Help Needed!!

" Academy of Management /ournal 1995. Vol. 38, No. 4. 107S-1104

ESPOUSED VALUES AND ORGANIZATIONAL CHANGE THEMES

BORIS KABANOFF ROBERT WALDERSEE

MARCUS COHEN University of New South Wales

Using computer-aided content analysis, we assessed the Ktrength of nine values—authority, performance, reward, normative, commit- ment, participation, leadership, teamwork, and affiliation—in 88 large Australian organizations. Cluster analysis was used to catego- rize each organization as having one of four value structures: elite, leadership, meritocratic, or roltegial. Organizations classified as hav- ing din'ereiit value structures difTered in a theoretically consistent way in how their members described organizational change. We discuss the Implications of these findings for the values theory and the use of con- tent analysis for the study of organizational values.

Greenberg called justice "a basic requirement for the effective func- tioning of organizations and the personal satisfaction of tbe individuals they employ" (1990: 399). An impressive body of empirical research sup- ports Greenberg's contention. Stndies have shown that perceptions of dis- tributive justice, which is our major focus here, are related to people's sat- isfaction with their jobs (Scarpello, 1988; Witt & Nye, 1992) and tbeir su- pervisors (Farh. Podsakoff. & Organ. 1988). to their levels of organizational commitment (Price & Mueller. 198G), and to organizational citizenship be- havior (Moorman, 1991; Organ, 1988), as well as to individual, collabo- rative, and organizational productivity (Gowherd & Levine. 1992; Lord & Hohenfeld, 1979; Pfeffer & Langton, 1988, 1993). Yet the very importance of distributive justice for both task and affective outcomes creates an in- herent distributive dilemma within organizations. The goal of this study was to describe this dilemma and to explore its implications for organ- izational values. Two characteristics of this study distinguish it from most of the existing distributive justice literature: (1) it is concerned with the conflict between different principles of distributive justice ratber tban with demonstrating the influence of a single principle, such as equity and (2) it deals with the implications of this conflict at an organizational, or sys-

This research was supported by grants from the Australian Research Council and the Cen- tre for Corporate Change at the Australian Graduate School of Management. We wish to thank Jane Craig. Dexter Dunphy. Robert House, and Raymond Zammuto for their valuable com- ments.

1075

1076 Academy of Management Journal August I

temic, level rather than at the level of individuals' justice beliefs and per- ceptions {cf. Sheppard, Lewicki, & Minton, 1992).

THEORETICAL BACKGROUND

The Distributive Dilemma

All organizations need to solve a fundamental problem—how to main- tain internal cohesion while producing economic outputs. Polley de- scribed this "task versus person" (1987: 424) conflict as one of the more remarkable convergences within the broad field of organizational behav- ior and as a fundamental source of tension within tbe "deep structure" (Gersick, 1991: 13-16) of organizations. But what is it about tbe task ver- sus person contrast tbat creates tension? We argue tbat a key source of tbe tension lies in the competing distributive justice principles implicated in these two sets of concerns.

To meet tbe requirements for cobesion and performance, organizations adopt different patterns of structures, processes, and values; bowever, meeting the dual goals of cobesion and performance creates a justice dilemma involving tbe kinds of structures, processes, and values tbey sbould have (cf. Sheppard et al, 1992). Tbe cohesion requirement leads organizations to adopt solidaristic, egalitarian values and practices, but tbe need for productivity leads them toward efficiency- and equity-based ones (cf. Deutsch, 1985; Meindl, 1989). The more tbat systems adopt an effi- ciency orientation, the more equity values will predominate and organ- izational resources and rewards will be differentially or unequally allo- cated. Tbus, inequality and consequent threats to social cohesion or inte- gration become problematic in the system (Kabanoff, 1991a). The more tbat an egalitarian orientation predominates, however, tbe more egalitarian practices and allocations will tend to be. Economic inefficiency and in- equity emerge as pressures or problems here. In sbort, to meet tbe need for economic efficiency, organizations adopt values that mandate allocat- ing their resources equitably; to maintain internal cobesion and solidari- ty, however, they adopt values tbat mandate allocating tbeir resources equally (Meindl. 1989). The different ways in which organizations try to find a sustainable balance between these competing values shapes tbeir overall value structures.

Implications for Organizational Values: A Typology

Values can be defined as generalized, enduring beliefs about tbe per- sonal and social desirability of modes of conduct or "end-states" of exis- tence (Rokeacb, 1979; Schwartz, 1992). The values concept is a powerful one because it can be meaningfully employed at all levels of social analy- sis—cultural, societal, institutional, organizational, group, and individual. Williams, for example, defined an institution as "a set of institutional norms tbat cohere around a relatively distinct and socially important complex of values" (1951: 29).

1995 Kabanoff. Waldersee. and Coben 1077

A "complex of values" can also be conceived of as a value structure. Following Schwartz (1992), we defined a value structure as the overall pat- tern of relations among a set of values, containing elements of botb com- patibility and conflict among tbem. A value structure is different from a value bierarchy because tbe former includes compatibility and conflict, and the latter is simply a priority-based ordering. The notion of botb com- patible and conflicting value relationships is important for understanding iiow the distributive justice dilemma tbat has been described sbapes organizational value systems.

Kabanoff (1991a, 1991b, 1993) described a typology of four ideal types (cf. Doty & Glick. 1994: 233) of organizations representing distinct- ly different ways of dealing with the conflict between equity and equali- ty. The more unequal or centralized the power structure is in an organ- ization, the more equity values (or an equity orientation) predominate, and the more resources and rewards will tend to be unequally allocated (e.g., Hedstrom, 1988). In organizations with decentralized, democratic power structures, an egalitarian orientation and more equal allocations will tend to predominate. Although organizations' power structures underlie their orientation toward distributive equality or equity, organizational process- es tben eitber reinforce or compensate for this orientation. Thus, an or- ganization with an egalitarian orientation might reinforce this structural tendency through processes such as a pay policy tbat compresses salary differentials or compensate for it tbrougb processes tbat increase differ- entiation, sucb as a policy linking pay to performance appraisal. In effect, organizations can ignore tbe tensions inherent in their equal or unequal structures and use their processes to reinforce their dominant orientation, or tbey can recognize tbese inberent tensions and seek to balance tbem, using a set of compensatory processes.

The equity and equality ends of tbe two dimensions (structure and process) define tbe four ideal types. Figure 1 sbows tbis typology, de- scribing four value structures deduced from tbe tbeory and reflecting tbe particular structure-process combination characteristic of each ideal type.

For example, the elite value structure represents the pure unequal type in which unequal power values are combined witb inequality-oriented eq- uity values. Tbe elite value structure emphasizes unequal power rela- tions (authority), deemphasizes egalitarian power values {participation, normative), and deemphasizes cohesion values (affiliation, teamwork, commitment, leadersbip), while emphasizing performance and reward. In contrast, the leadership type represents a mixed, compensatory pattern that retains the elite typo's unequal power orientation but superimposes on it a set of cobesion values that are nevertheless consistent with unequal pow- er relations, namely, leadership, teamwork, commitment, and affiliation. As Kabanoff observed, "Leadersbip bas a paradoxical or dualistic quali- ty—it both glorifies inequality and the differences between the leader and the led, while at the same time it creates identification and cobesiveness between the leader and bis or her followers" (1991a: 433-434). In a sim-

1078 Academy of Management Journal August

FIGURE 1 A Justice-Based Typology of Organizational Value Structures'

Process reinforces structure UnequEil

Power STRUCTURE Equal Power

Equitable

PROCESS

Egalitarian

•̂ A Elite

Strong Values

Authority ^Performance

Reward

Weak Values

Teamwork Leadership Participation Commitment Normative Affiliation

Leadership ^'

Strong Values

Authority Leadership Teamwork Commitment Performance Reward Affiliation

Weak Values

Participation Normative

Meritocratic M

Strong Values

Teamwork Participation Commitment Performance Reward Normative Affiliation

Weak Values

Authority Leadership

-

~'A CoUegial

Strong Values

Teamwork Participation Commitment Normative Affiliation

Weak Values

Authority Leadership Performance Reward

\^

Process compensates for structure

"Strong values are those strongly or frequently espoused by organizations with Ihis type of value structure, and weak values are those weakly or infrequently espoused.

ilar way, the meritocratic type combines equity and equality concerns by superimposing a set of equity-oriented values on a pure, egalitarian col- legial type.

An Initial Test of Validity: Differences in the Change Themes of Organizational Types

We assume that, to the extent that the proposed distributive jus- tice-based typology is valid, it should be able to inform us about signifi-

1995 Kabanoff. Waldersee. and Cohen 1079

cant, value-related processes in organizations. Specifically, organizations that differ in their value structures should differ in ways that are consis- tent with their values when their members discuss an issue like organ- izational change. We focused on orgBnizational change because values are likely to he highly salient when organizations address this key organiza- tional process. Current organizational values are likely to influence the kinds of change goals and means that are considered and discussed (Trice & Beyer, 1993), and they can also influence the kinds of shared values and themes to vî hich change agents will appeal to create a motivation for change among organization members (Shamir, House, & Arthur, 1993). Kan- ter. Stein, and Jick put it this way: "Change implementers must be con- cerned not only about changing to what: they must also be concerned with changing/rom what" (1992: 503). To date, few empirical studies have ex- amined how organizational values are related to how change agents dis- cuss and portray organizational change. Therefore, the issue of how val- ues relate to the way in which change issues are portrayed in organizations has substantive interest as well as being a way of examining the validity of the typology.

To summarize, the major aim of the present study was to explore the ability of a typological theory of organizations and organizational values to account for the differences in value structures that we observed in a sam- ple of actual organizations. We accomplished this aim by using content analysis to describe the values that organizations espouse in their docu- ments and then clustering organizations according to their overall value profiles. We then investigated whether those profiles were related to por- trayals of change themes and issues in organizational documents. Our ex- pectation was that change issues would be portrayed differently in organ- izations with different types of value structures and that those differences would be consistent with our interpretation of organizations' value struc- tures,

METHODS

Content Analysis

Rokeach (1979) explored five different means for measuring organ- izational values, one of which was content analysis. The core assumption underlying the use of content analysis for describing organizational val- ues is that organizations leave traces of their distinctive value patterns in their documents and that these traces can be observed and measured. Measurement is accomplished by counting the frequency with which dif- ferent values are referred to in the text being analyzed. Frequent references are interpreted as an indication of the values' importance or centrality (Huff, 1990).

We used computer-aided content analysis (cf. Kabanoff, forthcoming; Stone, Dunphy. Smith, 8t Ogilvie, 1966; Weber, 1985: Wolfe, Johnson. & Gephart, 1993) rather than a manual method. A computer checks the

1080 Academy of Management Journal August

words in the text under analysis against the words specified in a content dictionary, which consists of meaning categories, each of which contains words that are held to refer to the value theme being measured. The text analysis program rates each sentence according to whether it contains one or more of the dictionary categories.

Both manual and computer-aided content analysis offer a number of advantages for studying organizational values. These include their nonob- trusive character, their use of naturally evoked verbal behavior as the source of data on values, their suitability for longitudinal research, which stems from the availability of different kinds of text over long periods of time, and their systematic and quantitative approach to dealing with qual- itative data. To these, computer-aided text analysis adds the following ad- vantages: (1) It offers perfect reliability—classification of text by multiple human coders permits the level of reliability achieved to be assessed, but classification by computer leads to perfect reliability since the coding rules are always applied in the same way. (2) It uses standard dictionaries; over time, a number of generic content-analytic dictionaries have been devel- oped containing categories useful in many social science contexts, and use of these can enhance both the comparability and validity of analyses. (3) It is efficient; a major drawback of manual methods is that they are labor- intensive and inflexible and thus highly unsuitable for exploratory work. A computer-based dictionary can, however, be fairly readily refined as a researcher's knowledge about a text increases, and the dictionary can be reapplied to the same or to new text.

A criticism of computer-aided methods is that they are less context- sensitive than manual analysis—that human coders can better judge a word's meaning within a context. However, research indicates that this ap- parent advantage of manual coding is less clear than has been supposed. Rosenberg, Schnurr, and Oxman (1990) found that computer-aided text analysis using categories from the Harvard IV Psycho-Social Dictionary outperformed human coders even in a task likely to be highly context-sen- sitive—using samples of speech to assign psychiatric patients to a diag- nostic category. As Rosenberg and colleagues concluded, "The most par- simonious explanation [for the superiority of computer-based methods] ap- pears to be that the reliability of computerized systems gives them an advantage in classifying texts, even if their classification schemata are based on the sheer number of utterance types in a text. . . . it may be that the reliability of the system becomes an overriding factor if quantification is desired" (1990: 307).

Sample

Eighty-eight large, well-known Australian organizations supplied suf- ficient information to be included in the analysis. This sample, in which median total assets in 1986 were approximately $0.73 billion in U.S. dol- lars and the median number of full-time employees in 1986 was 4,243, pro- vided a general cross-section of industries without being fully represen-

1995 Kabanoff, Waldersee, and Cohen 1081

tative. Organizations came from both the private and not-for-profit sectors and included publicly listed companies {N = 49); unlisted companies un- der foreign, private, or mutual ownership {N - 16); government depart- ments (N = 7); universities {N ~ 8); and government-owned businesses [N = 8). Those that had been taken over, undergone a merger, or been placed in receivership during the years 1986-90 were excluded since equivalent documents were not available for these companies over the whole study period.

Source of text data. Like most previous researchers (e.g.. Bowman, 1984; D'Aveni & MacMillan, 1990; Staw, McKechnie, & Puffer, 1983), we used public documents as the source of text data, although we used a wider range of documents than is typical. Four kinds of organizational documents were gathered: annual reports, company-wide internal magazines, mission or corporate-values statements, and any other organization-wide docu- ments produced for distribution to employees, such as regular or irregu- lar newsletters. Organizations were asked to provide all documents of these types or as many as they could for years 1986-90. Over 1,000 documents were collected.

Our concern was with organizational values rather than individual values. Documents such as annual reports are clearly likely to represent the value consensus among an organization's senior managers more than the personal values of individuals, though these are likely to play some role also. For analytic purposes, all sections in the documents that referred to organizational goals and values, such as human resources management policies, statements of corporate philosophy, management overviews, CEOs' annual reports or letters to shareholders, and so on, were identified, including any "human interest" stories that made mention of "what it means to work for this organization." All sections that were purely or large- ly financial, technical, or descriptive were excluded.

Development of a new dictionary. We found it necessary to develop a customized value dictionary to measure the majority of the values spec- ified by the theory, although two existing dictionaries provided us with content categories for measuring two of the focal values. From the Harvard IV Psycho-Social Dictionary mentioned previously, we used a category called "affiliation" (see Table 1), and from the Lasswell Value Dictionary we used a category called "rectitude ethics" to measure the frequency of normatively oriented terms (Ziill, Weber, & Mohler, 1989; see Table 1). We created and validated the other seven value categories using a text-analy- sis software package called Textpack 4 (Mohler & Zuell, 1990). The vali- dation was done by using a small proportion of the total text database used in this study to develop and pilot-test the dictionary (Kabanoff, 1991b). This process involved using Textpack 4 to retrieve all sentences identified as containing any word referring to a target value, ensuring that the cor- rect sense of the word was being used, and if necessary defining "exclu- sion categories" that represented occasions of inappropriate word use. In- appropriate uses were assigned to separate categories and then subtract-

1082 Academy of Management Jaumal August

t eta

u . 1 - 1

Q Vi

CA

CQ

• i d

3 a o U

c o — c °̂ M o " § 1

00 . -

m

u .t: -S c. C

•I I

5 (0 TO "H

CJ C

E S

C C B B IH (H [-• F H 0 1 <U <U m u u u u c! d B C o o o o U U U U

O 03

J3 -a

J : B < CD

J 3 ^ en oi

g = t -2 W r/\ CC C l . . - U

Q (0 '^^

TO CO

B OJ c

9 d 2 w —( CO O N E N . ^

DC g DO 2 O Cu O 3

CO J 3

X o i2

B B 01 0 CJ U U B B Bo o o U U U

3) lU

c 0 CO Q .

u *-» CO 1 ,

a

B S 6 o u

o B CO

B >-•o

o CO

<

>

CO

hi

0 2

1995 Kabanoff, Waldersee. and Cohen 1083

ed from the correct-use category to provide a final, valid count for each value category. Table 1 shows the nine value categories and provides ex- amples of words from the texts that were assigned to each category.

RESULTS

As Weber (1985) recommended, the unit of analysis was tbe sen- tence. A total of 40,593 sentences were analyzed. The mean number of sen- tences per organization was 448.42 (median = 375), with a maximum of 985, and a minimum of 216. Annual reports provided an average of 292.9 sentences; internal magazines. 150.8; other sources. 94.7; and mission statements, 44.1 sentences.

We noted tbe average frequency of occurrence per sentence—or tbe density—of eacb type of value within eacb organization and for each doc- ument type. Tbese average frequency scores were tben standardized to a Z-score form as a control for potential differences in value frequencies be- tween different document types tbat could otberwise disadvantage organ- izations that did not have a particular type of document. Finally, we cal- culated an overall mean for eacb organization for eacb value by averaging the standardized frequency scores across all of tbe document types. If a company bad no documents of a particular type, we simply ignored tbat type in calculating tbe average score. Tbese averaged Z-scores were once again standardized.

Identifying Organizations' Value Structures

Value profiles. On tbe basis of our specification of four value struc- tures (Figure 1), four prototypical value profiles were postulated. In eacb profile, the nine values were defined either as strong (frequently men- tioned) or weak (infrequently mentioned). Weak values are not unimpor- tant since tbe character of a value structure is determined by botb its strong and weak values. For example, as explained earlier, the egalitarian, col- legial value structure differs from tbe meritocratic structure only in tbat the former does not emphasize tbe equity-oriented values of performance and reward. Anotber example is tbe leadersbip value structure, which em- phasizes five of tbe same values as tbe meritocratic, sucb as teamwork and commitment, but within the context of authority and leadership values that imply an underlying inequality of power, rather than within tbe merito- cratic structure's egalitarian context of normative and participation values. A statistical technique suited to tbe task of identifying cases that bave sim- ilar patterns across a number of attributes is cluster analysis (Aldenderfer & Blashfield. 1984).

Cluster analysis. Typically, cluster structures or solutions are identi- fied by an exploratory, empirical process. However, it is possible and even preferable to use cluster analysis in a more theory-driven fashion (Dowling & Midgley, 1988: 835). This process involves a priori specifying tbe expected cluster solution and tben clustering the data to test the va- lidity of this tbeoretical structure, rather than using tbe clustering program

1084 Academy of Management fourna! August

itseif to test for the hypothesized structure. There are both theoretical and empirical reasons for preferring a more tbeory-based approach. Although the present study explores the ability of ideal types to describe differences in the value structures of actual organizations, organizations are not con- sidered to be actual examples of the ideal types (cf. Doty & Glick, 1994: 233); ratber, they are real-world approximations to tbe ideal types. The is- sue is not whether organizations are ideal types but whether researchers can identify groups of organizations tbat approximate different ideal types. Since actual organizations are not ideal types, an exact correspondence be- tween theoretical and actual value profiles is not likely. On theoretical grounds, ideal types represent forms tbat might exist rather than forms that actually exist, so empirical examples may be rare or nonexistent (cf. Doty & Glick, 1994; 233). Actual organizations' value profiles may only ap- proximate ideal types because organizations are subject to various conta- minating influences or are in transition, moving from one type toward an- other. Or it may even be that tbose espousing given values are not fully aware of or able to articulate tbeir underlying assumptions. Inductive clustering can fail to identify the theoretically significant structure of val- ues in a sample because of its susceptibility to empirically real but tran- sitory influences that cloak the underlying tbeoretical structure (cf. Ketcben, Tbomas, & Snow, 1993).

Quick Cluster (SPSS, 1988) is a procedure that can carry out a form of tbeory-driven clustering, witb its user defining initial cluster centers for different clusters on the clustering variables. A cluster center is an esti- mate of tbe average value of eacb clustering variable for the cases in a clus- ter, and as noted, is usually derived empirically by tbe clustering proce- dure Itself. In tbis case, we employed Quick Cluster to cluster organiza- tions on tbe basis of their value scores by setting initial centers for eacb of the four value structures on the nine clustering variables (tbe value scores) based upon their theoretical value profiles. That is, the prototyp- ical value profiles (Figure 1) were used to define templates for the selec- tion of organizations into different clusters.

For a relatively strong value, the selection criterion (the initial value center for the cluster) was set at 1.0; since scores had been turned into Z, tbe value score of organizations to be selected for that cluster was tbere- fore initially set to be one standard deviation above the overall mean. Sim- ilarly, we set the initial center for a weak value at -1.0, or one standard deviation below tbe sample mean. For example, tbe selection profile for organizations with an elite value structure was as follows: affiliation (-1.0), authority (1.0), commitment (-1.0). normative (-1.0). leadership [-1.0], participation [-1.0), performance (1.0), reward (1.0), and teamwork [-1.0). If no initial cluster centers are specified. Quick Cluster proceeds by selecting k cases with well-separated, nonmissing values as initial cen- ters witb wbich to begin classifying cases, where k is tbe number of clus- ters requested. However, when initial cluster centers are provided. Quick Cluster proceeds by searcbing for cases that best approximate tbe initial

1995 Kabanoff, Waldersee. and Cohen 1085

templates, using a squared Euclidean distance measure to assign an organ- ization to its nearest cluster. Tbus, all of the values in tbe template, weak and strong, are important for assigning organizations to clusters. Since the squared Euclidean distance measure equally weights all clustering vari- ables, standardized forms of the clustering variables are tbe appropriate ones to use. Here, as cases were added to each cluster, the program updated the selection criteria to tbe value means for the cases (organizations) tbus far placed in a cluster. The final step of tbe algoritbm was to reassign eacb case to the nearest updated cluster center. Tbis reassignment yielded the final clusters, and tbe final cluster centers (depicted in Figures 2-5) were tbe variable means for tbe cases in tbe final clusters.

Tbus, although the initial selection criteria for the clustering proce- dure were theory-based, as organizations were added to a cluster tbe se- lection profile became the average of tbe value profiles of the organizations selected into the cluster. Tbe final value profile of eacb cluster therefore represented the outcome of a selection process that was initially based solely on theory but then came under tbe influence of the actual value pro- files of organizations selected into clusters. Several aspects of tbis updat- ing procedure are important to note. Quick Cluster uses an iterative, "K- means" type of algorithm, also called nearest centroid sorting, to assign each observation to tbe cluster with tbe nearest centroid (Aldenderfer & Blasbfield, 1984). Punj and Steward (1983: 143) judged the K-means ap- proach to be superior to most otber clustering methods in terms of its over- all robustness. Tbe criterion it uses for the final assignment of cases to clus- ters is the minimization of the sum of the squared Euclidean distances of observations from tbe cluster means or cluster centers. Thus, updating is a requirement of the iterative, distance-minimizing algorithm of the K- means type procedure; assigning organizations on tbe basis of tbeir simi- larity to the theoretical templates without updating would not meet the minimizing criterion and therefore would not identify tbe tightest, most homogeneous clusters that it was empirically possible to identify. The up- dating procedure also played another important role, since our aim was to test a theoretically specified cluster solution. Updating helped find tbe optimal fit between tbe theoretical cluster structure and an actual cluster structure witbin the sample. It did tbis by initially assigning organizations to clusters on tbe basis of their similarity to the ideal templates and then progressively modifying tbe templates to reflect tbe profiles of actual organizations. Because tbe procedure updated tbe center of eacb cluster to the mean for cases thus far in the cluster, as cases were processed the centers migrated to concentrations of observations. If few or no organiza- tions could be found tbat matcbed the theoretical templates to any great extent, we expected that over tbe course of tbe clustering process the val- ue profiles of clusters would start to resemble one another or lose any re- semblance to the original templates, or both. On the other band, the ex- tent to wbicb tbe process succeeded in identifying clusters with value pro- files tbat retained a meaningful resemblance to the ideal templates was

1086 Academy of Management fournal August

evidence that the tbeoretical cluster structure provided a feasible, cluster solution for the sample.

The results of this updating procedure were compared with those ob- tained wben updating was not used. Doing the clustering witbout updat- ing meant that the theoretical templates were used throughout tbe initial classification process to assign organizations to clusters without being modified to reflect tbe value scores of organizations actually being assigned to clusters. We used updating only in tbe final assignment of cases to clus- ters, as the algorithm mandated. This procedure maximized the influence of the theoretical templates upon initial assignments and final cluster centers. Cross-tabulating the cluster memberships derived following these different procedures indicated that 81 percent (71/88] of tbe organizations were assigned to the same clusters, whicb gave us some confidence in the stability of our clustering results.

We further investigated the coherence and stability of tbe cluster so- lution by repeating the cluster analysis in several different ways and com- paring tbe cluster membersbip of organizations from eacb procedure. To the extent that cases were assigned to the same clusters wben different sub- sets of cases were used to develop the selection criteria for cluster as- signment, we could be more confident that the results were not reflecting tbe influence of a particular starting point (a particular case) or particu- lar subgroups.

We first repeated tbe cluster analysis (with updating) on two randomly selected 50 percent subsamples of tbe organizations. Ninety percent (43/48) and 88 percent (42/48) of the organizations were assigned tbe same clus- ter membersbip in the subgroup analyses as tbey had been in the overall analysis, indicating that cluster results did not depend upon particular sub- sample characteristics. Next, we used the final cluster centers found in each subgroup analysis as bases for reassigning tbe wbole sample into clus- ters. Similar assignments using tbe tbeoretical templates and tbe final clus- ter centers from the two subsamples would constitute further evidence of the stability of the cluster solution. This process resulted in 76 percent (67/88) and 70 percent (62/88) of the whole sample being assigned to tbe correct cluster—tbe same cluster they were assigned to by use of the the- oretical templates—against a chance likelihood of correct assignment of on- ly 20 percent.

Value differences between the clusters were then tested in several ways. Results of a multivariate analysis of variance (MANOVA) indicated that tbere was a significant overall difference between clusters [p < .001). Univariate F-tests of differences between clusters were in all cases signif- icant (p < .05). We examined tbe ability of the value dimensions to assign organizations to their correct clusters on tbe basis of their value scores us- ing discriminant analysis (Dowling & Midgley. 1988). Discriminant analy- sis using the original nine clustering variables in a single block assigned 90.28 percent (65/72) of tbe organizations to tbeir correct clusters, in-

1995 Kabatwff, Waldersee, and Cohen 1087

eluding 100 percent of the elite types and 85.7, 85.7, and 86.7 percent of the meritocratic, leadership, and collegial types, respectively.

These results demonstrated that even using different subsamples of organizalions and somewhat different procedures for clu.storing, assign- ments of organizations to clusters were relatively stable and reliable. Clus- ters were also quite well differentiated by the value measures, and most organizations could be reliably assigned to their respective clusters on the basis of their value scores. That this degree of reproducibility and stabil- ity was achieved despite the relative smallness of the sample was grounds for believing that the cluster structure we had identified was adequately robust and empirically and theoretically coherent. Although these results cannot be taken as proof that the particular cluster structure we identified really existed among the organizations (Aldenderfer & Blashfield, 1984), the results did not support the view that the cluster structure we identi- fied was an arbitrary and therefore unreliable solution forced upon the da- ta by the procedure we employed.

Interpreting Cluster Profiles i

Figures 2-5 summarize the mean value profiles—that is, final cluster centers—of the organizations assigned to each theoretical type.

Initial explorations of the data indicated that some organizations were consistently below the mean on all or virtually all of the value measures. That is, these organizations did not espouse any of the nine values with average or above-average frequency. Assigning these organizations to any of the four value structures was clearly not meaningful. Therefore, we spec- ified a fifth value profile for a nonclassifiable cluster with assigned ini- tial centers of -1.5 on all nine values. A total of 16 organizations were thus assigned and not considered further in the present study. The 72 remain- ing organizations were distributed across the four clusters as follows: elite. 22, leadership. 21, collegial, 15. and meritocratic, 14. There ap- peared to be a preponderance of elite and leadership types, but the type f"requencies were not significantly different according to a chi-square, one-sample test (x ,̂ = 2.78, p = .42, N - 72).

In terms of the direction of each value's deviation from the mean across all nine values, the theoretical and actual profiles matched quite well. The elite profile matched on eight of the nine values, with no em- phasis on performance found in the actual sample, and the highs and lows of this profile were also not as pronounced as for the other three types. The meritocratic type also matched on eight of the nine values, although the actual profile did not show the expected weak authority value. The strong authority and affiliation values predicted for the leadership profile were not found, whereas strong leadership, teamwork, and commitment values were. The expected strong commitment and weak authority values were not found in the collegial profile, and the teamwork value was quite neu- tral, but as expected there were strong normative, participation, and af-

1088 Academy of Management fournai August

Values

M > Teamwork -

\ ^. Reward —

\ ^ Performance —

Participation -

Leadership -

Normative -

Commitment -

. Authority -

Affiliatioii -

FIGURE 2 Value Profile of Elite Type

- n . 4 5 ^ ^ ^ 1

-0,15 H

-0.22 ^ H

^ ' -0.20 ^ ^ H

> / -0.35 ^ ^ H

-0.29 ^ ^ 1

-0.45 ^ ^ ^ 1

0,3

0.31

- 2

filiation values and a weak reward value. The value profiles of several clus- ters also suggested that some value profiles contained a smaller subset of values that provided a relatively tight and distinctive pattern for identi- fying a particular value structure, such as the comhination of strong nor- mative, participation, and affiliation and weak reward values for the col- legial group and the combination of strong leadership, teamwork, and com- mitment values for the leadership group.

Evidence of Demographic Differences Between Clusters

There was little indication that cluster membership was associated with size, either in terms of total assets or numbers of employees; how- ever, there was evidence of an association with industry. The cross-tabu- lation of organizations' cluster membership score by industry classifica- tion resulted in a highly significant overall chi-square (x̂ jg - 91.45, p = .001). The two strongest associations were highly compatible with the the- ory underlying the typology. The majorities of both universities (6/8) and government departments (4/7) were classified as collegial. The meaning of the other associations was less clear; nevertheless, four out of five chemical companies were classified as in the leadership cluster; four of six mining companies were in the unclassified group; and the four largest

1995 Kabanoff. Waldersee, and Cohen

FIGURE 3 Value Profile of Leadership Type

of the six banks in the sample were classified as elite (cf. Kabanoff. 1993). Although there was evidence of industry-related differences in organiza- tions' value structures (cf. Chatman & Jehn, 1994), it was also evident tbat industry could not account entirely for the differences we identified. On- ly 5 of 16 industry groups showed a significant association with one or more type of value structure, and in no case were all organizations from one industry assigned to the same value structure, the two strongest as- sociations being found for the chemical group (4/5) and the universities (6/8). That is, industry was not a proxy for value structure, although there were some interesting associations that deserve future investigation.

We have provided evidence of our ability to empirically identify groups of organizations with value profiles that reasonahly match the pro- files specified by theory and have shown that these clusters have satis- factory empirical robustness and coherence. Aldenderfer and Blashfield (1984: fi6) argued that one of the best ways of providing evidence for the real, or construct, validity of a cluster solution was to show that clusters differed in theoretically or practically meaningful ways on variables that were not used to carry out the initial clustering. We argued earlier that to the extent the typology is valid, it should be able to inform us about a sig- nificant, value-related process in organizations—organizational change.

1090 Academy of Management fournal August

Values

Teamwork -

Reward —

Performance -

Participation -

Leadership —

Normative —

Commitment -

Authority -

Affiliation -

FIGURE 4 Value Profile of Meritocratic Type

-0.12

- 2

0.76

0.66

Sample size precluded the use of a holdout sample, or derivation of the cluster structure from one sample and testing of differences in change themes with another. Nevertheless, the procedure used here may not be particularly problematic because the clustering and the validating variables were conceptually quite distinct. Such a procedure is commonly used in marketing: analysts identify different groups or clusters of consumers (organizations) using a set of demographic [values) variables, then at- tempt to show that these groups vary in terms of a number of other atti- tudinal (change) variables that are conceptually distinct from the cluster- ing variables.

Cluster Membership and the Portrayal of Organizational Change

Change was defined broadly in this study as including all words that indicated variance from an existing condition. Typical change words were change, revise, vary, and alter. We read a sample of sentences to ensure that use of these change words actually related to some aspect of change that was of relevance to the organization, such as change in structure, products, markets, customers, and strategies. The frequency of change words was then correlated with the frequency of a number of other content categories taken from the Harvard and Lasswell dictionaries. Between them, these

1995 Kabanoff. Waldersee, and Cohen

FIGURE 5 Value Profile of Collegial Type

1091

general dictionaries contain around 150 meaning categories, but most of these have little relevance to organizations and change. Thus, on an a pri- ori basis, all the categories (a total of 24) that were potentially relevant for descrihing change in organizations were identified. Categories referring to goals, means, and failure were obviously potentially relevant (see Table 2 for a full list of the categories), and we also included six categories based upon Osgood's (Osgood, Suci, & Tannenbaum, 1957) well-known seman- tic differential dimensions (i.e., good-bad, weak-strong, passive-active). These were considered useful for exploring differences in general affec- tive tone associated with change in different organizations. The frequen- cy of cbange references was then correlated with these other content themes within each cluster to determine whether change in different val- ue structures was associated with different themes. Correlations were in- terpreted as an indication of a connection between cbange and these oth- er meaning categories (Huff, 1990).

Table 3 shows Spearman rank-order correlations between change con- cerns and each of the other content themes for the sample as a whole (ex- cluding the 16 nonclassifiables) and separately for each of the four clus- ters. Spearman correlations were preferred because of the relatively small number of organizations in each group.

1092 Academy of Management fournai August

u

lA £

3 > m 03

^ m O O

§•?:

03 X I

Q U

d " fcv-=: -a

' C

g S (- a

d o

a (C V Q

d c d

C d d o

' ^ "^ ^

T i r̂ -^^ C

< u

.« to

< U

|

.i § a B

QJ 0 0

5I a 1

> CO T ].5 - d >- aJ

U (0 u ,!2 CO

01 P Z3 H S " CO 03 00 h n t ; . « • " .M > w* (J Ĵ O

.;:; K 03 u H- * . * tC Tn . ^ j . ;3 "0.-6

w CO

3 d

O

a a

•— " i J CO

o

^ d

03' 03

•« d 03 O

Q U

cu -^ n

d .-̂ o a 'r! , 03 to

§ ? c ai e e 03 S > V

d' —

"̂ 1 o

0

0

p ro

_^

d 03

>. to'

f l .

03

t o

m al

l

t n

n o

r,

a 0

03 ^~*

0 t x>

at e.

d

rd i

0

a

M

C<]

CO en

-d

li g

e

Q

ti es

• >

'if-

1 "o u

a0 « u •a

a 0

u

d Q

- s p —

u u.

al s

0

\ \

— 03

d «CO to a 13

a h

\ \ » \CO d to 03

s

03

•. 10 " i ^ •-T3 to •ffl DO 5 03

2 ^

UI >H

3

03

n

er si

>

03

CO w to

IX

3 en to 0

K.

la u

" 0 OH

03

0 0 .

1995 Kabanoff, Waldersee, and Cohen

\

1093

TABLE 3 Spearman Correlations Between Change and Other Content Categories

by Value Structure

Theme

Activt! ColiBclivities Communication C^imptetion Failure Goai.s rtuman [ncrease S^eans Meeds SJegalive Ours Overstate Passive Pleasure Political Positive Power Strong Time Transaction gain Understate Wntik

Total

.48*"* ,26* ,16

- . 1 0 .18 .02 .27*

- . 2 3 * .40*** .40***

" ,26** - . 2 8 *

.01

.18* - . 2 1 *

.36***

.27*

.44***

.32'* - . 1 5

.04

.11

.09

Elite

,10 .04

- . 1 4 . 0 7

.13 -.12 - . 0 4

.18

.14 - . 0 1

.37* - . 1 0 - . 0 5

.32

.05 - . 0 8 - . 2 1

.06

.44*

.17

.18

.08

.21

Leadership

.18

.25

.33 - . 0 8

.23

. 0 5

.27 - . 4 2 *

.05

.18

.39'

.02

.08

.10 - . 0 7

, 1 9

, 1 8

. 1 6

- . 0 2 . 0 7

. 1 9

. 3 6

.53*

Meritocralic

,61* .39 .11

-.22 .44 .44 .27

- . 1 7 .44 .68** .46"

- . 0 3 .55* .40

- . 0 5 .31 .42 .38 .37

- . 0 9 - . 1 2

.20 - , 0 6

Collegial

.68*

.31

.03

. 2 0

.45*

. 4 1

.26 - . 3 9

.73**

.46*

.09

- . 7 9 * * * - . 1 6 - . 0 6 - . 6 7 * *

.47* ,48" •60** .38

- . 4 5 * .14

- . 3 0 ,03

•• Ns are 72, 22, 21, 14, and 15 for the value structure types in the order given above. * p < ,05 two-tailed test

•* p < .01 two-taili'd test •** p < .001 two-tailed test

Table 3 generally supports the view that change is portrayed and communicated differently in organizations with different types of value structures. Although some themes were significantly associated with change in each type, no theme was associated with change in all four types. For example, although negative themes were associated with more change references in three clusters, they were uncorrolated with change in the col- legial group (see Table 3). Weber stated that "interpretations of statistical manipulations based on quantified text musl be validated by reference to the text itself" (1985: 62; emphasis in the original).

Relating Correlations to Actual Text

A key-word-in-context (KWIC) analysis was conducted using Gener- al Enquirer, a program, to list all sentences that contained references to hoth change and the themes with which change was correlated in one or more of the clusters. For example, all sentences containing both a change reference and a means reference were retrieved. One of us then read these

1094 Academy of Management fournal August

sentences separately for organizations from different clusters and sum- marized the main themes separately for each type. Using examples of how words were used from the text, this author then wrote a general nar- rative summary of how change was portrayed in the documents of a par- ticular group of organizations. The other two authors then studied this summary plus 10 percent of all the sentences that were retrieved. This ap- proach allowed us to develop a qualitative understanding of how change was perceived in organizations with different value types. This is a struc- tured form of qualitative analysis that is guided by empirically estah- lished associations within a text hut also involves use of a degree of in- terpretive skill and suhtlety in characterizing differences. In general, neg- ative correlations were not interpreted qualitatively. Negative correlations can reflect the more frequent use of alternative words or can indicate that certain concepts are particularly irrelevant to the organization describing change. KWIC analysis cannot he meaningfully used to select sentences in which a word is not present. However, in some instances negative corre- lations were interpretable by studying patterns of correlations and the text, and these instances are reported here. In the qualitative interpretation of the correlations in Table 3, reported next, we will italicize the names of the actual content categories.

How change was portrayed in elite organizations. Elite organiza- tions tended to portray change as heing necessitated by negative circum- stances, such as declining profitability and tough competition; an exam- ple is "Declining profitability was contained with a management restruo ture." Furthermore, change tended to he portrayed as a negative or potentially negative experience: "would have left deep scars upon the com- pany." Change involved concerns and fears about people's ability to cope with the threat and insecurity attendant upon change: "I'm clearly aware lof] the areas which will be most sensitive and most concerned and won- dering 'what's going to happen to me?' " Changes also tended to he de- scribed somewhat combatively and harshly: "To attack the chronic level of overstaffing." Elite organizations also associated change with strength. offering reassurance, and giving a sense of scale, size, and importance to change. Examples were: "I believe the foundations for a secure, commer- cially based business have heen laid down with the change in direction" and "The new program represents a major shift in the philosophy of train- ing one of the State's largest workforces."

How change was portrayed in leadership organizations. Leadership organizations also viewed change negatively; it was described as "diffi- cult," "confronting," "hard-hitting," "stressful," "tense," "painful," "un- comfortable," and so on. As in elite organizations, the impetus for change was generally viewed as negative; examples include "cost conscious," "competitive," "cost pressures," "difficult trading environment," "costly," and "uncertain economy." Managers were portrayed as important change agents and models; an example is "Employees are unlikely to change their attitudes to productivity and competitive performance until those of their

1995 Kabanoff, Waldersee, and Cohen 1095

managers change" and "Only management can change the system and most problems are in the system."

The correlation between cbange and weait terms indicated several is- sues. First, leadership organizations explicitly focused their change con- cerns upon employees; exemplary statement.s include "The employees' role in improving quality, and the company's commitment to making it hap- pen" and "The involvement of all employees in training and quality im- provement processes." A second feature was that leadership organiza- tions, more than others, seemed to differentiate explicitly between man- agers and other employees; for example: "an ongoing program for all employees . . . line managers, and middle managers is seen as the most pro- ductive means" and "line managers are involved in the preparation and delivery of course material to their subordinates." That is, although elite systems focused on organizational needs and tended not to distinguish be- tween managers and nonmanagers, leadership sy.stems distinguished be- tween managers and nonmanagers and assigned them different roles in change.

How change was portrayed in meritocratic organizations. Merito- cratic organizations appeared to try to energize and challenge employees to overcome difficulties and achieve goals for the good of the organization and themselves. Like elite and leadership organizations, meritocratic organizations framed change negatively; it was regarded as involving "strain," "stress," and "unpleasantness." However, meritocratic systems fo- cused more on people as participants in the change process; for example: "Of course, some people will feel uneasy but this should be less of a prob- lem if the process and its context are fully understood."

Indicators of needs, wishes, or intent were also positively correlated with change. Needs tended to be framed positively in this value system. Change tended to be associated with softer need words that implied some degree of participant choice and discretion; exemplary statements include "If we totally rely on someone else to change it, we might end up with something we do not really want." This element of discretion and choice differentiated meritocratic systems from elite organizations in particular. Change was also portrayed in an active or dynamic way in meritocratic sys- tems and tended to be related to goal statements; for example: "The ques- tion now is: How are we going to do this most effectively and how quick- ly?"

Elite organizations used words focusing on strength, but meritocrat- ic ones tended to rely more on overstatement or exaggeration, using words such as "amazing," "many," "complete," "most," and so on; for example: "IWel were able to see the amazing public acceptance of our changed sta- tus." Although elite systems offered reassurance and a sense of security through the use of strength words, meritocratic cultures tended more to emphasize employee achievement and accomplishment; an exemplary statement is "The most far-reaching elements in the process concern the way in which we are channeling the energy and skills of our staff."

1096 Academy of Management Joumal August

How change was portrayed in collegial organizations. Collegial organ- izations were the only ones that associated cbange with positive rather than negative terms: "We can make some changes in institutional structures which can lead to a more flexible, responsive, and effective organization" and "These changes represented a significant increase in educational op- portunity in the State" are examples. In addition, active words were fre- quently used when discussing change. Reference was made to "action," "implementation," "enhanced efficiency," "take advantage," "breathing new life," "develop potential," and so on.

Change was also portrayed as approached enthusiastically, and there was a basic confidence in the capabilities and role of the organization: "I am however confident that [the organization] is well placed to take ad- vantage of these changes." Change was described in soft ways that implied consideration, concern, and creating opportunities: "It is a process which helps everyone." Collegial organizations related change much more strong- ly to means: "The accident reporting system was changed to ensure that senior management are rapidly involved in the notification and investi- gation of all lost time accidents." The focus on means also involved a con- cern with research into and discussion of change, with frequent references to "reviews," "discussion papers," "stages," and "strategies"; for example: "A review of the unit is to take place in 1988" and "A two-stage manage- ment of change strategy was introduced." Although these organizations es- poused a thorough and democratic approach to change, there were bu- reaucratic overtones also, which may reflect the fact that the majority of organizations in this group were from the public sector.

The needs that were mentioned in collegial organizations belonged to a variety of stakeholders such as the "society," "community," and "econ- omy." Two of the strongest correlates of change were politics and power, referring primarily to governments (state and federal) and their activities, and reflecting the public sector character of the collegial group noted ear- lier.

Both our words—words suggesting we-ness or solidarity—and plea- sure words (satisfaction, confidence, pleased) decreased in collegial sys- tems as change words increased. These negative correlations had a fairly clear interpretation. Despite their enthusiastic depiction of cbange, it seemed that change had a pronounced, negative association with what could be considered core aspects of collegial systems: tbeir expressions of collective solidarity, confidence, and satisfaction.

DISCUSSION

Studies of organizational as opposed to individual values have become more common than they had been (e.g., Chatman & Jehn, 1994; Hofsteda, Neuijen, Ohayv, & Sanders. 1990; O'Reilly, Chatman, & Caldwell 1991; Rousseau, 1990); the present study differs from most previous ones, how- ever, in several important respects. We defined four ideal types of organ- izational value structures based upon a theory that is rooted in distribu-

1995 Kabanoff. Waldersee, and Cohen 1097

tive justice concepts and used content analysis to measure the values that a relatively large group of organizations espoused over a five-year period. We then used these data to search for organizations whose actual value structures echoed those of the theoretical types and examined how organ- izations that resembled one of the four theoretical types represented issues related to organizational change in their documents. The study thus offers insight into the nature of organizational value systems and their relation- ship to organizational change as well as describing an approach that is rarely used to study these issues.

Organizational values have been identified as potentially central for understanding the structure, functioning, creation, and change of organ- izations and their cultures (e.g., Rousseau, 1990; Trice & Beyer, 1993). Dun- phy and Stace put the case for the importance of values as follows: "Struc- ture defines tbe key working and power relationships in an organization while culture defines the key values that support those relationships. The close connection of these two aspects of organizational life is demon- strated by la survey! which found that changes in structure were . . . pos- itively and significantly (p < .001) related to changes in corporate goals and objectives, strategy and philosophy/values" (1990: 49). O'Reilly and colleagues (1991) found that better fit between individuals' and organ- izations' values predicted higher levels of satisfaction and commitment and lower turnover resulting from voluntary departures (cf. Sheridan, 1992). Nevertheless, research into organizational values has had a number of shortcomings.

One of these is an absence of theory. Although a variety of value di- mensions have been identified via researchers' intuition, surveys of the lit- erature, and factor analysis, there is an absence of well-specified theories of organizational values. The absence of theory bas resulted in an inabil- ity to deal with higher-level value concepts, such as value structures, that researchers (e.g., Schwartz, 1992) believe are tbe key to understanding dif- ferences in social groups' and systems' values. Chatman and jehn [1994] may appear to have heen discussing value structures when they described differences in the value profiles of organizations in different industries; in fact, they focused on single dimensions of the profile rather than on overall value patterns. In the present study, we drew upon distrihutive jus- tice concepts to develop a theory of organizational types and value struc- tures. We noted that most of the considerable research into organization- al justice has been at the level of individuals' perceptions and reactions. Studies have rarely considered justice as a systemic property of organ- izations and as part of their value structures. The four theoretical value structures we identified reflected different solutions to the organization- al problem of distrihutive justice—how to balance equality and equity con- cerns (Kabanoff. 1991a, 1991b; Leventhal, 1976).

The value patterns found were broadly consistent with the predicted value profiles. However, there were some deviations between the pre- dicted and actual value profiles. Elite value structures differed from our

1098 Academy of Management Joumal August

theoretical expectations in that performance was a relatively minor con- cern, perhaps because it was seen as primarily dependent upon the deci- sions, personal qualities, and influence exercised by the elite. Conse- quently, an elite may be less concerned with communication about per- formance issues because it attributes others' performance to compliance with the elite's own decisions and directions [Kabanoff, 1993; Kipnis, 1976: 186-195). Leadership value structures differed from our theoretical expectations in an interesting way. Although emphasizing commitment, as expected, affiliativeness was not important, which suggests that the sense of relationship in this value system is centered on the importance of loyalty and respect that have an element of compliance rather than on a sense of closeness or similarity. Both collegial and meritocratic value pro- files showed a rather convincing fit with our theoretical templates, though the collegial types' average teamwork (expected high) and performance (ex- pected low) scores perhaps reflected that this group included a consider- able number of public, professional bureaucracies under pressure to im- prove performance.

The differences we found in how groups of organizations portrayed organizational change supported the validity of our classification of organ- izational value structures and the underlying typology. Both the correla- tions and our qualitative interpretations of the texts indicate variations in how change is viewed across the four organizational types. Importantly, the different change themes are compatible with theoretical descriptions of the four types.

The portrayal of change in elite cultures appears to have a strong top- down flavor to it. Change tends to be viewed from the perspective of the organization itself as tbe major stakeholder. The "change-tone" is relatively demanding of employees at all levels, and there appears to be underlying concern that not everyone will be able to cope. There are reassurances about the strength of the organization and how change will reinforce it.

Leadership organizations have a tough, but not harsh, view of change. They focus on accountability for actions and give some emphasis to the discussion of goals and means. Employees are seen as being directed and guided through the change process by the encouragement and persuasion of their managers. Although these organizations are still basically com- pliance-oriented, their documents portray the change process less imper- sonally and more persuasively, seeking to encourage employees to com- ply with requirements rather than simply expecting it.

In the meritocratic value structure, there appears to be a much greater emphasis on motivating employees to play a constructive role in change. This emphasis involves explaining both the goals of change and the means for bringing it about. Meritocratic structures can be characterized as try- ing to challenge or energize employees. Change, although difficult, is as- sociated with achieving important goals, and the organization signals that people's efforts and achievements are recognized and appreciated. Thus,

1995 Kabanoff, Waldersee, and Cohen 1099

the change schema associated with meritocratic organizations is charac- terized by themes of striving, effort, goals, achievement, motivation, and recognition. This focus contrasts with the compliance-oriented schema more characteristic of elite and leadership systems.

Of organizations espousing the four types of value systems, only col- legial organizations view change in a positive way and emphasize em- ployee participation. In contrast to meritocratic systems, collegial ones do not challenge their employees to achieve organizational goals: instead, they emphasize the benefits change brings lo internal and external stakehold- ers and depict an enthusiastic, responsive orientation to change. But there is an interesting contradiction within the collegial schema. Enthusiastic espousal of change is associated with indications of a declining sense of collective solidarity and collective satisfaction.

We have used the term organizational values throughout to describe the themes we measured in organizational documents and, clearly, sever- al caveats need to he noted concerning our use of this term. As stated pre- viously, we believe the values we measured are likely to reflect the views of senior managers more than those of lower-level employees. Neverthe- less, the possibility that not everyone in an organization shares given val- ues does not, in our view, preclude us from calling them organizational values, since senior managers have both the right and responsibility to ex- press a set of values on bebalf of their organizations. Rokeach (1979) con- sidered information about the values attributed to organizations by their gatekeepers particularly important. The extent to which the values ex- pressed in organizational documents are shared or universal is an empir- ical question, and it is likely that organizations vary in the extent to which their values are universal (Trice & Beyer. 1993).

A second caveat is that, although we interpreted our measures as in- dicators of organizational values, it could be that the values captured hy the analysis of organizational documents may be no more than examples of managerial image or impression management. It is likely tbat managers do use such documents to seek to influence stakeholders toward a favor- able interpretation of managerial performance (e.g.. Staw et al., 1983), but that likelihood does not make such documents mere impression manage- ment. Previous content analyses of organizational documents have shown they can be used to identify business strategies (Bowman, 1984), differ- ences in how CEOs of surviving and nonsurviving firms respond to a husiness crisis (D'Aveni & MacMillan, 1989), and levels of corporate so- cial responsibility (Wolfe, 1991). These types of documents are highly vis- ible and, in many cases, closely examined by internal and external stake- holders; in general, this bigh visibility would seem to put some limits on bow unrealistic or misleading they can be. Therefore, it seems reasonable to believe that the words used in organizational documents contain both image and substance. Nevortbeless, it is clear that further research needs to investigate whether the organizations that have been identified as hav-

1100 Academy of Management foumal August

ing different types of value structures differ from each other in actual processes and policies and do so in ways that are consistent with the ty- pology.

The present study also suggests a number of additional research is- sues. Organizational values have been shown here to have some relation to industry. As industry effects can reflect differences in dominant tech- nologies (Chatman & Jehn, 1994), in competition and environmental un- certainty (Gordon, 1991), and so on, the sources of the relationships we found need to be teased out by future research. Such research may be a valuable means for exploring how and why different organizational val- ue patterns develop in different contexts, a topic about which little is known at present (Trice & Beyer, 1993: 427-428). As was noted earlier, organizational change is a process that is generally assumed to have an in- timate relationship to organizational values, and these results lend some tentative empirical support to this view. Nevertheless, future research needs to show that the apparent differences in how change is portrayed in organizations have implications for actual change strategies and tactics. Even more convincing evidence for the importance of values in organiza- tional change would come from demonstrating that the outcomes of change were hetter in organizations in which strategies and tactics matched es- poused values.

There is also a need to understand how organizational value systems are related to organizations' human resource management (HRM) process- es and policies. Researchers do not have a well-developed understanding of the major dimensions of HRM systems (Greenwood & Hinings, 1993: 1055) and how they are configured (Meyer, Tsui, & Hinings, 1993). The pre- sent typology clearly offers some help for developing hypotheses about how organizational reward, socialization, and control systems might he configured differently to reflect different underlying value structures. A topic of perennial interest to organizational researchers is whether values influence performance (Siehl & Martin. 1990). Kabanoff (1991b) argued that the performance implications of a particular type of value structure depend upon the kind of business strategy an organization pursues. Siehl and Mar- tin (1990) observed that there are almost insurmountable difficulties to in- vestigating value-strategy fit models of organizational performance, stem- ming in large part from the need to ohtain longitudinal data on values from a sufficiently large sample. Although the methodology we used does not solve all the difficulties Siehl and Martin identified, it does offer one po- tential solution to the problem of obtaining longitudinal data about organ- izational values.

Finally, what implications does this research have for managers? At the broadest level, the typology identifies the value configurations that may underlie and support (Dunphy & Stace, 1990) different kinds of organiza- tional structures. These value structures help give organizations their co- herence, strength, and stability and their particular characters. They also offer managers a framework for conceptualizing the nature and purpose of

1995 Kabanoff Waldersee. and Cohen UOI

organizational change. Evidence of existence of different change themes in organizations with different value structures, although still tentative at this stage, suggests that change agents tailor their change messages to a giv- en values context. Tbis suggestion is consistent with the views of leader- ship researchers (e.g., Shamir et al., 1993) who have suggested that a key aspect of leading large-scale change is making followers aware of their deeply held values and conscious that other members of the group share those values. Lastly, evidence of existence of distinctive types of value structures among organizations supports Gartwright and Gooper's (1993) contention that one possible explanation for the high failure rate of com- pany mergers and acquisitions is "culture incompatibility." As those au- thors observed, "Tbe cost of 'culture collisions' resulting from poor inte- gration may typically be as high as twenty five to thirty percent of the per- formance of the acquired organization. This certainly makes culture fit of equal, if not greater, importance than 'strategic fit.' " We would expect that organizations that resembled the same ideal type would find integration easier than organizations that resembled very different ideal types. The pre- sent method again suggests one means for unobtrusively assessing the de- gree of cultural compatibility between organizations for botb research and practical purposes.

REFERENCES

AldenderfHr, M. S., & Blashfield, R, K. 1984. Cluster analysis. Beverly Hills, CA: Sage.

Bowman, E. H. 1984. Content analysis of annual reports for corporate strategy and risk. In- terfaces, 14(1): 61-71.

Cartwright, S.. & Cooper, C. I... 1993. The role of culture compatibility in successful organ- izational marriage. Academy af Management Executive, 7(2); 57-70.

Chatman, ]. A., & Jebn, K. A. 1994. Assessing the relationship between industry character- istics and organizational culture: How different can you be? Academy of Management Journal, 37: 522-553.

Cowherd, D. M., & Levine, D. I. 1992. Product quality and pay equity between lower-level nmplnyens and top management: An investigation of distributive justice theory. Ad- ministrative Science Quarterly, 37: 302-320.

D'Aveni. R. A., & MacMillan, I. 1990. Crisis and the content of managerial communication: A study of the focus of attention of top managers in surviving and failing firms. Ad- ministrative Science Quarterly, 35: 634-B57.

Deutsch, M. 1985. Distributive justice: A social psychological perspective. New Haven, CT: Yale University Press.

Doty, D. H., & Glick. W. H. 1994. Typologies as a unique form of theory building; Toward im- proved understanding and modeling, Academy of Monogement Review, 19: 230-251.

Dowling, G. R.. & Midgley. D. F. 198a. Identifying the coarse and fine structure of market seg- ment. Decision Sciences, 19; 830-847.

Dunphy, D., & Stace, D. 1990. Under new management: Australian organisations in tran- sition. Sydney; McGraw-Hill.

Farh. }.. Podsakoff. P. M., & Organ. D. W. 1988. Organizational citizenship behavior in the Pacific Rim: A constructive replication and extension of U.S. studies. Paper present- ed at the annual meeting of the Academy of Management. Anaheim. CA.

1102 Academy of Management fournal August

Gersick, C. J. G. 1991. Revolutionary change thuories: A multilevel exploration of the punc- tuated equilibrium paradigm. Academy of Management Review, Ui: 10-36.

Gordon, G. G. 1991. Industry determinants of organizational culture. Academy of Manage- ment Review, 16: 396-415.

Greenberg, J. 1990. Looking fair vs. being fair: Managing impressions of organizational Jus- tice. In L. L. Cummings & B. M. Staw (Eds.), Research in organizational behavior, vol. 12: 111-157. Greenwich, CT: JAI Press.

Greenwood, R., & Hinings, C. R. 1993. Understanding strategic change: The contribution of archetypes. Academy of Management Journal, 36: 1052-1081.

Hedstrom, P. 1988. Structures of inequality: A study of stratificatian within work organ- izations. Sweden: Kristianstads Boktryckeri.

Hofstede, G,. Neuijen, B., Ohayv, D. D., & Sanders. G. 1990. Measuring organizational cul- tures: A qualitative and quantitative study across twenty cases. Administrative Science Quarterly, 35: 286-316.

Huff, A. S. 1990. Mapping strategic thought. Chichester: Wilny.

Kahanoff, B. 1991a. Equity, equality, power and conflict. Academy of Management Review, 16:416-141.

Kabanoff, B. ]991h. Distrihutive cultures in organisations and their relation to strategy, compensation, and change. Working Paper Series. 91-020, University of New South Wales, Australian Graduate School of Management, Sydney.

Kahanoff, B. 1993, An exploration of espoused culture in Austnilian organisations (with a closer look at the hanking sector). Asia Pacific Journal of Human Resources, 31(3): 1-29.

Kabanoff, B. Forthcoming. Computers can read as well as count: How computer aided text analysis can benefit organizational research. In C. L. Cooper & D. M. Rousseau (Eds.), TYends in organizational behavior, vol. 3. Chichester. UK: Wiley.

Kanter, R. M., Stein, B. A., & Jick, T. D. 1992. The challenge of organizational change. New York: Free Press.

Ketchen. D. ].. Jr., Thomas, J. B., & Snow, C. C. 1993, Organizational configurations and per- formance: A comparison of theoretical approaches. Academy of Management Journal, 36: 1278-1313.

Kipnis, D. 1976. The powerholders. Chicago: University of Chicago Press.

Leventhal, G. S. 1976. The distrihution of rewards in groups and organizations. In L. Berkowitz & E. Walster (Eds.), Advances in experimental social psychology, vol. 9: 91-131. New York: Academic Press.

Lord, R. G., & Hobenfeld, J. A. 1979. Longitudinal field assessment of equity effects on the performance of major league baseball players. Joumal of Applied Psychology, 64: 19-26.

Meindl, J. R. 1989. Managing to he fair: An exploration of values, motives and leadership. Administrative Science Quarterly, 34: 252-276.

Meyer, A. D., Tsui, A. S., & Hinings, C. R. 1993. Configurational approaches to organizational analysis. Academy of Management Journal, 36: 1175-1195.

Mohler, P. P., & Zuell. C. 1990. Textpack P. C. Mannheim. Germany: ZUMA, the Center for Surveys, Research and Methodology.

Moorman, R. H. 1991. Relationship between organizational justice and organizational citi- zenship behaviors: Do fairness perceptions influence employee citizenship? Journal of Applied Psychology, 76: 845-855.

O'Reilly, C. A., Chatman, J., & Caldwell, D. F. 1991. People and organizational culture: A pro-

Kabanoff, Waldersee, and Cohen 1103

fiĥ comparison approacb to assessing person-organization fit. Academy of Management Journal, 43:487-516.

Organ, D. W. 1988. Organizational citizenship behavior: The good soldier syndrome. Lex- ington, MA: Lexington Books.

Osgood. G E,, Suci, G. J,, & Tannenbaum, P. H. 1957. The measurement of meaning. Urbana: University of Illinois Press.

Pfeffer, J. M., & Langton. N. 19H8. Wage inequality and tlif organization cif work: The case of academic departments. Administrative Science Quarterly, 33: 588-606.

Pfeffer, j . M,, & Langton, N. 1993. The effect of wage dispersion on satisfaction, productivi- ty, and working colUthoratively: Evidence from college and university faculty. Admin- istrative Science Quarterly, 38: 382-407.

Polley, R. B, 1987. Exploring polarization in organizational groups. Group and Organization Studies, 12: 424-444.

Price, J. L., & Mueller. G. W. 19BG. Absenteeism and turnover of hospital employees. Green- wich, CT: JAl Press.

Pun), G., & Stewart, D, W, 19H3. Cluster anaiysi.s in marketing research: Review and sugges- tions for applications. Journal of Marketing Research, 20: 134-148.

Roktiach, M. 1979. From individual to institutional values: With special reference to the val- ues of science. In M. Rokeach (Ed.). Understanding human values: 47-70. New York: Free Press.

Rosenberg, S. D., Schnurr, P. P.. & Oxman, T. E. 1990. Content analysis: A comparison of man- ual and computerized systems. Joumal of Personality Assessment, 54: 298-310,

Rousseau, D. M. 1990, Asse.ssing organizational culuirts: The i:ase for multiple methods. In B. Schneider (Ed.). Organizational climate and culture: 153-192. San Francisco: Jossey- Biiss,

Scarpello, V, 1988. Pay satisfaction and pay fairness: Are they the same? Paper present- ed at the meeting of the Society ft)r InduHtri:iI and Organizational Psychology. Dallas.

Schwartz, S. H. 1992. Universals in tha content and structure of values: Theoretical advances and empirical tests in 20 countries. In M. P. Zanna (Ed,), Advances in experimental so- cial psychology, 25: 1-65, Orlando, FL: Academic Press.

Shamir, B,, House, R. I.. & Arthur. M. B. 1993. The motivational effects of charismatic lead- ership: A self-concept hased theory. Organization Science, 4: 577-593.

Sheppard, B. H., Lewicki, R. J., & Minton, f. W. 1992. Organizational justice: Thesearchfor fairness in the workplace. New York: Lexington.

Sheridan. ]. E. 1902. Organizational culture and employee retention. Academy of Manage- ment journal, 35:1036-1056.

Siehl, C , & Martin, J. 1990. Organizational culture: A key to financial performance? In B. Schneider (Ed.), Organizational climate and culture: 241-281, Newbury Park, GA: Sage,

SPSS Inc. 1988. SPSS-X user's guide (3d ed.). Chicago: SPSS Inc.

Staw, B. M,, McKnchnic!. P. 1., & Puffer, S. M. 1983. The justification of organizational per- formance. Administrative Science Quarterly, 28: 582-600.

Stone, P. J., Dunphy, D. C . Smith, M. S., & Ogilvie, D. M. 1966. The Genera/ Enquirer: A com- puter approach to content analysis. Cambridge, MA: MIT Press.

Trice, H, M., & Beyer,), M. 1993. The cultures of work organizations. Knglewood Cliffs, NJ: Prentice-Hall,

Weber, R. R 1985. Basic content analysis. Beverly Hills, GA: Sage.

1104 Academy of Management Joumal August

Witt, L. A., & Nye, L. G, 1992. Gender and the relationship between perceived fairness of pay or promotion and job satisfaction. Journal of Applied Psychology, 77: 910-917.

Williams, R. M. Jr. 1951. American society: A sociological interpretation. New York: Knopf.

Wolfe, R. A. 1991. The use of content analysis to assess corporate social responsibility. In J. Post (Ed.), Research in corporate social performance and policy: 13: 281-307. Green- wich, CT: JAI Press.

Wolfe, R. A., Johnson, T. A., & Gephart, R. P. 1993. Computer-facilitated qualitative data analysis: Potential contributions to management research. Journal of Management, 19: 633-660,

Ziill, C , Weber, P. W., & Mohler, P. P. 1989. Computer-aided text classification for the so- cial sciences: The General Enquirer III. Mannheim. (Jermany: ZUMA, the Center for Surveys, Research and Methodology.

Boris Kabanoff is an associate professor at the Australian Graduate School of Man- ageniBjit in the University of New South Wales. He holds a Ph.D. degree in indus- trial/organizational psychology from the Flinders University of South Australia, His major interests are in organizational values and justice, and computer-aided text analysis as a research methodology.

Robert Waldersee is a research fellow in the Gentre for Corporate Change at the Aus- tralian Graduate School of Management in the University of New South Wales. He hold.s a Ph,D. degree in industrial/organizational psychology from the University of Nebraska. His major interests are in the influence of leadership on organizational change and service quality.

Marcus Cohen is a doctoral candidate in the Australian Graduate School of Man- agement in the University of New South Wales. He holds honours and masters de- grees in industrial/organizational psychology from the University of Cape Town. His doctoral work is examining leadership and influence processes in volunteer organ- izations.