Vous êtes sur la page 1sur 18

ORGANIZATION SCIENCE

V(il.4. No. 2. May 1993


Printedin U..S.A.

ORGANIZATIONAL CULTURE AND PERFORMANCE:


PROPOSING AND TESTING A MODEL*
GEORGE A. MARCOULIDES AND RONALD H. HECK
Department of Managemeni Science, California State University at Fullerton,
Fullerton. California 92634
University of Hawaii at Manoa, Honolulu, Hawaii 96822
The purpose of this article is twofold: first, to propose and test a model concerning how an
orgnnization's culture affects organizational performance: and second, to demonstrate the
application of LISREL modeling methodology to estimate and test this model. Organizational culture is hypothesized to consist of three interrelated dimensions: a sociocultural
system of ihe perceived functioning of the organization s strategies and practices, an
organizational value system, and the collective beliefs of Ihe individuals working within the
organization. Organizational culture is operationalized hy several latent variables: organizational structure and purpose, organizational values, task organization, climate, and individual
values and beliefs. These variables, in turn, are hypothesized to affeet organizational
performance. Analysis of data from 392 respondents who participated in the study confirms
Ihc tit of the proposed model to the data. The model presented in the study represents an
initial attempt to describe and evaluate the effects of various dimensions of organizational
culture. It appears that the comparison of visible aspects of culture across and within
organizations can provide useful information for guiding the directions of organizations. By
investigating the variables defined in this study further, it may eventually be possible to
explain why some organizations are not performing at desired levels of productivity. A
methodological Kwl has also been presented in this article. It is clear that the application of
structural equation modeling techniques can provide organizational scientists with powerful
analytic totals for furlhering theory testing and development. Structural modeling can be used
10 construct, estimate, and test a variety of models in organization science.
(ORGANIZATIONAL MODELING; STRUCTURAL EQUATION MODELING)

The intensification of research on organizational effectiveness has led to the


formulation of theories about factors within an organization that can make a
difference in performance. Organizational culture is one such variable that has
received much attention in organizational behavior literature (Amsa 1986; Hofstede
1986; Hofstede, Neuljen, Ohayv and Sanders 1990; Jelinek, Smircich and Hirsch 1983;
Kilman, Saxton and Serpa 1985; Ouchi 1981; Owens 1987; Schein 1990; Trice and
Beyer 1984). This attention is mainly beeause researchers has postulated that cultural
factors play a key role in determining levels of organizational outcomes. A common
hypothesis about this role suggests that if an organization possesses "strong culture"
by exhibiting a well-integrated and effective set of specific values, beliefs, and
behavior patterns, then it will perform at a higher level of productivity (Dennison
1984). The development of theory to guide the definition of organizational culture,
therefore, is of primary importance to improving organizational performance, especially beeause the variables which comprise culture have been postulated to be under
the control of organizational leaders (Deal and Kennedy 1982, Ouchi 1981 Owens
1987, Siepert and Likert 1973).
Despite eoncern with achieving improved organizational productivity through focusing on the development of cohesive organizational culture, determining the
*Accepted by Kenneth D. Mackenzie; received November 27, 1989. This paper has been with the
authors for four revisions.
209

1047-7039/93/()402/0209/$0l.25
Copyright 1W3. The lnNlilute of ManagcniL-m Scicntcs

210

GEORGE A. MARCOULIDES AND RONALD H. HECK

parameters of this construct has been problematic. The literature on organizational


culture taps essential ideas, but the theory and technology to utilize the theory in
improving organizations have remained fuzzy (Mackenzie 1986). As Trice and Beyer
(1984) have argued, previous research on organizational culture has tended to focus
on single, discrete elements of culture, while ignoring the multidimensional nature of
culture, that is, a construct composed of several intimately interrelated variables
(Schein 1990).
Another problem has been that researchers are still not sure whether the association between culture and organizational performance reflects a "cause-effect" type of
relationship (Saffold 1988). In fact, researchers have not really identified what specific
variables comprise an effective organizational culture, nor have they provided convincing empirical evidence to suggest that if leaders in organizations increased the
amount of time and quality of energy devoted to developing a particular type of
organizational culture, then an organization would perform at a higher level of
productivity (Barney 1986). There is presently little agreement, therefore, about what
the concept of organizational culture means or how it should be observed and
measured (Schein 1990).
Because of the lack of agreement concerning theoretical formulations about
organizational culture, its delineation, and its possible relationship to performance
outcomes, no significant body of empirical research exists. Instead, researchers have
primarily focused on defining and describing the variables of organizational culture
and cautiously suggested a possible relationship between organizational culture and
outcomes (Owens 1987). As Mackenzie (1986) argues, organizational culture as a
concept may be a useful means of assessing the congruency of the organization's
goals, strategies and task organization, and resulting outcomes. Without valid and
reliable measures of the critical aspects of organizational culture, however, statements about its importance and effect on performance will continue to be based on
speculation, personal observations, and case studies (Uttal 1983). As a consequence,
management strategies and programs to create organizational change through understanding the organization's environment and strategically manipulating aspects of its
culture will continue to be poorly focused and difficult to implement and evaluate.

Developing an Exploratory Model

Organizational culture is a difficult concept to examine and numerous arguments


have been presented in the literature concerning our ability to observe and measure it
(Hofstede et al. 1990, Trice and Beyer 1984). As Schein (1990) has argued, however,
if we are to study seriously the concept of culture we must adopt a systematic
approach that enables researchers to identify those dimensions and variables that are
amenable to empirical measurement and hypothesis testing. Nevertheless, any discussion of determining the parameters of organizational culture must begin with some
realistic admissions of inadequacy.
As Mackenzie (1986) indicates, because society and hence organizations are continually evolving, there is really no universal paradigm or theory for examining organizational behavior that is valid in all contexts. What is relevant today as theory may be
found wanting tomorrow in terms of describing and explaining social phenomena. In
addition, proposed theories often become problematic when they attempt to model
the actual detail and richness of real organizations because of the complexity of
interrelationships between organizational processes. Researchers often lack the ability to isolate and measure many important organizational variables that are needed
for theoretical hypothesis testing. Clearly, concerns can be raised about any attempt

ORGANIZATIONAL CULTURE AND PERFORMANCE

211

to quantify and measure the conceptual components of a theory of organizational


culture. While this reduction of reality can be considered a limitation, there is a
definite usefulness to such an approach, since it makes a fuzzy field somewhat more
accessible (Hofstede et al. 1990, Mackenzie 1986, Schein 1990).
Despite some inconsistency in the literature, previous theoretical formulations
about organizational factors and their infiuence on performance provide some solid
foundations to test the hypothesis that information about organizational culture can
be an important indicator in explaining an organization's level of productivity. In this
section we outline several conceptual relationships derived from the literature that
were used to develop the constructs of organizational culture in the exploratory
theoretieal model tested in this study. While we acknowledge that organizational
culture is holistic and socially constructed by its members (Hofstede et al. 1990, Louis
1983. Rousseau 1990, Smircich 1983), we believe that at least some visible aspects ean
be measured and can yield important information about the infiuences on an
organization's performance.
Organizational culture has been defined as patterns of shared values and beliefs
over time which produce behavioral norms that are adopted in solving problems
(Owens 1987, Schein 1990). Schein (1985) has also noted that culture is a body of
solutions to problems which have worked consistently and are therefore taught to
new members as the correct way to perceive, think about, and feel in relation to those
problems. In faet. these shared philosophies, assumptions, values, expectations,
attitudes, and norms bind an organization together (Kilman et al. 1985). Thus, the set
of integrated concepts becomes the manner or strategies through whieh an organization achieves its specific goals. It can therefore be postulated that an organization's
collective culture influences both the attitudes and subsequent behaviors of its
employees, as well as the level of performance the organization achieves.
Schein (1990) has noted that culture is thought to permeate the organization on at
least three fundamental levels. At the surface, one may observe visible artifacts of the
organization, that is, its structure, technology, rules of conduct, dress eodes, records,
physical layout, stories, and rituals. Beneath this dimension is a second level!
organizational values, and, finally, underlying assumptions about the nature of organizational "reality" that are deeper manifestations of values. Of a)urse, investigating
processes of culture at the latter level is more difficult, as they cannot be directly
observed and measured.
Considerable work in organizational theory has also recently been devoted to
specifying the efi'ccts of visible aspects of the organization such as the nature of its
hierarchical structure and teehnieal complexity on the process of administration
(Bossert et al. 1982; Heck and Marcoulides 1989; Mackenzie 1976a,b, 1991; Perrow
1972; Thompson 1967). Several studies (Blau 1979; Heck and Mareoulides 1989Mackenzie 1976a, b, 1986, 1991; Owens 1987) indicate that organizational structure
and process mfluenee, both directly and indirectly, the achievement of organizational
outcomes. Together, these studies suggest that some of these structure/process
variables sueh as organizational size, complexity of administrative hierarchy, or
presence of eontrol mechanisms, may be linked to, or intertwined with, administrative
practices, worker attitudes, and levels of produetivity (Blau 1979, Heck and
Marcoulides 1989). In addition, in a study that may have implications for the business
community, Heek et al. (1990) found that leadership efforts aimed at organizing the
task structure and building a positive elimate in sehools were directly related to
higher or lower performance outcomes.
Allaire and Firsirotu (1984) have also suggested that three interrelated sets of
systems help identify organizational culture, whieh were incorporated into our exploratory theoretical model. These systems correspond roughly to Schein's (1990)

212

GEORGE A. MARCOULIDES AND RONALD H. HECK

typology of culture. The first is a sociocultural system which is composed of the


perceived functioning of formal organizational structures, strategies, policies, and
management practices relative to the work structure in the organization. Because
these activities have been consistently successful, they become regular practice. This
aspect of organizational culture reflects the orientation of classical management
theory toward the goals of the organization and how tasks must be best organized to
meet those goals (Mackenzie 1986, Thompson 1967). Leaders make important decisions about the work structure of organizations that fundamentally shape how
workers go about doing their jobs. In managing the core technology of the organization, leaders help crystyallize production goals, achieve more certainty in how to
achieve goals, and develop strategies to organize the work force to translate goals into
outcomes (Bossert et al. 1982, Mackenzie 1986, March and Simon 1959).
A second domain of culture, according to Allaire and Firsirotu (1984), is an
organizational belief system that embodies the myths, values, and idealogies of the
organization. Researchers have suggested that developing this aspect of culture is a
powerful function of organizational management. Leaders attempt to clarify values
and to define the organization's purposes. This domain emphasizes communication of
goals and production results, both internally and externally (Heck, Larsen, and
Marcoulides 1990; Reynolds 1986). For example, Bolman and Deal (1984) have noted
the "mythological" roles often played by leaders as ceremonial heads of organizations, in an effort to clarify role responsibilities, teach organizational values, and
promote the organization's mission. These symbolic behaviors may represent attempts
to transmit organizational values at deeper levels of inculturation.
As Owens (1987) notes, one co-effect of the socialization of organizational values is
organizational climate, or the perceptions held by participants as to the nature of the
organization. Employees have a variety of perceptions about how well the work
environment of the organization is functioning including the quality of social interactions, recognition of their work-related efforts, the types of communication channels
open to them, access to teehnology and resources, and demands or stress placed upon
them by the organization. Besides managing the core technology of the organization,
leaders also may have some impact on building a productive organizational climate
through the emphasis of particular sets of organizational values, and the amount of
effort expended in this domain is predictive of organizational outcomes (e.g.. Heck
et al. 1990, Owens 1987). Part of the leader's role in managing this subsystem may
also include "buffering" the organization from outside influences, so that goals may
be attained and input from the environment is channeled constructively.
The last domain of culture suggested by Allaire and Firsirotu (1984) is the
collective individuals, with their unique experiences, belief systems, goals, and personalities. While such attitudes and values may be at least partially shaped by the
organization's belief system, individuals may be thought of as possessing a wide range
of perceptions about social, political, and work-related issues. Though the goal is to
socialize employees to perpetuate the organizational culture, the process does not
always have uniform effects on individuals. Fuller et al's (1982) concept of organizational efficacy further suggests that management behavior may to some extent be
eonstrained by the often divergent goals, attitudes, and decisions of individual
workers. The attitudes and values of employees individually, as well as about
organizational climate, therefore, may be important variables that affect organizational performance.
The role that culture plays in at least partially determining an organization's level
of performance without doubt has received considerable attention in the organizational behavior literature. An equal amount of attention, however, has also recently

ORGANIZATIONAL CULTURE AND PERFORMANCE

213

been given to the limitations associated with the study of these variables (Saffold
1988; Schein 1990; Weick 1985). Critics point out the inadequate methodologies used
to examine and describe the complex relationship among the organization's environment, its structure and purposes, organizational culture, and resulting outcomes. For
example, Mackenzie (1986) indicates the need to develop a methodology for studying
organizations that can trace or model cause and effect relationships in what is
essentially a dynamic, open process. Most models in the literature that try to identify
potential causal relationships, however, have yet to be adequately tested (Saffold
1988). In fact, Saffold suggests that the applications of mathematically sophisticated
models have been "inadequate" and that this area must receive more attention in the
near future.
It appears, therefore, that most research on organizational processes and performance, while suggesting some important and useful theoretical models, continues to
be problematic beeause of definitional and methodological constraints (Rousseau
1990). Thus, statements about the importance of organizational culture and its
possible effects on organizational performance must be better defined theoretically
and methodologically, if researchers hope to measure the relative contribution of
variables comprising organizational culture to levels of performanee. Of course, the
goal of such research is to provide managers with information to assess key organizational processes and design strategies to improve overall effectiveness.
The purpose of this study is twofold: (a) to propose and test an exploratory
structural model eoncerning how visible aspects of an organization's culture may
affect organizational performanee, and (b) to demonstrate the application of a
methodology to estimate and test this proposed model. This methodology is based on
structural equation modeling (Joreskog and Sorbom 1986; Heck, Marcoulides and
Glasman 1989; Marcoulides 1989). We believe the research methodology used in this
study can avoid many of the pitfalls associated with previous research on organizational culture and provide a reasonable "test" of many of the theoretical propositions
associated with the relationship between at least some more visible aspects of culture
and measures of performance. We do not, however, attempt to probe the deepest
level of culturethat of finding underlying organizational assumption.s.
The remaining paper is divided into three sections. The first section deals with the
participants and the instrumentation used to gather the data. The second section
presents the exploratory model and the methodology used to test this model. In the
third section we discuss the results of the analysis and their implications for understanding how organizational culture may be useful in assessing higher or lower levels
of organizational performanee, as well as evaluating organizational eongruency.
Consistent with Schein (1990), our goal is to begin to develop a "roadmap" that
suggests possible relationships among variables comprising organizational eulture and
to estimate their relative effects on performance.
Method
Participants

Three hundred ninety-two (392) participants were randomly selected from strata
within 26 seleeted organizations. The 26 organizations were randomly selected from
recent regional directories listing all organizations in two geographic regions on the
basis of type, size, earnings and growth over a multi-year period. The geographic
regions represented the midwestern and western parts of the United States. The
sample covered a wide range of organizations. For example, there was great variety
among the selected organizations relative to output type (product or service), size

214

GEORGE A. MARCOUHDES AND RONALD H. HECK

(small, medium, or Iarge), resource type (capital intensive or labor intensive), ownership (public or private), and objective (for profit or not-for-profit).'
j '

Instrumentation
The participants in this study responded to a structured interview and a follow-up
questionnaire (Cotlar 1987). The instruments were developed hased on a framework
adapted from Hackman and Oldham (1976). Each of these instruments included
questions regarding the variables that measured important criteria and variables
perceived to comprise organizational culture. For example, participants were asked:
"Please indicate the extent to whieh supervisors and subordinates tend to work
cooperatively in your organization." The variables were generally constructed as
five-point Likert-type scales with responses ranging from "strongly agree" to "strongly
disagree." Data about the indieators of organizational productivity were also collected from organizational records.
Each interview took approximately one hour to complete, and each questionnaire
took approximately one-half hour to complete. Both instruments consisted of three
sections: demographic characteristics of the respondent; descriptors about organizational variables relating to organizational culture as seen by the respondent; and
personal preferences with respect to a wide variety of important societal and
work-related issues.
The final selection of observed and corresponding latent variables was carried
forward in successive stages. First, a preliminary roster of items was prepared
aecording to the constructs considered in the literature to be predictive of organizational performance. The starting list was quite long, containing over 200 items.
Several preliminary models of the interrelationships between the observed variables
and the hypothesized constructs of organizational eulture were developed and tested
via confirmatory factor analysis before a final model was proposed (Marcoulides 1990,
Marcoulides and Heck 1991).
Testing the Model Using Structural Equation Modeling
Structural equation modeling is a useful tool in theory development because it
allows the researcher to propose and subsequently test theoretical propositions about
the interrelationships among variables in a multivariate setting (Heck et al. 1990). A
structural model may be viewed as a guide that allows the researcher to assess the
relative strength of each variable included in explaining a desired set of outcomes. It
is easy to get carried away and become too grandiose, however, when using structural
equation modeling, especially since most substantive theories are quite complex.
Although it is easy to hope that most of the eomplexity of a theory can be studied
within a single model, in reality this is usually not the case. Thus, it must be
recognized that one of the greatest weaknesses with structural equation modeling lies
in potentially excluding important variables that influence a system. As such, one is,
in principle, always subject to criticism for having omitted an important variahie.
Since in reality all plausible causes cannot possibly be entered into a single model,
the problem is reduced to constructing a model so that al! known relevant causes are
included, while recognizing that such a procedure is guaranteed to be theoretically
incomplete.
The exploratory model tested in this study was posited a priori to address the
question of whether visible aspects of organizational eulture affect levels of organizational performance. In proposing our model, we drew on several theoretical models
'Researchers inieresied in the full set of data should contact the first author for further infonnation by
writing to Ungsdorf Hall 540, Fullerton, CA 92634.

ORGANIZATIONAL CULTURE AND PERFORMANCE

215

FtGURE 1. Proposed Model of Variables Influencing Organizational Performance.

proposed in the organizational behavior literature as summarized in Trice and Beyer


(1984), Schein (1990), and Allaire and Firsirotu (1984), as well as our own interpretation of the research to date on the relationship between organizational culture and
performance.
Figure 1 presents the proposed theoretical predictive model for organizational
productivity. The outcome of interest is level of organizational performance (OP),
measured in terms of capital gains versus costs. The proposed model posits the
existence of several latent, or hypothetical, variables which together comprise visible
aspects of an organization's culture and, in concert, are believed to influence
organizational performance. It is hypothesized that culture is composed of five
interrelated latent variables. These latent variables include organizational
structure/purpose (OS) and organizational values (OV), which are seen in the model
as exogenous, in that their variability is determined hy factors outside the causal
model, and three endogenous variables (whose variability is determined by other
variables in the model). These latter variables include organizational climate (OC),
task organization (TO), and worker attitudes/goals (AT). The exogenous variables
are viewed as indirectly affecting organizational performance through the endogenous
variables in the model. The latent variables defined in the model are postulated as
loosely comprising the three subsystems of organizational culture (sociocultural,
organizational belief, and individual belief) suggested by Allaire and Firsirotu (1984).
Structural equation modeling (sometimes referred to as causal modeling or covariance structure modeling) can be used to estimate and test a variety of proposed
models. Statistical estimates of the parameters of the models can be generated using
a computer program called Linear Structural RELationsLISREL (Joreskog and
Sorbom 1986). This computer program is the most widely available, although other
packages are beginning to gain popularity (e.g., Bentler's EQS, McDonalds' COSAN,
and Muthen's LISCOMP). The LISREL program is specifically designed to estimate
parameters and test the validity of a wide variety of models including those that
contain measurement errors, reciprocal causation, variables measured at several
points in lime, and latent variables.
The use of path diagrams to represent a structural equation model facilitates the
specification and understanding of the LISREL methodology, ln Figure 2 we illus-

216

GEORGE A. MARCOULIDES AND RONALD H. HECK

trate the LISREL representation of our model of variables hypothesized to influence


organizational performance. Observed variables are enclosed in squares in the
diagram. Latent variables (unmeasured, hypothetically-existing constructs) are enclosed in circles. Symbols representing measurement errors and residuals are not
enclosed. Measurement errors are denoted by lower-case Greek letters: epsilon (e)
corresponding to a set of y variables, and delta (5) corresponding to the set of x
variables. Residuals in equations are denoted by zeta iOAs can be seen in Figure 2, there are six latent variables in the model. Each of
these latent variables was defined in terms of observable variables (and in some cases
the observed variables were also defined as combinations of specific observable
variables). The careful specification of a structural model thus allows the exploration
of a variety of relationships between variables through both direct and indirect paths
in the model. The structural equation model describes how we believe organizational
performance is influenced both directly and indirectly by the exogenous and endogenous latent variables, and how these latent variables are related to the observed
variables. Descriptions of the observed variables grouped according to the constructs
they are posited to measure are provided in Appendix A. The observed variables are
paraphrased from the original questionnaire and interview items used in the study.
In summary, we believe that at least some aspects of organizational culture may be
directly or indirectly assessed and their effects on organizational performance can be
measured. Organizational culture is clearly a multidimensional construct, comprised
in this model of several underlying, interconnected processes including organizational
structure/purpose, values, task organization, climate, and employee attitudes about
their work (Heck et al. 1990; Owens 1987; Schein 1990). The composite of these
interaction-influence systems creates an organizational culture consisting of values
and beliefs that are manifested In organizational behavior. Ultimately, along with
environmental processes (which are outside the present model), an organization's
culture generates an impact on its achievement of desired outcomes. We believe this
model to be consistent with previous organizational theory and research.
This model can be translated in the LISREL program into a mathematical model
with two fundamental parts. The first is known as the measurement model. This is the
model that specifies the relationships between the observed variables and the underlying constructs they are hypothesized to measure, referred to as latent variables.
Although the latent variables are not measured directly, they are related to the
observed variables through the measurement model. At the measurement stage of the
model, specification is confirmatory, in that the relationships between observed
variables and latent variables are defined a priori. It is important to note, therefore,
that a contribution of the LISREL method to organizational researchers is that it
allows latent variables to be tested for relationships that may be of interest to theory
building, even though the latent variables are not directly observed.
Once the latent variables have been satisfactorily defined, the second component
model, known as the structural equation model of LISREL, estimates the intercorrelations among the latent variables. A relationship between two latent variables can
either be a causal relationship, which is represented in the diagram by a straight line
between two variables, or a correlation, which would be represented by a curved line
between two variables (correlations have no causal interpretation). Causal relationships are denoted by gammas (y), and betas (/3), and correlations are denoted by
phi i(f>).

The logic of causal modeling forces researchers to clarify the conceptual reasons
for allowing variables to interrelate. That is, in using the LISREL program to specify
an exploratory model (as was shown in Figures 1 and 2), a researcher must consider

ORGANIZATIONAL CULTURE AND PERFORMANCE

217

218

GEORGE A. MARCOULIDES AND RONALD H. HECK

in detail why and how the variables are to be interrelated. Though the present
methodology is not inherently free from conceptual ambiguity and complexity, any
problems stemming from this ambiguity and complexity should be reduced by allowing previous theory to guide the definition of variables to be included in the model.
Results
Estimating the Parameters and Model Fit

The proposed model was tested using LISREL VL The use of unweighted least
squares (ULS) fitting function for estimating the parameters of the model was called
for because of the presence of several dichotomous variables in the sets of measures
(Joreskog and Sorbom 1986). In the specification of the model, Lomax's (1982)
recommendations were followed: parameters relating reference observed variables to
underlying constructs were constrained to equal one (LO). Similarly the error measurement parameters for those variables were fixed at zero (0.0).
Figure 3 presents the LISREL parameter estimates of the proposed model. These
parameter estimates are indices that represent the simultaneous contribution of each
observed and latent variable to the overall model. While the estimates provide
important information that can be used to describe the latent variables and their
interrelationships, they do not provide any indication of the assessment of the fit of
the overall model. Since we posited an a priori defined model to be tested, our
primary interest is in the model fit. Once this model fit is determined, then the
importance of the parameter estimates can be more clearly evaluated. Without a
significant model fit, we would have to reconceptualize our model.
The assessment of the fit of the proposed model can be estimated from various
criteria. These include both statistical and practical criteria. Statistical criteria include
the coefficient of determination (COD), the goodness of fit index (GFl), and the root
mean square residual (RMS). Practical criteria include the Bentler and Bonett (1980)
normed index (BBI).
Table 1 presents the criteria describing the fit of the proposed causal model. The
coefficient of determination for the measurement model is 0.84, indicating a relatively
strong relationship between the observed variables and the latent variables included
in the model. This coefficient is basically a generalized measure of reliability for the
whole measurement model. It shows how well the observed variables jointly serve as
instruments for measuring the latent variables. This coefficient is quite high, indicating that the measurement model is good.
The assessment of fit of the overall model is also revealed by the goodness of fit
index (GFI) and the Bentler and Bonett normed index (BBI). It is generally recognized that GFI and BBI values close to or above 0.90 indicate a good model fit. For
this model the GFI is 0.86 and BBI is 0.82, indicating a reasonably good model fit.
The GFI and BBI can be considered measures of the relative amount of variance and
covariance in the data accounted for by the proposed model. On the other hand, the
root mean square (RMS) residual is a measure of the average unexplained variances
and covariances in the model. This index should be close to zero if the model fits the
data well. The observed RMS is 0.08, indicating that very few of the variances and
covariances in the data are left unexplained by the hypothesized model. The estimates of the direct and indirect effects of the variables in the model were also tested
through / tests, and all parameters were found to he significant ip < 0.01). We would
thus consider that the model fairly accurately accounts for the variability observed in
the data.
Our practical experience with structural modeling suggests that, when model fit is
close to 0.9, the structure of the model is reasonably accurate with respect to number

ORGANIZATIONAL CULTURE AND PERFORMANCE

219

220

GEORGE A. MARCOULIDES AND RONALD H. HECK


TABLE 1
Measures of Model Fit
Coefficient of Determination (COD)
Goodness of Fit Index (GFI)
Bentler and Bonett Normed Index (BBI)
Root Mean Square Residual (RMS)

0.84
0.86
O.2
0.08

of variables, directions of causality, and errors in equations. Model improvement


comes primarily from more closely specift'ing errors of measurement (e.g., correlating
individual error terms on observed variables). For purposes of this study, we selected
to follow the recommendations of Marsh et al. (1988) that testing models should not
he totally motivated by goodness of fit improvement unless justified on theoretical
grounds. Although we could possibly ohtain more statistically fit models, the results
would make little sense theoretically and conceptually. Therefore, we presented our
results based only on the model that we posited.
Discussion
In this paper we proposed an exploratory theoretical model concerning how an
organization's culture might be defined in terms of observable variables and latent
constructs, and how these variables in turn help explain its level of performance. In
addition, we demonstrated the application of the LISREL paradigm to estimate and
test this model. In this section we discuss the usefulness (and cautions) of this
methodology which can be used to investigate complex interrelationships among
variables within organizations, as well as the implications of our research for the
effective design and assessment of organizational processes to improve performance
and overall effectiveness.
Throughout our presentation and estimation of the model, we have tried to
emphasize the importance of using substantive theory as a guide for the appropriate
application of this powerful (yet easily misused) analytic tool. Caution must be
exercised in employing structural equation models to test substantive theories, as
omitted variables and poor measurement of variables included are common sources
of misspecification in a model and can produce misleading resuits. Thus, even though
the observed variables in the present study showed acceptable reliability and were
well grounded in previous research, the results should be considered limited to the
variables we studied within the context of the proposed model. Structural equation
models provide a powerful analytic tool to test the adequacy of proposed theories. As
researchers agree, however, structural models should only be used when they are
guided by strong substantive theory and when psychometically sound data are
gathered to test a model (Bentler and Bonett 1980, Heck et al. 1989).
The psychometric quality of behavioral measurements is generally evaluated in
terms of two propertiesreliability and validity. Reliability concerns the consistency
of observed data, and validity involves the interpretation of observed data as representative of some external property. As such, validity is generally considered more
important than reliability, since a precise estimate of the wrong behavior is less useful
than a relatively imprecise estimate of the intended behavior. Thus, an important
requirement for the evaluation of models via structural equations lies in using
theoretically appropriate operationalizations of both observed and latent variables.
Of course, the evidence for validity is generally less obvious than the evidence for
reliability. For example, an individual's verbal report measure of perceived pressure
on the job may, or may not, adequately capture a key idea of organizational climate in

ORGANIZATIONAL CULTURE AND PERFORMANCE

221

the model. And even if it does, the way the individual's reply is coded into a score
may destroy its meaningfulness (Bentler and Bonett 1980). As previously stated,
concerns can be raised about any attempt to quantify and obtain valid measures of
the conceptual components of a theoretical framework. While this reduction of
reality can be considered a limitation in the present study, there is usefulness to such
an approach for both organizational designers and practitioners, because it makes the
concept of organizational culture somewhat more accessible (Hofstede et al. 1990).
The correct usage of LISREL methodology must be considered carefully within the
context of the following guidelines before any type of model testing can be accomplished; (i) the theoretical foundation of the overall model to be tested, (ii) the
validity of the measured variables, (iii) the nature of the relationships between the
observed variables and the latent variables in the model, (iv) the direction of
causality, and Iv) errors in the measurement of variables, and errors in the overall
model.
The fit of the measurement model to the data (which indicates the relationship of
the observed variables to the latent variables) lends support to the assertion that
organizational culture is an interconnected web of relationships which may be reliably
measured (Tierney 1988). The overall fit of the structural model further suggests that
the Information contained in the latent variables is important in determining whether
organizations are performing at higher or lower levels. We are now able to assess
more thoroughly the significance of the empirical validation of our proposed model.
The results of our analysis indicate that all of the latent variables included as
comprising organizational culture in the theoretical model have some direct (or
indirect) effect on the level of organizational performance observed, with the largest
direct effects being from worker attitudes (0.93) and the task organization activities
(0.71). Relatively large direct effects were also noted between organizational values
and organizational climate (0.88) and between organizational values and employee
attitudes (0.78).
The model of organizational culture supported in this research suggests specific
factors that managers can manipulate at the organizational level. Because the kinds
of observed variables measured in this study are to some extent influenced by
management actions (e.g., methods of recruitment and retention, evaluation, decision-making processes, attitudes toward employees), an organization might improve
performance by attempting to manage strategically particular aspects of its culture.
To illustrate, eonsider the direct effects of the properties of organizational
structure/purpose on the task organization processes (0.78). This direct effect indicates that the organization's overall purpose as manufacturing or service (A'3), how it
is structured in terms of complexity of administrative hierarchy (A'2), and its resource
and communication flow patterns (A'l) together are directly related to the type of
task organization and strategies adopted to achieve goals. Clearly, the assessment of
such information provides managers with options to help guide their organization,
that is, to enhance cultural elements that are viewed as critical to improving
performance and to discard those elements that are viewed as increasingly dysfunctional (Schein 1990). Management options might include, for example, decentralizing
the organizational hierarchy, which should also improve resource and communication
flow patterns.
As shown in the model, the task organization developed in the organization
exercises a strong direct effect on the organization's level of performance (0.71). The
most important contributors to this domain are the types of methods used to select
new employees (Vl), the quality of methods used to evaluate employee performance
{Y2), and the criteria and practices for remuneration (K3). The relatively large
parameter estimates for these variables indicate that an organization's selection

222

GEORGE A. MARCOULIDES AND RONALD H. HECK

practices, monitoring of performance, and incentives for performance are strongly


related to organizational performance. Of secondary importance to performance
outcomes are decision-making practices (>'6), mentoring iY5), and providing opportunities for employees to pursue interesting and challenging work (Yl). The sum of
these variables suggests emphasizing a view of organizational management that is
oriented toward employees and is evaluation based in order to promote organizational performance. Our results are therefore congruent with the human resource
management perspective, which has long argued the need to develop effective
methods of recruiting, evaluating, and compensating employees to enhance organizational performance.
As expected, a suhstantial direct effect was also observed between organizational
values and employees' perceptions about the climate of the organization in which
they work. All of the observed variables comprising the construct of organizational
values loaded highly (0.85 or above). The variables include placing emphasis on risk
taking (A'4), on safety in the workplace (.X5), on productivity and efficiency (A'6, XI),
on rapid response to market opportunities and image building iXS), and on creating
new outputs and improving existing ones (A'9). Organizational climate is dominated
by how employees view their organization's primary purposes (YS), the quality of the
organization's social life (VIO), and employees' perceptions of the organization's use
of technology and adoption of new ideas (Vl 1).
Besides affecting employees' perceptions of climate in the workplace, it can be
further stated that an organization's value system affects organizational performance
indirectly (through the climate, task organization, and individual attitudes paths in
the model). Utilizing the model, then, managers might attempt to reshape the
normative structure of poorly-pcrforniing organizations, as well as changing task
organization processes to improve both climate and performance.
A final domain that must be considered in conceptualizing organizational culture is
that of the attitudes, beliefs, and ethics of the workers who comprise the organization. The results of the study suggest that these attitudes may be strongly predicted
from organizational values (0.78), as one might expect, and moderately predicted
from organizational climate (0.38) and task organization (0.30). In turn, this domain
exercises a strong direct effect on defining the level of organizational performance
(0.93).
Conclusions
The results are consistent with previous research that has suggested that variables
associated with organizational culture are predictive of organizational performance
(Hofstede et al 1990, Kanter 1983, Moos 1979, Peters and Waterman 1982). As
Bolman and Deal (1984) argue, effective organizations share a variety of attributes
involving their cultures. Because of the complexity of interrelationships between the
environment and organizational processes, however, it is difficult to link manager
behavior in a causc-cffcct manner to performance outcomes. While acknowledging
that environmental variables impact organizational performance both directly and
indirectly, if we hold the environment constant at any moment in time, there may be
great variation in how organizations function internally to affect performance (Heck
et al. 1990).
Our findings therefore imply that management efforts directed toward the determination of the particular profile of organizational culture present can provide Information that outcomes can be improved by managing the determinant variables. Our
results support the possible importance of management's use of strategic organizational design as an activity that can improve organizational performance (Mackenzie

ORGANIZATIONAL CULTURE AND PERFORMANCE

223

1986). Organizational design is generally concerned with developing the appropriate


strategies and task organization that most effectively convert goals into desired
performance outcomes. Our beginning efforts to model some of these processes
within the organization and to estimate their effects on performance give managers
some options to improve both overall cfTcctivcness and interpersonal relationships,
which should improve the organization's ability to maintain itself through time.
The concepts presented in this research study represent an initial step to describe
and evaluate the effects of various dimensions of organizational culture as a means to
develop or extend theories that can explain why some organizations outperform
others. It appears that the comparison of visible aspects of culture across and within
organizations can provide useful information for guiding the directions of organizations. Assessing the organization's culture may be viewed as a possible means of
checking the congruency of where the organization is presently (i.e., at one point in
time when the observations are taken) versus its desired goals, strategies, and results.
Expanding on this rudimentary beginning might be a profitable goal of future
research. It suggests the need for further refinement and extension. As Tierney (1988)
argues, the methodological tools and skills for such cultural studies also need
elaboration. Structural equation modeling can make a significant contribution to a
class of research questions concerning the interrelationships between organizational
processes, such as culture and performance, where theoretical mapping has been
problematic. The application of structural equation modeling techniques can provide
organizational scientists with powerful tools for furthering theory testing and development.
Acknowledgements
The authors wish to acknowledge Morton Cotlar for his kind support with the data.

Appendix A: Observed Variables Included in the Model


Organizational Structure
This construct reflects the structure and operational processes of an organization. As conceived in this
study, the construct includes the relative size and complexity of the administrative hierarchy, as well as
whether the organization is primarily manufacturing or service-or len led. The construct is niea.surcd by the
complexity evident in the organization's resource and communication Row patterns (A'l: Ctiniplexity), the
extent of breadth and depth evident in the organization s hierarchy (A'2: Sophistication), and the relative
focus of the organization with respect to its output (A'3: Product/Service Line).
Organizational Values
This construct or latent variable (the tenns can be used interchangeably) describes the principles,
idealogies, and values that an organization holds as desirable in the practice of serving its clients. It is
meiisured by Ihe emphasis the organization places on risk-taking, and its acceptance of occasional failure
as a result (A'4: Risk), the emphasis the organization places on protecting its employees in the workplace
(A'5: Satcty), the emphasis the organization places on productivity and efficiency (Xb: Efficiency), on
integrity and orderliness of performance (A'7: Professionalism), on rapid response to expanding market
opportunities, advertising, public image, and community service (A'8: Marketing & Image), and on creating
new outputs and improving existing ones (A'9: Research & Development).
Task Organization
This construct represents the typical strategies, policies, and actions used by the organization in
achieving its production goals. The construct is measured hy the diversity and intensity of methods used in
selecting new employees (Vl: Selection), the quality and diversity of methods used in judging employee
performance (72: Evaluation), the propriety of criteria and practices for renumeration iVy. Comjjensation), the extent that members are dedicated to performing their organizational roles (74: Performance),
the extent that managers take personal interest in the welfare and performance of their enfiployees (V5:
Mentoring) and utilize effective methods of selecting decision-making alternatives (K6: Decision Making),

224

GEORGE A. MARCOULIDES AND RONALD H. HECK

and the extent that employees have opportunities to pursue interesting and challenging work {Y7:
Challenge).
Organizationat Climate
This construct is described by the perceptions of workere about a variety of conditions concerning the
work environment. It is measured by the awareness among employees of the nature of the organization's
output as a blend of product and service (VS: Industry Role), the opinion among employees of the ease
with which communicatians and resources are transmitted amung the organization's elements (VJ: Flow
Mechanisms), the perceptions among employees of the quality of interactions and recognilion within and
across organizational levels, as well as how the needs of individuals are a concern to the organization (VIO:
Organizational Life), tbe awareness among employees of the organization's use of availahle technology and
adoption of new ideas (711: Technology), and the perceptions of how much the organization exerts
pressures on individuals yet is sensitive to the effect of stress (Y\l: Stress).
Employee Attitudes
This construct reflects the beliefs of employees about a variety of issues related to social, political and
organizational concerns, some of which may be influenced by the organization and some of which may be
separately determined. In the present study, it is measured by Ihe extent thai employees resent recent
organizational policies in acceptance of mimmties (V13: Prejudice/Tolerance), regard nationalism as
important (V14: Nationalism), regard common courtesy and punctuality as important work attributes (V15:
Social Amenities), regard dedication and commitment to the organization as important (Y \(y\ Commitment), and tbe extent to which employees perceive that management involves them in the decision-making
process (Vl?: Involvement).
Organizational Performance
This construct reflects the extent of goal achievement in the organization's workforce, capital, marketing, and lisciil matters. It is measured by the extent of sales fultillment. as measured by gross revenue
relative to the value of the product (or service) line (VIS: Volume), the extent of penetration into the
organization's potential customer base (yi9: Share), tbe extent of revenue surplus over expense resulting
from organizational operations (>'2O: Profit), and the extent of surplus in relation to risk, as measured by
gross prolit relative to assets and equity invested (721: Return).

References
Allaire, Y. and M. E. Firsirotu (1984), "Theories of Organizational Culture," Organizational Studies, 5, 3,
193-226.
Amsa, P. (1986), "Organizational Culture and Work Group Behavior; An Empirical Study," Journal of
Management Studies, 23, 3. 347-362.
Barney, J. (1986), "Organizational Culture: Can It Be a Source of Sustained Competitive Advantage?,"
Academy of Management Review, 11, 656-665.
Bentler, P. and D. Bonett (1980). "Significance Tests and Goodness-of-Fit in the Analysis of Covariance
Structures," Psychological Bulletin. 88, 588-606.
Blau, J. (1979), "Expertise and Power in Professional Organizations," Society of Work and Occupations, 6,
I, 103-123.
Bolman, L. and T. Deal (1984), Modem Approaches to Understanding and Managing Organizations, San
Francisco: Jossey-Bass.
Bossert, S.. D. Dwyer, B. Rowan, and G. Lee (1982). "The Instructional Management Role of the
Principal," Educational Administration Quarterly. 18, 3, 34-64.
Cotlar, M. (1987). "Profiles of Organizations," College of Business Administration. University of Hawaii at
Manoa.
Deal, T. and A. Kennedy (1982). Corporate Cultures, Reading, MA: Addison-Westey.
Dennison, D. (1984), "Bringing Corporate Culture to the Bottom Line," Organizational Dynamics, 13. 2,
5-22.
Fuller, B., K. Wood, T. Rapaport. and S. Dombusch (1982), "The Organizational Context of Individual
Efficacy," Review of Educational Research. 52, 1, 7-30.
Ilackman, J. R. and G. Oldham (1976), "Motivation through the Design of Work: Test of a Theory."
Organizational Behavior and Human Performance 16, 2. 250-279.
Heck, R. H. and G. A. Marcoulides (1989), "Examining the Generalizability of Administrative Personnel
Allocation Decisions," The Urban Review 21, 1. 51-62.
,
, and N. S. Glasman (1989), "Tbe Application of Causal Modeling Techniques to
Administrative Decision Making: The Case of Teacher AlkK-ation." Educational Administration
Quarterly, IS, 3, 15-31.

ORGANIZATIONAL CULTURE AND PERFORMANCE

225

Heck. R, H.. T. J. Larsen, and G. A. Marcoulides (1990), "Instructional Leadership and School Achievement: Validation of a Causal Model," Educational Administratifm Quarterly, 26, 2, 94-125.
Hofstede, G. (1986), "The Usefulness of the Organizational Culture Concept." Journal of Managemeni
Studies. 23, 3, 253-258.
, B. Neuijen. D. Ohayv, and G. Sanders (1990), "Measuring Organizational Cultures: A
Qualitative and Ouantitative Study across Twenty Cases." Administrative Science Quarterly, 35,
286-316.
Jelinek, M., L Smircich, and P. Hirsch (1983), "Introduction: A Code of Many Colors," Administrative
Science Qtiarterly, 28, 31, 331-338.
Joreskog, K. G. and D. Sorbom (1984-1986), LISREL VI, Mooreville, IN: Scientific Software.
Kanter, R. (1983). The Change Masters: Innovation and Entrepreneurship in the American Corporation, New
York: Simon & Schuster.
Kilman, R., M. J. Saxton, and R. Serpa (1985), "Introduction: Five Key Issues in Understanding and
Changing Culture," in R. Kilman et al. (Eds.), Gaining Control of the Organizational Cidture. San
Francisco: Jossey-Bass, 1-16.
Lomax. R. G. (1982), "A Guide to Lisrel-Type Structural Equation Modelling," Behavior Research Methods
& Instrumentation. 14, 1, 1-8.
Mackenzie, K. D. (1976a), A Theory of Group Structures. Vol. I. Basic Theory, New York: Gordon and
Breach Science Publishers.
(1976b). A Theory of Group Structures. Vol. II. Empirical Tests, New York: Gordon and Breach
Science Publishers.
(1986). Organizational Design: The Organizational Audit and Analysis Technology, Norwood, NJ:
Ablex Publishing Corptiration.
.(1991), TIK Organizational Hologram: The Effective Management of Organizational Change,
Norwell. MA: Kluwer Academic Publishers.
March, i. and H. Simon (1959), Organizations. New York: Wiley and Sons.
Marcoulides, G. A. (1989), "Structural Equation Modelling Techniques for Scientific Researeh," Journal of
Business and Society, 2, 2. 130-138.
(1990). "Organizational Culture and Performance: An Empirical Study Using Structural
Equation Modelling." Proceedings of the Decision Sciences Itistitute, 1 (November), 1414-1416.
and R. H. Heck (1991), "Organizational Culture and Productivity: Assessing Theoretical
Models, Frweedings of the Western Decision Sciences Institute, 1 (March), 496-498.
Marsh, H. W.. J. R. Balla, and R. P. McDonald (1988). "Goodness-of-Fit Indexes in Confirmatoiy Factor
Analysis: The Effects of Sample Size," Psychological Bulletin. 103. 3, 391-410.
Moos, R. (1979), Evaluating Educational Environments, Palo Alto, CA: Consulting Psychologists Press.
Ouchi. W. (1981), Theory Z, Reading, MA: Addison-Wesley.
Owens, R. (1987), Organizational Behavior in Education, Englewood Cliffs, NJ: Prentice-Hall.
Perrow, C. (1970). Organization^^ Analysis: A Sociological View, Monterey, CA: Brooks/Cole Publishing
Co.
Peters, T. and R, Waterman (1982), In Search of Excellence: Lessons fmm America's Best-Run Companies,
New York: Harper & Row.
Reynolds. P. D. (I98fi), "Organizational Culture as Related to Industry, Position and Performance:
A Preliminary Report," Journal of Management Studies, 23. 3, 333-345.
Rousseau, D. (1990), "Assessing Organizational Culture: The Case for Multiple Methods," in B. Schneider
(Ed.), Organizational Climate and Culture. San Francisas: Jossey-Bass Publishers.
Saffold. G. S. (1988), "Culture Traits. Strength, and Organizational Performance: Moving Beyond "Strong"
Culture," Academy of Management Review. 13, 4. 346-558.
Schein, E. (1985), Organizational Culture and Leadership, San Francisco: Jossey-Bass.
(1990), "Organizational Culture," American P.'tycholngist, 45, 2, 109-119.
Siepert, A. and R. Likert (1973). "The Likert School Profile Measurements of the Human Organization,"
Paper presented at the Annual Meeting of the American Educational Research Association.
Smircich, L. (1983), "Concepts of Culture and Organizational Analysis," Administrative Science Quarterly,
28. 3, 339-358.
Thompson, J. (1967), Organizations in Action. New York: McGraw-Hill.
Tierney, W. (1988), "Organizational Culture in Higher Education," Journal of Higher Education, 59, 1,
2-21.
Trice, H. M. and J. M. Beyer (1984), "Studying Organizational Cultures Through Rites and Ceremonials,"
Academy of Management Review, 9, 4, 653-699.
Uttal, B. (1983), "The Corporate Culture Vultures," Fortttne (October), 66-72,
Weick. K. (1985), "The Significance of Corporate Culture," in P. Frost, L. Moore, M. Louis, C. Lundberg,
and J. Martin (Eds.), Organizational Culture, Beverly Hills, CA: Sage, 381-389.

Vous aimerez peut-être aussi