-
HD28
.M414
J^eWEy
i^AUG 181988
ALFRED
P.
WORKING PAPER
SLOAN SCHOOL OF MANAGEMENT
The Concept of Fit in Strategy Research:
Towards Verbal and Statistical Correspondence
by
N.
Revised June 1988
October 1986
Venkatraman
WP
//
1830-86
MASSACHUSETTS
INSTITUTE OF TECHNOLOGY
50 MEMORIAL DRIVE
CAMBRIDGE, MASSACHUSETTS 02139
The Concept of Fit in Strategy Research:
Towards Verbal and Statistical Correspondence
by
N.
Revised June 1988
October 1986
Venkatraman
WP
//
1830-86
The Concept
of Fit in Strategy Research:
Towards Verbal and
N.
Statistical
Correspondence
VENKATRAMAN
Massachusetts Institute
of
Technology
E52-553, Sloan School of Management
Cambridge, MA
02139
(617) 253-5044
June 1988
Acknowledgements
:
expanded version of
Management Best Paper
Discussions during the seminars at the Kellog Graduate
This
a
is
my paper which appeared
Proceedings
,
1987.
significantly revised and
in
the
Academy
of
School of Business Administration, Northwestern University, Graduate School
of Business Administration,
New York University, and Sloan School of
Management, Massachusetts Institute of Technology were useful in preparing
this version. Gregory Dess, Oscar Hauptman, John Henderson, John Prescott,
Vasu Ramanujam, Harbir Singh, and Meera Venkatraman read previous drafts
and made useful comments and suggestions, but the usual disclaimer applies
.
MIT. LIBRARIES
AUG
1
^
388
RECEIVH)
|
I
I
The Concept of Fit in Strategy Research:
Towards Verbal and Statistical Correspondence
Abstract
management
Strategic
invoking
a
general
relationships
conceptualization
are
of
increasingly
fit'
specified
by
termed
as
(alternately
contingency, congruency, coalignment, consistency, etc.), but with inadequate
correspondence between
This
schemes.
paper develops
different perspectives
fit
as
gestalts;
implying
distinct
analytical
-- fit as
as
fit
This
correspondence between
it
a
statements
a
and
operationalization
conceptual framework and
moderation;
fit
meanings and
highlights
particular
the
identifies
six
as mediation; fit as matching;
and
profile-deviation;
theoretical
schemes.
scheme(s), but,
theoretical
fit
as
requiring
covariation
the
isomorphic
conceptualization
--
each
use of
specific
nature
of
and
testing
its
the
appears that researchers have used these perspectives
interchangeably, often invoking one perspective
and employing another within empirical
in
research.
the theoretical discussion
Since such
a
research
practice considerably weakens the critical link between theory-development
and theory-testing, this paper argues for increased explicit linkage between
the verbalization of theoretical propositions and their operational tests
strategic
management research.
in
Introduction
The concept
of
theory construction
served as an important building block for
has
'fit'
several areas of management research (Aldrich, 1979;
in
Fry & Smith, 1987; Galbraith, 1977; Katz & Kahn, 1966; Lawrence
1967;
Thompson; 1967; Van de Ven
strategic
Miles
&
£.
Drazin, 1985; Woodward, 1965) including
management research (Chakravarthy
Snow,
However,
Snow
1978;
major problem
a
t Miles,
in
1983;
Thus,
while
using
relationships
phrases
such
'consistent with' or more simply,
guidelines
at
observational
performing
are
level
empirical
convenient) mathematical form,
examining
the
validity
of
for
1984).
this concept lies in
postulate
to
'contingent
with,'
to
the
choose
implicitly
an
statistical
assumptions
and
precise
and
researchers
available
(often,
tests without really
underlying
1985;
'
upon,'
analytical
Consequently,
Van de Ven,
operationalization (see Drazin I
L Camillus,
theoreticians
statements
and perform
the
Osborn, 1981;
'congruence,' or 'coalignment,
provided.
often
in
£•
operationalization and statistical
'matched
verbal
seldom
tests
its
as
'fit,'
such
translating
1982; Jauch
Venkatraman
uncommon
not
is
it
,
testing theories rooted
the lack of corresponding schemes for
testing.
Lorsch,
L
their
choice
of
Schoonhoven, 1981
for exceptions)
Poor
operationalizations
[injvalidity"
fit-based
testing)
(Blalock,
rise
to
problems
of
"methodological
(Brinberg £ McGrath, 1982), and could seriously
of fit as a central
Specifically,
give
a
concept for theory construction
lack
of
and
weakens the
link
mathematical
its
between
and could be
research streams rooted
the use
strategic management.
correspondence between the conceptualization
propositions)
1965),
in
limit
in this
a
theory-building
major
concept.
formulation
and
(and
(of
statistical
theory-testing
reason for inconsistent
results
in
Indeed, Galbraith and Nathanson's
following observations nearly
the concept of
test
a
fit
a
useful one,
theory
Van
tone.
form
the
specifying
itself
is
central
strategic
Ven
de
of
both
to
(b)
domain
relationships);
of
and
operationalization of
fit
while
exist
the
conceptualizations
array
with
fit-based
of
(a) the
research
concept of
research
in
severely
is
several
general
Miller,
1982;
fit
and
relationships)
statistical
(i.e.,
(see especially,
neither address several issues of
on
to
1985).
and empirical
strategy
in
(verbalizing
theory-testing
(c)
attention
appropriate linkages between the conceptual
of
Joyce, Slocum, & Von Glinow,
focus
& Drazin,
discussions
role
its
theory-building
operational
inadequate
that
in
fundamentally alter the meaning of the
theoretical
management;
of
or not" (1979; p. 266).
it
developed on the following premise:
is
handicapped by the absence
domain
ha?;
argued
has
"although
lacks the precise definition needed to
it
could
fit
equally valid today:
(Van de Ven, 1979; Van de Ven
This paper
nor
is
and recognize whether an organization
similar
'fit
is
decade ago
a
testing
such
of
discussions
on
Drazin L Van de Ven,
1982;
the
the
1985;
Schoonhoven, 1981), they
particularly relevant to strategy research
testing
statistical
empirical
tests.
issues
Thus,
central
this
paper
linking
to
links
the
conceptualization of fit-based propositions with the statistical tests of such
relationships by highlighting that several alternate conceptualizations (with
differing theoretical content) of
fit
exist within strategic
that they imply specific operational and testing schemes.
attention
to
the
isomorphic
particular conceptualization and
nature
its
of
the
management, and
The aim
correspondence
is
to call
between
a
testing scheme(s), and to contribute to
the critical link between theory-building and theory-testing.
Alternative Perspectives of Fit
in
Strategic Management
Following Mohr (1982), strategic management research studies can be
divided
into
broad
two
paradigms
The former
'variance' paradigm.
--
the
paradigm;
'process'
and
the
characterized by the research studies of
is
Mintzberg (1978), Mintzberg and Waters (1985); Miles and Snow (1978), and
Burgelman (1983). The distinguishing characteristics
lie
their ability to
in
describe and explain the process of strategy formation (Mintzberg & Waters,
Snow, 1978), and internal corporate venturing (Burgelman,
1985;
Miles
1983)
rather than
&
explain
performance) using
variables (say,
variance
in
the dependent
paradigm and aim
set of theoretically-related independent,
a
growing number of studies subscribe
a
at
explain
reflecting
variations
both
(say,
explanatory
to the 'variance'
theory-testing through confirmation/disconfirmation of
specific array of hypotheses. For instance, several
to
variable
strategic choice variables).
contrast,
In
the
PIMS-based research seek
business performance through
in
strategic
resource
characteristics (see for instance,
a
deployments
as
set
a
well
of
variables
environmental
as
Hambrick, 1983; Prescott, 1986).
Within the variance paradigm, one can distinguish between the 'main
model'
effects
and
the
fit
model.'
The former seeks
variance explained by the chosen set of variables
explanatory power.
overall
In
contrast,
the
fit
model
explanatory power of an equation,
specifically
concerned with the general
paradigm that are specified
research
in
a
is
fit
fit.
maximize
the
terms of their collective
concerned with the
less
but the focus
significance of the impact of the specification of
is
in
to
is
more on the
The ensuing discussion
model'
multitude of ways
within the variance
in
empirical
strategy
A Classificatory Frampwork
reviewing
In
and
conceptualizations
tests
the specification of the functional form of
in
general form of congruency among
general
fit.
For example,
choice
specify and test the concept of
independent
of
to
fit
(and tests).
any criterion, thus reflecting
identify
of these
dimensions
perspectives
distinct
The former
latter
is
more
to
the
Some researchers
with explicit (and idiosyncratic) link to
while
performance or effectiveness),
power and usefulness
them
set of strategy variables.
anchoring the conceptualization
(e.g.,
it
between
fit
The other characteristic pertains
conceptualization.
its
of
criterion
a
more precisely understood functional form, while the
in
the
is
strategy and structure) and the specification of some
two variables (say,
a
in
fit
One
important to distinguish between the specification and tests of
invokes
of
relationship indicating the level of
specificity of the theoretical
of
precision
is
various
management research, two characteristics stand out
strategic
degree
the
is
of
others
view
it
a
as
more universal view. The
a
best demonstrated by employing
fit
as
they
apply
to
strategic
management research.
Figure
to
maps
six
(f)
follows:
fit
is
a
conceptual framework that employs the two dimensions
perspectives of
mediation; (c)
and
1
fit
fit.
as matching;
(d)
fit
are;
(a)
as gestalts;
fit
is
as moderation;
(b)
fit
as
(e) fit as profile-deviation;
The ensuing discussion
as covariation.
each perspective
These
is
discussed independently
issues of conceptualizations and statistical testing.
paper
in
is
organized as
terms of the central
Subsequently,
a
set of
issues pertaining to theory-construction are discussed to develop implications
for future research.
(Insert Figure
1
About Here)
Perspective One:
Fit as
Conceptualization
.
Moderation
A general axiom
is
superior irrespective of the environmental
that no strategy
universally
is
and/or organizational context.
Consequently, the contingency perspective has dominated our thinking (Hofer,
Steiner,
1976;
operationalized within
Schoonhoven:
by
Harrigan,
1979;
a
1983;
Ginsberg
Venkatraman,
t
1985),
moderation (i.e., interaction) perspective. As noted
"when
contingency
relationship between two variables
.
.
.
theorists
which predicts
assert
there
that
a third variable.
.
.
.
a
is
,
they
are stating that an interaction exists between the first two variables" (1981;
thus highlighting the popularity of the moderation perspective
p. 351),
in
organizational research.
Moderation
variable on
perspective
criterion variable
a
that
specifies
the
a
is
viewed either as
stages
product
of
predictor
a
a
between predictor and
fit
A moderator can be
significant determinant of performance.
a
of
fundamentally dependent on the level of
is
third variable, termed as the moderator. Thus, the
moderator
impact
categorical variable (e.g., different types of environment,
or organizational types)
cycle,
life
or
quantitative,
a
interval-level variable (e.g., degree of business-relatedness or competitive
intensity)
between
that
(e.g.,
moderator
is
a
direction
predictor variable (e.g.
a
variable
the
affects
performance).'
,
a
where
=
A formal
f(X,
Z,
a
is
is
a
relation
dependent
that
Z
is
X and
a
Y,
mathematically represented as:
X.Z)
contextual variable that
is
representation
the
the relationship between two (or more) variables, say
if
(1)
(for illustrative purposes),
Figure 2
of
strategy characteristic) and
function of the level of Z, which
Y
and/or the strength
fits'
Y
=
performance, X
=
strategy, and Z = the
with strategy for performance improvement.
schematic representation of this perspective.
(Insert Figure 2 About Here)
Analytical
Issues
Four
.
analytical
construction point of view are addressed:
(b) the role
moderation;
and impact
of main versus interaction effects;
from
important
issues
theory-
a
the form versus strength of
(a)
of multicollinearity
;
(c)
and (d) the requirement
the comparison
of partialling out
quadratic effects for testing the moderating effects.
specifying
In
fit
as moderation,
it
is
important to distinguish between
the form and the strength of moderation (Arnold,
because the mathematical representation
be tested through
(a)
(1)
This
1982).
important
is
can be specified such that
the subgroup analysis,
it
can
and/or (b) the moderated
regression analysis (Arnold, 1982; Sharma, Durand, & Gur-Arie, 1981) with
differing theoretical meanings and empirical support.
The subgroup
the
case,
sample
variable (Z)
test
.
focuses
analysis addresses the strength of moderation.
split
is
into
For example,
on
the
groups generally based on the contextual
Z represents different environments, then the
if
moderating
role
of
environment
performance relationships (Prescott, 1986). The hypothesis
in
strength
the
differences exist
of
in
In this
moderation
supported when
is
on
strategy--
of fit as reflected
statistically
significant
the value of correlation coefficients between strategy
and performance across the environmental groups'^.
The complementary moderated regression analysis (MRA)
more commonly used
for testing the form of moderation,
and
management research (e.g., Gupta
Govindarajan
1982;
t Palia,
Hitt,
Ireland,
L
&
Stadter,
is
is
1982;
,
appropriate
in
strategic
1984; 1986; Hitt, Ireland,
Prescott,
1986).
MRA
can be
some cases, it might make sense to split the sample on the basis
moderating role of X on the relationship between Z and
either case, the analytical issues are the same, although significant
in
of
X
to evaluate the
Y. In
differences are likely to exist
^
as
Arnold,
the theoretical conceptualization.
The differences in the correlation coefficients can be tested as a twhen there are two groups (see Bruning and Kintz, 1987; pp. 226-8);
(see
a chi-squared statistic for when there are multiple groups
statistic
and
in
1982;
p. 152).
represented through
set of two equations as follows:
a
Y
= aQ
^
a^X
*
a2Z
*
e
Y
=
*
a^X
*
a2Z
*
a3X.Z
aQ
The moderation
(2)
e
*
hypothesis
(3)
supported
is
coefficient 33 differs significantly from zero,
between (X) and (Z) on performance,
the
if
unstandardized
attesting to the effects of
fit
(Y).
The differences between the form and strength
of moderation are best
by Cohen and Cohen (1975):
illustrated
"Investigators should be aware, however, that the questions answered
these comparisons [of regression coefficients and correlation
coefficients] are not the same. Comparisons of correlations answer the
question "does X account for as much variation in Y in group E as in
group F?" comparisons of regression coefficients answer the question
"does a change in X make the same amount of score difference in Y
in group E as in group F?" (1975: p. 66).
by
Given that
(Arnold,
1982),
particular data set
a
conceptualization
critical
is
it
researchers
that
explicitly
not the other
articulate
and justify their choice
moderation,
of
may support one and
of
their
analytical
technique (subgroup analysis versus moderated regression analysis) to ensure
correspondence between their theoretical propositions and
Recognizing this distinction,
moderating
Prescott
(1986)
statistical tests.
evaluated the nature of the
environment, and concluded that environments moderate
role of
the strength,
but not the form of strategy- -performance relationship. The
implication
that
fit-based
is
is
it
(simply
relationship
subgroup analysis
no longer adequate to theorize
as
in
a
general form of
MRA
and
testing
the
terms of moderation) and treat
substitutable
analytical
schemes
for
proposition
The second
multicollinearity
analytical
--
a
issue
statis.tical
pertains
problem
to
that
the
arises
role
and
impact
of
when correlations
between independent variables are extremely 'high,' producing large standard
errors
of
regression
coefficients,
and
unstable
coefficients.
This
is
a
relevant issue for estimatinc equation (3) since the cross-product term (X.Z)
is
be strongly correlated with X, and Z. Some researchers explicitly
likely to
invoke multicollinearity as
perspective (see Dewar
we
While
measurements,
of origin
the
ex istence
moderation
has been shown that
it
reduces the level
of
unstandardized
is
it
1985).
posed
by
not problematic for
For
effects.
interval-level
simple transformation of the scale
a
(X,Z)"^.
Such
transformation does not
a
although
affect the test for the unstandardized coefficient a3,
for
moderation
a
correlation between the cross-product term
(X.Z) and the original variables
tests
within
problems
estimation
statistical
of
fit
Drazin t Van de Ven,
1979;
important to recognize that
is
it
the
establishing
Werbel,
L
recognize
multicollinearity,
reason for not testing
a
coefficients,
and
a-],
as
32
it
well
affects the
as
all
the
standardized coefficients (see Southwood, 1978; pp. 1166-1167; 1199-1201).
Thus,
contrary
multicollinearity
1985; p. 519),
also
Allison,
it
to
the
problems
is
a valid
1977;
for
call
(Dewar
&
I
MRA
for
Werbel,
1979;
Drazin
method for testing
analytical
Arnold
rejecting
Evans,
1979).
The
fit
testing
&
fit
due
to
Van de Ven,
as moderation (see
implication
is
that
if
researchers conceptualize their theories within the moderation perspective,
they should not dismiss the use
of
MRA
for multicollinearity reasons without
evaluating this important property for interval-level measurement.
The third
(fit)
issue pertains to the comparison of main versus interaction
For example,
effects within the moderation perspective.
may be interested
in
the
performance
as
additional
as
well
effects
of
chosen
effects
of
business-unit
the
fit
a
researcher
strategy
on
between strategy and
administrative characteristics, such as structure (Chandler, 1962), systems
Southwood (1978; p. 1198) demonstrates that 3ny .equation (X^, X,2,
X3 =Xi Xo ), where X^ =
X^
X3=Xi .X2) can .be transformed into one (X-j
(X^*c), and X2 = (X2*k); that is the two equations express the same surface
in the three dimensions (Y,X^, and X2), except for a shift in the axes of X^,
and X2 However, if the variables are measured using ratio scales, such a
transformation alters the meaning of the measures and can not be applied.
•^
,
,
.
and Vancil,
(Lorange
shown
for
in
a-]
or
rrunagerial
traits
(Gupta,
be accomplished
not
note
and
given
the
Thus, within
an
at
interval-level
arbitrary
a
interval-level measurements),
of
measurement.
As
can not be accomplished along with tests
testing for a3=
2,
aj'
measurement.
scale
origin
of
for
interval-level
moderated regression analysis (with
typical
one can test for the existence of interaction
not compare the relative influence of main and interaction
effects but can
effects because the standardized coefficients are meaningless.
is
Such
1984).
require the joint assessment of main and interaction effects,
theorizations
which can
1978),
However,
if
it
possible to develop ratio scales for X, and Z, then one can compare their
relative effects (Allison, 1977;
implication
is
that
if
Southwood, 1978; Friedrich, 1982). Thus, the
theory
the
for
calls
comparison
a
may be
interaction effects, the moderation perspective
main
of
and
limited to those cases
that accommodate ratio-levels of measurements.
The fourth
issue focuses on
the requirement of partialling out the
quadratic effects of the original variables (X) and (Z) for establishing the
presence (or, absence) of multiplicative effects
to be
sure that the relationship
curvilinearity,
and X.Z
it
is
is
.
Southwood (1978) argued that
one of interaction rather than parabolic
necessary to control by using the terms, X, Z,
simultaneously
.
A
test
of
the
partial
correlation
and
9
Z'^
Z^
between the
dependent variable (Y) and (X.Z), after partialling out the effects
9
X'',
X'^,
of X, Z,
provides support for multiplicative interaction, while tests of the
partial correlation coefficients
the effects of X,
Z,
(i)
between (Y) and (X^), after partialling out
X.Z., and Z^ and
partialling out the effects of X,
Z,
(ii)
between (Y) and (Z^), after
X.Z., and X^ provide evidence of the
occurrence of curvilinearity. A review
of strategy
and organizational research
adopting the moderation perspective indicates that extant research has not
incorporated such controls, which weakens the interpretations.
that future attempts at modeling
fit
It
is
hoped
as moderation reflects the need to rule
10
out
the
explanation
rival
when establishing the existence
variables
Limi tations
at
fit
curvilinear effects
plausible
of
The use
.
the
of
original
multiplicative effects.
of joint,
of the moderation perspective of
fit
is
limited in
two ways. One concerns the inability to separate the existence of
least
from
the
framework
effects
of
involves
--
equation (3)
which
interaction term)
is
the
assessment
the
either
that
Recall
fit.
the
of
significance
an indication of the effect of
on the criterion variable,
Y
--
within
tests
statistical
of
this
in
33
(depicted by the
fit
or the difference
in
the
correlation coefficients involving the dependent variable, (Y) across various
groups defined
moderation
intricately
by either
with
is
tied
(X)
or
(Z).
reference
to
a
the
to
specific
Since the
variable,
criterion
performance
the
performance effects
of
incidence, or, proportion) of
but
fit,
fit in
instead
one
if
on
its
the
is
of
fit
as
meaning
is
may
and
variable,
generalizable to other performance measures. Thus,
in
specification
be
not
not interested
existence
(e.g.,
different sub-samples, this perspective
is
inappropriate.
The second
meanings
pertains
limitation
inability
sets
where strategy
measured using
is
captured through
a
set of
theoretical
content.
environment
is
However,
m
may not
in
a
set of n
reference to
in
a
the
study
variables and environment
the
terms of
moderating
a
set of
effects
is
of
strategy
and
(nXm) interaction variables.
implied
by
the
set
of
components may not adequately represent the nature
1985).
multiple sets of
collectively reflect the moderating effects because
relationships
implying an error of
(a)
theoretical
variables, with each variable reflecting specific
Then,
specified
this set
theoretically,
can be illustrated with
interactions
of
attach
The problem inherent
interactions and (b) higher-order interactions.
multiple
to
especially involving
interaction terms,
to the
the
logical typing'
individual
interaction
of systemic interaction,
(Bateson, 1979; Van de Ven
&
Drazin,
11
by Jauch, Osborn, and Glueck (1980), which
Let us consider the study
examined the financial performance implications
fit
using
multiplicative (i.e.
a
a
they did) or note,
the
(as
fit
more appropriately, that the empirical results do not
support the hypothesis when tested using
conclude
Should we interpret this result
.05 level.
performance effects of environment-strategy
of the
rejection
as a
fit
but none of the 72 possible interactions were
statistically significant at the p <
as
environment-strategy
moderation) perspective. They modeled
,
interaction terms,
set of 72
of the
latter,
one
then
a
should
moderation perspective.
go
further
question
to
we
If
the
appropriateness of this perspective to test the underlying theory. This
because one can argue that the moderation view of
multiplying
has
variables)
Jauch
a
individual
little
strategy
variables
and
individua
and hence,
theoretical basis,
(viewed
fit
it
environmental
l
not surprising that
is
Indeed,
careful review of the research literature arguing for environment--strategy
coalignment indicates that the proponents of this view (e.g.
Bourgeois,
1980;
metaphorically,
Miles
and Snow,
1978)
invoke the
,
Andrews, 1980;
notion
of
and hardly specify the specific functional form
multiplicative effects.
a
terms of
in
any significant multiplicative effects.
et al (1980) did not find
is
It
is
the empirical researcher
fit
of
rather
joint,
who has translated such
theoretical position into a moderation perspective for analysis.
The problem inherent
ambiguity
Although
attaching
in
it
in
higher
order
interactions
any clear theoretical meaning
has been shown
to
relate
these
to
the
terms.
1977) that the three-way interaction
(Allison,
terms as well as higher-order terms are not susceptible to multicollinearity
problems
theoretical
(for
in
measurements),
interval-level
inherent
the
limitation
is
nature rather than one of statistical estimation. This requires
that researchers have to be particularly persuasive
those terms than they would
Given these
in
in
attaching meaning to
the case of two-way interactions.
critical limitations,
it
is
important to recognize that as we
12
inappropriate linkage between the theoretical
strive to avoid the error of
and the empirical domain, we should move away from viewing moderation as
scheme
only
the
to
multiplicative terms
fit-based
conceptualize
in
and
fit,
treating
the
introduction
of
regression analysis as the only means of testing for
a
Towards
relationships.
end,
this
other
five
perspectives
are
discussed below.
Perspective Two:
Conceptualization
a significant
Mediation
Fit as
The mediation perspective
.
intervening mechanism (say, organizational structure) between
an independent variable (say,
Thus,
performance).
independent variable on
variable,
effects.
an
specifies the existence of
this
strategy) and the dependent variable (say,
moderation
while
dependent variable
a
varying
specifies
effects
as a function of the
an
of
moderating
perspective specifies the existence of intervening (indirect)
Stated differently, this accommodates the case of
independent
variable
well
as
as
a
combinatory,
a
direct effect of
indirect
on
effects
performance.
Strategy researchers have embraced the moderation perspective more
but the latter offers the capability to
often than the mediation perspective,
recognizes the differences
Our
position
is
the various stages of
in
that neither
is
a
system of relationships.
universally superior, and that the choice
is
to
be explicitly predicated on theoretical considerations. For example, Hambrick,
MacMillan,
and Day (1982) treated relative market share
as a
moderator
in
their analysis of the effectiveness of strategies across the four cells of a
business portfolio matrix
the
portfolio
matrix.
--
which
is
consistent with the theory underlying
Alternatively,
in
examining
the
nature
of
the
relationship (direct versus spurious) between market share and profitability,
Prescott, Kohli, and Venkatraman (1986) treated market share as
in
a
system
different
of
relationships
environmental
between
contexts.
Both
strategy
studies
and
a
mediator
profitability
generally
across
addressed
the
13
strategy -- market-share
fit
on performance using different theoretical and
analytical perspectives.
Similarly,
one
if
embracing the classical industrial organization (10)
is
economics paradigm (Scherer, 1980; Porter, 1981) of Structure --> Conduct
--> Performance to test the role
influencing the
any) of firm-level strategic actions
(if
in
between market structure characteristics and
relationship
firm performance, one could use both perspectives with differing theoretical
The mediation perspective decomposes the
meanings.
effects
market
of
structure characteristics on firm-level performance into direct effects (i.e.,
performance levels attributable
versus
indirect
effects
domain choice,
to
performance levels attributable
(i.e.,
navigation, or competitive strategies)
is
or corporate strategy)
.
useful to evaluate the differences
domain
to
In
contrast, the moderation perspective
in
the relationships between strategy
and performance across various environments (Prescott, 1986).
We view mediator
proportion
Z
formally,
the
of
a
is
relation
between
X
(i.e.,
where X,
Y=f[Z]),
(i.e.,
variable
a
that
the
Z=f[X]) and Y
Figure 3
as mediation involving three variables,
is
(say,
firm-level
relation
is
a
a
significant
Stated
criterion.
Y=f(X),
Z
if
is
a
probabilistic function of
performance),
schematic representation of
a
where, Z (say, market-share
acts as a mediating mechanism (fit) between
Y
and
for
and Y have different distinct theoretical
Z,
content (see Roseboom, 1956).
accounts
predictor
mediator of the probabilistic
probabilistic function of
Z
as
X
fit
level)
(say, strategic decisions) and
and supported by the following set
of
equations
Y
= ag
*
a^X
*
a2Z
Z = bQ
*
b^X
*
e
*
e
(4)
(5)
(Insert Figure 3 About Here)
Comparison
(4)
and
(5)
of Figures 2
reveal
important
and
3,
as well as equations (2)
distinctions
between
the
and
(3) with
two perspectives
14
thus
discussed
restrictive
While
far.
moderating variable,
and
variable
between
distinguishing
in
perspective. This
is
perspective
independent
the
generally
is
less-
and
variable
the
more precise distinction between the independent
a
mediating
the
moderation
the
variable
required
is
because the system of equations
the
mediation
critically
dependent
within
is
on the theoretical specification of the nature of relationships, especially the
ordering of the variables.
Analytical Issues
within
a
The
.
tests of
fit
as mediation
typically carried out
is
path-analytic framework (for details, refer sources such as: Alwin
£•
Hauser, 1972; Blalock, 1971; Duncan, 1971; Heise, 1975; Kenny, 1979). Forthe
two issues are important:
present discussion,
complete versus
impacts of
The
mediation;
partial
and
(b)
the distinction between
(a)
the test for the performance
fit.
on
issue
first
the
distinction
between complete versus partial
mediation has important theoretical implications for strategy research.
instance,
it
has
the
potential
address
to
the
dilemma
on
the
For
relative
importance of market-structure' factors and
strategic choice' decisions. Let
us consider Figure 3.
equation (4)
If
the coefficient
a-]
in
is
not statistically
different from zero, the strongest support for the mediating effects of Z
obtained.
This
implies
that
the
presence
transmission of effects of X on Y, and
model.
Thus,
testing
in
Structure --> Conduct
firm-conduct
(i.e.,
-->
is
of
propositions
theoretical
Performance paradigm,
strategy)
plays
a
is
is
necessary
critical
role
rooted
in
the
implication
in
translating
market structure
(i.e.,
implication
is
Z,
that partial effect
domain definition or choice,
implying
is
that
is
On the other hand,
and some indirect effect between X and Y through
The
the
Market
the
not zero, there exists some direct effect between
mediational model.
for
termed as the complete mediational
structure opportunities to firm-level performance.
coefficient a^
Z
is
market
if
the
X and Y,
a
partial
obtained due to
reflecting the famous
15
--
question
"what
our business?'
is
)
and partial effect due
decisions given the market structure characteristics (i.e.
performance
that firm
position
is
a
factors as well as strategic choice (e.g., Andrews,
& Hofer,
Schendel
1979).
Finally,
if
for firm-level conduct (i.e.,
given
important
attributable
issue
Thus,
fit.
dependent on the
that there
is
theories
the
seek
usefulness
of
test
specify
to
of
statistic
this
for
other words, while the test for 33
In
systematic
statistical
corresponding test for
evidence
for
as mediation.
fit
given that the impact of
a-]
is
fit
structural
Bourgeois, 1980;
much stronger
is
insignificant role
strategic choice).
availability
mediation).
1980;
the test for the performance effects of
most
that
to
is
function of
the direct effect,
than the indirect effects, the implication
The second
domain navigation,
"how do we compete?"). Such tests directly address
reflecting the impact of
the theoretical
,
strategic
to
is
in
moderating
However,
This
fit.
is
performance effects
perspective
the effects
is
of
partly
fit
(as
equation (3) provided
we
effects,
this issue
is
need
a
complicated
given by the product of two path-coefficients
Based on "Simon-Blalock decompositional technique (see Duncan,
(b-j.ai).
1972), the following equation can be specified'^:
''xy
~
where (r^y)
^1
is
*
^2'-'1
*
spurious effects
(6)
the observed zero-order correlation between
X and
Y^. The
relative proportion of the two effects -- indirect (32. b1) versus direct (al)--
provides an
index of the relative effect of
versus the direct effects. More formal,
through
the
a test of
mediator.
fit
(namely,
indirect effects)
statistical corroboration
is
provided
the statistical significance of the indirect effects through
Since
this
involves
testing
a
product of two
regression
See Prescott, Kohli, and Venkatraman (1986) for a strategy research
study that uses this technique for estimating the various coefficients.
It is useful to restrict this discussion to those cases, where the zeroorder correlation is statistically significant, and 'sufficiently large' to make
the decompositions worthwhile.
16
coefficients,
standard
a
t-test
cannot
we
Hence,
adopted.
be
an
use
approximation due to Sobel (1988) as follows;
t
where
=
a2
(ai.b2)/sqrt((b22.sea2^)
and
b-j
(a2.seb2^))
*
are as defined earlier, and (se) refers to the standard error
of estimates.
Before we conclude the discussion on this perspective,
to note that while the discussion
involving
three
variables
(X,
focused on
and
Z,
a
Y),
important
simple system of relationships
can
it
be
(see Duncan,
analytically extended to multiple variables
is
it
and
conceptually
1972;
Heise,
1975;
Kenny, 1979), although estimation and interpretation problems could be more
pronounced
Perspective Three:
Fit as
Conceptualization
Matching
perspective
This
.
conceptualizations that view
fit
is
invoked
those
for
as a theoretically defined
strategy
match between two
theoretically-related variables. A major point of departure from the previous
two perspectives
variable,
is
that
fit
is
specified
reference to
although one could subsequently examine
criterion variables. Stated differently,
is
without
developed independent
of
a
measure
its
of fit
a
effect on
criterion
a
set of
between two variables
any performance anchor, unlike the previous two
perspectives
Let us consider Chandler's (1962) classical thesis that a diversification
strategy requires
a
multi-divisional structure, while a geographical expansion
strategy requires field units; and that the absence of such
a
match leads to
administrative inefficiency. Thus, the measure of strategy-structure
be
derived
efficiency.
reference
without
Such
Yasai-Ardekani,
and
theoretical
a
operationalized within
any
to
a
particular
proposition
view
is
of
most
matching perspective (see Rumelt,
Al-Bazzaz,
1981),
and
fit
can
administrative
appropriately
1974;
independent of the
Grinyer,
specific
conceptualization of performance. Similarly, Chakravarthy (1987) defined
fit
17
within
use
planning system context as existing when the planning system
a
matches
required
the
system
ideal
to
whether
test
such
in
match
a
enhances system performance.
Analytical Issues
The matching perspective has been supported by
.
three somewhat related analytical schemes: (a) the deviation score analysis;
and
(b) the residual analysis;
the analysis of variance
(c)
The deviation score analysis
interested
is
based on
(ANOVA).
premise that
a
one
if
is
quantifying the degree of match between two variables, then
in
the absolute difference between the standardized scores of two variables
an indication of the lack of
fit
IX-ZI
earlier,
performance implications
of
fit
is
the three variable system discussed
tested
The forma!
performance.
variable on
in
an indication of the lack of
is
Alexander & Randolph,
(see for instance,
1985; Bourgeois, 1985). For instance,
is
fit
between X and
and the
Z,
by examining the impact
specification
of
of
equation
the
this
as
is
follows
Y
If
= ao
aiX
*
the coefficient a3
then
^
is
a2Z
*
agdX-ZI)"''
^
e
positive, and statistically significant in equation (7),
hypothesis of performance effects of
a
(7)
fit
is
supported.
This analysis has intuitive appeal, but the problems pertaining to the
difference scores
potential
(Johns,
unreliability
difference score IX-ZI
parts (X) and (Z);
variable -- even
spuriously
original
--
if
related
1981)
the
of
is
(b)
less
fit
are to be
recognized.
--
measure
since
These include
the
reliability
than the average reliability of
its
of
a
component
possibility of spurious association with an external
the difference score has acceptable reliability,
to
(a)
the
criterion
components (X) and (Z); and
variable
(c)
through
the
it
effects
may be
of
the
generally weak discriminant validity
given that the transformed variable may not be differentiable form
its
component
variables.
measure
within this perspective do not discuss their response to such
fit
Since
researchers
employing
difference
scores
to
18
issues,
results, but
it
is
the use of difference scores.
in
analysis
residual
operationalization of
subsequently
proponents,
the
related
criterion
to
universalistic and contingency predictions.
of
.
436).
p.
earlier
analysis
theory
is
in
approach
multiplicative
shown
But,
in
for
arguing
for
to
be
which can
According
(Y).
of
simultaneously
to
its
both
test
(and) one can obtain some idea
.
this
multicollinearity
paper
this
(1979)
by comparing the coefficients"
the magnitude of the different effects
(1979;
Werbel
to reflect fit,
variable,
advantages are "one can
main
and
method, the residuals from the regression
In this
fit.
Dewar
the
reflects
one variable (say, X) on the other (say, Z) are used
be
to the
hoped that future studies devote increased sensitivity
potential problems
The
research
impact on
possible to systematically assess the
not
is
it
they
approach,
ill-founded.
Our
appropriate to test fit-based relationships
position
--
if
the
have been
which
--
reasons
reject
that
is
(a) the
this
underlying
conceptualized as deviation, and (b) one can demonstrate that the
is
limitations
of
deviation
are
scores
inappropriate to conceptualize
not
as
fit
serious.
It
is,
however,
moderation and then reject
clearly
MRA
and
favor the residual analysis, because they are not directly interchangeable--
given the fundamental differences
in
the underlying theoretical positions.
However, the residual approach has the following
Werbel,
1979;
1982) that are to be recognized:
Miller,
choosing an appropriate base
confounding
(c)
of error
variance
line
in
both reflect the
fit
of
X on
Z,
residuals,
in
Z.
At
this analysis should discuss the relevance
&
the problems of
and (d) the need
a
(Dewar
(b) the
addition to measurement error;
and the residuals
between X and
(a)
model to calculate the residuals;
the arithmetic sign of the residuals;
between the residuals
limitations
of Z
to distinguish
on X, although they
minimum, researchers employing
and implications
of these issues in
their specific research context.
The third
analytical approach within the matching perspective
is
the
19
analysis of variance
(ANOVA). This scheme
interaction
but Joyce,
effects,
useful tests of
between
typically viewed as tests for
and Von Glinow (1982) developed
They distinguish
reflecting the matching perspective.
fit
forms
three
Slocum,
is
of
--
congruence)
(or,
fit
and
general,
effect,
functional -- where the general congruency model hypothesizes interaction
effects,
but emphasizes the similarity and matching levels of independent
A
variables.
competing perspectives
tested within
advantage of
parenthetical
a
framework (see Joyce
analytical
The three perspectives thus
far share one
that they are appropriate for specifying bivariate
various
forms
functional
perspective
has
the
scheme
analytical
moderation versus matching) of
(e.g.,
common
this
between
potential
two
to
that
is
fit
can be
1982).
et.al.,
common feature, namely
fit
(i.e.,
specified
fit
although
variables),
accommodate
the
second
system
larger
a
in
of
relationships. In contrast, the next three perspectives are appropriate for the
specification
and testing
among multiple variables simultaneously.
of fit
Perspective Four: Fit as Gestalts
Conceptualization
The
.
d'etre for this perspective
raison
from Van de Ven's (1979) articulation of one of the meanings of
derived
is
fit
"that
as:
characteristics of environmental niches and organizational forms (that) must
be joined together
description
together
in
causation
is
of
a
particular configuration to achieve completeness
a
in
system
social
ways
certain
implied
.
.
.
to
and
a
--
like
obtain
a
pieces
of
complete
a
puzzle
image.
conceptual explanation of
a
must be put
Here,
"fit"
in
no
direct
found
is
in
a
hierarchical theory on the holistic nature of an open social system." (1979;
p. 323).
A
similar call
is
made by
Miller
(1981),
who reacted against the
dominance of bivariate contingency perspectives and called for
a
movement
towards the development of gestalts.
The need
on
the
for a gestalt view
plausibility
of
internal
is
highlighted by Child's note of caution
inconsistencies
(or,
mutually
conflicting
20
directiois)
He
among multiple pairwise contingencies.
happens when
a
problem may be found
partial solution to this
which
viewed
be
can
perspective through
The
in
Miller's
at
linear
as
extension
logical
a
which are
in
a
temporary state
a
number
and ordered about the pattern
provide
is
bivariate
"
insights
useful
into
(1981;
'feasible sets of internally consistent
Analytical Issues.
need attention, namely
The analysis
(a)
to
to
p. 5).
Along
represent
There
find
set
a
.
situations
is
something
.
....attributes (1977; p. 264), and
of
powerful
a
.
1984).
best described
of balance, the
of gestalts
frt
few variables or
a
and Friesen noted: "Archetypes appear
which are described seem to form
could
fit
"Instead of looking at
frequently recurring clusters of attributes or gestalts
holistic
the
of
among such variables we should be trying
associations
A
p. 175).
multi-tiered taxonomical approach (Hambrick,
a
(1981) terms as follows:
similar lines, Miller
(1975;
the development of gestalts
in
role of gestalts in reflecting the multivariate
of relationships
each
configuration of different contingencies are found,
having distinctive implications for organizational design?'
--
"What
remarked.
concept
of
--
equifinality
and equally effective configurations."
of fit as gestalts raises
two issues that
the descriptive validity of the gestalts; and (b)
the predictive validity of gestalts.
The descriptive
interpretable
especially
in
validity
requires
that
gestalts
terms of the theoretical positions implied by
important
since
most
analytical
development are inductive, which has given
as a
the
schemes
available
rise to this activity
be
obtained
fit.
for
This
is
gestalt
being called
fishing exercise.' Miller (1981) argues that these gestalts are "relatively
both
few and very different from one another,
and relationships among, variables," but
it
is
in
terms of the scores
of,
important to underscore that
the types and characters of the gestalts are sample-specific, which require
that
external
validity
generalizability to
a
criteria
(e.g.,
cluster
stability;
robustness
and
different sample or population) should be satisfied.
21
Further, there exists
between the treatment
subtle, but nevertheless, important distinction
a
of fit as gestalts
and strategy taxonomies. The
latter
represents an empirical identification of naturally-occurring strategy types
(Galbraith
t
Schendel,
Hambrick,
1985;
1983;
Miller
Friesen,
&
1984)--
which do not invoke the concept of internal congruence, except indirectly to
name the
inquiry
gestalts.
is
coverage.
The
selection of the underlying variables for taxonomic
guided by the need to balance parsimony and exhaustiveness of
In
contrast, the specification testing of
more careful enumeration
fit
as gestalts requires a
of the underlying dimensions
such that the gestalts
can be ordered along their levels of
and Snow's (1978) theory
consistencies
internal
engineering;
or congruence.
fit
of strategic adaptation
is
For example. Miles
rooted
among three problem domains
and administrative) and
is
it
in
the notion of
(entrepreneurial;
the pattern of coalignment that
results in the four strategy types (prospectors, defenders, analyzers, and
reactors).
A rigorous
problem domains as
in
a
test
of
this
theory requires the use of the three
basis to develop four gestalts that can be interpreted
the light of their theory.
This ensures
a
tighter linkage between the
underlying theory, choice of dimensions, as well as the interpretations of the
clusters
typology
in
is
the light of the theory. Thus, although the Miles and
a
popular one,
it
is
unfortunate that
its
use
in
Snow
(1978)
empirical research
studies has not yet been explicitly predicated on the pattern of
fit
among
the three dimensions.
The predictive
validity
is
important
given
our
interest
in
(a)
establishing performance implications of congruency; and/or (b) demonstrating
the existence of generic strategy types or multiple configurations of equally
successful strategies (i.e., equifinality)
.
For instance, Etzioni argued that
"Congruent (organizational) types are more effective than incongruent types"
(1961; p. 14, emphasis added); and Child reported that the high performing
organizations had internally consistent structural configurations, which were
22
absent
poorer performi ig organizations.
in
directly tested within
gestalt perspective,
a
performance variable(s)
Hambrick
is
to incorporate
we concur with
suggestion of identifying subsamples of high- and low- performing
s
businesses to identify profiles of
fit
develop distinct profiles of
to (a)
the challenge
For this task,
the analysis
into
such propositions are to be
If
within each subsample. This enables one
across the performance categories; and
fit
(b) assess the possibility of discovering patterns of equifinality within low-
and high- performing businesses. Such an approach enabled Hambrick (1983)
and
and
Miller
unsuccessful
Friesen
terms of
the conceptualization of
supported by an array
to test the
among
perspective views
,
and
The
analysis.'
role
following case:
(say,
in
a
fit
While
.
set
of
is
is
ably
numerical taxonomy.
The
to carefully articulate the theoretical
is
not an end
let
and use
gestalts
variables
degree
as the
akin to what
is
of
itself,
in
but
of
the
reflect
adherence
to
performance,
Conversely,
a
such
viewed as
a
that
form
adherence
of this
perspective
is
a
then
multi-dimensional profile
its
taxonomy,
the
high
deviation
this
to an externally-specified
as
pattern
best introduced through the
we can specify an
particular environment,
internal
of
Van de Ven and Drazin (1985) term
us suppose that
given
the
is
degree
ideal strategy profile
terms of the level of resource deployments along
dimensions) for
in
Profile-Deviation
Fit as
Conceptualization
profile
powerful
is
theory of multivariate congruence.
Perspective Five:
congruence
gestalts
congruence and employ appropriate numerical methods
internal
such that the analytical exercise
means
as
fit
methods
analytical
of
challenge for strategy researchers
position of
and
'successful'
nature of coalignment, and
ability to retain the holistic
its
generic
isolate
to
gestalts.
summary,
In
(1978)
a
set of strategy
a
business unit
will
s
degree
of
be positively related to
environment- -strategy coalignment.
level
of
from
this
profile
implies
a
weakness
in
23
environment- -strategy coalignment,
performance. Figure
six
4
a
is
resulting
negative
a
in
on
effect
schematic representation of this perspective using
dimensions of strategy.
(illustrative)
This perspective
for testing
useful
is
the effects of environment--
strategy coalignment (Andrews, 1980; Bourgeois, 1980; Hofer and Schendel,
1978; Miles and Snow,
1978)
environment-specific
in
the sense that deviations
profile
ideal'
should
be
strategy from an
in
negatively
related
to
performance. Venkatraman and Prescott (1988) argued that this perspective
best
the
reflects
proposition
theoretical
coalignment,
environment-'Strategy
and
of
performance
employed
it
on
effects
of
sample
of
a
PIMS-based businesses across eight environments over two different timeperiods.
since
They found strong, and consistent support
deviations
in
resource
environment-specific
allocation
were
profiles
patterns
strongly
for their hypotheses,
corresponding
from
(negatively)
related
to
performance.
(Insert Figure 4 About Here)
Analytical
Issues
conceptualization
development
of
Three
.
fit
as
critical
issues
profile-deviation
of the 'ideal profile,'
dimensions; and (c) the need for
a
interplay
and
its
between
tests:
(a)
the
the
(b) differential weights for the multiple
baseline model to assess the power of the
test.
The
issue focuses
first
profile that serves
businesses
in
a
as the
study.
on the approach to the development of the
benchmark for calibrating the strategies
Two obvious
choices
exist.
One
is
a
specification of the profile along a set of dimensions that can be
be most appropriate for
a
of the
theoretical
argued
to
particular environment. While intuitively appealing,
the operational task of developing such
a profile
with numerical scores along
24
a
set of dimensions
difficult^.
is
The other option
Van de Van,
to empirically
is
Ferry,
1985;
calibration sample. In
a
develop this profile (see Drazin L
Venkatraman and Prescott, 1988) using
1979;
a
strategy research context, Venkatraman and Prescott
(1988) developed their profile of ideal strategic resource deployments using
the mean scores of
in
calibration sample -- defined as the businesses that
a
the top 10% on the performance scale within an environment
--
subsequently not used for the study sample. Given the importance
at
a
valid
sample,
ideal
profile'
jackknifing
for
purposes,
calibration
as
it
is
important to explore the
different
well
as
proportions
subsample
use of
replications
to
that was
of arriving
hold-out
used
for
increase
the
sample
the
of
lie
robustness of the profile specified for calibration purposes.
The second
issue
concerns the development
of
multi-dimensional
a
profile with either equal weights to the dimensions or differentially weighing
the dimensions based on their relative importance to the context. Drazin and
Van de Ven
(1985)
Drazin
that
note
introducing
deviation
we argue
.
.
the
in
adopted the unweighted scheme, but Van de Ven and
the
assumption
possibility
of
equal
of
weights
weighing
differentially
determining performance.
"
(1985; p. 351
that the assumption of equal weights to
dimensions
is
all
)
.
be
"can
the
relaxed
importance
by
of
For strategy research,
the underlying strategy
untenable since an effective package of strategic
resource
deployments should reflect differential emphasis depending on the importance
of
a
particular dimension to the particular environment.
Adopting such
a
view, Venkatraman and Prescott (1988) derived the weighing scheme from the
beta
weights
of
the
regression
equations
using
variables on performance for each environment
in
the
array
of
strategy
the calibration sample.
" For example. Porter (1980) provides a conceptual base for identifying
the ideal strategies for different generic environments, but it is not possible
to translate these verbal statements into numerical scores of ideal resource
deployments for each environment.
25
The third
issue pertains to the power of the test, which requires the
specification of a baseline model to demonstrate that the predictive
measure
the
of
coalignment
(calculated
as
issue
and
ability to discriminate
For this
variables.
compared against
a
This
profile.
by drawing an analogy
explained
where the power
analysis -its
best
is
significantly better than
is
random
as deviation from a
among certain groups using
a
is
non-trivial
demonstrated by
is
a set of
discriminating
the ciassificatory accuracy of the model
'chance' or
naive' model (Morrison, 1969).
and Drazin's (1985) pattern analysis,
In
negative and significant correlation
a
demonstrate the performance effects of
unknown,
it
weakens the
is
Van de ven
between the coalignment measure and the criterion serves
coefficient
a
discriminant
the
to
of the discriminant function
purpose,
of
weighted euclidean distance
a
using the profile of the calibration sample)
measure calculated
power
Since the power of their test
fit.
interpretive
ability
of
the
empirical
However, more powerful tests that discount other plausible
rival
to
is
result.
hypothesis
(e.g., one that argues that deviations from any random profile would exhibit
equally strong and significant negative correlation coefficient) are needed as
we begin
to
Development
deviation
is
employ
of
schemes
this
.
externally-specified
is
a
in
in
empirical
strategy
the power of modeling
assess
fit
research.
as
profile-
Covariation
Fit as
Conceptualization
reflected
to
perspective
an important area of inquiry.
Perspective Six:
an
this
While
fit
profile,
as profile-deviation signifies
this
perspective
the pattern of covariation among
response to an oft-repeated
call to
a
views
adherence
coalignment
set of dimensions.
to
as
Indeed
develop notions of meqastrateqy
(Mintzberg, 1978) or to treat strategy as "a pattern or stream of major and
minor decisions" (Miles & Snow,
'pattern
of
decisions'
are
best
1978).
The notions
represented
as
of
'megastrategy' and
covariation
among the
constituent dimensions rather than as any of the earlier perspectives (with
26
the possible exception of gestalts).
explication
the
requires
dimensions. General linear models
like
have
central thread or internal logic
a
among the various independent variables"
The concept
of
megastrategy
(1980;
p. 571).
predicated on the notion of internal
is
strategic resource deployments (Mintzberg, 1978) and
in
operationalized
as
among the various dimensions
covariation
deployments. The theoretical support for such
a
view
covariation
of
among
constituent
the
differs from the fourth perspective
It
sense that
has the capability to specify the notion of
it
resource
of
intelligence
dimensions
intelligence.
best
is
derived from Weeks'
is
view of operationaiizing the notion of general
pattern
the
but may indicate no apparent logical linkage
significance,
statistical
(1980)
among
linkage
logical
strategy" (Hambrick, 1980) as the "regression coefficient may
a
consistency
internal consistency
the regression analysis are of limited
use given that they miss "the concept of
underlying
because,
is
underlying
the
of
This
(fit
human
of
as gestalts)
fit
the
as
the
in
theoretically (as
an unobservable construct) rather than derived empirically as an output of
methods.
taxonomic
numerical
approach
appropriate
in
appropriate
in
the
of
fit
as
it
follows
opposed
deductive
a
to the inductive
cluster analysis. Thus, while the fourth perspective
early
theories
as
differently,
and testing
to the specification
approach reflected
Stated
stages
get
theory
of
refined
development,
this
is
is
more
by successively building from prior
theorizing and empirical results.
Analytical Issues
testing
of
as
fit
.
Two
critical
covariation.
issues emerge
These
are:
confirmatory approach to the specification of
existence and impact of
The
age old
fit;
the
the specification and
exploratory
versus
and (b) the tests
is
in
social
rooted
in
of the
fit.
distinction between exploratory and confirmatory approach
one
covariation
(a)
in
science
research.
The operationalization
the basic principles of factor analysis
--
of
is
an
fit
as
that seeks
27
among
to explain covariation
set of indicators
a
terms of
in
a
among the
factors (i.e., first-order factors) and explain the covariation
order factors
terms
in
second-order factor(s).
of
specified as a second-order factor,
as
Fit
smaller set of
first-
covariation
is
where the first-order factors represent
the dimensions to be coaligned (see Venkatraman, 1986).
Given
the
choice
basic
in
exploratory
versus
confirmatory
specification, the pattern of covariation can be modeled either as exploratory
(EFA)
analysis
factor
specification
analysis
factor
analysis
second-order factor within EFA involves
of
with
confirmatory
or
oblique
rotations
with
correlations
a
(CFA)
.
The
common factor
among the first-order
factors used to examine the possibility of second-order factors (see Cattell,
1978; Chapter 9 for details)
1979)
seeks
test
to
one
.
In
(or,
contrast, the
more)
CFA
(see Joreskog and Sorbom,
theoretically-specified
factor(s) against the observed data. Thus, unlike
EFA
second-order
that uses predefined
mathematical criteria to determine the optimal rotation which may result
CFA
uninterpretable theoretical meaning, the
evaluate
provides
basis to explicitly
a
second-order factor model against an alternative model
a
that may, for example,
Given
the
in
--
one
specify only inter-correlated first-order factors.
relative
benefits
of
CFA over EFA
for
theory
testing
purposes (see Mulaik, 1972; Joreskog and Sorbom, 1979; Bagozzi, 1980), and
the need to distinguish this perspective from other inductive perspectives
(e.g., gestalts as
in
modeling
fit
in
perspective four), the use of
as covariation.
Figure 5
is
a
CFA
is
preferred over EFA
schematic representation of this
perspective with two models that are pitted against each other. Figure 5(A)
specifies the base model of direct effects of four dimensions on a criterion,
which
is
limited
covariation
specifies
in
terms
of
among the four dimensions.
the
covariation
any
providing
In
insights
contrast.
into
in
impact
of
Figure 5(B) explicitly
among the four dimensions
internally-consistent business strategy, which
the
as
reflecting
an
turn has an effect on the
28
The coalignment among them
criterion.
unobservabie theoretical construct
functional dimensions
but derives
at
formally
is
specified
an
as
higher plane than the individual
a
This construct has no directly observable indicators,
meaning through the first-order factors that are directly
its
operationalized using observable indicators reflecting the resource allocation
patterns to the functions (say, indicative of functional strategy).
(Insert Figure 5 About Here)
The second
covariation.
issue
pertains
Without getting
into
to
testing
the details
LISREL framework (readers are directed
L
to
the
of
impact of
fit
model estimation within
used:
(a)
a
comparison
a
Bagozzi & Phillips, 1982; Joreskog
Sorbom, 1978; 1979), we note that three complementary test
to be
modeled as
statistics are
of the coefficients of determination;
(b) the
calculation of the Target Coefficient, T (Marsh & Hocevar, 1985); and (c) the
statistical
significance of the
(Venkatraman
second-order factor loadings
1986).
Within CFA,
the other)
it
is
if
the models are nested (namely that one
possible to use the difference
in
is
a
chi-squared statistic (which
Since the
follows a chi-squared distribution) for assessing model superiority.
two models (5a
b 5b)
are not nested,
a
chi-square difference test criterion
(Joreskog, 1971) cannot be employed here. Thus,
the model superiority
is
subset of
a
preliminary indication of
provided by the comparison
of the coefficients of
determination of the two models (analogous to R^). A more formal treatment
requires
the
calculation
of
Target
Coefficient
(T).
Following
Hocevar (1985), we note that the second order factor model
to
explain
the
among
covariation
the
first-order
is
Marsh and
merely trying
factors
in
a
more
parsimonious way. Consequently, even when the second order factor model
able to explain effectively the covariation
goodness
of fit
among the
first
is
order factors, the
can never exceed that of the first order factor model. Marsh
and Hocevar (1985) propose
a
Target coefficient (T) that reflects the ratio
29
of
chi-square of
the
the
first-order
model
chi-square
the
to
of
the
second-order model, and has an upperbound of 1.00. A value close to 1.00
can be used to support the covariation model (5b)
in
preference to the main
effects model (5a), and can be interpreted similar to the Bentler and Bonett
(1980) delta index, until the distributional properties of (T) are established.
The third
criteria
is
to assess the statistical significance of the four
loadings on the second-order factor, termed as coalignment. In addition, the
significance of the effect of coalignment on performance provides
statistical
a direct test of
the performance impact of
Towards Achieving Verbal and
The Concept
These
in
Table
1.
is
It
illustrative
intended to highlight the fundamental differences
theoretical
testing
well
is
as
proposition,
the
illustrative
as
number
more than
references.
in
themes
the verbalization of an
fit,
the nature of measurement of
fit,
fit,
General Metaphor
a
perspectives are compared along several key characteristics
specification of
concept
Beyond
the underlying conceptualization of
such as:
Correspondence
Statistical
of Fit in Strategy Research:
six
viewed as covariation.
fit
fit,
Thus,
of
variables
in
the
analytical schemes for
it
is
clear that the
general metaphor, but has specific theoretical and
a
analytical meanings.
(Insert Table
Future theorizing
in
away from treating
fit
as
Van de Ven noted
in
his
a
strategic
1
About Here)
management
is
dependent on
a
movement
general metaphor that has universal applicability.
review of Aldrich's (1979) book:
"Central to the population ecology model (as well as to contingency
theories in general) is the proposition that organizational forms must
fit their environmental niches if they are to survive.
There are at
least four different conceptual meanings of "fit" in this proposition,
each of which significantly alters the essence of the theory" (1979; p.
322).
As
this
paper has argued, the same criticism applies equally
well to
strategy propositions and/or theoretical statements that have invoked
many
fit at a
30
metaphorical
rather
with
level,
Bourgeois, 1980; Chakravarthy
Recognizing
1983).
by
accompanied
1982; Jauch L
,
interdependence
the
phrases
theory-testing,
guidance
little
guidelines
descriptive
Osborn, 1981; Snow
between
"congruence",
like
operationalization
at
I Miles,
and
theory-building
or "alignment" should be
"fit
that,
(see
at
minimum,
specify
their
specific functional forms(s).
We need
be
used
move beyond treating
to
signify
to
array
wide
a
as a general
fit
meanings
of
congruency, matching, interactive effects,
scheme
classificatory
perspectives of
well
as
specific
classification
discussions,
fit
--
(Figure
each implying
as
contingency,
This paper has provided
1)
rooted
in
six
and that future work
a
different
distinct theoretical position as
a
serve
least
at
Table
such
frameworks. The expectation
set of analytical
would
and
1
etc.
metaphor that can
as
a
common
reference
is
that this
point
for
exhibit increased sensitivity to the
will
widely-differing theoretical and empirical meanings of
fit.
Explicit Justification of the Specification of Fit
A corollary
to the
above issue
justify their specification of
is
within
fit
Random and convenient choice
of
a
a
call
for researchers to explicitly
particular research context.
statistical
methods should be abandoned
with an explicit recognition that the research results are sensitive to this
choice.
link
As we expect that strategy researchers routinely demonstrate the
between conceptual definition and operational measures (Venkatraman &
Grant, 1986),
it
should be that researchers justify not only the measures of
their constructs,
but also their specific scheme of
fit
between (or among)
the constructs.
Specifically,
this
requires that empirical research studies discuss the
rationale for their specification and demonstrate the correspondence between
theory and statistical method(s)
developed their hypothesis
in
.
For example, Gupta and Govindarajan (1984)
such
a
way
that they reflected the perspective
31
of
perspective of
be
fit
in
well
argue that the theoretical
between strategy and managerial characteristics could also
operationalized
invoke
However, one could
moderation.
as
fit
within
matching
a
perspective
(which
they
implicitly
their theoretical discussions supporting their hypotheses), as well
as a mediational perspective (namely, managerial characteristics facilitate the
translation
of
strategic
and
decisions
empirical results would have been richer
choices
by
if
to
performance).
Their
they had argued for their choice
explicitly pitching the moderation perspective against relevant alternative
perspectives.
A related question pertains
study that adopts one
of the
to
whether the results
an empirical
of
perspectives (without discounting other relevant
ones) are sensitive to the particular operationalization scheme.
is
the
empirical
strategy-manager
fit
demonstration
of
the
performance
of
can we generalize the
'difference-score' approach?
if
we
test the
same
the matching and mediational perspectives.
linkage using say,
Similarly,
impacts
reported by Gupta and Govindarajan (1984) generalizable
beyond the moderation perspective? What would happen
theoretical
For example,
results
of
Bourgeois
(1985)
Also, would Chakravarthy's (1987)
beyond the
results on
the absence of performance impacts of tailoring strategic planning systems to
the context hold up
when the contingent theory
be tested within the perspective
Such questions are
theoretical
(Denzin,
using
relationships
1978;
Jick,
of profile-deviation (i.e.
the spirit of
in
1979),
planning system design
of
triangulation'
multiple
measures
and are relevant since
perspective four)?
,
for testing
that
calls
and
multiple
a
methods
general theoretical
perspective of performance effects of structure-context
fit
was supported
using some analytical perspectives and not others (see Drazin and Van de
Ven, 1985; Joyce
et.
al.,
1982). Similarly, the theory of performance impacts
of environment- -strategy coalignment
reflecting
the
strength
of
was supported using sub-group analysis,
moderation
(Prescott,
1986),
and with
fit
as
32
(Venkatraman
profile-deviation
and
Prescott,
moderated regression perspective (Jauch
et
.
use of multiple specifications of
extant literature
fit
unequivocal
is
area.
this
in
preference for the
a
and Slocum (1984)
Tosi
suggesting guidelines for contingency theory research noted:
techniques have frustrated researchers
being
effects
Researchers
modeled
need
to
because
technique
utility
technique using the same data set " (1984;
p.
of
16;
has
each
"Statistical
these
of
"
.
.
(and) Examining multiple approaches to
and relating the findings
development
of
and
Van
de
of
fit
as
contingency studies
fit in
unique sample characteristics can greatly aid the
mid-range theories
(1985; pp. 358-360).
Drazin
to
statistical
studies should
:
be designed to permit comparative evaluation of as many forms
.
biases.
implied
emphasis added). A similar
recommendation comes from Van de Ven and Drazin (1985)
possible.
in
attempts to test for the interaction
each
compare the
Indeed,
single empirical study.
a
a
A 'Triangulation-Trap?'
there exists
within
within
net
1980).
al.,
Multiple Specifications of Fit Within a Dataset:
In the best spirit of triangulation,
but
1988J,
of
what approach
to
applies where"
fit
Specifically, the use of multiple specifications enabled
Ven
to
(1985)
structural contingency theory that
uncover
may
insightful
nuances
of
the
not have been otherwise possible.
Philosophically, one can not quarrel with such recommendations.
But,
before we get carried away into testing our theoretical relationships with
multiple perspectives within
a
single research study,
we need
to evaluate the
appropriateness of the alternate perspectives to the theory and the research
design that guided the data collection exercise. Not
all
theoretical statements
are amenable to the six perspectives identified here. Similarly, measurement
schemes may
limit
the use of some perspectives.
Operationally, there exists
results
a
danger
across multiple statistical tests.
across multiple perspectives, evidence of
If
of not obtaining
convergence of
one obtains converging results
robustness'
is
provided. However,
33
the converse
is
Van de Ven
For example, given the same data set, Drazin and
not true.
reported
(1985)
that
but not an
'systems' approach,
their
was
theory
supported
interaction' approach.
such
In
a
using
a
we
case,
need to evaluate whether the results
reflect
competing theoretical perspectives or
fundamental nature of (mathematical
and
statistical)
Thus,
relationships
probability of accepting
fit
across
of
statistical tests?
given
is:
a
(random) data
set,
what
is
the
theoretical relationship of performance impacts of
a
subset of these perspectives that are comparable. Let us suppose
a
that three different
theory
among the
relevant question
a
a
support for the
differential
researchers are seeking to test the same underlying
performance impact
of fit
between two variables,
(X),
and (Z),
with three different perspectives: the moderation perspective, the matching
perspective, and the mediation perspective. Further
independently
obtain
pattern
similar
a
of
Z.
What
converge irrespective
of
the specific values
Answers
X,
Y,
this
to
between
relationships
independent
question
of
and
is
us suppose that they
correlations
and
variables,
let
among the three
the probability that their results would
explicate
the
in
the correlation
fundamental
nature
among these three perspectives
the specific theory being tested;
matrices?
of
the
that
are
and they provide useful
pointers on the appropriateness of generally pursuing multiple tests of
Until
we show
that there exists no
should be cautious
in
a
a
efficient
approach
to
a
one test, we
data-set, post-hoc
answer
this
question
.
is
Monte Carlo simulation design that accommodates varying patterns
of correlations
In
priori preference (bias) of
following triangulation within
An appropriate and
through
a
fit.
separate
among X, Y, and Z and performing the
technical
note
,
I
tests of convergence.
demonstrate that the probability
A separate note titled, "Testing the Concept of
Research: The Triangulation-trap" is under preparation.
Fit
in
of
Strategy
34
convergence across
mediation)
three
the
(Tosi and Slocum,
hoc attempts
(moderation,
and
matching,
implying that recommendations calling for post-
close to zero,
is
perspectives
triangulation
to assess
1984)
limited
of
is
use.
Longitudinal Operationalizations of Fit
Most
static,
within
(if
not
all)
perspectives discussed
cross-sectional
However,
dynamic and never-ending task,
a
moving target
but every organization
forces
issues
explaining
equilibrium
(fit);
are
is
as
and testing
of
Thompson noted, coalignment
whereby the organization
fit
is
a
continually
is
coalignment" (1967; p. 234). According to this
of
view, no organizational system
theoretical
paper have focused on
approaches to the specification
strategy research.
"shooting at
in this
is
in
moving towards
rooted
dynamic coalignment,
state of perfect
a
this
state.
the dynamic perspective of
in
interesting
Several
such as:
fit,
movement towards versus away from
the organizational
short-term benefits of
fit
versus long-term benefits of
versus long-term
fit;
While these
possible trade-offs between short-run
fit
are interesting theoretical issues,
unclear whether the six perspectives
it
is
fit.
identified here are most appropriate for testing such issues or not.
it
is
However,
important to note that this paper has not treated this issue,
recognizes
that
appropriate
a
promising
mechanisms
to
area
specify
research
of
and
test
is
fit
but
development
the
it
of
a
longitudinal
strategic
management
within
perspective.
Summary
The general
research.
concept
But, the progress
be limited unless we seek
of
in
fit
is
central
to
theory construction
to clarify the specific
in this
meaning
area
is
of fit as
likely to
used
in
various settings, develop strong linkage between verbalizations and statistical
testing schemes, and test theoretical relationships using multiple comparable
statistical
schemes where necessary and feasible.
Towards
this end,
this
35
paper
identified
management
gestalts;
--
as
fit
perspective,
fit
it
six
different
conceptualizations
as moderation;
fit
profile-deviation;
as
and
of
mediation;
fit
as
fit
fit
within
strategic
as matching;
covariation.
For
fit
as
each
discussed the various available statistical schemes to highlight
the isomorphic nature of the link between theory and testing schemes; and
demonstrated that the
statistical
methods are not freely interchangeable
without confounding the underlying theoretical meaning of
fit.
36
References
Aldrich, H. E. (1979) Organization and environments
Prentice Hall.
Alexander, J. W., and W. A. Randolph. (1985) The
and structure as a predictor of performance
Academy of Management Journal 28:844-859.
Englewood, N. J.:
.
fit
in
between technology
nursing subunits.
,
Testing for interaction in multiple regression.
P. D. (1977)
American Journal of Sociology 83 1 44 - 1 53
Allison,
:
,
Althauser, R.
models.
Chicago:
Multicollinearity and non-additive regression
(1971)
H. M. Blalock (Ed.) Causal models in the social sciences
P.
In
,
Aldine-Atherson
Alwin, D. F., and R. M. Hauser (1972) The decomposition of effects
analysis.
American Sociological Review 40, 37-47.
in
path
,
Anderson, C. R., and ZeithamI, C. P. (1984) Stages of product life cycle,
business strategy and business performance. Academv of Management
Journal
Andrews,
27:5-24.
,
R.
K.
(1980)
The concept
of corporate strategy
.
New York: Dow
Jones-Irwin
Moderator variables:
A clarification on conceptual,
(1982)
analytic and psychometric issues. Organizational Behavior and Human
Arnold, H.
Performance, 29:143-174.
Arnold, H. J., and Evans, M. G. (1979) Testing multiplicative models does
Organizational Behavior and Human
not require ratio scales.
Performance, 24:41-59.
Bagozzi,
R.
(1980)
P.
Causa! models
and Phillips,
Bagozzi, R. P.
organizational theories:
Quarterly 27:459-489.
in
marketing
Wiley.
Representing and testing
Administrative Science
L.
W.
A
holistic construal.
(1982)
New York:
.
,
Bateson, G.
(1979) Mind and Nature
,
New York, NY:
E.
P.
Dutton.
M., and Bonett, D. G. (1980) Significance tests and goodness
the analysis of covariance structures. Psychological Bulletin,
of
88:588-606.
Bentler,
P.
fit in
Theory building and the statistical concept of
Blalock, H. M., Jr. (1965)
interaction.
American Sociological Review 30:374-380.
,
H. M., Jr. (1969)
Theory construction: From verbal to
Prentice-Hall.
Englewood Cliffs, NJ
mathematical formulations
Blalock,
:
.
Bourgeois, L.
J.
Ill
integration.
Bourgeois,
L.
J.
economic
Ill
A conceptual
5:25-39.
Strategy and environment:
(1980)
Academv
of
(1985)
performance
Management Review
,
Strategic goals, perceived uncertainty, and
Academy of
volatile environments.
in
37
Management Journal
.
28:548-573.
Brinberg, D., and McGrath, J. E. (1982)
A network of validity concepts
within the research process. In D. Brinberg and L. H. Kidder (Eds.)
New directions for methodology of social and behavioral science: 12
San Francisco: Jossey-Bass.
.
L.. and Kintz, B. L.
(1987)
Computational
Glenview, III.:
Scott, Foresman, and Co.
Statistics
Bruning,
handbook
J.
of
.
Burgelman. R. H. (1983) A model of the interaction of strategic behavior,
corporate context, and the concept of strategy.
Academy of
Management Review 8:61-70.
.
R.
Cattell,
B.
(1978)
The
logic
and use
of factor analysis
.
San Francisco:
Jossey-Bass
Chakravarthy
B. S. (1982)
strategic management.
,
A promising metaphor for
Management Journal 7:35-44.
Adaptation:
Academy
of
,
Chakravarthy, B. S. (1987). On tailoring a strategic planning system to
context:
Some empirical evidence. Strategic Management Journal
,
its
8,
517-534.
Chandler, A.
D.
Strategy and structure
(1962)
.
Cambridge, Mass:
MIT Press.
Child, J.
(1972) Organization structure, environment, and performance:
Sociology 6:1-22.
role of strategic choice.
Child,
J.
The
,
(1975) Managerial and organization factors associated with
company performance - Part II, A contingency analysis.
Management Studies 12:12-27.
Journal of
,
Cohen, J., and Cohen,
Applied multiple regression/correlation
P. (1983)
Hillsdale, NJ: Lawrence Eribaum
analysis for the behavioral sciences
Associates
.
.
Denzin,
N.
K.
(1978)
Dewar, R., andWerbel,
of
The research
J.
(1979)
employee
Ouarterlv
,
satisfaction
24:426-448.
act
.
New York:
McGraw
Hill.
Universalistic and contingency predictions
and conflict.
Administrative Science
An examination of alternative
Drazin R., and Van de Ven, A. H. (1985)
forms of fit in contingency theory. Administrative Science Quarterly.
PP
Duncan, O. D. (1972) Path analysis: Sociological examples (plus addenda)
AldineIn
H.M. Blalock (Ed.) Causal models in social sciences
.
Atherson, pp. 115-138..
Egelhoff, W. G. (1982) Strategy and structure in multinational corporations:
An information processing approach. Administrative Science Quarterly
27:435-458.
,
38
Etzioni,
A.
A comparative analysis of complex organizations
Free Press.
(1961)
York:
.
Friedrich, R. J. (1982) In defense of multiplicative terms in multiple
regression equations. American Journal of Political Science 26,
833.
,
New
797-
Fry, L. W. and Smith, D. A. (1987) Congruence, contingency, and theory
Academy of Management Review 12:117-132.
building.
,
,
Galbraith,J. R. (1977) Organization design
.
Reading, Mass
:
The role of
Galbraith, J. R., and Nathanson, D. (1979)
structure and process in strategy implementation. In
A new
C. W. Hofer (Eds.), Strategic management:
Boston, Little Brown and Co.,
policv and planning
,
Ginsberg, A. and Venkatraman, N.
organizational strategy:
Academv
Grant,
of
A
Addison-Wesley.
organizational
D. Schendel and
view of business
pp. 249-283.
Contingency perspectives on
review of the empirical research.
5:25-39.
(1985)
critical
Management Review
,
The logic of strategic planning
J. H., and King, W. R. (1982)
Boston: Little, Brown and Company.
.
Contingency linkages between strategy and general
(1984)
characteristics:
A conceptual examination.
Academv of
Gupta, A. K.
manager
Management Review
,
9:399-412.
Gupta, A. K. and Govindarajan, V. (1984) Business unit strategy, managerial
characteristics, and business unit effectiveness at strategy
implementation.
Academv of Management Journal 27:25-41.
.
Hambrick, D. C.
strategy
in
Operationalizing the concept of business-level
research.
Academv of Management Review 5:567-575.
(1980)
.
C, MacMillan, I. G., and Day, D. L. (1982) Strategic
attributes and performance in the BCG matrix -- a PIMS-based
analysis of industrial product businesses.
Academv of Management
Journal 25:510-531.
Hambrick, D.
,
Hambrick, D. C. (1983a) An empirical typology of mature industrial product
environments.
Academv of Management Journal 26:213-230.
,
Hambrick,
D. C.
industries:
(1983b)
High profit strategies
A contingency approach
.
in
Academy
of
mature capital goods
Management Journal
.
26:687-707.
Hambrick, D. C. (1984) Taxonomic approach to studying strategy: Some
conceptual and methodological issues.
Journal of Management,
10:27-42.
Harrigan, K. R. (1983) Research methodologies for contingency approaches
to business strategy.
Academy of Management. 8:398-405.
Fundamentals of concept formation in empirical
(1952)
science. In International Encvclopedia of Unified Science. 11. Chicago:
University of Chicago Press.
Hempel, C. G.
39
Hitt, M. A.
Ireland, R. D.
,
and
,
K. A. (1982)
Pal ia,
I
ndus trial firm
s
ground
strategy and functional importance: Moderating effects of technology
and uncertainty. Academy of Management Journal 25:265-298.
.
Hitt,
M A
,
Ireland, R. D.
and Stadter, G. (1982) Functional importance and
Moderating effects of grand strategy and
,
company performance:
Strategic Management Journal
industry type.
C. W.
Hofer,
Hofer, C. W.
Toward a contingency theory
Management Journal 18:784-810.
(1975)
Academy
of
of
business strategy.
,
and D. E. Schendel (1978) Strategy formulation:
West Publishing.
,
concepts
3:315-330.
,
,
Analytical
Horovitz, J H
and Thietart, R A (1982) Strategy, management design, and
firm performance.
Strategic Management Journal 3:67-76.
.
.
,
.
.
,
C.G.
The organization and environment research program:
In A. H. Van de Ven and W. F. Joyce
(Eds.) Perspectives on organization design and behavior
New York:
Hrebiniak,
(1982)
An overview and
critigue.
.
Wiley,
Jauch,
pp. 338-345.
L. R., Osborn, R. N., andOlueck, W. F. (1980) Short-term financial
success in large business organizations:
The environment-strategy
1:49-63.
connection.
Strategic Management Journal
,
Jauch,
Toward an integrated theory of
L. R., and Osborn, R. N. (1981)
strategy.
Academy of Management Review 6:491-498.
,
Jick,
T.D.
(1979) Mixing qualitative and quantitative methods:
in action. Administrative Science Quarterly
24, 602-611.
Triangulation
,
Joreskog, K. G. (1971) Statistical analysis of congeneric tests.
Psvchometrika, 36:109-133.
K. G., and Sorbom, D. (1978)
Analysis
relationships bv method of maximum likelihood
Educational Resources.
Joreskog,
of
Joreskog, K. G., and Sorbom, D. Eds. (1979) Advances
Mass:
Abt Books.
structural equation models
structural
National
linear
Chicago:
.
in
factor analysis and
.
Slocum, J, and Von Glinow, M,
Joyce, W.
interaction:
Competing models of
Behavior 3:265-280.
,
(1982)
fit.
Person-situation
Journal of Occupational
,
Katz,
D.,
and R.
New York:
L.
Kahn (1966)
The
social
psychology of organizations
Wiley.
Lawrence, P., and Lorsch, J. (1967) Organization and environment Boston:
Graduate School of Business Administration, Harvard University.
.
Lorange, P., and Vancil, R.
Englewood Cliffs, NJ
:
Strategic planning systems
(1977)
Prentice Hall.
F.
.
.
40
Guidelines for the empirical classification of
Administrative Science Quarterly 21, 571-597.
McKelvey, B.
(1975)
organizations.
,
and Snow, C. C. (1978) Organizational strategy, structure, and
New York, NY: McGraw-Hill.
process
Miles, R. E.
,
.
Miller,
Toward
(1982)
P.
J.
Paper presented
Miller,
Miller,
clarity in contingency hypothesis analysis.
Academy of Management Meeting, New York.
at the
Toward a
D. (1981)
organizational gestalts.
new contingency approach: The search for
Journal of Management Studies 18:1-26.
,
A quantum view.
Organizations:
D., and Friesen, P. H. (1984)
Prentice-Hall.
New Jersey:
Miller,
D., and Friesen, P. H. (1978) Archetypes
24:921-933.
Management Science
of strategy formulation.
.
Miller,
Innovation in conservative and
D., and Friesen, P. H. (1982)
entrepreneurial firms: Two models of strategic momentum. Strategic
Management Journal 3:1-25.
,
Patterns
Mintzberg, H. (1979)
The structuring
in
of organizations
Mintzberg, H., and Waters, J. (1985) Of strategies:
emergent.
Strategic Management Journal.
Mohr,
L.
(1982)
Management Science
strategy formation.
Mintzberg, H. (1978)
24:934-948.
Explaining organizational behavior
NJ
.
,
Prentice Hall.
:
Deliberate and
San Francisco:
.
Jossey-Bass
Neter, J.
P.,
and Wasserman, W.
Homewood,
III.
:
(1974)
Applied linear statistical models
.
Irwin.
V., and Toulouse, J. M. (1977) Toward a multi-level
congruence theory of organization. Administrative Science Quarterly,
22:264-280.
Nightingale,
D.
A framework for the comparative analysis of
(1967)
32:194-208.
organizations.
American Sociological Review
Perrow, C.
.
Peters, T., and Waterman, R. (1982)
America's best run business
.
In
search of excellence:
New York:
Lessons from
Harper and Row.
Pinder, C. C, and Moore, L. F. (1979) The resurrection of taxonomy to aid
the development of middle range theories of organizational behavior.
Administrative Science Quarterlv 24:99-117.
,
Competitive strategy
Porter, M.
E.
(1980)
Porter, M.
E.
(1985) Competitive advantage
New York:
.
.
New York:
The Free Press.
Free Press.
Prescott, J. E. (1986) Environment as a moderator of strategy-performance
relationship.
Academy of Management Journal 29:329-346.
,
41
Venkatraman, N
1986) The market share -An empirical assessment of assertions and
7, 377-394.
contradictions. Strategic Management Journal
and
Kohli, A. K.
profitability relationship:
Prescott,
J
.
E
.
,
,
(
.
,
Roseboom, W.W. (1956) Mediation variables
Review, 63, 249-264.
in
scientific theory. Psychological
Strategy, structure and economic performance
Rumelt, R. P. (1974)
Boston: Harvard Graduate School of Business Administration.
.
Schendel, D. E., and C. W. Hofer (1979) Strategic management: A new yiew
Little, Brown and Co.
Boston:
of business policy and planning
.
Scherer, F. M. (1980) Industrial market structure and economic performance
Rand McNally.
Chicago:
Schoonhoven, C. B. (1981) Problems with contingency theory;
assumptions hidden within the language of contingency
Administrative Science Quarterly
.
Testing
"theory".
26:349-377.
.
Identification and
Sharma, S., Durand, R. M., and Gur-Arie, O. (1981)
Journal of Marketing Research
analysis of moderator variables.
,
18:291-300.
Snow, C. C, and Miles, R. E. (1983) The role of strategy in the
In R. Lamb (Ed.)
development of a general theory of organizations.
Vol. 2 (pp. 231-259), New York,
Advances in Strategic Management
JAI Press.
NY.
.
:
Southwood, K.
Substantive theory and
American Journal of Sociology
Stonich,
P.
(1982)
J.
Implementing strategy:
Tosi, H.
J.
(1967)
D.
Organizations
and Slocum,
L., Jr.,
suggested directions.
Van de Ven,
H.
A.
(1979)
and environments.
,
83:1154-1203.
Making strat egy happen.
Ballinger.
Cambridge, Mass:
Thompson,
interaction:
statistical
(1978)
E.
Five models.
in
action
.
New York:
McGraw-Hill.
W. Jr. (1984) Contingency theory:
Journal of Management 10:9-26.
J.
,
Some
,
book
Quarterly
Administrative Science
Review
Organization
24:320-326.
--
of Aldrich's (1979)
,
and Drazin, R. (1985) The concept of fit in contingency
Research in
In L. L. Cummings and B. M. Staw (Eds.).
theory.
JAI Press.
New
York:
333-365).
Organizational Behavior, Vol. 7 (pp.
Van de Ven, A.
H.
,
Venkatraman, N. (1986) Performance implications of strategic coalignment.
Revised version of the paper presented at the Academy of Management
meetings at Chicago, August.
Venkatraman, N., and Camillus, J C. (1984) Exploring the concepts of "fit"
Academy o f Management Review, 9:513-525.
in strategic management.
.
Venkatraman, N
research:
11:71-87.
.
,
and Grant, J H. (1986) Construct measurement in strategy
A critigue and proposal. Academy of Management Review
.
,
42
Venkatraman, N. and Prescott, J.E. (1988) Environment--Strategy
Coalignment An empirical examination of its performance implications.
Academy of Management Best Paper Proceedings in press.
:
,
Walker, G., and Thietart, R. A. (1986)
Strategic consistency, synergy, and
Paper 1756-86, Sloan School of
Working
business performance.
Management, MIT, Cambridge.
Waterman, R. H., Peters, T. J., and Phillips, J. R. (1980)
organization. Business Horizons, 23(3), 14-26.
Structure
is
not
A second-order longitudinal model of ability
(1980)
15:353-365.
structure. M ultivariate Behavioral Research
Weeks, D. G.
,
Strategies of effective low share
Woo, C. Y., and Cooper, A. C. (1982)
Strategic Management Journal 2, 301-308.
businesses.
,
Woodward, J. (1965) Industria organization
Oxford University Press.
l
Zedek,
:
Theory and practice. London:
Problems with the use of moderator variables.
S. (1971)
Psvchological Bulletin 76:295-310.
,
o
Figure
2:
A Schematic Representation
Fit
X
(e.g.,
Strategy)
Z
(e.g.,
Context)
of
Moderation
as
a1
a2
Y
(Performance)
a3
X.Z
(Mult. Term)
Figure
3:
A Schematic Representation
Fit
as
X
Mediation
(e.g.,
Strategy)
a1
of
Figure
4:
A Schematic Representation
Fit
as
of
Profile-deviation
Standardized Scale for
Measuring the Dimensions
Importance
Strategy
Dimensions
+1
-1
Xi
bi
X2
b2
X3
b3
X4
b4
X5
b5
XS5
\
Profile of Unit
X6
(i)
XS5
be
The measure
"Ideal Profile"
'
of profile-deviation for unit
(i) is
X C6
calculated as follows:
6
PD =
^
i
Xsj
=
(bj(Xsj-Xcj))2,
where
i
= The score for business unit
(j)
in
the study sample.
Xq = The score (average) for the calibration sample,
bj
= Standardized importance scores.
c
1/1
c
Q)
c
^>
%
'*->
>
(U
a
<u
a;
n
c
a;
o
c
o
TO
a
E
o
u
<
<V
>
CI
tn
c
a
u
C
o
u
TO
lo
-•T
Date Due
OC
1
7 '89
Lib-26-67
MIT LIBRARIES
3
TDflD
DOS 223 141