Description
The adoption of new manufacturing practices such as just-in-time (JIT) and total quality management
(TQC) is only a first step to improving manufacturing performance. Even more critical is the fit between
manufacturing practices and organizational design, structure and processes. Using archival and survey
data, this paper reports the results of a field study within a Fortune
Pergamon Accounti ng, Organfzati ons and Soci ety, Vol. 20, No. 7/ S, pp. 665-684, 1995
Copyright 0 1995 Elsevier Science Ltd
Printed in Great Britain. All rights reserved.
0361-3682/95 $9.50+0.00
ASSESSING THE ORGANIZATIONAL FIT OF A JUST-IN-TIME MANUFACTURING
SYSTEM: TESTING SELECTION, INTERACTION AND SYSTEMS MODELS OF
CONTINGENCY THEORY*
FRANK H. SELTO
University of Colorado at Boulder
CELIA J. RENNER
University of Northern I owa
and
S. MARK YOUNG
University of Southern California
Abstract
The adoption of new manufacturing practices such as just-in-time (JIT) and total quality management
(TQC) is only a first step to improving manufacturing performance. Even more critical is the fit between
manufacturing practices and organizational design, structure and processes. Using archival and survey
data, this paper reports the results of a field study within a Fortune 500 company that tests three
operationalizations of contingency theory as discussed by Van de Ven and Drazin (1985) [The Concept of
Fit in Contingency Theory, Research i n Organi zutfonal Behavi or, pp. 333-3651. Results show that the
misfit between worker empowerment required by JIT/ TQC practices and existing authoritarian manage-
ment partially explain relative Workgroup performance as do other conflicts within workgroups and
between operators and supervisors.
Successfully adopting Japanese manufacturing how management controls should be designed
practices such as the just-in-time (JIT) manufac- to be consistent with organizational structure
turing and total quality control (TQC) systems and context (Hopwood, 1976; Otley, 1980; Gor-
has been problematic for many organizations don & Narayanan, 1984; Govindarajan, 1984,
(Wilkinson & Oliver, 1989; Young & Selto, 1988; Govindarajan & Gupta, 1985; Chenhall
1991; Young, 1992). In particular, there has & Morris, 1986; Abernethy & Stoelwinder,
been a lack of attention to needed changes in 1991) little research has directly addressed
organizational structure, context, and control. the issues of the best “fit” of organizational
While considerable literature has discussed structure and controls for JIT/TQC systems.’
l We thank Shannon Anderson, Dale Jasinski, Chris Koberg, participants at the First International Conference on Account-
ing, Taipei, Taiwan (January 1994) the Fourth Biennial Management Accounting Conference, Sydney, Australia (September
1994) and the Sixth Tokyo Keizai University International Symposium, Tokyo, Japan (November 1994), as well as the
anonymous reviewers for their helpful comments on this paper.
i Hereafter, we refer to the joint JIT/ TQC system since most observers agree that it is not possible to implement JIT
without TQC.
665
This paper addresses the issue offit among can be produced, inventories build up in push
the central organizational elements in a JIT/ systems. These inventories may be used to
TQC manufacturing firm by studying the rela- shield the push process from breakdowns and
tions among organization structure, JIT/ TQC quality failures in previous process stages
manufacturing methods, and management con- (referred to as a “just-in-case” system - Hayes
trols at the Workgroup level within a Fortune et al . (1988). JIT/ TQC means that operators do
500 fi rm. Fit also is tied directly to the perfor- not have or cannot build buffer inventories
mance of workgroups. Data are tested against (though some JIT/ TQC facilities allow minimal
three contingency theory models as summar- inventories at each stage or at a central location
ized by Van de Ven and Drazin (1985). for severe problems (Schonberger, 1987). In
The paper proceeds as follows. The next order to maintain production flow, JIT/ TQC
section presents a theoretical framework in operators should continually improve the pro-
which several key dimensions of the JIT/ TQC
cess to prevent stoppages. If breakdowns or
system are discussed within a review of the
quality failures do occur, JIT/ TQC operators
contingency literature in management account-
must communicate and interact to solve what-
ing. Also discussed are three operationaliza-
ever problems arise to keep production flowing.
tions of contingency fit within the context of
These new job responsibilities require consider-
JIT/ TQC manufacturing. The third section dis-
able process knowledge, flexibility, communi-
cusses the research site and describes measures
cation, control, and authority at the worker or
of various dependent and independent vari-
operator level - characteristics that are not
ables. Statistical results are presented, and the
typical in traditional, push manufacturing.
three contingency theory approaches are com-
JIT/ TQC requires that operators are able to
pared. The final section concludes the paper.
detect and correct process problems and
defects when they occur and to seek for ways
to prevent process variation and product
THEORETICAL FRAMEWORK
defects, without waiting for central or end-of-
process quality assurance to issue change
J I T/TQC manufacturi ng
orders. Therefore, JIT/ TQC also requires opera-
In theory, JIT/ TQC makes significantly differ-
tors to have the flexibility, knowledge, and
ent demands on operators than traditional man-
authority usually reserved for managers in tradi-
ufacturing methods. JIT/ TQC is often
tional systems. In other words, operators need
characterized as apull approach to manufactur-
to be able to manage themselves. Managers are
ing, as opposed to a more traditional push sys-
more like “coaches” and may even work
tem (Schonberger, 1982, 1986). JIT/ TQC
directly on products alongside operators. It
operators produce a part or product only in
seems likely, then, that the relative roles of
response to an order from an internal or exter-
operators and low-level managers can be criti-
nal customer. JIT/ TQC tasks are highly interde-
cal to the successful functioning of JIT/ TQC.
pendent; each succeeding task depends on the Undoubtedly there is considerable variation
quality and timeliness of the preceding task. in JIT/ TQC implementations among manufac-
Thus, no excess parts or products are pro- turing firms (for example, Schonberger,
duced, unlike traditional systems where opera- 1987). These variations may reflect manage-
tors push parts and products through the ment ability, external and internal constraints,
process as quickly as possible. If parts are not and, perhaps lack of understanding of the
used and products are not sold as quickly as they implications of JIT/ TQC.’ The objective of the
666 F. H. SELTO et al.
’ Variations in JIT/TQC implementation themselves may explain differences in performance across manufacturing firms,
which is worthy of future study.
CONTINGENCY THEORY
667
present study is to investigate whether, given a
particular JIT/ TQC implementation, variation
in fit among other elements of an organization
that are under managers’ control explain cross-
sectional performance within a firm. Implica-
tions of JIT and TQC allow us to hypothesize
what these elements and effects might be.
Conti ngency theory
The central theme of contingency theory
(applied here to JIT/ TQC settings) is that alI
components of an organization must “fit” well
with each other or the organization will not
perform optimally (Perrow, 1967). Within a
single organization, contingency theory also
predicts that variations in subunits’ perfor-
mance (for example, at the Workgroup level)
are due to variations in fit of their various com-
ponents (March & Mannari, 1981; Tushman,
1979). An extensive contingency theory
literature has defined essential organizational
elements and has explored their interrelation-
ships and effects on performance. All of these
elements of contingency theory are compatible
with JIT/ TQC concerns, as will be shown.
We investigate the role of JIT/ rQC fit in a
contingency theory framework because: (1)
no other theory of which we are aware
directly concerns fit; (2) despite criticism, the
intuition behind the theory continues to be
appealing; and (3) several recent operationali-
zations of fit have overcome some previously
voiced criticisms (see Otley, 1980; Schoonha-
ven, 1982). To our knowledge, no study has
extended the contingency literature descrip-
tions or predictions to organizations that
employ manufacturing methods such as JIT
and TQC. These manufacturing practices have
been adopted widely throughout the world in
the last decade, yet many observers note that
expected benefits of these methods have rarely
materialized (for example, Jacobs, 1993; Bleak-
ley, 1993; Young, 1992). The hypothesis we
investigate is that poor fit among critical orga-
nizational elements explains lack of success of
JIT and TQC methods.
Drazin and Van de Ven’s (1985) review of
the literature suggests that contingency theory
has considerable power to explain Workgroup
performance within organizations. They cate-
gorize the primary components of an organiza-
tion into organizational context, organizational
structure, and organizational process or con-
troZ.3 Context includes factors such as size,
culture, environment, technology, tasks, and
methods such as JIT/ TQC. Structure refers to
an organization’s formality (mechani sti c) or
flexibility (organi c), incentives, degrees of
authority, and bureaucracy. Control includes
communications and information flow, includ-
ing explicit management controls. Variable
definitions used in this study are in Table 1.
Tbeoly of fit. Theoretically, each organiza-
tion has its own optimal configuration or best
jit of context, structure, and control. Deviation
from that ideal fit (which is mi sfi t) should
cause lack of coordination, miscommunica-
tion, misunderstanding, poor morale, and
poor motivation, which, in turn, should lead
to poor performance. Concerns for the misfit
induced by outmoded, irrelevant control infor-
mation in today’s manufacturing firms are con-
sistent with contingency theory arguments. As
discussed earlier, JIT/ I’QC manufacturing sys-
tems (theoretically) are characterized by high
levels of worker autonomy and decision mak-
ing, worker flexibility, and quick response to
variable product and process demands (IIayes
et al ., 1988). Furthermore, this environment
may depend more on horizontal communica-
tion among operators than on vertical commu-
nication from managers. Thus, though one
means of achieving JIT is to standardize ele-
ments of work, JIT/ TQC should be more suc-
cessful in organic, rather than mechanistic
organizations. These profiles are shown in
Table 2.
In addition, an organization’s context may be
’ Because we use the termprocess to refer to the manufacturing steps or procedures, hereafter we will refer to organ&a-
tional process as control . The term control is also more consistent with the management control literature.
668 F. H. SELTO et al
TABLE 1. Definitions of contineencv constructs*
Sfundurd~zulion - Extent to which standard operating procedures and performance expectations are formalized and
followed and the degree to which the roles and task assignments that make up a job are listed in the job description
and other procedures.
WorkerAuthorfty - Includes personal decision making by employee as well as decision making by members of the unit as
a group. Also includes amount of discretion that the job incumbent has in making decisions about what tasks and
assignments make up the job, how the job is to be done, and how problems are handled.
Vertical Communication - The amount of feedback the job incumbent receives from his supervisor about performance,
work-related messages sent by writing, i.e. memos, reports and letters, and amount of explanation of changes in group
operating procedures.
Hortzontul Communication - Frequency of group or staff meetings among three or more personnel in the Workgroup
and difficulty in getting the Workgroup together to accomplish work.
Workgroup ConfZict - Frequency of disagreements or disputes among members of the group and extent to which conflict
hinders others’ work.
Tusk Dffj?culty/Variubilfty - Includes analyzability and predictability of work done by the unit and number of exceptions
encountered in the work.
J ob Dependency on Supervisor - Refers to how much each person’s job depends on activities performed by the
supervisor and the hierarchical authority exercised by the supervisor in terms of decisions about what tasks are to be
performed, what performance evaluation criteria are, and operating rules established to coordinate and control unit
activities.
J ob Dependency on Workgroup - Relates to intensity of work-interdependence among unit personnel. Refers to how
much each person’s job depends on activities performed by other unit members. Also includes how specialized the
jobs are within the unit and degree to which unit members can do each other’s work.
l Adapted from Van de Ven and Ferry (1980)
TABLE 2. Contingencies and design patterns: a summary of
fit and misfit*
Mechanistic
Design dimension pattern
sTRucTuRE:
Standardization High
Worker authority Low
MANAGEMENT CONTROL PROCESSES:
Vertical communication High
Horizontal communication Low
Workgroup conflict Low
CONTEXT:
Task difficulty and variability
Low Fit
High Misfit
Job Dependency on Supervisor
(Vertical Dependence)
Low Misfit
High Fit
Job Dependency on Work Group
(Horizontal Dependence)
Low Fit
High Misfit
Organic
pattern
Low
High
Low
High
Low
Misfit
Fit
Fit
Misfit
Misfit
Fit
l Adapted from Gresov (1989)
internally inconsistent, and optimal perfor-
mance may not be possible with any pattern
of structure and control (Gresov, 1989). Speci-
fically as found in this study, pervasive depen-
dence on managers may conflict directly with a
JIT/TQC process that requires empowered
operators.
Empi ri cal test resul ts. Empirical tests of con-
tingency theory have produced mixed results,
with as many studies finding significant correla-
tions between measures of fit and performance
as not. These mixed results have caused many
researchers to criticize past operationalizations
of fit itself (for example, Schoonhaven, 1982;
Van de Ven & Drazin, 1985) as ad hoc. Further,
a few studies have used objective measures of
firm performance when assessing the relation
of fit to performance. Most studies have ana-
lyzed self-assured performance as a function
of fit among a limited (but differing) set of
organizational variables (for example, Gordon
& Narayanan, 1984; Ito & Peterson, 1986;
Chenhall & Morris, 1986; Macintosh & Daft,
1987; Abernethy & Stoelwinder, 1991). Govin-
darajan (1988) considered multiple measures of
fit among strategy, organizational structure,
CONTINGENCY THEORY 669
managers’ characteristics, and control systems,
but again using self-assessed performance.
To summarize, contingency theory has pros
and cons. Advantages include a rich descriptive
framework, plentiful opportunities for mea-
surement and observation, explicit linking of
organizational characteristics and perfor-
mance, and an extensive, supportive theoreti-
cal literature. Disadvantages, though, are
significant and include lack of standard mea-
sures, ambiguity in the operationalization of
the key construct, fit, and equivocal past
empirical results. Any contingency study,
then, must exploit a rich framework and miti-
gate threats to validity. Unfortunately, past con-
tingency studies in the aggregate suffer from
lack of standardization, which impedes cross-
study comparisons. There is an opportunity,
therefore, to add to the contingency literature
by building on past developments and by con-
sidering objective measures of performance.
Al ternati ve approaches to conti ngency
theory j i t
A valid interpretation of contingency theory
must operationalize the concept of fit (or mis-
fit). To enhance the validity of work on fit and
performance we turned to the work of Van de
Ven and Drazin (1985) and Van de Ven and
Ferry (1980). These studies have built a com-
mon framework for studying contingency fit
and have shown that at least one operationali-
zation of fit can explain objectively measured
Workgroup performance, Van de Ven and Dra-
zin (1985) critique three current approaches to
contingency theory based on alternative
notions of fit: (1) selection; (2) interaction;
and (3) systems. They point out that these
three approaches are not mutually exclusive,
and may provide complementary information
about an organization. Since each approach
addresses different aspects of fit and since the
most appropriate definition and measure of fit
is still unclear, this study uses all three
approaches.
Sel ecti on. In the selection approach, organi-
zational context drives organizational design.
Fit is defined in terms of predictable correla-
tions between pairs of organizational vari-
ables. Natural sel ecti on predicts that all
structure and control variables are correlated
with context since anything less would lead
to exti ncti on in a competitive environment.
Manageri al sel ecti on predicts correlations
between context and only those control and
structural characteristics managed by the orga-
nization. In general, we hypothesize that JIT/
TQC context variables are correlated with
Workgroup structure and control. For exam-
ple, worker authori ty should be positively cor-
related with hori zontal communi cati on but
negatively correlated with verti cal communi -
cati on. Likewise, task d@hdty and vari abi l i ty
should be positively correlated with worker
authori ty. In all, 15 such relationships are pre-
dicted by the selection approach in this study.
Note that explaining performance is not an aim
of the selection approach since it is assumed
that only good performers survive to be
observed.
I nteracti on. The interaction form of contin-
gency theory hypothesizes jit as the i nterac-
ti on of pairs of context-structure or context-
control factors on performance. The basic con-
cept is that none of context, structure, or con-
trol alone should affect performance; it is the fit
among them that affects performance. In a
regression explaining performance, therefore,
an interaction term of context and Workgroup
structure (and control) should be significant
while main effects should not. Statistically,
main effects may be significant, of course, but
such results detract from the theory. For exam-
ple, in a JIT/ TQC organization task di j @ul ty
and vari abi l i ty should interact with worker
authori ty to have a positive effect on perfor-
mance, but their individual contributions
should be insufficient to beneficially affect per-
formance. Likewise, standardi zati on should
interact with hori zontal communi cati on to
have a negative impact on performance, but
without main effects. As shown in Table 3,
there are 30 such interactions possible in this
study (3 context X 5 control and structure X 2
performance variables). Though it would be
possible to reduce context, control, and struc-
670 F. H. SELTO et al.
TABLE 3. Alternative contingencv tests
Outcomes tested
Basic tests
Number of control or
structural variables in each
test
Number of tests to cover all
structure, context, process,
and outcome variables
Selection
No
Correlations
interaction
Yes
Regressions
Systems
Yes
Regressions
Pair Pair
Eight, reduced to Euclidean
distance
15 Correlations 30 Regressions
3 x 5” 3X5X2b 2 Euclidean distance tests=
’ This test included 3 context variables and 5 control and structure variables.
’ This test included 3 context variables, 5 control and structure variables, and the 2 outcome measures,
’ Each test included 3 context, 2 structure, and 3 control variables with one performance measure. In addition, the 2
covariates were included in the Workgroup effectiveness test.
ture to composite variables and reduce the
number of reported tests, this aggregate
approach would reduce the descriptive power
of this approach considerably.
Systems. The systems approach, the most
recent and least-tested form of contingency the-
ory, is a hol i sti c approach to studying the inter-
dependencies in organizations. In concept,
optimal systems fit occurs when all design ele-
ments of structure, control, and context are
congruent. Variations in performance result
from variations in this overall systemic fit.
The further a Workgroup’s design is from opti-
mal, then the lower its performance should be.
Identifying the optimal organization or work-
group is the primary drawback to this
approach, since it is possible there are many
equally effective, feasible sets of organizational
design elements (equifinality). Defining optimal
fit is problematic, as well, and is usually defined
in a somewhat circular manner: optimal fit is
the configuration of the optimally performing
organization.
Van de Ven and associates operationalized
deviations from optimal systems fit as the dif-
ference from the set of characteristics (along
each critical dimension) of the top performing
Workgroup(s) in an organization. Drazin and
Van de Ven (1985) found that of the alternative
definitions of fit, only systems fit explained per-
formance in their large sample of state employ-
ment security workgroups Van de Ven and
Drazin (1985) recommend the summary mea-
sure of Euclidean distance (ED) as the most
appropriate operationalization of systems fit.
Following their lead, we selected the work-
group that performed the highest on each out-
come measure. That Workgroup became the
benchmark for each test of cross-sectional out-
comes.5
Misfit, then, is measured in the jth work-
group’s Euclidean distance, (ED,, from the
* Govindarajan (1988) also found that systems fit measures explained self-assessed performance by business unit managers
in a number of large companies. Gresov (1989) took the systems approach one step further by investigating multiple
contingencies (i.e. multiple contexts). In addition to the context dimension of task uncertainty, he tested for the effects of
horizontal dependence (i.e. dependencies within workgroups) and the interaction of these two context variables. He
proposed that when the Workgroup faces conflicting contexts, the result may be unavoidable design misfit and lower
performance. He tested his approach on the data from Drazin and Van de Ven (1985) and did find support for his theory
that Workgroup design is affected by multiple contingencies and that conflicting contexts are associated with suboptimal
Workgroup performance.
5 From here on, we drop the terms i deal or opti mal and use the term benchmark since the notion of ideal is so arbitrary.
One might construct an hypothetically ideal Workgroup that has the highest independent variable scores. There is no
indication that such a Workgroup would be feasible since tradeoffs among the independent variables may be required in a
particular workplace.
CONTINGENCY THEORY 671
benchmark Workgroup’s organizational profile
along each dimension (structure, process, con-
text),’ Xi*:
Euclidean Distance: EDi =(& (Xi* - Xc)2)1’2.
RESEARCH SITE
Our test site was a major manufacturing divi-
sion of a Fortune 500 company. Before collect-
ing any systematic data, three months of
negotiations were required to establish both
entry to the division and the scope of the
study. Once there was a general agreement,
we spent parts of approximately 20 days over
six months inside the firm gathering data, inter-
viewing employees, and learning about the
manufacturing processes, product lines, organi-
zational culture, and new control information.’
The physical location of the host division
was in one of many large buildings clustered
around the firm’s international headquarters.
The firm competes in an industry where suc-
cess depends on cost, reliability, innovation,
and speed-to-market. Deficiencies in reliability
and speed-to-market nearly led to the firm’s
demise five years earlier, and management
resolved to achieve significant improvements
to regain competitiveness. Top management
identified JIT and TQC manufacturing pro-
cesses as the best chance to effect the desired
changes throughout its operations. As a starting
point the firm selected its manufacturing opera-
tions division, which performs final assembly
and shipment of the firm’s major electronic
products, to be the first to install new sys-
tems. Management believed that implementa-
tion should begin at the point where
customer orders were satisfied and changes
eventually would permeate backwards
through the entire organization.
The division had approximately 800 employ-
ees divided into 31 direct labor (operator)
workgroups and 11 indirect labor (engineer-
ing, production support, etc.) workgroups,
each with its own first-level manager. Each
Workgroup (direct and indirect) was responsi-
ble for some stage of one of the assembly pro-
cesses. In addition, there were second-level
managers with responsibility for each of the
major assembly processes, incorporating sev-
eral stages. For example, one of the major
assembly processes supervised by a second-
level manager was assembling all circuit
boards for the final products. Workgroups
who had responsibilities for basic circuit-board
assembly or end-of-assembly testing accom-
plished this process. Different workgroups
assembled each of the major product lines,
and other workgroups were responsible for
final testing, packaging, and shipping. Other
workgroups were solely responsible for
rework of defective products pulled out during
test phases.
Each Workgroup began its shift with a team
meeting to discuss assembly schedules for the
day. In addition, the Workgroup reviewed any
feedback from production control or engineer-
ing regarding earlier work. This feedback,
which was directed originally to the first-level
manager dealt with maintenance, defectives,
design, and process issues. Posted in promi-
nent places was graphical and tabular evidence
regarding the division’s recent performance on
several key factors (discussed below), but there
were relatively few displays for workgroup-
level performance. The operators on average
were relatively low-paid and not highly edu-
cated. At JIT/ TQC implementation, the divi-
sion spent several days training employees
and considerably longer reconfiguring the
work-flow to conform with JIT/ TQC.
The division assembled multiple variations of
” Each dimension is measured as the mean for the Workgroup; that is, this analysis is at the Workgroup level.
’ This site was used for another study (see Young and Selto, 1993), but the instruments used to collect data on the
independent variables were different from that of the previous study.
672 F. H. SELTO et al
three major electronic products. Several work-
groups assembled all circuit boards on the
same line in response to final product orders.
Though operators performed quality testing as
they assembled boards, other workgroups per-
formed extensive end-of-assembly testing.
Completed circuit boards then channeled into
several assembly lines or cells where other
workgroups completed product assembly.
After further environmental-stress testing, sev-
eral workgroups installed covers on (or boxed)
completed products and moved them to ship-
ping bays. The division’s final step was placing
completed products in shipping cartons at the
shipping bay.
The division was constantly experimenting
with different assembly configurations, as
some products phased out, as others were com-
ing on-line, and as new processes were intro-
duced. Change (or “turmoil”, as one manager
termed it) seemed the rule rather than the
exception. Whether this rate of change indi-
cated continuous improvement was a matter
of some internal debate. For instance, the
firm had dramatically improved a key external
quality indicator, overall mean time to field fail-
ure; however, the firm still had a long way to
go to meet industry standards. Regardless, it
appeared that managers were continually revis-
ing Workgroup responsibilities. This made it
difficult for the control system to keep up,
especially since revised control information
always seemed to be an afterthought.
Only after a year from the start of the JIT/
TQC manufacturing process did the firm initi-
ate changes to the division’s accounting sys-
tem, and only within the past several years
did the system produce regular reports. At
the time of the study the accounting system
produced both cycle-time and costing reports
on a weekly and monthly basis. The typical
cycle-time or cost report distributed to the divi-
sion manager was a reproduction of a large
spreadsheet, filled to the margins with col-
umns and rows of numbers. Columns were
comparisons of actuals to budgets - for the
period and year-to-date. Rows corresponded
to major assembly processes, aggregated
across a number of work groups. At least one
engineering technician in the division was
responsible for transforming information from
these reports and other sources into bar charts,
pie charts, and control charts as requested for
monthly management meetings. However,
none except the cost accounting staff showed
much facility with these reports. Production
and quality control personnel monitored other
types of control information, and produced at
least monthly compilations in the form of
graphs, also for monthly meetings. Except at
a few operator locations, there was very little
real-time control information in this division,
in contrast to the daily (and usually more
often) revisions to production schedules and
kanban orders.
The 3 1 direct labor workgroups and 11 indir-
ect labor workgroups faced a similar environ-
mental context and generated similar,
objective performance measures, Note that
only direct labor workgroups had objective per-
formance measures at this site. We expected,
therefore, that this test site would yield suff-
cient data (i.e. 31 observations of appropriate
variables) to test various forms of contingency
theory. The Workgroup performance informa-
tion available to us includes: (1) cycIe time;
(2) cost; (3) yield; (4) defects; (5) schedule
adherence; (6) process problems; and (7) engi-
neering changes. The company maintains this
information on critical success factors outside
the financial accounting system, devoting a sig-
nificant amount of time and effort to data collec-
tion and maintenance. This set of performance
measures is consistent with the firm’s strategy
to be a low cost, high quality, quick response
manufacturer. These measures also are consis-
tent with the objectives of JIT/ TQC manufactur-
ing processes (for example, Chase and
Aquilano, 1993), which are to improve effec-
tiveness and efficiency through reducing
defects, increasing yields, and meeting custo-
mer expectations for quality and shipments.
We did have numerous opportunities to
informally interview members of this division
and the level of discord was palpable. Direct
laborers would speak to us only if their man-
CONTINGENCY THEORY
673
agers were not present. Engineers made depre-
cating remarks about the competence and fore-
sight of management, particularly with regard
to relationships with suppliers; and managers
insisted that operators needed more discipline
and appreciation of their jobs in a depressed
local economy. We also observed a number of
departures from textbook descriptions of JIT/
TQC. For example, operators had little or no
authority to identify and fix manufacturing pro-
cess problems; instead they relied on engineer-
ing and quality control staff. These and other
examples led us to expect that conflicts both
within and among workgroups, between work-
groups and managers and staff, and between
workgroups’ assigned objectives and manage-
ment processes were commonplace. Thus,
Workgroup conflict may be a serious problem
for this organization. This organization had
implemented JIT and TQC systems, and both
of these require that the operators be empow-
ered to make decisions about the work flow.
Interviews and surveys of managers and
operators to determine the organizatonal struc-
ture, context, technology, and management
controls at a field site, which was one division
of a large electronics manufacturer were con-
ducted. The ability of alternative theories of fit
to explain Workgroup performance within the
division was then tested. We found consistent
evidence that the organization we studied con-
tained a significant inconsistency in organiza-
tional design, which has impeded attainment
of its goals. That is, the managers exerted
almost complete control of Workgroup tasks.
In addition, we found significant levels of
intragroup conflict. There was additional sup-
port for hypothesized effects on Workgroup
performance from many variables representing
structure, context, and management controls.
MEASUREMENT OF CONTINGENCY
CONSTRUCTS
Van de Ven and Ferry (1980) developed and
tested the organizational assessment instru-
ment (OAI), which measures contingency the-
OV
constructs. The OAI assesses the
dimensions of organizational context, struc-
ture, and control as perceived by individual
members of the organization. Though the OAI
was not specifically designed to assess fit in
modern manufacturing firms, it was a relatively
straightforward task to adapt it to reflect the
terminology and characteristics of JIT and
TQC and the specifics of the research site.
We pre-tested the revised OAI on two groups
of first- and second-level managers and on two
groups of direct laborers (called operators at
our site) to insure that we meaningfully modi-
fied the survey instrument for the local condi-
tions. The pretest resulted in rewording, some
additions to reflect the complexities of the divi-
sion, and some shortening to remove several
questions senior managers found objection-
able (i.e. we were not allowed to ask about
incentives). The full set of contingency tests
is described in Table 3. Since our study is direc-
ted at workgroups within one division, envir-
onmental variables were not examined (Hayes,
1977).
We administered the revised OAI (available
on request) in person on a single day in Decem-
ber 1990, to 406 direct labor operators and 19
managers at our test site. There were 81 ques-
tions on the worker form and 92 questions on
the manager form. On average, respondents
spent 45 minutes responding. Anonymity was
guaranteed to operators and confidentiality to
managers. Constructs were developed from the
questions as designed by Van de Ven and Ferry
(1980). Descriptive statistics for these mea-
sures are listed in Table 4, along with their
reliabilities tested with Cronbach’s alpha. Vari-
able measures by Workgroup are available in
the Appendix.
The OAI as designed by Van de Ven and
Ferry contains up to 42 contingency constructs
plus several other, related constructs added for
this study (for example, job dependence on
cost accounting). Factor analysis revealed that
some of the original items loaded onto con-
structs as designed, but a relatively large num-
ber either did not load convincingly as
designed or loaded with other, intendedly
674 F. H. SELTO ei al.
TABLE 4. Contingency variables: descriutive statistics
Cronbach’s
alpha operators Managers Standard
(N = 406) (N = 19) N Mean deviation
Outcomes:
Job Satisfaction
Workgroup Effectiveness
Organization Structure:
Worker Authority
Standardization
Management Control Process:
Horizontal Communication
Vertical Communication
Workgroup Conflict
Manufacturing Context:
Task Difficulty/Variability
Job Dependency on Supervisor
Job Dependency on Workgroup
Covariates:
Engineering Changes
Process Changes
0.81 0.69 384 3.15 0.82
n/a 0.77 14 0.03 3.18
0.84 0.80 343 2.94 0.79
0.86 0.90 332 3.36 0.58
0.75 0.82
344 2.58 0.63
0.81 0.71 363 2.07 0.79
0.77 0.72 371 2.21 1.08
0.72 0.81 357 1.81 0.53
0.70 0.86 347 3.19 0.74
0.78 0.62 399 3.39 1.00
19 12.11 6.88
19 1.37 1.44
unrelated items. For the most part, we could
not know whether these unforeseen loadings
were systematic or random occurrences.
Therefore, we conservatively retained for this
study nine constructs whose items loaded as
designed and met minimum reliability levels
(including job satisfaction); that is, Cronbach’s
alphas greater than 0.60. We did reason that
observed common loadings of items for task
difficulty and task variability could represent a
credible common factor, which we named task
difficulty/ variability.
Nothing in the literature suggested that item
mortality from the OAI would be so high. Dra-
zin and Van de Ven (1985) did not report simi-
lar difficulties with the OAI and reported
Cronbach’s alphas for the nine OAI constructs
they used of 0.40-0.85. Though the OAl was
inefficient in this study, it did generate a suffi-
cient number of reliable independent and
dependent measures. Tentative explanations
for OAI performance in this study are:
1. In retrospect, expanding several constructs
related to Workgroup interactions with sev-
eral functional groups (for example,
accounting, logistics) unnecessarily Iength-
ened the OAI. We had thought these inter-
actions could be important effects on
outcomes. The increased length and per-
haps irrelevant items may have reduced sub-
jects’ attention to their responses.
2. The subject pool on average may not have
the educational level necessary to respond
to the complex and lengthy questionnaire.
Because managers selected the operators for
pretests, a random sample of subjects did
not scrutinize the early versions of the ques-
tionnaire. Thus we were not alerted to
potential communication problems for a
proportion (about 10%) of the workforce
for whom English is a second language.
In this paper, we use objective measures of
JIT/ TQC performance as our dependent perfor-
mance measure. Objective performance data
were available for up to 19 workgroups from
the fourth quarter (October-December) of
1990. All objective data as reported to us
were designed, gathered, and reported consis-
tently across workgroups, either by the central
cost accounting staff or by the production con-
trol staff. Cost efficiency, operationally defined
as the ratio of standard cost to actual cost,
CONTINGENCY THEORY 675
ranged from 0.67481 to 1.80270 for the work-
groups.8 Defects, defined as the number of
defects per time period to units tested, ranged
from 0.0413 to 0.3550. Yield, the number of
good units to total units per month, varied
from 0.6880 to 0.9876. All of these measures
showed, to us, surprisingly large variations for
a JIT/ TQC facility, perhaps because implemen-
tation was incomplete or because the measures
were faulty. Schedule, the proportion of times
actually met schedule to budgeted times met
schedule, ranged from 0.93 to 1.03, after two
outliers measured at 0.10 were eliminated. The
much narrower range on this variable indicates
possibly more attention paid to schedule adher-
ence than other dimensions of performance.
The fewer than expected observations of
objective performance measures, unfortu-
nately, limit the statistical power of some of
our tests. All in all, the firm’s performance mea-
surement system was much less advanced than
we expected. Factor analysis of the standar-
dized performance measures determined that
cost efficiency, yield, defects, and schedule
measured a common factor. Using factor
weightings we combined these into a summary
measure called Workgroup effecti veness.’ The
OAl includes the self-reported outcome mea-
sure,job sati sfacti on. Since little empirical evi-
dence links job satisfaction with performance,
we used job satisfaction as another dependent
variable.
Two of the archival measures were covari-
ates in the various tests since they could affect
outcomes but were not controlled by work-
groups. These were the number of engineering
changes (EC) and the number of problems
identified with the process in the time period
(Process). Engineering changes were exogen-
ous to workgroups and attempt to improve
the manufacturability of the product or to cor-
rect field-performance defects. These usually
were results of quality assurance testing or cus-
tomer complaints or field failures. Surprisingly
(to us) workgroups also did not have authority
to correct process problems at this site but
relied on a staff of production engineers who
responded on call to accommodate machine
failures, stoppages, or an EC requiring process
changes.
RESULTS
We consider each of the contingency tests in
turn, beginning with the selection approaches.
Sel ecti on tests
The selection approach predicts correlations
of context variables with organizational struc-
ture or control variables. The results of testing
the selection approaches at the i ndi vi dual
l evel are in Table 5. The data do not point to
any differences in fit solely due to managerial
discretion, so that distinction is dropped. Eight
of the 14 significant correlations (57%) are evi-
dence of misfit; that is, their signs are incorrect.
So much misfit indicates potentially serious
coordination and communication problems
within and across workgroups, inconsistent
with JIT/ TQC. A JIT/ TQC shop should be
smooth-running and well-coordinated, but the
data tell a different story.
Mi sfi t and j i t wi th organi zati on structure.
Surprisingly, worker authori ty is not corre-
lated with any context variables. In a JIT/ TQC
shop one expects workers to have high job
authority for difficult/ variable tasks, but they
perceive none at this site. Likewise one
expects worker authori ty to have a negative
’ We did not use cycle-time efficiency, the ratio of standard to actual cycle time, because its range, from 0.451 to 2.498,
suggested serious measurement problems. Nearly all tests using cycle-time efficiency were statistically insignificant,
perhaps due to measurement error. In some workgroups, cycle time was measured using barcodes, but in others measure-
ment was visual and haphazard.
9 We also ran regressions for alI tests using each of the performance measures as a dependent measure. The results of these
numerous tests were similar to those reported here.
676 F. H. SELTO et al
TABLE 5. Selection auuroach - correlation coefficients
Context variables
subject to
management Task difficulty/ Job dependency Job dependency
Contingency variables discretion Variability on supervisor on Workgroup
Organization Structure:
Worker Authority Yes ns ns ns
Standardization Yes
_ .,,**.
.2y*** ,,..
:
Management Control System (process):
Vertical Communication Yes ns .35*** .ll”
Horizontal Communication Yes .25*” 18”’
:
-.12***
Workgroup Conflict No .27”* ns - .09’
Manufacturing Context:
Task Difficulty/Variability n/a
-
-.lO’ - .23***
Job Dependency on Supervisor n/a -.10*
-
18”’
;
Job Dependency on Workgroup n/a -.23*** 18”’
-
:
Covariates:
Engineering Changes ns ns ns
Process Changes ns ns ns
l =p
1
7
6
.
1
1
8
1
.
9
1
8
0
.
7
1
5
6
5
.
1
2
.
0
?
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
-
)
1
6
1
.
8
1
6
8
.
8
1
8
1
.
0
1
5
0
6
.
3
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
-
)
1
8
0
.
0
1
8
3
.
8
1
8
9
.
2
1
5
9
3
.
3
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
-
)
1
7
5
.
2
1
7
8
.
5
1
8
7
.
9
1
5
7
2
.
9
H
o
r
i
z
o
n
t
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
-
)
1
8
8
.
2
1
9
0
.
9
1
9
3
.
6
1
6
2
2
.
3
W
o
r
k
g
r
o
u
p
c
o
n
f
l
i
c
t
(
+
>
1
7
3
.
0
1
8
0
.
2
1
7
8
.
3
1
6
3
6
.
7
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
+
)
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
+
)
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
H
o
r
i
z
t
o
n
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
1
5
2
.
5
1
5
3
.
8
1
7
6
.
8
1
4
1
1
6
8
.
1
1
7
0
.
7
1
8
5
.
7
1
4
9
1
6
0
.
0
1
6
2
.
4
1
8
4
.
8
1
4
6
1
.
1
2
.
3
2
.
2
8
.
7
?
3
.
9
K
5
.
6
2
2
.
3
z
2
.
5
R
1
1
.
0
7
.
7
1
1
.
1
1
7
4
.
8
1
7
6
.
2
1
8
8
.
4
1
5
5
1
.
2
5
.
9
W
o
r
k
g
r
o
u
p
c
o
n
t
l
i
c
t
(
+
)
1
6
5
.
6
1
6
5
.
8
1
6
7
.
3
1
5
8
0
.
2
0
.
8
l
T
e
s
t
t
h
a
t
i
n
t
e
r
a
c
t
i
o
n
t
e
r
m
a
d
d
s
s
i
g
n
i
f
i
c
a
n
t
e
x
p
l
a
n
a
t
i
o
n
b
e
y
o
n
d
m
a
i
n
e
f
f
e
c
t
s
:
F
0
0
1
(
1
,
n
-
3
)
=
6
.
8
;
F
o
.
o
s
(
1
,
n
-
3
)
=
3
.
9
.
l
*
T
e
s
t
t
h
a
t
m
a
i
n
e
f
f
e
c
t
s
a
d
d
s
i
g
n
i
f
i
c
a
n
t
e
x
p
l
a
n
a
t
i
o
n
b
e
y
o
n
d
i
n
t
e
r
a
c
t
i
o
n
t
e
r
m
:
F
0
.
0
,
(
2
,
n
-
3
)
=
4
.
7
;
&
O
S
(
2
,
t
l
-
3
)
=
3
.
0
.
T
A
B
L
E
6
(
c
o
n
r
.
)
.
P
a
n
e
l
B
:
j
o
b
s
a
t
i
s
f
a
c
t
i
o
n
P
e
r
f
o
r
m
a
n
c
e
m
e
a
s
u
r
e
C
o
n
t
e
x
t
v
a
r
i
a
b
l
e
I
n
t
e
r
a
c
t
i
o
n
w
i
t
h
s
t
r
u
c
t
u
r
e
o
r
c
o
n
t
r
o
l
v
a
r
i
a
b
l
e
(
a
c
t
u
a
l
s
i
g
n
o
f
i
n
t
e
r
a
c
t
i
o
n
)
B
S
S
a
l
l
B
S
S
B
S
S
m
a
i
n
i
n
t
e
r
a
c
t
i
o
n
F
F
m
a
i
n
o
n
l
v
o
n
l
v
?
I
i
n
t
e
r
a
c
t
i
o
n
*
e
f
f
e
c
t
*
*
J
o
b
S
a
t
i
s
f
a
c
t
i
o
n
T
a
s
k
D
i
f
f
i
c
u
l
t
y
/
V
a
r
i
a
b
i
l
i
t
y
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
+
)
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
+
)
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
H
o
r
i
z
o
n
t
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
-
>
W
o
r
k
g
r
o
u
p
C
o
n
l
l
i
c
t
(
-
)
J
o
b
D
e
p
e
n
d
e
n
c
y
o
n
S
u
p
e
r
v
i
s
o
r
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
+
)
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
+
>
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
H
o
r
i
z
o
n
t
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
>
W
o
r
k
g
r
o
u
p
C
o
n
f
l
i
c
t
(
-
>
J
o
b
D
e
p
e
n
d
e
n
c
y
o
n
W
o
r
k
g
r
o
u
p
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
+
)
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
+
>
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
H
o
r
i
z
o
n
t
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
>
W
o
r
k
a
r
o
u
n
C
o
n
f
l
i
c
t
(
-
)
1
7
4
.
3
1
7
4
.
3
1
9
4
.
7
2
9
3
0
.
0
1
5
2
.
1
1
5
2
.
1
1
8
4
.
9
3
0
1
0
.
0
1
6
4
.
4
1
6
4
.
6
1
9
5
.
0
3
0
3
0
.
5
1
9
7
.
7
1
9
9
.
5
2
0
4
.
1
3
1
3
2
.
8
1
9
0
.
4
1
9
3
.
1
1
9
2
.
1
3
1
7
4
.
4
1
8
9
.
0
1
8
9
.
1
1
9
1
.
9
3
0
4
0
.
1
1
7
2
.
1
1
7
2
.
1
1
7
8
.
9
3
1
7
0
.
1
1
7
6
.
6
1
7
6
.
7
1
8
8
.
6
3
1
9
0
.
2
2
1
3
.
7
2
1
3
.
8
2
1
4
.
8
3
2
9
0
.
2
1
9
7
.
5
1
9
7
.
5
2
1
0
.
3
3
2
8
0
.
0
1
8
5
.
4
1
8
6
.
6
1
9
3
.
6
3
0
0
1
.
9
1
6
5
.
5
1
6
6
.
1
1
7
4
.
3
3
0
3
1
.
0
1
6
8
.
4
1
6
8
.
4
1
8
2
.
4
3
1
1
0
.
0
2
1
5
.
4
2
1
5
.
6
2
1
6
.
9
3
2
4
0
.
3
2
0
3
.
8
2
0
4
.
3
2
1
5
.
1
3
2
6
0
.
7
1
6
.
9
3
2
.
2
2
8
.
0
5
.
0
1
.
4
8
s
2
.
3
3
6
.
2
z
1
0
.
7
r
l
0
.
8
1
0
.
5
2
8
6
.
6
z
8
.
0
1
2
.
8
1
.
1
8
.
9
*
T
e
s
t
t
h
a
t
i
n
t
e
r
a
c
t
i
o
n
t
e
r
m
a
d
d
s
s
i
g
n
i
f
i
c
a
n
t
e
x
p
l
a
n
a
t
i
o
n
b
e
y
o
n
d
m
a
i
n
e
f
f
e
c
t
s
:
F
,
.
,
,
(
1
,
n
-
3
)
=
6
.
8
;
F
0
.
0
s
(
1
,
n
-
3
)
=
3
.
8
5
.
l
*
T
e
s
t
t
h
a
t
m
a
i
n
e
f
f
e
c
t
s
a
d
d
s
i
g
n
i
f
i
c
a
n
t
e
x
p
l
a
n
a
t
i
o
n
b
e
y
o
n
d
i
n
t
e
r
a
c
t
i
o
n
t
e
r
m
:
F
,
,
,
(
2
,
r
z
-
3
)
=
4
.
7
;
F
o
.
o
s
(
2
,
n
-
3
)
=
3
.
0
.
680 F. H. SELTO et al.
bility) interact with structure or control (work-
group conflict) to affect outcomes (job satisfac-
tion) as predicted (alpha = 0.05). In the few
cases where the interaction was significant,
main effects also were significant. It was far
more common that main effects rather than
interactions significantly explain outcomes.
Since the interaction approach tests fit in a
piecemeal manner and does not consider likely
tradeoffs among design elements, it may be too
simplistic to explain JIT/ TQC outcomes.
Systems tests
According to the systems approach to con-
tingency theory, the overall fit of context,
structure, and control explains the level of per-
formance. Misfit is measured as the overall
deviation from a benchmark configuration,
and the more a Workgroup departs from the
benchmark in any direction, the worse its per-
formance is expected to be. We used two mea-
sures of fit to test this theory, Euclidean
distances and absolute values of differences in
means on each dimension of fit between the
benchmark Workgroup and each other work-
group. Each approach has major drawbacks.
Computation of Euclidean distances ignores
quite reasonable and effective tradeoffs that
managers and workgroups may make that also
result in good performance. Using differences
in Workgroup means is an ad hoc approach
since each difference is implicitly hypothe-
sized to be an equally important, negative
impact on outcomes.
The benchmark workgroups for measuring
differences were:
Group 71° Highest performance for
Workgroup effectiveness.
Group 48 Highest level of job satisfac-
tion.
EDs were calculated for all the other work-
“’ We numbered the groups randomly.
I’ Workgroup effectiveness: R* = 0.02, F,,,, = 0.238.
Job satisfaction: R2 = 0.03, F,,,s = 0.555.
groups, and then outcome measures were
regressed on EDs. Neither of these regressions
was significant, which refutes this version of
contingency theory. ’ *
As a second test of the systems approach, for
each context, structure, and control variable
we regressed the absolute values of differences
of each Workgroup’s mean from the mean of
the benchmark Workgroup on job satisfaction.
This regression indicated that differences from
the benchmark explained levels of job satisfac-
tion at a significance of p = 0.005, adj-R* =
0.42. However, the signs of only several differ-
ences were negative, as predicted. The abso-
lute values of differences from Workgroup 7
also were regressed on Workgroup effective-
ness. These results were not significant, and
the degrees of freedom were relatively small.
Viewing these tests conservatively, we find lit-
tle support for the systems approach from
either form of test.
Summary of test results
Our tests of the selection approach reveal
interesting evidence of JIT/ TQC misfit using
the context variables of task dtj$?culty/varia-
bility and job dependency on supervisor or
Workgroup as correlates. These tests indicate
perhaps irreconcilable conflict between ele-
ments of context. The interaction and the sys-
tems approaches seek to explain performance,
but we found no support for the interaction
approach, either from regressions that explain
job satisfaction or Workgroup effectiveness.
We also found no support for the systems
approach. In this study, the simpler selection
approach was the most descriptive.
CONCLUSIONS
We attempted to respond to Hopwood’s
(1976) call to study management control issues
CONTINGENCY THEORY 681
by simultaneously considering organizational
structure, context, and processes. By adopting
a contingency theory framework, with its
numerous variables and multiple approaches,
and by focusing at the operator level of the
organization we had the raw material to
develop a plausible explanation of Workgroup
performance. The strongest results point to
internal inconsistency of organizational con-
text, perhaps regardless of management pro-
cesses. At our field site, many of the
management control measures and devices
thought to be consistent with JIT/ TQC were
in place. However, conflicts between opera-
tors’ needs for empowerment required by
JIT/ TQC systems and a management approach
better suited to mechanistic work quite prob-
ably negate beneficial influences of appropriate
controls. As indicated by the selection tests,
this firm has strong vertical dependence with
a management firmly in control, which is not
compatible with the concept of worker
empowerment. We believe this result indicates
the role of supervisors must be altered if work-
groups are to meet JIT/ TQC objectives at this
site.
It was disappointing to be unable to explain
Workgroup effectiveness satisfactorily with the
interaction and systems versions of contin-
gency theory after applying what we believed
to be powerful and reliable descriptive tools.
Retrospection yields some insights that may
inform future work, however. A key reduction
in power was due to the relatively few work-
groups with independent, objective measures
of performance. We had been led to believe
that the host company “managed by the num-
bers”, but if so the firm did not seem to have a
full set of information. This provides some evi-
dence that implementation of JIT/ TQC had not
been successful at this site. A company that
routinely generated periodic variance reports
had not made the transition to decentralized
information gathering and use. The company
reorganized the workforce into nominally
empowered workgroups, but they had neither
the information nor the authority to manage
themselves.
Another source of low power in the regres-
sion models is due to noise in the dependent
variable from measurement error in the compo-
nents of Workgroup effectiveness. In some
cases measurement errors would offset, but
there is no guarantee of that, particularly
within a single organization. We suspect con-
siderable error in all the measures, but were
especially surprised to find so much apparent
error in cycle-time measures, which we
decided not to use. The surprise was the result
of the build-up given to us by top management
that cycle time was the cornerstone of the divi-
sion’s new management control system, to be
replicated throughout the company. This mea-
sure would have to be improved to be useful.
In retrospect, a study of measurement pro-
blems and errors in assessing JIT/ TQC perfor-
mance would be a worthwhile contribution to
the manufacturing control literature. It is our
suspicion that managers within the division
and the firm have been making critical deci-
sions based on faulty JIT/ TQC performance
data. This may be a general problem and may
prevent full realization of the gains promised
by JIT/ TQC in other firms.
A third source of low power is the possibility
that the contingency theory used may not be as
descriptive as we would like to think. There
have been enough equivocal or negative con-
tingency results that can led to the question of
whether these can be attributed to uneven
implementation of the theory (a fourth source
of low power), or to narrow operationaliza-
tions of the theory. These either did not cap-
ture the richness of contingency theory and so
findings were minimal, or happened to focus
on the one or few aspects that came through
in a specific test.
As a final point, contingency theory does
have intuitive appeal as a method for testing
the fit of management controls, but it has weak-
nesses as well as strengths. One may obtain a
big picture of the organization using many fea-
sible predictor variables, but tight focus
depends on the empirical relationships that
emerge. Though significant relationships may
be found, these may not provide reliable
682 F. H. SELTO el al
guidance either for future research or for the
practitioner in the field attempting to improve
his or her process. Any change in circum-
stances may invalidate observed relationships.
This may be especially problematic in a manu-
facturing firm that is continuously revising pro-
ducts and manufacturing systems, in some
cases in subtle ways. Furthermore, we feel
there is a distinct possibility that both interac-
tion and systems implementations of contin-
gency theory are too simplistic to describe
performance in a complex workplace, espe-
cially at the Workgroup level. However, if we
had not tested multiple operationalizations of
contingency theory (as recommended by Van
de Ven and Drazin (1985)) we might not have
identified intragroup and structure conflicts as
the most likely impediments to superior perfor-
mance. Though most other investigators have
not found elementary contingency theories (for
example, selection approaches) to be informa-
tive, we found them to be crucial to under-
standing the dynamics of this workplace.
BIBLIOGRAPHY
Abemethy, M. & Stoelwinder, J., Budget Use, Task Uncertainty, System Goal Orientation, and Subunit
Performance: a Test of the “Fit” Hypothesis in Not-For-Profit Hospitals, Accounti ng, Organi zati ons
and Soci ety (1991) pp. 105-120.
Bleakley, F., The Best Laid Plans: Many Companies Try Management Fads Only To See Them Flop, Wal l
StreetJ ournal (6 July 1993) p. Al.
Chase, R. & Aquilano, N., Producti on and Operati ons Management (Homewood, IL: Irwin, 1993).
Chenhali, R. H. & Morris, D., The Impact of Structure, Environment, and Interdependence on the
Perceived Usefulness of Management Accounting Systems, 7I?e Accounti ng Revi ew (1986) pp. 16-35.
Drazin, R. & Van de Ven, A., Alternative Forms of Fit in Contingency Theory, Admi ni strati ve Sci ence
Quarterl y (1985) pp. 514-539.
Gordon, L. & Narayanan, V., Management Accounting Systems, Perceived Environmental Uncertainty and
Organization Structure: an Empirical Investigation, Accounti ng, Organi zati ons, and Soci ety (1984) pp.
33-47.
Govindarajan, V., Appropriateness of Accounting Data in Performance Evaluation: an Empirical Examina-
tion of Environmental Uncertainty as An Intervening Variable, Accountfng, Organi zati ons and Society
(1984) pp. 125-135.
Govindarajan, V., A Contingency Approach to Strategy Implementation at the Business-unit Level: Inte-
grating Administrative Mechanisms with Strategy, Academy ofManagemeniJournal(1988) pp. 826-
853.
Govindarajan, V. & Gupta, A., Linking Control Systems to Business Unit Strategy: Impact on Performance,
Accounti ng, Organi zati ons and Soci ety (1985) pp. 51-66
Gresov, C., Exploring Fit and Misfit with Multiple Contingencies, Admi ni stratfve Scfence QuarterZy
(1989) pp. 431-453.
Hayes, D., The Contingency Theory of Managerial Accounting, The Accounti ng Revi ew (1977) pp. 22-
39.
Hayes, R. H., Wheelwright, S. & Clark, K. B., DynamfcManufacturfng (New York: The Free Press, 1988).
Hopwood, A., Accounti ng and Human Behavi or (EngIewood CIifs, NJ: Prentice-Hall, 1976).
Jacobs, R., JIT More Than a Dying Fad? Wal l StreetJ ournal (18 October 1993) pp. 66-72.
Johnston, J., Economelric Methods (New York McGraw Hi& 1984).
ho, J. K. & Peterson, R. B., Effects of Task Difticulty and Interunit Interdependence on Information
Processing Systems, Academy of Management J outnai (1986) pp. 139-149.
Macintosh, N. & Daft, R., Management Control Systems and Departmental Interdependencies: an Empiri-
cal Study, Accounti ng, Organizations and Soci ety (1987) pp. 49-61.
March, R. M. & Mannari, H., Technology and Size as Determinants of the Organizational Structure of
Japanese Factories, Admi ni strati ve Sci ence Quartet@(1981) pp. 33-57.
CONTINGENCY THEORY
683
Otley, D., The Contingency Theory of Management Accounting: Achievement and Prognosis, Account-
ing, Organizations and Socteety (1980) pp. 413-428.
Perrow, C., A Framework for the Comparative Analysis of Organizations, American Sociological Review
(1967) pp. 194-208.
Schoonhoven, C. B., Problems with Contingency Theory: Testing Assumptions Hidden within the Ian-
guage of Contingency Theory, Administrative Scfence Quarterly (1982) pp. 349-377.
Schonberger, R. J., J apanese Manufacturing Techniques - Nine Hi&en Lessons in Simplicity (New
York: The Free Press, 1982).
Schonberger, R. J., World Class Manufacturfng: i’be Lessons of Stmpliciry Applied (New York: The Free
Press, 1986).
Schonberger, R. J., World Class Manufacrurtng Casebook: I mplementing J I T/TpC and /I T-TQC (Flew
York: The Free Press, 1987).
Tushman, M. L., Work Characteristics and Subunit Communication Structure: a Contingency Theory
Analysis, Admintstrartve Science Quarterly (1979) pp. 82-98.
Van de Ven, A. & Drazin, R., The Concept of Fit in Contingency Theory, Research in Organizational
Bebavtor (1985) pp. 333-365.
Van de Ven, A. & Ferry, D., Measuring and Assessing Organizarions (New York: Wiley, 1980).
Wilkinson, 9. & Oliver, N., Power, Control and the Kanban, J ournal of Management Studies (1989) pp.
47-58.
Young, S. M., A Framework for Research on Successful Adoption and Performance of Japanese Manu-
facturing Practices in the United States, Academy of Management Review (1992) pp. 677-700.
Young, S. M. & Selto, F., Explaining Crosssectional Workgroup Performance Differences in a JIT/TQC
Facility: a Critical Appraisal of a Field-based Study, J ournal of Management Accounting Research
(1993) pp. 300-326.
Young, S. M. & Selto, F., New Manufacturing Practices and Cost Management: a Review of the Literature
and Directions for Research, J ournal of Accounting Literature (1991) pp. 320-351.
684 F. H. SELTO et al.
Appendix
VariableMeans-By Workgroup
93 2. 65 [ O. ?? 0. 63 2. 87 I O. 531 3. 39 [ 060] 1. 78[ 0. 53] Z. M[ O. rZQ] 2. 26 11. 06] 1. 77 [ 0. 42] 2. 71[ 0. 59] 2. 52 CO. 771
97 3. 07 10971 n/ a 3. 17 I O. 461 3. 63 ( 0. 441 2. 44 I O. 851 2. 88[ 0. 51] 2. 58 [ I . 271 l . SO[ O. Cd] 3. 24[ 0. 38] 3. 30[ 1. 00]
*n/ a=AUc~~ponenl sof worl rgroupcf f ccl i venessnotavai l abl e
doc_683572303.pdf
The adoption of new manufacturing practices such as just-in-time (JIT) and total quality management
(TQC) is only a first step to improving manufacturing performance. Even more critical is the fit between
manufacturing practices and organizational design, structure and processes. Using archival and survey
data, this paper reports the results of a field study within a Fortune
Pergamon Accounti ng, Organfzati ons and Soci ety, Vol. 20, No. 7/ S, pp. 665-684, 1995
Copyright 0 1995 Elsevier Science Ltd
Printed in Great Britain. All rights reserved.
0361-3682/95 $9.50+0.00
ASSESSING THE ORGANIZATIONAL FIT OF A JUST-IN-TIME MANUFACTURING
SYSTEM: TESTING SELECTION, INTERACTION AND SYSTEMS MODELS OF
CONTINGENCY THEORY*
FRANK H. SELTO
University of Colorado at Boulder
CELIA J. RENNER
University of Northern I owa
and
S. MARK YOUNG
University of Southern California
Abstract
The adoption of new manufacturing practices such as just-in-time (JIT) and total quality management
(TQC) is only a first step to improving manufacturing performance. Even more critical is the fit between
manufacturing practices and organizational design, structure and processes. Using archival and survey
data, this paper reports the results of a field study within a Fortune 500 company that tests three
operationalizations of contingency theory as discussed by Van de Ven and Drazin (1985) [The Concept of
Fit in Contingency Theory, Research i n Organi zutfonal Behavi or, pp. 333-3651. Results show that the
misfit between worker empowerment required by JIT/ TQC practices and existing authoritarian manage-
ment partially explain relative Workgroup performance as do other conflicts within workgroups and
between operators and supervisors.
Successfully adopting Japanese manufacturing how management controls should be designed
practices such as the just-in-time (JIT) manufac- to be consistent with organizational structure
turing and total quality control (TQC) systems and context (Hopwood, 1976; Otley, 1980; Gor-
has been problematic for many organizations don & Narayanan, 1984; Govindarajan, 1984,
(Wilkinson & Oliver, 1989; Young & Selto, 1988; Govindarajan & Gupta, 1985; Chenhall
1991; Young, 1992). In particular, there has & Morris, 1986; Abernethy & Stoelwinder,
been a lack of attention to needed changes in 1991) little research has directly addressed
organizational structure, context, and control. the issues of the best “fit” of organizational
While considerable literature has discussed structure and controls for JIT/TQC systems.’
l We thank Shannon Anderson, Dale Jasinski, Chris Koberg, participants at the First International Conference on Account-
ing, Taipei, Taiwan (January 1994) the Fourth Biennial Management Accounting Conference, Sydney, Australia (September
1994) and the Sixth Tokyo Keizai University International Symposium, Tokyo, Japan (November 1994), as well as the
anonymous reviewers for their helpful comments on this paper.
i Hereafter, we refer to the joint JIT/ TQC system since most observers agree that it is not possible to implement JIT
without TQC.
665
This paper addresses the issue offit among can be produced, inventories build up in push
the central organizational elements in a JIT/ systems. These inventories may be used to
TQC manufacturing firm by studying the rela- shield the push process from breakdowns and
tions among organization structure, JIT/ TQC quality failures in previous process stages
manufacturing methods, and management con- (referred to as a “just-in-case” system - Hayes
trols at the Workgroup level within a Fortune et al . (1988). JIT/ TQC means that operators do
500 fi rm. Fit also is tied directly to the perfor- not have or cannot build buffer inventories
mance of workgroups. Data are tested against (though some JIT/ TQC facilities allow minimal
three contingency theory models as summar- inventories at each stage or at a central location
ized by Van de Ven and Drazin (1985). for severe problems (Schonberger, 1987). In
The paper proceeds as follows. The next order to maintain production flow, JIT/ TQC
section presents a theoretical framework in operators should continually improve the pro-
which several key dimensions of the JIT/ TQC
cess to prevent stoppages. If breakdowns or
system are discussed within a review of the
quality failures do occur, JIT/ TQC operators
contingency literature in management account-
must communicate and interact to solve what-
ing. Also discussed are three operationaliza-
ever problems arise to keep production flowing.
tions of contingency fit within the context of
These new job responsibilities require consider-
JIT/ TQC manufacturing. The third section dis-
able process knowledge, flexibility, communi-
cusses the research site and describes measures
cation, control, and authority at the worker or
of various dependent and independent vari-
operator level - characteristics that are not
ables. Statistical results are presented, and the
typical in traditional, push manufacturing.
three contingency theory approaches are com-
JIT/ TQC requires that operators are able to
pared. The final section concludes the paper.
detect and correct process problems and
defects when they occur and to seek for ways
to prevent process variation and product
THEORETICAL FRAMEWORK
defects, without waiting for central or end-of-
process quality assurance to issue change
J I T/TQC manufacturi ng
orders. Therefore, JIT/ TQC also requires opera-
In theory, JIT/ TQC makes significantly differ-
tors to have the flexibility, knowledge, and
ent demands on operators than traditional man-
authority usually reserved for managers in tradi-
ufacturing methods. JIT/ TQC is often
tional systems. In other words, operators need
characterized as apull approach to manufactur-
to be able to manage themselves. Managers are
ing, as opposed to a more traditional push sys-
more like “coaches” and may even work
tem (Schonberger, 1982, 1986). JIT/ TQC
directly on products alongside operators. It
operators produce a part or product only in
seems likely, then, that the relative roles of
response to an order from an internal or exter-
operators and low-level managers can be criti-
nal customer. JIT/ TQC tasks are highly interde-
cal to the successful functioning of JIT/ TQC.
pendent; each succeeding task depends on the Undoubtedly there is considerable variation
quality and timeliness of the preceding task. in JIT/ TQC implementations among manufac-
Thus, no excess parts or products are pro- turing firms (for example, Schonberger,
duced, unlike traditional systems where opera- 1987). These variations may reflect manage-
tors push parts and products through the ment ability, external and internal constraints,
process as quickly as possible. If parts are not and, perhaps lack of understanding of the
used and products are not sold as quickly as they implications of JIT/ TQC.’ The objective of the
666 F. H. SELTO et al.
’ Variations in JIT/TQC implementation themselves may explain differences in performance across manufacturing firms,
which is worthy of future study.
CONTINGENCY THEORY
667
present study is to investigate whether, given a
particular JIT/ TQC implementation, variation
in fit among other elements of an organization
that are under managers’ control explain cross-
sectional performance within a firm. Implica-
tions of JIT and TQC allow us to hypothesize
what these elements and effects might be.
Conti ngency theory
The central theme of contingency theory
(applied here to JIT/ TQC settings) is that alI
components of an organization must “fit” well
with each other or the organization will not
perform optimally (Perrow, 1967). Within a
single organization, contingency theory also
predicts that variations in subunits’ perfor-
mance (for example, at the Workgroup level)
are due to variations in fit of their various com-
ponents (March & Mannari, 1981; Tushman,
1979). An extensive contingency theory
literature has defined essential organizational
elements and has explored their interrelation-
ships and effects on performance. All of these
elements of contingency theory are compatible
with JIT/ TQC concerns, as will be shown.
We investigate the role of JIT/ rQC fit in a
contingency theory framework because: (1)
no other theory of which we are aware
directly concerns fit; (2) despite criticism, the
intuition behind the theory continues to be
appealing; and (3) several recent operationali-
zations of fit have overcome some previously
voiced criticisms (see Otley, 1980; Schoonha-
ven, 1982). To our knowledge, no study has
extended the contingency literature descrip-
tions or predictions to organizations that
employ manufacturing methods such as JIT
and TQC. These manufacturing practices have
been adopted widely throughout the world in
the last decade, yet many observers note that
expected benefits of these methods have rarely
materialized (for example, Jacobs, 1993; Bleak-
ley, 1993; Young, 1992). The hypothesis we
investigate is that poor fit among critical orga-
nizational elements explains lack of success of
JIT and TQC methods.
Drazin and Van de Ven’s (1985) review of
the literature suggests that contingency theory
has considerable power to explain Workgroup
performance within organizations. They cate-
gorize the primary components of an organiza-
tion into organizational context, organizational
structure, and organizational process or con-
troZ.3 Context includes factors such as size,
culture, environment, technology, tasks, and
methods such as JIT/ TQC. Structure refers to
an organization’s formality (mechani sti c) or
flexibility (organi c), incentives, degrees of
authority, and bureaucracy. Control includes
communications and information flow, includ-
ing explicit management controls. Variable
definitions used in this study are in Table 1.
Tbeoly of fit. Theoretically, each organiza-
tion has its own optimal configuration or best
jit of context, structure, and control. Deviation
from that ideal fit (which is mi sfi t) should
cause lack of coordination, miscommunica-
tion, misunderstanding, poor morale, and
poor motivation, which, in turn, should lead
to poor performance. Concerns for the misfit
induced by outmoded, irrelevant control infor-
mation in today’s manufacturing firms are con-
sistent with contingency theory arguments. As
discussed earlier, JIT/ I’QC manufacturing sys-
tems (theoretically) are characterized by high
levels of worker autonomy and decision mak-
ing, worker flexibility, and quick response to
variable product and process demands (IIayes
et al ., 1988). Furthermore, this environment
may depend more on horizontal communica-
tion among operators than on vertical commu-
nication from managers. Thus, though one
means of achieving JIT is to standardize ele-
ments of work, JIT/ TQC should be more suc-
cessful in organic, rather than mechanistic
organizations. These profiles are shown in
Table 2.
In addition, an organization’s context may be
’ Because we use the termprocess to refer to the manufacturing steps or procedures, hereafter we will refer to organ&a-
tional process as control . The term control is also more consistent with the management control literature.
668 F. H. SELTO et al
TABLE 1. Definitions of contineencv constructs*
Sfundurd~zulion - Extent to which standard operating procedures and performance expectations are formalized and
followed and the degree to which the roles and task assignments that make up a job are listed in the job description
and other procedures.
WorkerAuthorfty - Includes personal decision making by employee as well as decision making by members of the unit as
a group. Also includes amount of discretion that the job incumbent has in making decisions about what tasks and
assignments make up the job, how the job is to be done, and how problems are handled.
Vertical Communication - The amount of feedback the job incumbent receives from his supervisor about performance,
work-related messages sent by writing, i.e. memos, reports and letters, and amount of explanation of changes in group
operating procedures.
Hortzontul Communication - Frequency of group or staff meetings among three or more personnel in the Workgroup
and difficulty in getting the Workgroup together to accomplish work.
Workgroup ConfZict - Frequency of disagreements or disputes among members of the group and extent to which conflict
hinders others’ work.
Tusk Dffj?culty/Variubilfty - Includes analyzability and predictability of work done by the unit and number of exceptions
encountered in the work.
J ob Dependency on Supervisor - Refers to how much each person’s job depends on activities performed by the
supervisor and the hierarchical authority exercised by the supervisor in terms of decisions about what tasks are to be
performed, what performance evaluation criteria are, and operating rules established to coordinate and control unit
activities.
J ob Dependency on Workgroup - Relates to intensity of work-interdependence among unit personnel. Refers to how
much each person’s job depends on activities performed by other unit members. Also includes how specialized the
jobs are within the unit and degree to which unit members can do each other’s work.
l Adapted from Van de Ven and Ferry (1980)
TABLE 2. Contingencies and design patterns: a summary of
fit and misfit*
Mechanistic
Design dimension pattern
sTRucTuRE:
Standardization High
Worker authority Low
MANAGEMENT CONTROL PROCESSES:
Vertical communication High
Horizontal communication Low
Workgroup conflict Low
CONTEXT:
Task difficulty and variability
Low Fit
High Misfit
Job Dependency on Supervisor
(Vertical Dependence)
Low Misfit
High Fit
Job Dependency on Work Group
(Horizontal Dependence)
Low Fit
High Misfit
Organic
pattern
Low
High
Low
High
Low
Misfit
Fit
Fit
Misfit
Misfit
Fit
l Adapted from Gresov (1989)
internally inconsistent, and optimal perfor-
mance may not be possible with any pattern
of structure and control (Gresov, 1989). Speci-
fically as found in this study, pervasive depen-
dence on managers may conflict directly with a
JIT/TQC process that requires empowered
operators.
Empi ri cal test resul ts. Empirical tests of con-
tingency theory have produced mixed results,
with as many studies finding significant correla-
tions between measures of fit and performance
as not. These mixed results have caused many
researchers to criticize past operationalizations
of fit itself (for example, Schoonhaven, 1982;
Van de Ven & Drazin, 1985) as ad hoc. Further,
a few studies have used objective measures of
firm performance when assessing the relation
of fit to performance. Most studies have ana-
lyzed self-assured performance as a function
of fit among a limited (but differing) set of
organizational variables (for example, Gordon
& Narayanan, 1984; Ito & Peterson, 1986;
Chenhall & Morris, 1986; Macintosh & Daft,
1987; Abernethy & Stoelwinder, 1991). Govin-
darajan (1988) considered multiple measures of
fit among strategy, organizational structure,
CONTINGENCY THEORY 669
managers’ characteristics, and control systems,
but again using self-assessed performance.
To summarize, contingency theory has pros
and cons. Advantages include a rich descriptive
framework, plentiful opportunities for mea-
surement and observation, explicit linking of
organizational characteristics and perfor-
mance, and an extensive, supportive theoreti-
cal literature. Disadvantages, though, are
significant and include lack of standard mea-
sures, ambiguity in the operationalization of
the key construct, fit, and equivocal past
empirical results. Any contingency study,
then, must exploit a rich framework and miti-
gate threats to validity. Unfortunately, past con-
tingency studies in the aggregate suffer from
lack of standardization, which impedes cross-
study comparisons. There is an opportunity,
therefore, to add to the contingency literature
by building on past developments and by con-
sidering objective measures of performance.
Al ternati ve approaches to conti ngency
theory j i t
A valid interpretation of contingency theory
must operationalize the concept of fit (or mis-
fit). To enhance the validity of work on fit and
performance we turned to the work of Van de
Ven and Drazin (1985) and Van de Ven and
Ferry (1980). These studies have built a com-
mon framework for studying contingency fit
and have shown that at least one operationali-
zation of fit can explain objectively measured
Workgroup performance, Van de Ven and Dra-
zin (1985) critique three current approaches to
contingency theory based on alternative
notions of fit: (1) selection; (2) interaction;
and (3) systems. They point out that these
three approaches are not mutually exclusive,
and may provide complementary information
about an organization. Since each approach
addresses different aspects of fit and since the
most appropriate definition and measure of fit
is still unclear, this study uses all three
approaches.
Sel ecti on. In the selection approach, organi-
zational context drives organizational design.
Fit is defined in terms of predictable correla-
tions between pairs of organizational vari-
ables. Natural sel ecti on predicts that all
structure and control variables are correlated
with context since anything less would lead
to exti ncti on in a competitive environment.
Manageri al sel ecti on predicts correlations
between context and only those control and
structural characteristics managed by the orga-
nization. In general, we hypothesize that JIT/
TQC context variables are correlated with
Workgroup structure and control. For exam-
ple, worker authori ty should be positively cor-
related with hori zontal communi cati on but
negatively correlated with verti cal communi -
cati on. Likewise, task d@hdty and vari abi l i ty
should be positively correlated with worker
authori ty. In all, 15 such relationships are pre-
dicted by the selection approach in this study.
Note that explaining performance is not an aim
of the selection approach since it is assumed
that only good performers survive to be
observed.
I nteracti on. The interaction form of contin-
gency theory hypothesizes jit as the i nterac-
ti on of pairs of context-structure or context-
control factors on performance. The basic con-
cept is that none of context, structure, or con-
trol alone should affect performance; it is the fit
among them that affects performance. In a
regression explaining performance, therefore,
an interaction term of context and Workgroup
structure (and control) should be significant
while main effects should not. Statistically,
main effects may be significant, of course, but
such results detract from the theory. For exam-
ple, in a JIT/ TQC organization task di j @ul ty
and vari abi l i ty should interact with worker
authori ty to have a positive effect on perfor-
mance, but their individual contributions
should be insufficient to beneficially affect per-
formance. Likewise, standardi zati on should
interact with hori zontal communi cati on to
have a negative impact on performance, but
without main effects. As shown in Table 3,
there are 30 such interactions possible in this
study (3 context X 5 control and structure X 2
performance variables). Though it would be
possible to reduce context, control, and struc-
670 F. H. SELTO et al.
TABLE 3. Alternative contingencv tests
Outcomes tested
Basic tests
Number of control or
structural variables in each
test
Number of tests to cover all
structure, context, process,
and outcome variables
Selection
No
Correlations
interaction
Yes
Regressions
Systems
Yes
Regressions
Pair Pair
Eight, reduced to Euclidean
distance
15 Correlations 30 Regressions
3 x 5” 3X5X2b 2 Euclidean distance tests=
’ This test included 3 context variables and 5 control and structure variables.
’ This test included 3 context variables, 5 control and structure variables, and the 2 outcome measures,
’ Each test included 3 context, 2 structure, and 3 control variables with one performance measure. In addition, the 2
covariates were included in the Workgroup effectiveness test.
ture to composite variables and reduce the
number of reported tests, this aggregate
approach would reduce the descriptive power
of this approach considerably.
Systems. The systems approach, the most
recent and least-tested form of contingency the-
ory, is a hol i sti c approach to studying the inter-
dependencies in organizations. In concept,
optimal systems fit occurs when all design ele-
ments of structure, control, and context are
congruent. Variations in performance result
from variations in this overall systemic fit.
The further a Workgroup’s design is from opti-
mal, then the lower its performance should be.
Identifying the optimal organization or work-
group is the primary drawback to this
approach, since it is possible there are many
equally effective, feasible sets of organizational
design elements (equifinality). Defining optimal
fit is problematic, as well, and is usually defined
in a somewhat circular manner: optimal fit is
the configuration of the optimally performing
organization.
Van de Ven and associates operationalized
deviations from optimal systems fit as the dif-
ference from the set of characteristics (along
each critical dimension) of the top performing
Workgroup(s) in an organization. Drazin and
Van de Ven (1985) found that of the alternative
definitions of fit, only systems fit explained per-
formance in their large sample of state employ-
ment security workgroups Van de Ven and
Drazin (1985) recommend the summary mea-
sure of Euclidean distance (ED) as the most
appropriate operationalization of systems fit.
Following their lead, we selected the work-
group that performed the highest on each out-
come measure. That Workgroup became the
benchmark for each test of cross-sectional out-
comes.5
Misfit, then, is measured in the jth work-
group’s Euclidean distance, (ED,, from the
* Govindarajan (1988) also found that systems fit measures explained self-assessed performance by business unit managers
in a number of large companies. Gresov (1989) took the systems approach one step further by investigating multiple
contingencies (i.e. multiple contexts). In addition to the context dimension of task uncertainty, he tested for the effects of
horizontal dependence (i.e. dependencies within workgroups) and the interaction of these two context variables. He
proposed that when the Workgroup faces conflicting contexts, the result may be unavoidable design misfit and lower
performance. He tested his approach on the data from Drazin and Van de Ven (1985) and did find support for his theory
that Workgroup design is affected by multiple contingencies and that conflicting contexts are associated with suboptimal
Workgroup performance.
5 From here on, we drop the terms i deal or opti mal and use the term benchmark since the notion of ideal is so arbitrary.
One might construct an hypothetically ideal Workgroup that has the highest independent variable scores. There is no
indication that such a Workgroup would be feasible since tradeoffs among the independent variables may be required in a
particular workplace.
CONTINGENCY THEORY 671
benchmark Workgroup’s organizational profile
along each dimension (structure, process, con-
text),’ Xi*:
Euclidean Distance: EDi =(& (Xi* - Xc)2)1’2.
RESEARCH SITE
Our test site was a major manufacturing divi-
sion of a Fortune 500 company. Before collect-
ing any systematic data, three months of
negotiations were required to establish both
entry to the division and the scope of the
study. Once there was a general agreement,
we spent parts of approximately 20 days over
six months inside the firm gathering data, inter-
viewing employees, and learning about the
manufacturing processes, product lines, organi-
zational culture, and new control information.’
The physical location of the host division
was in one of many large buildings clustered
around the firm’s international headquarters.
The firm competes in an industry where suc-
cess depends on cost, reliability, innovation,
and speed-to-market. Deficiencies in reliability
and speed-to-market nearly led to the firm’s
demise five years earlier, and management
resolved to achieve significant improvements
to regain competitiveness. Top management
identified JIT and TQC manufacturing pro-
cesses as the best chance to effect the desired
changes throughout its operations. As a starting
point the firm selected its manufacturing opera-
tions division, which performs final assembly
and shipment of the firm’s major electronic
products, to be the first to install new sys-
tems. Management believed that implementa-
tion should begin at the point where
customer orders were satisfied and changes
eventually would permeate backwards
through the entire organization.
The division had approximately 800 employ-
ees divided into 31 direct labor (operator)
workgroups and 11 indirect labor (engineer-
ing, production support, etc.) workgroups,
each with its own first-level manager. Each
Workgroup (direct and indirect) was responsi-
ble for some stage of one of the assembly pro-
cesses. In addition, there were second-level
managers with responsibility for each of the
major assembly processes, incorporating sev-
eral stages. For example, one of the major
assembly processes supervised by a second-
level manager was assembling all circuit
boards for the final products. Workgroups
who had responsibilities for basic circuit-board
assembly or end-of-assembly testing accom-
plished this process. Different workgroups
assembled each of the major product lines,
and other workgroups were responsible for
final testing, packaging, and shipping. Other
workgroups were solely responsible for
rework of defective products pulled out during
test phases.
Each Workgroup began its shift with a team
meeting to discuss assembly schedules for the
day. In addition, the Workgroup reviewed any
feedback from production control or engineer-
ing regarding earlier work. This feedback,
which was directed originally to the first-level
manager dealt with maintenance, defectives,
design, and process issues. Posted in promi-
nent places was graphical and tabular evidence
regarding the division’s recent performance on
several key factors (discussed below), but there
were relatively few displays for workgroup-
level performance. The operators on average
were relatively low-paid and not highly edu-
cated. At JIT/ TQC implementation, the divi-
sion spent several days training employees
and considerably longer reconfiguring the
work-flow to conform with JIT/ TQC.
The division assembled multiple variations of
” Each dimension is measured as the mean for the Workgroup; that is, this analysis is at the Workgroup level.
’ This site was used for another study (see Young and Selto, 1993), but the instruments used to collect data on the
independent variables were different from that of the previous study.
672 F. H. SELTO et al
three major electronic products. Several work-
groups assembled all circuit boards on the
same line in response to final product orders.
Though operators performed quality testing as
they assembled boards, other workgroups per-
formed extensive end-of-assembly testing.
Completed circuit boards then channeled into
several assembly lines or cells where other
workgroups completed product assembly.
After further environmental-stress testing, sev-
eral workgroups installed covers on (or boxed)
completed products and moved them to ship-
ping bays. The division’s final step was placing
completed products in shipping cartons at the
shipping bay.
The division was constantly experimenting
with different assembly configurations, as
some products phased out, as others were com-
ing on-line, and as new processes were intro-
duced. Change (or “turmoil”, as one manager
termed it) seemed the rule rather than the
exception. Whether this rate of change indi-
cated continuous improvement was a matter
of some internal debate. For instance, the
firm had dramatically improved a key external
quality indicator, overall mean time to field fail-
ure; however, the firm still had a long way to
go to meet industry standards. Regardless, it
appeared that managers were continually revis-
ing Workgroup responsibilities. This made it
difficult for the control system to keep up,
especially since revised control information
always seemed to be an afterthought.
Only after a year from the start of the JIT/
TQC manufacturing process did the firm initi-
ate changes to the division’s accounting sys-
tem, and only within the past several years
did the system produce regular reports. At
the time of the study the accounting system
produced both cycle-time and costing reports
on a weekly and monthly basis. The typical
cycle-time or cost report distributed to the divi-
sion manager was a reproduction of a large
spreadsheet, filled to the margins with col-
umns and rows of numbers. Columns were
comparisons of actuals to budgets - for the
period and year-to-date. Rows corresponded
to major assembly processes, aggregated
across a number of work groups. At least one
engineering technician in the division was
responsible for transforming information from
these reports and other sources into bar charts,
pie charts, and control charts as requested for
monthly management meetings. However,
none except the cost accounting staff showed
much facility with these reports. Production
and quality control personnel monitored other
types of control information, and produced at
least monthly compilations in the form of
graphs, also for monthly meetings. Except at
a few operator locations, there was very little
real-time control information in this division,
in contrast to the daily (and usually more
often) revisions to production schedules and
kanban orders.
The 3 1 direct labor workgroups and 11 indir-
ect labor workgroups faced a similar environ-
mental context and generated similar,
objective performance measures, Note that
only direct labor workgroups had objective per-
formance measures at this site. We expected,
therefore, that this test site would yield suff-
cient data (i.e. 31 observations of appropriate
variables) to test various forms of contingency
theory. The Workgroup performance informa-
tion available to us includes: (1) cycIe time;
(2) cost; (3) yield; (4) defects; (5) schedule
adherence; (6) process problems; and (7) engi-
neering changes. The company maintains this
information on critical success factors outside
the financial accounting system, devoting a sig-
nificant amount of time and effort to data collec-
tion and maintenance. This set of performance
measures is consistent with the firm’s strategy
to be a low cost, high quality, quick response
manufacturer. These measures also are consis-
tent with the objectives of JIT/ TQC manufactur-
ing processes (for example, Chase and
Aquilano, 1993), which are to improve effec-
tiveness and efficiency through reducing
defects, increasing yields, and meeting custo-
mer expectations for quality and shipments.
We did have numerous opportunities to
informally interview members of this division
and the level of discord was palpable. Direct
laborers would speak to us only if their man-
CONTINGENCY THEORY
673
agers were not present. Engineers made depre-
cating remarks about the competence and fore-
sight of management, particularly with regard
to relationships with suppliers; and managers
insisted that operators needed more discipline
and appreciation of their jobs in a depressed
local economy. We also observed a number of
departures from textbook descriptions of JIT/
TQC. For example, operators had little or no
authority to identify and fix manufacturing pro-
cess problems; instead they relied on engineer-
ing and quality control staff. These and other
examples led us to expect that conflicts both
within and among workgroups, between work-
groups and managers and staff, and between
workgroups’ assigned objectives and manage-
ment processes were commonplace. Thus,
Workgroup conflict may be a serious problem
for this organization. This organization had
implemented JIT and TQC systems, and both
of these require that the operators be empow-
ered to make decisions about the work flow.
Interviews and surveys of managers and
operators to determine the organizatonal struc-
ture, context, technology, and management
controls at a field site, which was one division
of a large electronics manufacturer were con-
ducted. The ability of alternative theories of fit
to explain Workgroup performance within the
division was then tested. We found consistent
evidence that the organization we studied con-
tained a significant inconsistency in organiza-
tional design, which has impeded attainment
of its goals. That is, the managers exerted
almost complete control of Workgroup tasks.
In addition, we found significant levels of
intragroup conflict. There was additional sup-
port for hypothesized effects on Workgroup
performance from many variables representing
structure, context, and management controls.
MEASUREMENT OF CONTINGENCY
CONSTRUCTS
Van de Ven and Ferry (1980) developed and
tested the organizational assessment instru-
ment (OAI), which measures contingency the-
OV
constructs. The OAI assesses the
dimensions of organizational context, struc-
ture, and control as perceived by individual
members of the organization. Though the OAI
was not specifically designed to assess fit in
modern manufacturing firms, it was a relatively
straightforward task to adapt it to reflect the
terminology and characteristics of JIT and
TQC and the specifics of the research site.
We pre-tested the revised OAI on two groups
of first- and second-level managers and on two
groups of direct laborers (called operators at
our site) to insure that we meaningfully modi-
fied the survey instrument for the local condi-
tions. The pretest resulted in rewording, some
additions to reflect the complexities of the divi-
sion, and some shortening to remove several
questions senior managers found objection-
able (i.e. we were not allowed to ask about
incentives). The full set of contingency tests
is described in Table 3. Since our study is direc-
ted at workgroups within one division, envir-
onmental variables were not examined (Hayes,
1977).
We administered the revised OAI (available
on request) in person on a single day in Decem-
ber 1990, to 406 direct labor operators and 19
managers at our test site. There were 81 ques-
tions on the worker form and 92 questions on
the manager form. On average, respondents
spent 45 minutes responding. Anonymity was
guaranteed to operators and confidentiality to
managers. Constructs were developed from the
questions as designed by Van de Ven and Ferry
(1980). Descriptive statistics for these mea-
sures are listed in Table 4, along with their
reliabilities tested with Cronbach’s alpha. Vari-
able measures by Workgroup are available in
the Appendix.
The OAI as designed by Van de Ven and
Ferry contains up to 42 contingency constructs
plus several other, related constructs added for
this study (for example, job dependence on
cost accounting). Factor analysis revealed that
some of the original items loaded onto con-
structs as designed, but a relatively large num-
ber either did not load convincingly as
designed or loaded with other, intendedly
674 F. H. SELTO ei al.
TABLE 4. Contingency variables: descriutive statistics
Cronbach’s
alpha operators Managers Standard
(N = 406) (N = 19) N Mean deviation
Outcomes:
Job Satisfaction
Workgroup Effectiveness
Organization Structure:
Worker Authority
Standardization
Management Control Process:
Horizontal Communication
Vertical Communication
Workgroup Conflict
Manufacturing Context:
Task Difficulty/Variability
Job Dependency on Supervisor
Job Dependency on Workgroup
Covariates:
Engineering Changes
Process Changes
0.81 0.69 384 3.15 0.82
n/a 0.77 14 0.03 3.18
0.84 0.80 343 2.94 0.79
0.86 0.90 332 3.36 0.58
0.75 0.82
344 2.58 0.63
0.81 0.71 363 2.07 0.79
0.77 0.72 371 2.21 1.08
0.72 0.81 357 1.81 0.53
0.70 0.86 347 3.19 0.74
0.78 0.62 399 3.39 1.00
19 12.11 6.88
19 1.37 1.44
unrelated items. For the most part, we could
not know whether these unforeseen loadings
were systematic or random occurrences.
Therefore, we conservatively retained for this
study nine constructs whose items loaded as
designed and met minimum reliability levels
(including job satisfaction); that is, Cronbach’s
alphas greater than 0.60. We did reason that
observed common loadings of items for task
difficulty and task variability could represent a
credible common factor, which we named task
difficulty/ variability.
Nothing in the literature suggested that item
mortality from the OAI would be so high. Dra-
zin and Van de Ven (1985) did not report simi-
lar difficulties with the OAI and reported
Cronbach’s alphas for the nine OAI constructs
they used of 0.40-0.85. Though the OAl was
inefficient in this study, it did generate a suffi-
cient number of reliable independent and
dependent measures. Tentative explanations
for OAI performance in this study are:
1. In retrospect, expanding several constructs
related to Workgroup interactions with sev-
eral functional groups (for example,
accounting, logistics) unnecessarily Iength-
ened the OAI. We had thought these inter-
actions could be important effects on
outcomes. The increased length and per-
haps irrelevant items may have reduced sub-
jects’ attention to their responses.
2. The subject pool on average may not have
the educational level necessary to respond
to the complex and lengthy questionnaire.
Because managers selected the operators for
pretests, a random sample of subjects did
not scrutinize the early versions of the ques-
tionnaire. Thus we were not alerted to
potential communication problems for a
proportion (about 10%) of the workforce
for whom English is a second language.
In this paper, we use objective measures of
JIT/ TQC performance as our dependent perfor-
mance measure. Objective performance data
were available for up to 19 workgroups from
the fourth quarter (October-December) of
1990. All objective data as reported to us
were designed, gathered, and reported consis-
tently across workgroups, either by the central
cost accounting staff or by the production con-
trol staff. Cost efficiency, operationally defined
as the ratio of standard cost to actual cost,
CONTINGENCY THEORY 675
ranged from 0.67481 to 1.80270 for the work-
groups.8 Defects, defined as the number of
defects per time period to units tested, ranged
from 0.0413 to 0.3550. Yield, the number of
good units to total units per month, varied
from 0.6880 to 0.9876. All of these measures
showed, to us, surprisingly large variations for
a JIT/ TQC facility, perhaps because implemen-
tation was incomplete or because the measures
were faulty. Schedule, the proportion of times
actually met schedule to budgeted times met
schedule, ranged from 0.93 to 1.03, after two
outliers measured at 0.10 were eliminated. The
much narrower range on this variable indicates
possibly more attention paid to schedule adher-
ence than other dimensions of performance.
The fewer than expected observations of
objective performance measures, unfortu-
nately, limit the statistical power of some of
our tests. All in all, the firm’s performance mea-
surement system was much less advanced than
we expected. Factor analysis of the standar-
dized performance measures determined that
cost efficiency, yield, defects, and schedule
measured a common factor. Using factor
weightings we combined these into a summary
measure called Workgroup effecti veness.’ The
OAl includes the self-reported outcome mea-
sure,job sati sfacti on. Since little empirical evi-
dence links job satisfaction with performance,
we used job satisfaction as another dependent
variable.
Two of the archival measures were covari-
ates in the various tests since they could affect
outcomes but were not controlled by work-
groups. These were the number of engineering
changes (EC) and the number of problems
identified with the process in the time period
(Process). Engineering changes were exogen-
ous to workgroups and attempt to improve
the manufacturability of the product or to cor-
rect field-performance defects. These usually
were results of quality assurance testing or cus-
tomer complaints or field failures. Surprisingly
(to us) workgroups also did not have authority
to correct process problems at this site but
relied on a staff of production engineers who
responded on call to accommodate machine
failures, stoppages, or an EC requiring process
changes.
RESULTS
We consider each of the contingency tests in
turn, beginning with the selection approaches.
Sel ecti on tests
The selection approach predicts correlations
of context variables with organizational struc-
ture or control variables. The results of testing
the selection approaches at the i ndi vi dual
l evel are in Table 5. The data do not point to
any differences in fit solely due to managerial
discretion, so that distinction is dropped. Eight
of the 14 significant correlations (57%) are evi-
dence of misfit; that is, their signs are incorrect.
So much misfit indicates potentially serious
coordination and communication problems
within and across workgroups, inconsistent
with JIT/ TQC. A JIT/ TQC shop should be
smooth-running and well-coordinated, but the
data tell a different story.
Mi sfi t and j i t wi th organi zati on structure.
Surprisingly, worker authori ty is not corre-
lated with any context variables. In a JIT/ TQC
shop one expects workers to have high job
authority for difficult/ variable tasks, but they
perceive none at this site. Likewise one
expects worker authori ty to have a negative
’ We did not use cycle-time efficiency, the ratio of standard to actual cycle time, because its range, from 0.451 to 2.498,
suggested serious measurement problems. Nearly all tests using cycle-time efficiency were statistically insignificant,
perhaps due to measurement error. In some workgroups, cycle time was measured using barcodes, but in others measure-
ment was visual and haphazard.
9 We also ran regressions for alI tests using each of the performance measures as a dependent measure. The results of these
numerous tests were similar to those reported here.
676 F. H. SELTO et al
TABLE 5. Selection auuroach - correlation coefficients
Context variables
subject to
management Task difficulty/ Job dependency Job dependency
Contingency variables discretion Variability on supervisor on Workgroup
Organization Structure:
Worker Authority Yes ns ns ns
Standardization Yes
_ .,,**.
.2y*** ,,..
:
Management Control System (process):
Vertical Communication Yes ns .35*** .ll”
Horizontal Communication Yes .25*” 18”’
:
-.12***
Workgroup Conflict No .27”* ns - .09’
Manufacturing Context:
Task Difficulty/Variability n/a
-
-.lO’ - .23***
Job Dependency on Supervisor n/a -.10*
-
18”’
;
Job Dependency on Workgroup n/a -.23*** 18”’
-
:
Covariates:
Engineering Changes ns ns ns
Process Changes ns ns ns
l =p
1
7
6
.
1
1
8
1
.
9
1
8
0
.
7
1
5
6
5
.
1
2
.
0
?
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
-
)
1
6
1
.
8
1
6
8
.
8
1
8
1
.
0
1
5
0
6
.
3
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
-
)
1
8
0
.
0
1
8
3
.
8
1
8
9
.
2
1
5
9
3
.
3
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
-
)
1
7
5
.
2
1
7
8
.
5
1
8
7
.
9
1
5
7
2
.
9
H
o
r
i
z
o
n
t
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
-
)
1
8
8
.
2
1
9
0
.
9
1
9
3
.
6
1
6
2
2
.
3
W
o
r
k
g
r
o
u
p
c
o
n
f
l
i
c
t
(
+
>
1
7
3
.
0
1
8
0
.
2
1
7
8
.
3
1
6
3
6
.
7
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
+
)
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
+
)
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
H
o
r
i
z
t
o
n
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
1
5
2
.
5
1
5
3
.
8
1
7
6
.
8
1
4
1
1
6
8
.
1
1
7
0
.
7
1
8
5
.
7
1
4
9
1
6
0
.
0
1
6
2
.
4
1
8
4
.
8
1
4
6
1
.
1
2
.
3
2
.
2
8
.
7
?
3
.
9
K
5
.
6
2
2
.
3
z
2
.
5
R
1
1
.
0
7
.
7
1
1
.
1
1
7
4
.
8
1
7
6
.
2
1
8
8
.
4
1
5
5
1
.
2
5
.
9
W
o
r
k
g
r
o
u
p
c
o
n
t
l
i
c
t
(
+
)
1
6
5
.
6
1
6
5
.
8
1
6
7
.
3
1
5
8
0
.
2
0
.
8
l
T
e
s
t
t
h
a
t
i
n
t
e
r
a
c
t
i
o
n
t
e
r
m
a
d
d
s
s
i
g
n
i
f
i
c
a
n
t
e
x
p
l
a
n
a
t
i
o
n
b
e
y
o
n
d
m
a
i
n
e
f
f
e
c
t
s
:
F
0
0
1
(
1
,
n
-
3
)
=
6
.
8
;
F
o
.
o
s
(
1
,
n
-
3
)
=
3
.
9
.
l
*
T
e
s
t
t
h
a
t
m
a
i
n
e
f
f
e
c
t
s
a
d
d
s
i
g
n
i
f
i
c
a
n
t
e
x
p
l
a
n
a
t
i
o
n
b
e
y
o
n
d
i
n
t
e
r
a
c
t
i
o
n
t
e
r
m
:
F
0
.
0
,
(
2
,
n
-
3
)
=
4
.
7
;
&
O
S
(
2
,
t
l
-
3
)
=
3
.
0
.
T
A
B
L
E
6
(
c
o
n
r
.
)
.
P
a
n
e
l
B
:
j
o
b
s
a
t
i
s
f
a
c
t
i
o
n
P
e
r
f
o
r
m
a
n
c
e
m
e
a
s
u
r
e
C
o
n
t
e
x
t
v
a
r
i
a
b
l
e
I
n
t
e
r
a
c
t
i
o
n
w
i
t
h
s
t
r
u
c
t
u
r
e
o
r
c
o
n
t
r
o
l
v
a
r
i
a
b
l
e
(
a
c
t
u
a
l
s
i
g
n
o
f
i
n
t
e
r
a
c
t
i
o
n
)
B
S
S
a
l
l
B
S
S
B
S
S
m
a
i
n
i
n
t
e
r
a
c
t
i
o
n
F
F
m
a
i
n
o
n
l
v
o
n
l
v
?
I
i
n
t
e
r
a
c
t
i
o
n
*
e
f
f
e
c
t
*
*
J
o
b
S
a
t
i
s
f
a
c
t
i
o
n
T
a
s
k
D
i
f
f
i
c
u
l
t
y
/
V
a
r
i
a
b
i
l
i
t
y
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
+
)
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
+
)
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
H
o
r
i
z
o
n
t
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
-
>
W
o
r
k
g
r
o
u
p
C
o
n
l
l
i
c
t
(
-
)
J
o
b
D
e
p
e
n
d
e
n
c
y
o
n
S
u
p
e
r
v
i
s
o
r
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
+
)
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
+
>
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
H
o
r
i
z
o
n
t
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
>
W
o
r
k
g
r
o
u
p
C
o
n
f
l
i
c
t
(
-
>
J
o
b
D
e
p
e
n
d
e
n
c
y
o
n
W
o
r
k
g
r
o
u
p
S
t
a
n
d
a
r
d
i
z
a
t
i
o
n
(
+
)
W
o
r
k
e
r
A
u
t
h
o
r
i
t
y
(
+
>
V
e
r
t
i
c
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
)
H
o
r
i
z
o
n
t
a
l
C
o
m
m
u
n
i
c
a
t
i
o
n
(
+
>
W
o
r
k
a
r
o
u
n
C
o
n
f
l
i
c
t
(
-
)
1
7
4
.
3
1
7
4
.
3
1
9
4
.
7
2
9
3
0
.
0
1
5
2
.
1
1
5
2
.
1
1
8
4
.
9
3
0
1
0
.
0
1
6
4
.
4
1
6
4
.
6
1
9
5
.
0
3
0
3
0
.
5
1
9
7
.
7
1
9
9
.
5
2
0
4
.
1
3
1
3
2
.
8
1
9
0
.
4
1
9
3
.
1
1
9
2
.
1
3
1
7
4
.
4
1
8
9
.
0
1
8
9
.
1
1
9
1
.
9
3
0
4
0
.
1
1
7
2
.
1
1
7
2
.
1
1
7
8
.
9
3
1
7
0
.
1
1
7
6
.
6
1
7
6
.
7
1
8
8
.
6
3
1
9
0
.
2
2
1
3
.
7
2
1
3
.
8
2
1
4
.
8
3
2
9
0
.
2
1
9
7
.
5
1
9
7
.
5
2
1
0
.
3
3
2
8
0
.
0
1
8
5
.
4
1
8
6
.
6
1
9
3
.
6
3
0
0
1
.
9
1
6
5
.
5
1
6
6
.
1
1
7
4
.
3
3
0
3
1
.
0
1
6
8
.
4
1
6
8
.
4
1
8
2
.
4
3
1
1
0
.
0
2
1
5
.
4
2
1
5
.
6
2
1
6
.
9
3
2
4
0
.
3
2
0
3
.
8
2
0
4
.
3
2
1
5
.
1
3
2
6
0
.
7
1
6
.
9
3
2
.
2
2
8
.
0
5
.
0
1
.
4
8
s
2
.
3
3
6
.
2
z
1
0
.
7
r
l
0
.
8
1
0
.
5
2
8
6
.
6
z
8
.
0
1
2
.
8
1
.
1
8
.
9
*
T
e
s
t
t
h
a
t
i
n
t
e
r
a
c
t
i
o
n
t
e
r
m
a
d
d
s
s
i
g
n
i
f
i
c
a
n
t
e
x
p
l
a
n
a
t
i
o
n
b
e
y
o
n
d
m
a
i
n
e
f
f
e
c
t
s
:
F
,
.
,
,
(
1
,
n
-
3
)
=
6
.
8
;
F
0
.
0
s
(
1
,
n
-
3
)
=
3
.
8
5
.
l
*
T
e
s
t
t
h
a
t
m
a
i
n
e
f
f
e
c
t
s
a
d
d
s
i
g
n
i
f
i
c
a
n
t
e
x
p
l
a
n
a
t
i
o
n
b
e
y
o
n
d
i
n
t
e
r
a
c
t
i
o
n
t
e
r
m
:
F
,
,
,
(
2
,
r
z
-
3
)
=
4
.
7
;
F
o
.
o
s
(
2
,
n
-
3
)
=
3
.
0
.
680 F. H. SELTO et al.
bility) interact with structure or control (work-
group conflict) to affect outcomes (job satisfac-
tion) as predicted (alpha = 0.05). In the few
cases where the interaction was significant,
main effects also were significant. It was far
more common that main effects rather than
interactions significantly explain outcomes.
Since the interaction approach tests fit in a
piecemeal manner and does not consider likely
tradeoffs among design elements, it may be too
simplistic to explain JIT/ TQC outcomes.
Systems tests
According to the systems approach to con-
tingency theory, the overall fit of context,
structure, and control explains the level of per-
formance. Misfit is measured as the overall
deviation from a benchmark configuration,
and the more a Workgroup departs from the
benchmark in any direction, the worse its per-
formance is expected to be. We used two mea-
sures of fit to test this theory, Euclidean
distances and absolute values of differences in
means on each dimension of fit between the
benchmark Workgroup and each other work-
group. Each approach has major drawbacks.
Computation of Euclidean distances ignores
quite reasonable and effective tradeoffs that
managers and workgroups may make that also
result in good performance. Using differences
in Workgroup means is an ad hoc approach
since each difference is implicitly hypothe-
sized to be an equally important, negative
impact on outcomes.
The benchmark workgroups for measuring
differences were:
Group 71° Highest performance for
Workgroup effectiveness.
Group 48 Highest level of job satisfac-
tion.
EDs were calculated for all the other work-
“’ We numbered the groups randomly.
I’ Workgroup effectiveness: R* = 0.02, F,,,, = 0.238.
Job satisfaction: R2 = 0.03, F,,,s = 0.555.
groups, and then outcome measures were
regressed on EDs. Neither of these regressions
was significant, which refutes this version of
contingency theory. ’ *
As a second test of the systems approach, for
each context, structure, and control variable
we regressed the absolute values of differences
of each Workgroup’s mean from the mean of
the benchmark Workgroup on job satisfaction.
This regression indicated that differences from
the benchmark explained levels of job satisfac-
tion at a significance of p = 0.005, adj-R* =
0.42. However, the signs of only several differ-
ences were negative, as predicted. The abso-
lute values of differences from Workgroup 7
also were regressed on Workgroup effective-
ness. These results were not significant, and
the degrees of freedom were relatively small.
Viewing these tests conservatively, we find lit-
tle support for the systems approach from
either form of test.
Summary of test results
Our tests of the selection approach reveal
interesting evidence of JIT/ TQC misfit using
the context variables of task dtj$?culty/varia-
bility and job dependency on supervisor or
Workgroup as correlates. These tests indicate
perhaps irreconcilable conflict between ele-
ments of context. The interaction and the sys-
tems approaches seek to explain performance,
but we found no support for the interaction
approach, either from regressions that explain
job satisfaction or Workgroup effectiveness.
We also found no support for the systems
approach. In this study, the simpler selection
approach was the most descriptive.
CONCLUSIONS
We attempted to respond to Hopwood’s
(1976) call to study management control issues
CONTINGENCY THEORY 681
by simultaneously considering organizational
structure, context, and processes. By adopting
a contingency theory framework, with its
numerous variables and multiple approaches,
and by focusing at the operator level of the
organization we had the raw material to
develop a plausible explanation of Workgroup
performance. The strongest results point to
internal inconsistency of organizational con-
text, perhaps regardless of management pro-
cesses. At our field site, many of the
management control measures and devices
thought to be consistent with JIT/ TQC were
in place. However, conflicts between opera-
tors’ needs for empowerment required by
JIT/ TQC systems and a management approach
better suited to mechanistic work quite prob-
ably negate beneficial influences of appropriate
controls. As indicated by the selection tests,
this firm has strong vertical dependence with
a management firmly in control, which is not
compatible with the concept of worker
empowerment. We believe this result indicates
the role of supervisors must be altered if work-
groups are to meet JIT/ TQC objectives at this
site.
It was disappointing to be unable to explain
Workgroup effectiveness satisfactorily with the
interaction and systems versions of contin-
gency theory after applying what we believed
to be powerful and reliable descriptive tools.
Retrospection yields some insights that may
inform future work, however. A key reduction
in power was due to the relatively few work-
groups with independent, objective measures
of performance. We had been led to believe
that the host company “managed by the num-
bers”, but if so the firm did not seem to have a
full set of information. This provides some evi-
dence that implementation of JIT/ TQC had not
been successful at this site. A company that
routinely generated periodic variance reports
had not made the transition to decentralized
information gathering and use. The company
reorganized the workforce into nominally
empowered workgroups, but they had neither
the information nor the authority to manage
themselves.
Another source of low power in the regres-
sion models is due to noise in the dependent
variable from measurement error in the compo-
nents of Workgroup effectiveness. In some
cases measurement errors would offset, but
there is no guarantee of that, particularly
within a single organization. We suspect con-
siderable error in all the measures, but were
especially surprised to find so much apparent
error in cycle-time measures, which we
decided not to use. The surprise was the result
of the build-up given to us by top management
that cycle time was the cornerstone of the divi-
sion’s new management control system, to be
replicated throughout the company. This mea-
sure would have to be improved to be useful.
In retrospect, a study of measurement pro-
blems and errors in assessing JIT/ TQC perfor-
mance would be a worthwhile contribution to
the manufacturing control literature. It is our
suspicion that managers within the division
and the firm have been making critical deci-
sions based on faulty JIT/ TQC performance
data. This may be a general problem and may
prevent full realization of the gains promised
by JIT/ TQC in other firms.
A third source of low power is the possibility
that the contingency theory used may not be as
descriptive as we would like to think. There
have been enough equivocal or negative con-
tingency results that can led to the question of
whether these can be attributed to uneven
implementation of the theory (a fourth source
of low power), or to narrow operationaliza-
tions of the theory. These either did not cap-
ture the richness of contingency theory and so
findings were minimal, or happened to focus
on the one or few aspects that came through
in a specific test.
As a final point, contingency theory does
have intuitive appeal as a method for testing
the fit of management controls, but it has weak-
nesses as well as strengths. One may obtain a
big picture of the organization using many fea-
sible predictor variables, but tight focus
depends on the empirical relationships that
emerge. Though significant relationships may
be found, these may not provide reliable
682 F. H. SELTO el al
guidance either for future research or for the
practitioner in the field attempting to improve
his or her process. Any change in circum-
stances may invalidate observed relationships.
This may be especially problematic in a manu-
facturing firm that is continuously revising pro-
ducts and manufacturing systems, in some
cases in subtle ways. Furthermore, we feel
there is a distinct possibility that both interac-
tion and systems implementations of contin-
gency theory are too simplistic to describe
performance in a complex workplace, espe-
cially at the Workgroup level. However, if we
had not tested multiple operationalizations of
contingency theory (as recommended by Van
de Ven and Drazin (1985)) we might not have
identified intragroup and structure conflicts as
the most likely impediments to superior perfor-
mance. Though most other investigators have
not found elementary contingency theories (for
example, selection approaches) to be informa-
tive, we found them to be crucial to under-
standing the dynamics of this workplace.
BIBLIOGRAPHY
Abemethy, M. & Stoelwinder, J., Budget Use, Task Uncertainty, System Goal Orientation, and Subunit
Performance: a Test of the “Fit” Hypothesis in Not-For-Profit Hospitals, Accounti ng, Organi zati ons
and Soci ety (1991) pp. 105-120.
Bleakley, F., The Best Laid Plans: Many Companies Try Management Fads Only To See Them Flop, Wal l
StreetJ ournal (6 July 1993) p. Al.
Chase, R. & Aquilano, N., Producti on and Operati ons Management (Homewood, IL: Irwin, 1993).
Chenhali, R. H. & Morris, D., The Impact of Structure, Environment, and Interdependence on the
Perceived Usefulness of Management Accounting Systems, 7I?e Accounti ng Revi ew (1986) pp. 16-35.
Drazin, R. & Van de Ven, A., Alternative Forms of Fit in Contingency Theory, Admi ni strati ve Sci ence
Quarterl y (1985) pp. 514-539.
Gordon, L. & Narayanan, V., Management Accounting Systems, Perceived Environmental Uncertainty and
Organization Structure: an Empirical Investigation, Accounti ng, Organi zati ons, and Soci ety (1984) pp.
33-47.
Govindarajan, V., Appropriateness of Accounting Data in Performance Evaluation: an Empirical Examina-
tion of Environmental Uncertainty as An Intervening Variable, Accountfng, Organi zati ons and Society
(1984) pp. 125-135.
Govindarajan, V., A Contingency Approach to Strategy Implementation at the Business-unit Level: Inte-
grating Administrative Mechanisms with Strategy, Academy ofManagemeniJournal(1988) pp. 826-
853.
Govindarajan, V. & Gupta, A., Linking Control Systems to Business Unit Strategy: Impact on Performance,
Accounti ng, Organi zati ons and Soci ety (1985) pp. 51-66
Gresov, C., Exploring Fit and Misfit with Multiple Contingencies, Admi ni stratfve Scfence QuarterZy
(1989) pp. 431-453.
Hayes, D., The Contingency Theory of Managerial Accounting, The Accounti ng Revi ew (1977) pp. 22-
39.
Hayes, R. H., Wheelwright, S. & Clark, K. B., DynamfcManufacturfng (New York: The Free Press, 1988).
Hopwood, A., Accounti ng and Human Behavi or (EngIewood CIifs, NJ: Prentice-Hall, 1976).
Jacobs, R., JIT More Than a Dying Fad? Wal l StreetJ ournal (18 October 1993) pp. 66-72.
Johnston, J., Economelric Methods (New York McGraw Hi& 1984).
ho, J. K. & Peterson, R. B., Effects of Task Difticulty and Interunit Interdependence on Information
Processing Systems, Academy of Management J outnai (1986) pp. 139-149.
Macintosh, N. & Daft, R., Management Control Systems and Departmental Interdependencies: an Empiri-
cal Study, Accounti ng, Organizations and Soci ety (1987) pp. 49-61.
March, R. M. & Mannari, H., Technology and Size as Determinants of the Organizational Structure of
Japanese Factories, Admi ni strati ve Sci ence Quartet@(1981) pp. 33-57.
CONTINGENCY THEORY
683
Otley, D., The Contingency Theory of Management Accounting: Achievement and Prognosis, Account-
ing, Organizations and Socteety (1980) pp. 413-428.
Perrow, C., A Framework for the Comparative Analysis of Organizations, American Sociological Review
(1967) pp. 194-208.
Schoonhoven, C. B., Problems with Contingency Theory: Testing Assumptions Hidden within the Ian-
guage of Contingency Theory, Administrative Scfence Quarterly (1982) pp. 349-377.
Schonberger, R. J., J apanese Manufacturing Techniques - Nine Hi&en Lessons in Simplicity (New
York: The Free Press, 1982).
Schonberger, R. J., World Class Manufacturfng: i’be Lessons of Stmpliciry Applied (New York: The Free
Press, 1986).
Schonberger, R. J., World Class Manufacrurtng Casebook: I mplementing J I T/TpC and /I T-TQC (Flew
York: The Free Press, 1987).
Tushman, M. L., Work Characteristics and Subunit Communication Structure: a Contingency Theory
Analysis, Admintstrartve Science Quarterly (1979) pp. 82-98.
Van de Ven, A. & Drazin, R., The Concept of Fit in Contingency Theory, Research in Organizational
Bebavtor (1985) pp. 333-365.
Van de Ven, A. & Ferry, D., Measuring and Assessing Organizarions (New York: Wiley, 1980).
Wilkinson, 9. & Oliver, N., Power, Control and the Kanban, J ournal of Management Studies (1989) pp.
47-58.
Young, S. M., A Framework for Research on Successful Adoption and Performance of Japanese Manu-
facturing Practices in the United States, Academy of Management Review (1992) pp. 677-700.
Young, S. M. & Selto, F., Explaining Crosssectional Workgroup Performance Differences in a JIT/TQC
Facility: a Critical Appraisal of a Field-based Study, J ournal of Management Accounting Research
(1993) pp. 300-326.
Young, S. M. & Selto, F., New Manufacturing Practices and Cost Management: a Review of the Literature
and Directions for Research, J ournal of Accounting Literature (1991) pp. 320-351.
684 F. H. SELTO et al.
Appendix
VariableMeans-By Workgroup
93 2. 65 [ O. ?? 0. 63 2. 87 I O. 531 3. 39 [ 060] 1. 78[ 0. 53] Z. M[ O. rZQ] 2. 26 11. 06] 1. 77 [ 0. 42] 2. 71[ 0. 59] 2. 52 CO. 771
97 3. 07 10971 n/ a 3. 17 I O. 461 3. 63 ( 0. 441 2. 44 I O. 851 2. 88[ 0. 51] 2. 58 [ I . 271 l . SO[ O. Cd] 3. 24[ 0. 38] 3. 30[ 1. 00]
*n/ a=AUc~~ponenl sof worl rgroupcf f ccl i venessnotavai l abl e
doc_683572303.pdf