We use a pull system for production control.
The control of production is in the hands of
the workers.
Generally, workers on the production floor have the au-
thority to decide how to handle production problems.
We have low work-in-process inventory on the
shop floor.
When we have a problem on the production floor, we
can identify its location easily.
Repetitive nature Our master schedule repeats the same mix of products
of master schedule from hour to hour and day to day.
The master schedule is level-loaded in our plant from
day to day.
We make every model every day.
A fixed sequence of items is repeated throughout our
master schedule.
We are able to use a mixed model schedule because our
lot sizes are small.
Within our schedule, the mix of items is designed to be
similar to the forecasted demand mix.
Setup reduction We are aggressively working to lower setup times in our
plant.
We have converted most of the setup time to external
time while the machine is running.
We have low setup times of equipment in our plant.
Our crews practice setups to reduce the time required.
Our workers are trained to reduce set-up time.
Management emphasizes importance of set-up time re-
duction.
TQM α=0.89346 Customer involvement We frequently are in close contact with our customers.
Our customers seldom visit our plant.7

————
7 Indicates that the variable is reversed scored.
56 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

Our customers give us feedback on quality and

delivery performance.
Our customers are actively involved in the product design
process.
We strive to be highly responsive to our customers’ needs.
We regularly survey our customers’ requirements.
Rewards for quality Workers are rewarded for quality improvement.
Supervisors are rewarded for quality improvement.
If I improve quality, management will reward me.
We pay a group incentive for quality improvement ideas.
Our plant has an annual bonus system based on
plant productivity.
Non-financial incentives, such as jackets, coffee cups, etc.,
are used to reward quality improvement.
Supplier quality We strive to establish long-term relationships with suppliers.
management Our suppliers are actively involved in our new product de-
velopment process.
Quality is our number one criterion in selecting suppliers.
We rely on a small number of high quality suppliers.
We use mostly suppliers which we have certified.
We maintain close communication with suppliers about
quality considerations and design changes.
Top management All major department heads within our plant accept their
leadership for quality responsibility for quality.
Plant management provides personal leadership for quality
products and quality improvement.
The top priority in evaluating plant management is quality
performance.
All major department heads within our plant work towards
encouraging JIT production.
Our top management strongly encourages employee involve-
ment in the production process.
Plant management creates and communicates a vision fo-
cused on quality improvements.
Plant management is personally involved in quality improve-
ment projects.

Bagozzi, R.P., 1980. Causal Models in Marketing. Wiley, New

References York, NY.
Barley, S.R., 1990. The alignment of technology and structure
Adler, P.S., Shenhar, A., 1990. Adapting your technological through roles and networks. Administrative Science Quarterly
base: the organizational challenge. Sloan Management Review, 35, 61–103.
25–36, Fall Bollen, K.A., 1989. A new incremental fit index for general
Anderson, J.C., Gerbing, D.W., 1988. Structural equation modeling structural equation models. Sociological Methods and Research
in practice: a review and recommended two-step approach. 17, 303–316.
Psychological Bulletin 103, 411–423. Browne, M.W., Cudeck, R., 1993. Alternative ways of assessing
Arbuckle, J.L., Amos Users’ Guide Version 3.6, Small Waters, model fit, In: Bollen, K., Long, J.S. (Eds.), Testing Structural
Chicago, IL Equation Models. Sage, Newbury Park, CA, pp. 136–162.
 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58 57

Burnkrant, R.E., Page, T.J., 1982. An examination of the P., Dromgoole, T., Peppard, J. (Eds.), Operations Management:
convergent, discriminant and predictive validity of Fishbein’s Future Issues and Competitive Responses, Papers from the
behavioral intention model. Journal of Marketing Research 19, 5th International Conferences of the European Operations
550–561. Management Association, University of Dublin, Trinity College,
Campbell, J.L., 1994. Recent trends in institutional analysis: Dublin 2, Ireland. 304–309.
bringing culture back into political economy. Working Paper, Mardia, K.V., 1970. Measures of multivariate skewness and
Department of Sociology, Harvard University. kurtosis with applications. Biometrika 57, 519–530.
Choi, T.Y., Eboch, K., 1998. The TQM paradox: relations among Mardia, K.V., 1974. Applications of some measures of multivariate
TQM practices, plant performance, and customer satisfaction. skewness and kurtosis in testing normality and robustness
Journal of Operations Management 17, 59–75. studies. Sankhya, Ser. B 36, 115–128.
Cleveland, G., Schroeder, R.G., Anderson, J.C., 1989. A theory Mardia, K.V., 1985. Mardia’s test of multinormality. In: Kotz, S.,
of production competence. Decision Sciences 20, 643–655. Johnson, N.L., (Eds.), Encyclopedia of Statistical Sciences Vol.
Cochran, W.G., 1952. The χ 2 test of goodness of fit. Annals of 5 Wiley, New York, pp. 217–221.
Mathematical Statistics 23, 315–345. Marsh, H.W., Hocevar, D., 1985. Application of confirmatory factor
Ferdows, K., DeMeyer, A., 1990. Lasting improvements in analysis to the study of self-concept: first and higher-order
manufacturing performance: in search of a new theory. Journal factor models and their invariance across groups. Psychological
of Operations Management 9 (2), 168–184. Bulletin 97, 562–582.
Flynn, B.B., Schroeder, R.G., Sakakibara, S., 1994. A McArdle, J.J., Prescott, C.A., 1992. Age-based construct validation
framework for quality management research and an associated using structural equation modeling. Experimental Aging
measurement instrument. Journal of Operations Management Research 18 (3), 87–115.
11 (4), 339–366. McKone, K.E., Schroeder, R.G., Cua, K.O., 1999. Total productive
Flynn, B.B., Schroeder, R.G., Sakakibara, S., 1995. The impact of maintenance: a contextual view. Journal of Operations
quality management practices on performance and competitive Management 17 (2), 123–144.
advantage. Decision Sciences 26 (5), 659–691. Menda, R., Dilts, D., 1997. The manufacturing strategy
Flynn, B.B., Schroeder, R.G., Sakakibara, S., 1996. The
formulation process: linking multifunctional viewpoints. Journal
relationship between quality management practices and
of Operations Management 15 (4), 223–241.
performance: synthesis of finding from the World Class
Miller, J., Roth, A., 1994. A taxonomy of manufacturing strategies.
Manufacturing Project. In: Fedor, D., Gosh, S., (Eds.), Advances
Management Science 40 (3), 285–304.
in the Management of Organizational Quality Vol. 1 JAI Press,
Miller, W.B., Schenk, V.L., 1997. All I Need to Know About
Greenwich, CT, pp. 141–186.
Manufacturing I Learned at Joe’s Garage: World Class
Galbraith, J.R., 1977. Organizational Design. Addison-Wesley,
Manufacturing Made Simple. Bayrock Press, USA.
Reading, MA.
Mulaik, S.A., James, L.R., Van Alstine, J., Bennett, N., Lind,
Garwood, W.R., 1990. World class or second class. Vital Speeches
S., Stilwell, C.D., 1989. Evaluation of goodness-of-fit indices
of the Day October 1.
for structural equation models. Psychological Bulletin 105 (3),
Gerwin, D., 1976. A systems framework for organizational
430–445.
structural design. In: Kilman, R.H., Pondy, L.R., Slevin,
Nakajima, S., 1988. Introduction to TPM. Productivity Press,
D.P. (Eds.), The Management of Organization Design. North
Cambridge, MA.
Holland, New York, pp. 89–102.
Hayduk, L.A., Lisrel: Issues, Debates and Strategies. Johns Nunnally, J., 1978. Psychometric Theory. 2nd edn. McGraw-Hill,
Hopkins, Baltimore. New York.
Hayes, R., Pisano, G., 1996. Manufacturing strategy: at the Powell, T.C., 1995. Total quality management as competitive
intersection of two paradigm shifts. Production and Operations advantage: a review and empirical study. Strategic Management
Management 5 (1), 25–41. Journal 16, 15–37.
Hayes, R., Wheelwright, S.C., Clark, K.B., 1988. Dynamic Sakakibara, S., Flynn, B.B., Schroeder, R.G., 1993. A framework
Manufacturing. Free Press, New York. and measurement instrument for just-in-time manufacturing.
Jarrell, S., Easton, G., 1997. An exploratory empirical investigation Production and Operations Management 2 (3), 177–194.
of the effects of total quality management on corporate Sakakibara, S., Flynn, B.B., Schroeder, R.G., Morris, W.T., 1997.
performance. In: Lederer, P., Karmarkar, W. (Eds.), The Practice The impact of just-in-time manufacturing and its infrastructure
of Quality Management. Kluwer Academic Publishers, Norwell, on manufacturing performance. Management Science 43 (9),
MA, Chap. 2. 1246–1257.
Koelsch, J.R., 1993. A dose of TPM: downtime needn’t be a bitter Schonberger, R.J., 1986. World Class Manufacturing: The Lessons
pill. Manufacturing Engineering, 63–66, April. of Simplicity Applied. The Free Press, New York.
MACI, 1995. Four phase approach to planned maintenance. In: Schonberger, R.J., 1990. Building a Chain of Customers. The Free
Proceedings of the Sixth Annual Total Productive Maintenance Press, New York.
Conference and Exposition. Productivity Inc., Norwalk, CT, Schroeder, R.G., 1993. Operations Management, Decision Making
pp. 204–243. in the Operations Function. 4th edn. McGraw-Hill, New York.
Maier, F.H., Milling, P.M., Hasenpusch, J., 1998. Implementation Shirose, K., 1992. TPM for Workshop Leaders, Productivity Press,
and outcomes of total productive maintenance. In: Coughlan, Cambridge, MA.
58 K.E. McKone et al. / Journal of Operations Management 19 (2001) 39–58

Skinner, W., 1969. Manufacturing — missing link in corporate Tyer, M.J., Hauptman, O., 1992. Effectiveness of organizational
strategy. Harvard Business Review 47 (3), 136–145. responses to technological change in the production process.
Steinbacher, H.R., Steinbacher, N.L., 1993. TPM for America: Organizational Science 3 (3), 301–320.
What It Is and Why You Need It. Productivity Press, Cambridge, Van de Ven, A.H., Ferry, D.L., 1980. Measuring and Assessing
MA. Organizations. Wiley, New York.
Suzuki, T., 1993. New Directions for TPM. Productivity Press, Ward, P.T., Duray, R., Leong, G.K., Sum, C.C., 1995. Business
Cambridge, MA. environment, operations strategy and performance: an empirical
Tajiri, M., Gotoh, F., 1992. TPM Implementation: A Japanese study of Singapore manufacturers. Journal of Operations
Approach. McGraw-Hill, New York. Management 13, 99–115.
Teresko, J., 1992. Time Bomb or Profit Center. Industry Week, Wheaton, B., Muthen, B., Alwin, D.F., Summers, G.F., 1977.
52–57, March 2. Assessing reliability and stability in panel models. In: Heise,
Tsuchiya, S., 1992. Quality Maintenance: Zero Defects Through D.R., (Ed.), Sociological Methodology. Jossey-Bass, San
Equipment Management. Productivity Press, Cambridge, MA. Francisco, pp. 84–136.
Tunälv, C., 1992. Manufacturing strategy — plans and Westphal, J.D., Gulati, R., Shortell, S.M., 1997. Customiz-
business performance. International Journal of Operations and ation or conformity? An institutional and network pers-
Production Management 12 (3), 4–24. pective on the content and consequences of TQM
Tyer, M.J., 1991. Managing innovation on the factory floor. adoption. Administrative Science Quarterly 42,
Technology Review, 59–65, October. 366–394.