strategic groups: a cognitive perspectiveeprints.maynoothuniversity.ie/5473/1/strategic... · by...

Strategic Management Journal, Vol. 14, 103-124 (1993)

STRATEGIC GROUPS: A COGNITIVE PERSPECTIVERHONDA K. REGERDepartment of Management, Arizona State University, Tempe. Arizona. U.S.A.

ANNE SIGISMUND HUFFCollege of Commerce and Business Administration, University of Illinois, Champaign,Illinois, U.S.A.

The strategic group concept provides an attractive middle ground hetween firm and itiditsiryfor both theory development and empirical analysis. To date, this concept has been definedby researchers in terms of secondary accounting and financial data, and a number of criticshave questioned the validity of this work. Our research shows that industry participantsshare perceptions about strategic commonalities among firms, and thai participants clustercompetitors in subtle ways not reflected in extant academic research on .strategic groups.Decision makers' perceptions and cognitions are phenomena that can be expected toinfluence industry evolution. They are of research interest as an additional source of dataon firm commonalities which helps address concerns about previous strategic group research.

In 1972 Hunt introduced the term 'strategicgroup' to describe the 'symmetry of operations'he observed In the appliance industry. Huntnoted significant differences in characteristicsand strategies among the firms in his study, whileat the same time he found that many firmsfollowed similar strategies. Grouping firms clari-fied understanding of the apparently viablestrategic options in the industry.

Although no formal definition is universallyaccepted, a 'strategic group' continues to becommonly defined as a group of firms within thesame industry making similar decisions in keyareas (Porter, 1980: 129). If these possibilitiesare thought of as an n-dimensional graph,strategic groups identify clusters of firms in'strategic space,' and group membership definesthe essential characteristics of a firm's strategy.The identification of strategic groups has beenused primarily to explore systematic differencesin profitability among firms in the same industry

Key words: Strategic groups, managerial cognition,cognitive maps, industry evolution, repertory gridtechnique

(McGee and Thomas, 1986). The strategic groupsarea is quite active and continues to generatenumerous studies. Table 1 summarizes the mainstudies published since the McGee and Thomas(1986) review.

Recent reviewers, however, have voiced severalfrustrations with the state of strategic grouptheory and empirical research (Barney andHoskisson, 1990; Cool. 1985; Hatten and Hatten,1987; McGee and Thomas, 1986; Thomas andVenkatraman, 1988). Dissatisfactions includeinsufficient theoretic underpinnings for the con-struct itself, inadequate model specification,haphazard selection of strategic dimensions usedto form groups, and inconclusive results fromempirical research. Given these serious concerns,it is essential that we step back from research inthis area to examine fundamental questions aboutthe strategic group concept.

The most recent and potentially most devastat-ing attack on strategic group research has beenmade by Barney and Hoskisson (1990). whoclaim that two critical assertions of strategicgroup theory remain untested: '(1) that strategicgroups exist and (2) that a firm s performance

0143-2095/93/02010.V-22$ 16.00© 1993 by John Wiley & Sons, Ltd.

Received 1 June 1990Final revision received 15 September 1992

104 R. K. Reger and A. S. Huff

Table 1. Recent strategic groups studies

Study Industry Primary dala type

Cool and Schendel (1987)Fombrun and Zajac (1987)

Hatten and Hatten (1987)Johnson and Thomas (1987)Cool and Schendel (1988)Lawless, Bergh and Wilsted

(1989)Lawless and Finch (1989)Mascarenhas (1989)Mascarenhas and Aaker (1989)McNamee and McHugh (1989)Zajac and Shortell (1989)Barney and Hoskisson (1990)Fiegcnbaum and Thomas (1990)Kumar. Thomas and Fiegenbaum

(1990)Lewis and Thomas (1990)Porac and Thomas (1990)Schroeder (1990)

Ulrich and McKelvey (1990)Boeker (1991)Lawless and Tegarden (1991)Nohria and Garcia-Pont (1991)Sudharshan. Thomas and

Ficaenbaum (1991)Tallman (1991)Caves and Ghemawat (1992)

pharmaceuticalfinancial services

U.K. brewingpharmaceuticalmanufacturing firms

52 manufacturing finnsoil-well drillingoil-weli drillingNorthern Ireland clothing industryhospitals

insurance

U.K. retail grocery

foundry

Japanese electronicsbrewing47 four-digit SICautomobilepharmaceutical

automobilevariety from PIMS NRDB database

archivalsurvey questionnairevalidated with archivalconceptualarchivalarchivalarchival

archivalarchivalarchivalarchivalquestionnaireconceptualarchivalconceptual

archivalconceptualinterviewsmultiple case studiesarchivalarchivalarchivalarchivalarchival

archivalarchival

depends upon strategic group membership' (1990:187). Further, the assumption '[t]hat there aregroups of firms suggests that firms are notidiosyncratic in strategically relevant ways'(Barney and Hoskisson, 1990: 188). Strategicgroup research is also suspect, since, to date,groups have been defined using 'an almoststandard method. . .[which[ employs some formof cluster or factor analysis' (Barney and Hoskis-son. 1990: 189). Hatten and Hatten support thisbasic challenge with the assertion that strategicgroups are merely analytical conveniences usedby researchers, artifacts of our theories andtechniques, without any objective analogue inthe natural environment (1987: 329).

Our view is, first, that strategic groups canbe defined in a way that allows some strategicallyimportant variance among firms within eachgroup. This view is supported by resultspresented in this paper. Second, although theexplanation of performance differences among

firms is clearly elusive (Cool and Schendel,1988; Fiegenbaum and Thomas, 1990; McGeeand Thomas. 1986). we believe a modifiedtheory of group structure, along lines developedlater in this paper, may still make a contributionto the understanding of profitable strategicchoices. Predicting firm profitability, however,is not the key contribution of a cognitiveapproach to strategic groups.

Even if strategic groups are found to beuncorrelated with performance outcomes, ourposition is that managerial perception of similari-ties and differences among competitors influencestrategic decision making, and thus are worthyof study. In fact, strategists' grouping schemesmay prove to be more significant than researcher-defined groupings for understanding competitionand performance because, through enactmentprocesses (Weick, 1979), the way firms seethemselves and their competitors (Porac. Thomas,and Emme. 1987) is expected to have tangible

Strategic Groups 105

effects on strategy reformulation and subsequentindustry structure.

It seems logical that strategists will think interms of clusters of competitors to cognitivelysimplify a complex environment, Commonalitiesamong firms should also be expected, indepen-dent of individual precepts, as a result of broaderpressures creating institutional isomorphism(DiMaggio and Powell, 1983) and because ofthemany inertia factors inhibiting strategic change(Tushman and Romanelli, 1985). Recent work onisomorphism and inertia (Powell and DiMaggio.1991) have bolstered our understanding of whyfirms do not always achieve the kind of idiosyn-cratic positions Barney and Hoskisson (1990)rightly indicate are often associated with competi-tive advantage.

Finally, though we will make the argumentthat perceptual and economic factors canreinforce each other, fundamental questionsabout the real and the perceived cannot beanswered definitively by any research endeavor.The demise of logical positivist standards makeeven simple assertions such as "the sun will risetomorrow' unprovable (Kuhn, 1970). As allproofs become suspect, no research project cansatisfactorily say it has identified anything thatis objectively "real.* The study reported herefocuses on determining if strategists perceivedifferentiating commonalities among firms and ifso, to outline the nature of those perceptions. Ifgroups of firms are 'real' for strategists, thenresearch at the strategic group level is muchmore important than if groups are only the resultof researchers" analytical exercises. Empiricalevidence using strategists as data sources alsoprovides multimethod confirmation (Huff, 1982;Jick, 1979) for extant research using archivalsources.

In this paper, we summarize research fromcognitive psychology, organization theory andstrategy that supports two propositions. Thesepropositions, that strategists will perceive simi-larities among subgroups of firms in an industryand that strategists in the same industry willgroup participants in that industry in similarways, are examined with data from the Chicagobanking industry. The data show a shared senseof group structure in the industry, but also reveala range of agreement on the similarity offirm strategies. These subtleties in executiveperceptions add important dimensions to the

strategic group concept and increase its potentialutility for understanding competitive positioning.To foreshadow points elaborated upon in theconcluding sections of the paper, we believethose interested in describing and explainingstrategic group structure might well amplifyresults from archival studies with cognitivestudies. Those interested in predicting change instrategic position may be even more dependenton cognitive data to extend retrospective financialand accounting data.

STRATEGIC GROUPS AS THE RESULTOF PERCEPTION AND COGNITION

Strategic management research is just beginningto assess the importance of cognitive frameworksin strategic decision processes (Dutton. Fahey,and Narayanan, 1983; Huff, 1990; Mason andMitroff. 1981; Stubhart, 1986, 19S7). Porac andThomas' (1990) recent theoretical paper examinesmany of the cognitive rationales for mental modelsof competition. Two reasons why managers indemanding competitive situations might focus ongroups of firms rather than individual competitorsdeserve special note.

Simplification

Cognitive research suggests that simpliftcation isa cognitive necessity in cases of informationoverload. Research also shows that human beingscan be easily overloaded. Miller (1956), forexample, found that people cannot hold morethan about seven bits of data in mind at a time.Work on cognitive simplification processes amongmanagers (e.g., Schwenk, 1984; Simon, 1945)suggests that strategists simplify the complexcognitive problem of independently analyzing alarge number of competitors by grouping them.In fact, executives, journalists and industryanalysts often refer to groups of firms in publicstatements.

Elaboration

Cognitive elaboration involves filling in gaps(often unconsciously) when interpreting stimuli.The theoretical rationale for this activity isprecisely the opposite of that offered for simplifi-cation. Sometimes data are not rich or varied

106 7?. K. Reger and A. S. Huff

enough to allow interpretation without furtherelaboration. Often unconsciously, the unknownsabout a particular case are likely to be filled inwith information consistent with beliefs aboutother, better known examples that are believedto be similar (Rosch, 1981).

Prahatad and Bettis (1986) have given acompelling example of this kind of embellishment.They propose that managers make acquisitionsand divestments on the basis of a 'dominantlogic' or dominant paradigm developed throughpast experience. The result is a consistency infirm decisions over time, despite a wide varietyof acquisition opportunities. We submit thatexecutive perceptions may similarly 'gel' withrespect to the dominant logic ascribed to sur-rounding firms. It is possible that these attributesgo beyond the deliberate intentions of thecompanies involved, but the practical result isincreased facility with the difficult problems ofcomplex analysis.

While simplification and elaboration are verydifferent responses to competitive conditions, animportant result of both processes, we suggest,is that strategists will perceive similarities amongfirms. Our first proposition, then, based on bothcognitive simplification and elaboration is that:

Strategists^ perceptions of their competitors'strategies will be characterized by a groupstructure, rather than each firm's strategy beingperceived as unique or all firms' strategiesbeing perceived as similar.

Interaction

While cognitive simplification and cognitive elab-oration could lead to totally idiosyncratic group-ings, strategists who work in the same industryenvironment are expected to develop sharedperceptions of the competitive environment overtime. Company executives interact with eachother at industry associations and other gather-ings; they share similar sources of informationsueh as trade publications; they hire from thesame professional labor pool and frequentlyemploy the same consultants. As events shapethe industry, such commonalities in informationsources encourage shared interpretation of thepresent and shared expectations for the future,including shared perspectives on industry group-ings (Huff, 1982; Porac, Thomas, and Baden-

FuIIer, 1989). Thus, based on arguments ofshared information sources and interaction, wefurther propose that:

Strategic group structure will be widely sharedby strategists within an industry, rather thaneach strategist holding unique perceptions ofstrategic group structure.

COGNITIVE STRUCTURES

Given the potential importance of participantperspectives on the competitive environment, webelieve that inquiry on strategic groups shouldinclude research based on cognitive data. Thiswork appropriately begins with tests of theexistence of competitor groupings in the minds ofstrategists. Previous research has made importantfirst steps in this direction (Dess and Davis, 1984;Fombrun and Zajac, 1987). These researchers'purpose was to measure managerial perceptionsof theoretically significant constructs, not touncover cognitive structure perse. Questionnaireswere used to gather data in these studies, andtherefore it is not clear if the competitorgroups found represent the respondents' cognitivestructure of the industry or a structure inherentin the instruments used (Fransella and Bannister,1977).

Two streams of research in cognitive psychologyare particularly appropriate for overcoming thisproblem in studies of cognitive structure Includingthe study of cognitive strategic groups. Classifi-cation theories concentrate on categories ofconcepts and the hierarchical relationships amongthem (Johnson-Laird and Wason, 1977: 169-253;Lakoff, 1987; Rosch, 1978). Alternatively, per-sona! construct theory (Kelly, 1955; Fransellaand Bannister, 1977) focuses on the dimensionsof concepts. The research reported here drawsprimarily on personal construct theory, which isespecially germane given that extant strategicgroup theory assumes variation among firmsalong key dimensions of decision making (Porter,198(); Cool and Schendel. 1987; Porac andThomas. 1990) just as personal construct theoryconcentrates on the key dimensions, or constructs,of human cognition (Kelly, 1955). In addition,Thomas and Venkatraman (1988) advocated theuse of this approach in their review of strategicgroups research.


Personal construct theory was first postulated asa theory of personality (Kelly, 1955); but lateradherents assign it a more limited role as a theoryof cognition (Fransella and Bannister, 1977).Briefly, the theory posits that bipolar constructs (ordimensions) are the primary mechanism individualsuse to organize, simplify, and interpret the massof stimuli that constantly confront them. Constructsare defined in terms of similarities and differencesand are organized into systems of meaning whichindividuals use to develop theories about theenvironment, to make predictions and guide action.

Personal construct theory is predicated on thebelief that individuals act on their perceptionsof the objective world filtered through theirconstructive system. The individual does notpassively perceive the environment, rather he orshe actively construes (attaches meaning to)perceptions. Kelly suggests that it is relativelyeasy for an individual to move an object alonga cognitive construct, but much more difficult tothink of objects in terms that are not part of anexisting construct system. Constructs are thus seento form a somewhat flexible yet structured networkthat both facilitates and restricts an individual'sperceptions and actions (Kelly, 1955: 49).

Hundreds of studies have been published Inpsychology and related fields using personalconstruct theory (Fransella and Bannister, 1977),but application of the theory and its relatedmethodology, repertory grid technique, is quitenew to strategic management. Ginsberg (1987)and his colleagues (Dunn et al. 1986; Dunn andGinsberg, 1986) have recommended the theoryand methodology for the study of business andpublic policy issues. Empirical studies are limited,but initial studies illustrate the flexibility of itsapplication. Eden. Jones, and Sims (1979; 1983)combined personal construct theory with causalmapping to help organization members clarifyand alter their perceptions of managementsituations. Walton (1986) built profiles of proto-typical successful and unsuccessful firms usingthe repertory grid technique and Dutton (1987)studied the dimensions managers use to categorizeissues with the same method.

RESEARCH DESIGN

We used personal construct theory to addressthe research questions of this study. First, do

strategists group competitors in an industry?Second, are perceptions of competitor groupingswidely shared or are they idiosyncratic? The twopropositions were examined in the context ofU.S. banking industry—a quite demanding testsite, given that recent changes in regulation andincreased competition might be expected to makeclusters of competitors less stable and moredifficult to assess.

The research focused on the competitivestrategies of the 18 largest bank holding compa-nies (BHCs) headquartered in the Chicago areabetween 1982 and 1985 (Table 2). This timeframe was chosen because it was bracketed bymajor regulatory changes in Illinois. In 1982,Illinois bank holding companies were allowed toacquire additional banks. Previous to this date,Illinois was a unit banking state, meaning that abank holding company could only own one bank.Beginning in July, 1985, Illinois passed a regional,reciprocal banking pack with other midwesternstates. Both of these changes were expected toincrease the rate of environmental change andfacilitate major strategic repositioning in thebanking industry. While all 18 firms were notnecessarily direct competitors in all markets wheninterviews were conducted in 1986. we expected,and found, that most strategists would be awareof the total population of major firms underthese circumstances.

Interviews were conducted at 6 ofthe 18 BHCsin the study population. The six were chosen fortheir diversity, relying on industry observers notin the study and secondary sources. Two firmswere widely touted as well run, two werefinancially troubled and two had mixed perform-ance reviews from industry insiders. Two of thebanks were located 'in the loop,' two were inother downtown locations, and two were thoughtof as suburban banks. Two of the BHCs owneda single large 'flagship' bank, one was clearlyorganized as a multibank confederation. Interest-ingly, at least four of the six banks were rumoredto be acquisition oriented, while five of the sixwere rumored to be takeover candidates.

All six firms chosen for the sample agreed toparticipate in the study. The CEO at each bankidentified five strategic decision makers withinthe firm who were especially familiar withcompetition in the Chicago market from 1982through 1985. Four of the six CEOs personallyparticipated in the study. In all, 30 strategists


were contacted and 23 provided usable data fora 77% participation rate.

All Informants for the study were experiencedbankers. They were employed in the bankingindustry for an average of 18.5 years, withbanking experience ranging from 10 to 35 years.Respondents had worked for their current firmfor an average of 10.25 years. Their experiencevaried widely, with seven (30%) in generalmanagement, five (22%) in finance, three (13%)in strategic or corporate planning and one (4%)in marketing. Seven of the informants workedin major product areas, including two (9%) incommercial loans, two (9%) in retail or privatebanking and one each in correspondent banking,financial institutions and trust (4% each).

Data collection

Data were collected in Fall 1986 via semistruc-tured interviews using the minimum context formof the repertory grid technique (Dunn andGinsberg, 1986; Fransella and Bannister, 1977;Kelly, 1955; Reger, 1990b). This techniquerequires each informant to generate personaldimensions for describing the phenomenon of

study and to make similarity and differencejudgments about individual examples of thatphenomenon along those dimensions only. Thus,researcher-imposed structure on subject cognitionIs minimized.

Bach informant in the study was interviewedseparately. Following standard procedures forthe minimum context format, the 18 BHCs'names were presented to each Informant threeat a time. The sequence of presentation of triadswas random and the same for each informant.Informants were asked the way or ways in whichtwo of the BHCs were more similar strategicallyand how the third was different. In this way,bipolar dimensions used by strategists to differen-tiate among the strategies of competitors wereelicited.

Each informant personally identified dimen-sions, which were then used to elicit judgementabout competitors" strategies. Ideally, informantswould rate each focal BHC along all thedimensions they personally use to distinguishcompetitors. However, pretests indicated thatasking an individual to rate all 18 BHCs alongall defining dimensions was infeasible because ofthe time involved; pretest subjects reported their

Table 2. Number and percentage of informants who assignedeach fociil bank holding company to each cluster

First ChicagoNorthern TrustContinental [11.Harris

ABN-LaSalleExchange

First EvergreenGary WheatonAffiliated

Firs! UnitedUSA meri banesState NationalFirst Illinois

First ColonialCole-Taylor

BoulevardLake ShoreUnibancorp

Cluster 1No.

18171716

It8

000

1100

I2

44

%

100100l(X)100

7967

i77

1013

292427

Cluster 2No.

0000

02

15169

14139

10

n11

662

%

17

100100100

93939090

8073

433518

Cluster 3No.

0000

32

000

0011

I2

476

%

2117

109

1013

294155


boredom level rose substantially when so manyratings were required. Therefore, each informantwas randomly assigned to one of four groups torate an overlapping subsample of 12 of the 18focal BHCs. This sampling procedure, commonlyused in marketing (Green. 1977). allowed forcomparisons across informants without eachinformant rating all BHCs.

Each informant rated the BHCs along his (allsubjects were male) self-supplied dimensionsusing 11 point scales. The procedure yielded23 n X m rating grids, one for each informant,where n is the number of dimensions elicited foreach of the 23 informants and m represents thenumber of firms the informant felt knowledgeableassessing. The Appendix provides a sample ofthe elicited dimensions. Detailed dimension-levelresults of this research also are reported elsewhere(Reger and Palmer. 1990).

Data analysis

It is possible to ask informants to directly formgroups by sorting the population into subsets, avariant of repertory grid technique called the fullcontext form (Fransella and Bannister. 1977: 15).This sorting task was rejected for two reasons.First, extant strategic group theory is boththeoretically and empirically concerned with thedimensionality of competitive strategy; that is.how firms are positioned in multidimensionalcompetitive space. Therefore, we employedprocedures which mirrored current researchpractice. Second, most applications of personalconstruct theory have used the minimum contextform and, thus, reliability and validity measureshave been better established for this procedure(Fransella and Bannister. 1977).

Instead of directly asking informants for theirperceived groupings, we cluster analyzed eachinformant's rating grid using SAS CLUSTER.Cluster analysis particularly fits the objectives ofthis study, since it places objects into groupssuggested by the data. Clustering techniques canbe criticized, however, since by their nature theywill break the data available into subsets,however weak the association among data points.Therefore, three clustering methods wereemployed, each of which has been found to havesomewhat different strengths and weaknesses. Ifall three clustering methods give similar resultsusing multiple criteria, confidence that the groups

are an inherent part of the data and not anartifact of the particular clustering algorithm isincreased.

The first clustering technique used, the averagelinkage algorithm, forms clusters on the basis ofthe average distance between pairs of obser-vations. This method 'tends to join clusters withsmall variances and is slightly biased towardproducing clusters with the same variance' (SAS,1985: 263). The second, centroid method maxi-mizes the distance between group means, recom-puting the centroids each time a new observationis included in a cluster. This method 'is morerobust to outliers than most other hierarchicalmethods' (SAS, 1985: 264). Ward's method wasused as a third alternative. Groups are formedby minimizing the within cluster ANOVA sumof squares. With this technique Ward's clusteranalysis 'tends to join clusters with a smallnumber of observations and is strongly biasedtoward producing clusters with roughly the samenumber of observations' (SAS, 1985: 267).

Cluster analysis is not an inferential statisticaltechnique in which the researcher can link aspecified confidence level with a set of clusters(Everitt. 1980: 66); instead, the researcher mustspecify rules for choosing the number of clustersthat will be used for description and analysis ofthe data. Three decision rules determined thenumber of clusters from each algorithm's outputfor each informant's data: adopt cluster solutionsat large breaks in the dendrograms, avoid clustersolutions that produce one firm groups, andchoose solutions with high interpretability basedon qualitative comments made by informants(Everitt, 1980; Hartigan, 1975; Romesburg,1984). As a further assurance that the clustersrepresented 'real" groups in the data and notsimply artifacts of the methods, the first authorapplied the decision rules to the average linkageoutput, while the second author independentlyevaluated the centroid and Ward's results. Useof these decision rules was considered preferableto blind application of any one decision rule thatmight misrepresent the data (Everitt, 1980:Romesburg. 1984).

As a test of the correspondence betweenstrategic groups identified by industry participantsand economic 'reality,' we then looked at twooutcome indicators. Data on profitability andsurvival, independent variables of longstandinginterest in strategy research, were gathered on

R. K. Reger and A. S. Huff

the 18 firms in the study. Profitability data weregathered in 1986. Though mergers made itimpossible to compare profitability in subsequentyears, the merger pattern itself provides aninteresting applieation of strategic groupings.

RESULTS

Individual results

The decision criteria outlined above yieldedeither two or three eluster solutions in the datafrom all but one of the 23 respondents, for whomonly one of the three clustering algorithmsyielded a cluster solution, the other two did not.This informant, however, only felt comfortableranking 6 of the 12 BHCs presented, and theseBHCs were most often clustered together byother resp<indents. For 17of the other informants,the cluster results from the three methodsconverged. In one of the remaining five cases,one method created a group where ihe othertwo did not. In a second instance, one proceduredid not create a group where the other two did.Two other cases involved one mismatch amongthe three methods out of the 12 firms therespondents answered, and a third involved 2out of 12 discrepancies among the three methods.

The fact that three algorithms yielded exactlythe same clusters in data from 17 of 23 informants(74%) and that the worst case involved only a16% discrepancy argues for accepting the firstproposition of the study—that strategists cogni-tively group their competitors—and rejecting thealternative explanation that the clusters formedare mere artifacts of a clustering technique.Additional support for this proposition can befound in the strong similarities among clusters inthe six cases that showed any discrepancy at allamong the methods explored.

Commonality across informants

There is also strong commonality across respon-dents in the firms assigned to each cluster,which supports our second proposition: groupingschemas are not idiosyncratic, but widely sharedacross strategists. Since the results were similaracross the three clustering algorithms, for clarityof exposition analyses from the average linkagemethod only are presented below.

Table 2 summarizes the number and percentage

of informants who assigned each BHC to eachof the three clusters commonly observed in thedata. Both Cluster 1 and Cluster 2 include acore group of BHCs assigned to the same clusterby all informants rating that company and asecondary group of companies clustered with thiscore by more than 65% of the informants. Firmsin Cluster 1 generally followed a 'money center'strategy concentrating on wholesale and commer-cial activities in a national or even internationalmarket whereas firms in Cluster 2 generally werestronger in middle market commercial, smallbusiness and consumer banking in a more limitedgeographic area. Cluster 3 does not exhibit acore, nor as strong a secondary group, andappears to be more of a catch-all cluster thatcannot be as meaningfully compared acrossinformants. For example. Boulevard Bankcorp(BOU) and Unibancorp (UNI) are assigned mostoften to this cluster, but are also found in eachof the other clusters as well.

Robustness of results

Since cluster analysis requires researcherinterpretation and is not a statistical methodology,it is prudent to examine the robustness of resultsunder various assumptions. Thus, we took threeapproaches to evaluating the data shown in Table2. First, we assumed that a three cluster solutionrepresented common opinion about group struc-ture in ihc industry, and asked: How often dida given individual assign a given firm to the'right' cluster? Table 3, which summarizes the251 competitor assessments shown in Table 2,indicates (on the diagonal) that in 209 instances,or 83% of the total, individual assessmentsresulted in assigning the firm to the clusterexpected if group assignments are assumed apriori to be shared across alt industry participants.If each informant in this study had randomlyassigned firms to groups based on the relativesizes of the three groups, assignments would be'successful" only 40% of the time. The differencebetween the observed result and the randomexpectation (which is statistically significant atthe O.(Kll level) provides strong support for theproposition that cognitive groups are sharedamong participants in an industry.

It seems reasonable to suspect, given the lowlevel of common assignments to Cluster 3, thatthis is a residuat group which does not mirror


Table 3. Assignment of Bank Holding Com-panies to Clusters: Three Cluster Solution

3 groups

Cluster

1

2

3

of firms

A

87

2

5

B

5

105

5

C

11

14

17

Total

103

121

27

94 115 42 251

'outlier' firms from those in clusters. These tablesshow that ambiguity about the strategies ofBoulevard Bankcorp (BOU) and Unibancorp(UNI) occurs almost exclusively among thosewho see the industry as falling into two strategicgroups. There is little ambiguity among thosewho cluster the industry into three groups. Theyunanimously agree that BOU belongs in Group3, and only one individual disagrees on theplacement of UNI in Group 3.

This third view of the data suggests even morestrongly that there is considerable agreement onthe strategic location of the BHCs in thepopulation we studied. Commonality of groupmembership is supported under three differentapproaches to analyze the data, thus, propositiontwo is conclusively supported.

strategic commonalities. The second test ofshared strategic groups assumed that firmsassigned to the first two clusters followedespecially clear cut strategies. The data found inthe upper left square in Table 3 summarize thenumber of firms assigned to these clusters. Thisinner square shows only 7 'misassignments' outof 199 judgements made about firms belonging tothe two clear strategic groups in this industry; twoin which a firm that common wisdom said belongedin Cluster 1 was instead assigned to Cluster 2, andfive in which a firm common wisdom assigned toOuster 2 was instead placed in Cluster 1.' Thistest shows even stronger support for the idea thatgroup assignments are shared.

Further examination of the data shows moreagreement about the two frequently 'misassignedfirms' than at first appears. The approach herewas to relax the expectation that industryparticipants will share one view of competitorgroupings. Tables 4 and 5 divide respondents bythe number of clusters judged to best portraytheir assessments of competitors, and distinguish

' Given the strength of results showing assignments jn bolhthe three group case and the two group case, no furtherevidence is really necessary to support proposition two. Inmore complex cases, it might be appropriate lo compare theobserved assignment with assignment to any other cluster.(In our case, the number of as.stgnments lo Cluster I vs.assignments to Clusters 2 or 3: and then the number ofassignments to Cluster 2 vs. assignment to Clusters I or 3).A phi square, which corrects for the size of the sample(Reynolds. 1977; 27). or Cramer's V, which is particularlyuseful for assessing symmetric arrays (Reynolds. 1977; 32),can be used to describe data that are less clear eut.

Cognitive groups, performance and survival

The strong results of this study show thatcognitive data can be used to form strategicgroups. Are these groups useful for understandingperformance or other outcomes of interest? Table6 shows the 1986 ROA for each focal BHC,as well as whether the BHC has remainedindependent or was acquired subsequent to thestudy. Again, since the status of Cluster 3 asa cohesive strategic group is in doubt, weconcentrated our analysis on the first two clusters.Using a /-test of group mean differences forsmall samples, average ROA for the two groupswas significantly different in 1986(alpha = O.(K)5). Since many of the focal BHCswere acquired in 1987, and thus, did not reportindependent ROA figures, it is impossible tomeaningfully compare profitability differencesfor later years.

However, substantial differences exist betweenthe two groups identified with cognitive data asto whether the members remained independent,or if acquired, what type of firm made theacquisition up to 5 years after executive assess-ments were made. In Cluster 1. both Harris andLaSalle had been purchased by internationalbank holding companies when data were collectedin 1986. Subsequent to the interviews, LaSalle'sparent, ABN, also acquired Exchange. The threeother members of this cluster have remainedindependent. None of the core BHCs in Cluster 1have changed ownership since the research wasconducted.


Table 4. Assignment of bank holding companies to clusters by informants whose data resulted in a twocluster solution

Bank holding companiesInformants

A2 A3 A4 A5 A6 A7 Bl B4 CI C2 C5 C6 C7 DI D2

ABN-USalleAffiliatedBoulevardCole-TayiorContinental 111ExchangeFirst ChicagoFirst ColonialFirst EvergreenFirst IllinoisFirst UnitedGary WheatonHarrisLake ShoreNorthern TrustState NationalUnibancorpUSAmeribancs

1 1 1 1 1 1 1 1

1 1

1 1

1

11

2

221X

I

2

1

11

2

2211I

I

2

11

2

221I1

2

X

11

2

2r-l

1r-l1

2

2

11

2

22121

2

I1

1X

1

2121

22

2121

2

2I11

22

2I2I

1

1

1

X

2

X

1

122

2

11

2

2

1

I22

2

1X

2

2

1

11X

1

X

2

2

I

122

2

11

2

2

21

2

222

12IX

21

2

222

1212

X = outlier firm

Table 5. Assignment of bank holding companies to clusters by informants whosedata resulted in a three cluster solution

Bank holding companies

ABN-LaSalleAffiliatedBoulevardCole-TaylorContinental IIIExchangeFirst ChicagoFirst ColonialFirst EvergreenFirst IllinoisFirst UnitedGary WheatonHarrisLake ShoreNorthern TrustState NationalUnibancorpUSAmeribancs

B2

32I21

22

2131

3

B3

32111

22

2121

3

B.S

32I3I

22

2I31

2

informantsB6

3213I

22

2131

3

C3

123

1

122

2

13

2

2

D2

3

321

2

322

1332

D4

3

321

3

222

I232

D5

3

23I

2

222

1232


Table 6. Profitability and ownership changes

Cluster ICore Firms

First ChicagoNorthern TrustContinental IllinoisHarris

Secondary FirmsABN-LaSalleExchange

Mean ROA. Cluster 1

Cluster 2Core Firms

First EvergreenGary WheatonAffiliatedFirst UnitedUSAmeribancsState NationalFirst Illinois

Secondary FirmsFirst ColonialCole-Taylor

Mean ROA. Cluster 2

Cluster 3BoulevardLake ShoreUnibancorp

1986 ROA

0.710.570.50.7

0.520.53

0.59'

1.31.1

-1.0

-0.851.2

0.82

1.05

1.01.01.1

Ownership^

IndependentIndependentIndependentBank of Montreal

Algemene Bank NederlandAlgemene Bank Nederland

IndependentFirst ChicagoManufacturers (Michigan)First ChicagoNBD BancorpNBD BancorpBancOne (Ohio)

IndependentIndependent

IndependentIndependentOld Kent Financial

'Mean ROA for Cluster 1 and Cluster 2 arc significantly different at the 0.005level of significance using a Mest of differences between group means for smallsample sizes.'Ownership changes which occurred after the interviews were conducted andannounced as of October I, 1991 are indicated in bold. All of the focal BHCswere independent at the time of the study except for the following; LaSalle waspurchased by Algemene Bank Nederiand (ABN) on August 14. 1979; Harrisbecame a wholly-owned subsidiary of the Bank of Montreal on Sept. 4,1984; NBD Bancorp (headquarter in Michigan) announced the purchase ofUSAmeribancs on October 23. 1986. Interviews were conducted in November1986.

The BHCs in Cluster 2 have faced a quitedifferent fate. Five of the nine BHCs have beenacquired, all by BHCs headquartered in themidwestern United Slates. Only one of the coreBHCs in this cluster. Evergreen, has remainedindependent. The other two banks which haveremained independent in this group. First Colo-nial and Cole Tayior, were the members mostoften assigned to other clusters.

Performance at the time of data collection andsurvival 5 years later thus support the cognitivedata on strategic group assignment. Firms inCluster 1, especially those assigned unambigu-ously, did not change ownership. In fact, theonly member of this group that was acquired,Exchange, followed the pattern set by othergroup members—it was acquired by a foreignbank holding company. All except one of

114 R. K. Reger and A, S. Huff

the firms assigned by at least 90 percent ofrespondents to Cluster 2 were acquired byregionally-based U.S. bank holding companieswithin 5 years of our analysis. TTiese resultsare supportive of the other findings on thehomogeneity of firms within cognitive groups,however, the small sample sizes and the corre-lational nature of the analysis precludes theassertion of a causal link between these outcomesand strategic group membership.

DISCUSSION OF RESULTS

Given the amount of change in the Chicagobanking market from the early 1980s on, thesupport for the basic propositions of this studyis surprisingly strong, especially since respondentsused their own dimensions to describe firmstrategies rather than a common set of researcher-supplied categories. Nonetheless, idiosyncraticassessments in the data deserve further expla-nation. Why were some firms unambiguouslyassigned to a strategic group across informants,while placement of other firms is more uncertain?Further examination of the data suggest sixarguments which may explain disagreement aboutstrategic group membership for some firms.These arguments address the nature of strategicpositioning as well as the ability of researchersand competitors to perceive subtleties in thestrategic choices firms make.

1. Fuzzy Groups. Some BHCs that were notuniformly assessed by informants may haveexhibited strategies essentially in line with thoseof one or more 'core' BHCs in a particular group,but these firms also may have shared secondaryattributes with other groups or possessed idiosyn-cratic secondary traits. Zadeh (1965). a leadingcategorization theorist, has suggested that manyconceptual categories are "fuzzy" on their bound-aries, sueh that some members of the category arcbetter examples of the category than are others.Thus, a firm that shares many commonalities withthe category is likely to be widely perceived as agood example of the category. A firm that sharesfewer characteristics is likely to be associated witha wider variety of firms. Some informants mighthave concentrated on core aspects of this firm'sstrategy, while other informants focused on second-ary characteristics the firm happens to share withfirms in other clusters.

2. Obscured Strategy. Disagreement amongthis group of expert observers about the place-ment of a few BHCs also may be due to thedifficulty of observing some firms' strategicactions. Porter (1980) and others have suggestedthat firms are wise to obscure their intentions inorder to gain competitive advantage. Thus, it islikely that the strategies of some firms, eitherintentionally or unintentionally, will be difficultto interpret. Given the theoretical importance ofsurprise in establishing competitive advantage,our results are interesting not because a fewfirms' strategies were difficult to identify, butbecause respondents agreed on the basic strategyof almost all firms in the population. Obscuredstrategy appears to have played, at most, anextremely limited role in this industry.

3. Firm Repositioning. Even during the moststable periods of industry evolution, the stra-tegies of some BHCs may be in fiux due lofirm specific conditions such as new manage-ment, poor past performance or firm specificinnovation. Respondents in our study mighthave varied in their perceptions of such firms.Some observers may have focused on the BHC'sold strategic position, some on current positionand some may have been oriented towardanticipated future position.

4. Industry Realignment. Another possibleexplanation for variation in the categorization ofsome firms is tied to the structure of the industryas a whole. An industry in transition, such asthe one we observed, is likely to be movingtoward a larger or smaller number of viablegroup 'recipes' (Spender, 1980, 1989), or. if thekey dimensions of strategy are themselves in theprocess of change, some observers may havebeen more aware of this transformation thanothers. Inconsistencies in the data may thus bedue to some informants reporting an 'old"structure, while others anticipate a new industrystructure. Though similar in effect to internallymotivated repositioning of individual firms, indus-try realignment is conceptually different becauseit is more closely tied to external environmentalchange.

3. Reactors and Misfits. Some companies donot follow clear, constant strategies. Such firmshave been described as 'stuck in the middle'(Porter, 1980) or 'reactors' (Miles and Snow,1978); they change strategic direction often orsimply react to environmental events as they


occur. These firms might be best described asmisfits in that they do not exhibit a coherentstrategic pattern over time. It is not surprisingthat industry observers vary in the way theydescribe such firms.

6. Strategic Idiosyncracy. Some focal BHCsmay have adopted strategies which are sodramatically different from all others in theindustry that it is not easy to position them v'(.s-a-vis other firms along common dimensions.These firms are not aligned with other firms;they do not compete in similar ways. Such firmsmay have a coherent strategy and that strategymay fit the industry, but it does not fit thecognitive map that usefully describes almost allother competitors within the industry. WhenTimex first entered the watch industry, it wasmisaligned in this way. Its strategy was so radicallydifferent from existing watch manufacturers thatit could not be meaningfully mapped on dimen-sions which differentiated other competitors.New dimensions, and hence a new map, wereneeded. Lack of consensus among industryobservers may indicate analogous outliers in theChicago banking market.

These arguments are complementary, notmutually exclusive. Multiple explanations appearto account for disagreements about the placementof all six equivocal firms in our study. Forexample, the banking industry was undergoing amajor realignment at the time data were collectedfor this study (McKinsey and Company, 1985).Several informants indicated both Lake Shoreand Boulevard were pursuing strategies that wereradically different from other bank holdingcompanies at this time. ABN-LaSalle. on theother hand, was often described as both a misfitand a firm with a 'fuzzy" strategic position. Thisfirm had been acquired twice in the 7 yearspreceding the study. It had long operated undera consistent strategy which aligned it with firmsin Cluster 1, and at the time of the study itretained strengths associated with that strategy.New ownership had moved the company towardstrategies more distant from either Cluster 1 or2, but decisions in the 1986 time frame wereinterpreted by some observers as reasserting astrategy in line with Cluster I. The lack ofagreement about the association of ABN-LaSalleacross informants appears to be a good indicatorof genuine uncertainty about the strategic direc-tion of this firm.

A BROADENED THEORY OFSTRATEGIC GROUPS

While there is some ambiguity in the currentiiterature about the ontological status of strategicgroups, with the skeptic position expressedespecially by Hatten and Hatten (1987) andBarney and Hoskisson (1990), our study suggeststhat strategic groups are readily perceived bystrategists. The study thus provides evidencethat strategic groups are more than analyticalconveniences used by researchers; they are partof the way strategists organize and make sense oftheir competitive environment. The performancedifferences between Cluster 1 and Cluster 2 atthe time of the study as well as the utilityof these groups for differentiating acquisitionexperience 5 years after data collection suggestpromising avenues for future research. Explainingperformance was not the foeus of the currentresearch and these results, although suggestive,do not establish a causal link between groupmembership and outcomes. Nonetheless, thestrategic groups derived from the cognitivestructure of strategists in this industry demon-strate that executive perceptions have real poten-tial for competitive strategy research. Further,the results contribute to evidence from previousempirical studies using different methods (Poracet al., 1989; Porac et al, 1987). Accumulatingevidence suggests that strategic groups arephenomena with an ontoiogical status; they arenot the contrivance of a single method. This isparticularly important as industry level groupings,such as those based on SIC eode assignments,become more and more tenuous.

Disagreement on strategic group assignmentfor some firms suggests, however, that futureresearch might benefit from a revised viewof strategic group structure. Existing researchimplicitly assumes that all firms follow a strategy,that this strategy is knowable by outsiders, andthat all firms can be assigned unambiguously toa strategic group (even if the group has only onemember). Few of the studies reviewed by Cool(1985) or by McGee and Thomas (1986) removedeven a single firm from the analysis due tointerpretability problems (Hatten, Schendel andCooper, 1978, is an exception). The results ofthis study support the existence of strategic groupstructure, but suggest several ways in whichcurrent conceptualizations could be enriched.


1. Every firm does not have to have a strategicposition. The overwhelming majority of inform-ants in our study classified two-thirds of the BHCsinto one of two strategic clusters. Agreement waslower for the other 6 firms in the population.Three firms (Boulevard Bankcorp, Lake ShoreBankcorp and Unibancorp) appear in all threeclusters and, at the least, do not appear to befollowing a widely agreed upon strategy. Adistinction must be made between strategicsituation, which characterizes the unique environ-mental and internal circumstances challengingevery firm, and strategic position, which charac-terizes the firm's response to those circumstances.Every firm faces a strategic situation. These data,however, suggest that the largely unexaminedassumption of existing strategic group theory,that every firm has a strategic position, may betrue for only some firms in the industry. Thecompetitive positions of at least some firms arenot easily assessed and/or readily assigned to astrategic group. When strategy is conceptualizeda.s a pattern over time (Mintzberg, 1978), it isquite likely that some firms do not have acoherent strategy. Including such firms willdiminish the usefulness of studies observingstrategic position.

One particularly interesting implication ofthis argument for future research is that pastcorrelational studies might be usefully re-exam-ined. For example, the ambiguous results ofstudies attempting to link strategic group member-ship to profitability seem to have brought thisline of inquiry to an impasse. If problematicfirms are removed from group assignments, it ispossible that more conclusive results could beachieved.

2. Group membership is a matter of degree.Strategic groups were 'fuzzy sets' when aggregatedacross informants in this study. A more fine-grained view of membership within groups mightbe profitable for many strategic group studies.Our results suggest that a strategic group mightbe best conceptualized as a core group of firmsthat define the group position and secondaryfirms that are aligned with core firms in manyessential respects, though they also make someunique strategic decisions. Porter (1980: 154)implicitly acknowledges the possibility of ambigu-ously defined groups when he advises analyststo draw many strategic group maps using differentdimensions depending upon the analyst's pur-

poses. However, no published empirical workhas yet taken this approach.

The general point worth considering is thatsome firms are prototypical of a group's position,while others are not. Categorization theoristslabel this phenomenon 'membership gradience'(Berlin and Kay, 1969; Zadeh, 1965). However,it is also probable that the firms included in aparticular analysis will determine which firms arecore and which are secondary (Lakoff, 1987). Amap of the financial services industry (Fombrunand Zajac. 1987) provides different results thanone which includes commercial banks only(Passmore, 1985). Reger (199()a) pursues theramifications of this point for the design ofcompetitive positioning research.

3. Some groups may be tightly associatedwhile others are more diverse. One interestingobservation on the pattern of cluster formationfrom our dala analysis is that some firms arerapidly linked by clustering algorithms whileothers lie at greater distance and are incorporatedinto a group much later in the clusteringprocedure. The implication is that some firms inan industry are quite similar while others,though associated, are more disparate. Studiesof performance, and other potential correlatesof group membership, again might benefit fromdistinguishing 'tight' and 'loose" groups.

4. Overlapping group membership may charac-terize some industries. Our data suggest thatstrategic groups may be overlapping at theirperiphery, with a small set of firms sharing somestrategic characteristics with one group and otherstrategic characteristics with another. While thispossibility muddies the ideas of symmetry andsimplification that are basic attractions of thestrategic groups idea, allowing overlap may benecessary if we wish to more realistically mapcomplex industries. The existence of overlap alsoraises some interesting research questions. Firmsat the intersection of two existing groups mayindicate a potentially viable location for newgroup formation, for example, or they mayrepresent the problem of literally being 'stuck inthe middle' (Porter, 1980).

5. For periods of time, group structure maynot exist or may not be apparent. Cool (1985)and Fiegenbaum (1987) identify stable strategictime periods in an industry by stable strategicgroup membership. Taking the previous sugges-tions (1 through 4) to an extreme, it may be


necessary to allow not only for shifting groupmembership between stable time periods, butfor periods in which an industry can not beusefully defined in terms of strategic groupstructure at all. Mintzberg (1978) suggests thatfirms often go through periods of inchoatesearch before establishing a strategic directionor redirection. The same may be true forindustries as a whole. Economic conditions canbe in such flux that firms will not be able tofollow consistent strategies, and. as firms changeposition this movement will tend to furtherdestabilize the industry. However, we expectedless stability in this study of bank holdingcompanies than we found. Even in a period ofuncertainty and search for new strategies—likethe one that currently characterizes the bankingindustry—it may be that similar firms muddle insimilar ways and the strategic groups idea remainsuseful.

In summary, based on our findings from thebank study, we propose that strategic groupstructure be reconceptualized as graphicallydepicted in Figure 1. This figure provides ageneral picture of strategic groups composed ofthe following kinds of firms:

- Core firms that are tightly associated anddefine the basic 'recipe' of a strategic group.- Secondary Group Members that implementthe strategic group recipe less consistentlythan core firms.- Transient firms whose strategies are changingfrom one strategic position to another, butalong dimensions common to other firms inthe industry.

Left out of this schematic are two kinds of firms:

- Misfit firms whose strategies are inconsistentover time.- Idiosyncratic firms whose strategies cannotbe easily expressed in terms of the key strategicdimensions used to define the competitivepositions of most firms in the industry.

DIRECTIONS FOR FUTURE RESEARCH

If strategic group membership is reconceptualizedas a matter of degree, a number of interestingresearch possibilities become available.

^ P Core Rrms

C_J Secondary Firms

\^-—^^- Cj Transient FirmsMisfits and Idiosyncratic Firms (inappropriate lo map)

Figure 1. A reconceptualization of strategic grouprelationships

1. Strategic groups research provides an oppor-tunity to look more closely at viable strategicalternatives. If groups are defined solely ondimensions associated with the core membersof the group, clearer insights into strategicalternatives may be forthcoming. Deviation fromthe core might indicate either inability toimplement the group recipe completely or deliber-ate attempts to differentiate the firm from coremembers. The distinction between imitabiiityand differentiation might be addressed moredirectly by distinguishing core and secondaryfirm characteristics.

2. Intragroup dissimilarities may be clarifiedby exploring the roles different types of firms playin industry dynamics. For instance, the extent towhich core firms serve as referents for othermembers or tend to be first movers are promisingresearch avenues. We hypothesize that core firmswill vary in terms of innovativeness, but thatthey will serve as key referents and tend todominate external views of the industry.

Transients, misfits and idiosyncratic firms mayprovide important clues as to future strategicoptions or necessities in an industry, and aretherefore Important to those interested in industryevolution and competitive dynamics (Lewin andMinton, 1986). From a population ecologyperspective, these firms may represent the 'gene

l l g R. K. Reger and A. S. Huff

pool' for future variation in response to newconditions; they provide some evidence as towhat may be more or less attractive positions inthe future. These firms provide the counterbal-ance to what can be learned from core firms.Core firms reflect the current consensus ofopinion about viable modes of operating withinan industry context; outlier firms suggest futurepossibilities. This evidence is especially importantas exogenous forces impinge upon the industryor powerful competitors change the nature ofcompetitive dynamics within the industry.

3. Attention should be directed at more validmethods for determining group membership andstructure. Valid strategic groups should accountfor outcomes of interest in strategic managementsuch as strategic choice and performance. Ourresults tentatively suggest that cognitively-derivedstrategic groups are promising for exploringperformance outcomes and independent survivalin a consolidating industry. More research shouldfocus on the usefulness of strategic groups forpredicting other strategic outcomes. One focusof this inquiry might be on the relationshipsbetween 'competitive' groups (firms that competeagainst each other) and "strategic* groups (firmsthat compete using similar strategies). Theinteraction between changes in strategic po-sitioning as perceived by managers and changesevident in 'objective' data is also an alluringresearch area. Under what conditions do mana-gerial perceptions lead, or lag, economic andfinancial indicators of changing positions?

4. The study of cognitive strategic groups canenrich the theory of how analytically identifiedgroups form and why they tend to remainstable. We suggest that the 'analytic' strategicgroups formed by observable variations inaccounting and financial data (Cool, 1985; Fiegen-baum, 1987) owe their regularities, at least inpart, to the process of enactment (Weick, 1979).It is reasonable to expect that the groupsmanagers perceive will have real effects onstrategy reformulation, strategic action and sub-sequently on industry structure. Economic reali-ties make strategic similarities likely, but cognitiveprocesses also may be expected to reinforceeconomic influences and thus help maintain theexisting group structure. As Pfeffer and Salanciknote, "environments are not given realities, butcreated through a process of attention andInterpretation' (1978: 13). Enactment of strategic

groups may be aided by several processesincluding simplification, elaboration, interaction,borrowed experience, and expectations. In aggre-gate, these processes provide the mechanismswhereby cognitive and economic realities con-verge.

The effect of cognitive simplification, wesuggest, is not just that strategists will tend tothink of competitors in clusters, but that theywill tend to recognize similarities between theirown firm and a set of fellow firms and actaccordingly. While the link between thought andaction remains problematic in social scienceresearch, most strategy researchers assume someconnection. It then follows that cognitive simpli-fication should contribute to simplification ofaction. As firms recognize links between them-selves and others, and act based on these apparentsimplifications, groupings of firms within anindustry will tend to become even more distinct.

Cognitive elaboration, the process which cre-ates information about competitors, may alsolead to the enactment of strategic groups. Tothe extent strategists find similarities betweentheir firm and others, and act upon thesesimilarities, they can create further alignment ineconomic realities. For example, if firm Abelieves its competitor B faces similar productioncapacity constraints (whether this is true or not),it may monitor and react to B's capacity expansiondecisions in a number of ways that work to fulfillthat expectation. If B begins to act as if it mightincrease its capacity (by initiating exploratorystudies, for example), A may begin to contem-plate a similar move that it had not here-to-foreconsidered. Firm A interprets B's actions in away that leads it to imitate B's strategy. A'sassumption of the need for competitive reactionthus serves to bring the two firms' strategies intocloser alignment. Our point, again, is thatperceived similarities, whether or not they accu-rately reflect reality initially, channel actions intoa smaller number of alternatives than mightotherwise be the case and strategic groups arelikely to further coalesce.

Likewise, interaction among strategists andwith industry analysts, distributors and otherthird parties tends to result in similar thoughtpatterns (Porac et ai, 1989). Executive fraterni-zation makes it more likely that firms will reactin similar ways to threats and opportunitieswithin the industry—further clustering firms.


While shared information and interpretationare expected to promote similar decisions, cogni-tive clustering is even more likely when strategists'borrow experience' through observing otherfirms in the industry as a substitution for directexperimentation with strategic alternatives (Huff,1982). As firms experience success and failure.Iheir actions will tend to be copied and/oravoided by others in the industry. Over time weexpect this process of replication and circumven-tion to make a major contribution to clusteringfirms within an industry.

As strategists scrutinize other firms within anindustry and develop mental frameworks forinterpreting what they see. they come to expectcertain behavior and act on these expectations.Expectations thus are another force for chan-neling firms into a limited repertoire of behavior.If a firm is classified as aggressive, for instance,small changes in its pricing policy will be givenincreased importance and retaliatory responsesare more likely. Similar actions by another firm.perceived to be following a less aggressive path,will be taken less seriously. Such expectationsexist not only among competitors, but amongbuyers, suppliers and potential entrants. Theresult, particularly given limited ability to makedistinctions among firms due to cognitive over-load, is to further homogenize both the perceptionof strategic options within an industry and therepertoire of strategic actions taken.

LIMITS TO COGNITIVE GROUPRESEARCH

In our view, these are strong arguments forassuming that strategists' cognitive structures willhave a material effect on strategic choices. Aschoices channelled by managerial perceptioncontribute to the economic structure of ;mindustry, the external stimulus for future percep-tions of similarity are further reinforced. Inother words, cognitive groups tend to reinforceeconomic groups. They combine with forces atother levels that create institutional isomorphism(DiMaggio and Powell, 1983).

The spiral of mutual reinforcement should behalanced by evolutionary changes in technology,buyer demographics and so on. along with thedeclining profitability of overpopulated strategicpositions. These evolutionary agents force firms

to change their strategies and break groups apart,Research shows, however, that established mentalmaps lead individuals to ignore contradictorydata (e.g., Prahalad and Bettis, 198ft). Thus aproblem with cognitive associations is they maynot reflect evidence from a changing world.Cognitive structures also are inevitably based onincomplete knowledge, and even the simplestinferences are frequently biased (Schwenk, 1984).

These limitations on the accuracy of managerialperceptions limit the research utility of managerialohservations in general, and cognitive strategicgroups in particular, but they do not justifyabandoning cognitive data. In our view, strategyresearchers interested in competitive interactionshould treat participants in an industry as expertwitnesses and important sources of data. Manybiases and limitations can be mitigated, if notcanceled, by broadly seeking opinions amongindustry participants. Furthermore, accountingand financial data have their own categorizationand reporting biases that might be more easilyseen when compared with the categorizations ofindustry participants.

Our results must be interpreted with cautionsince the study focuses on only one geographicarea of one industry. The banking industryin Chicago was facing unique environmentalchallenges stemming primarily from state deregu-lation. Thus, future research will be necessaryto determine the generalizability of these resultsto other industries facing other environmentalconditions. Another limitation has to do withthe small number of informants; the intensiveinterviewing needed to collect rich data neces-sarily limits cognitive studies. Future researchmight be usefully directed to the optimal numberof subjects necessary to establish a "cognitivecommunity' (Porac et al.. 1989). Again, furtherresearch is required to examine these generaliz-ability issues.

CONCLUSION

Strategic groups offer a promising way tosummarize competitors' strategies in industriespopulated by so many competitors that individualconsideration of each firm is imptissible. Thisstudy theoretically argues and finds empiricallythat even in a changing industry strategistscognitively group their competitors' strategies.


In fact, there is strong uniformity across oursample of informants in the grouping of almostall 18 firms studied. These groups accuratelypredicted subsequent performance differencesand acquisition patterns.

Managers group firms in subtle ways notcaptured by economically oriented research.If more subtle distinctions in strategic groupmembership along the lines shown in Figure 1are identified, we believe new utility in thestrategic group concept may emerge. Even moreinteresting, in our view, is the potential of arefined concept of strategic groups for furtherunderstanding competitive positioning. What arethe differences between core and secondarygroup members? Are there barriers to mobilityeven within groups? How will firms perceived asfollowing aspects of more ihan one strategy fare?Are they stuck in the middle or are theydeveloping a new and viable niche? Oncequestions of positioning like these become thefocus for research, it becomes more and moreinteresting lo include practitioners as a source ofdata on strategic groups. The fact that we foundsuch strong consensus among informants, evenin a rapidly changing industry, suggests (hatpractitioners can be a reliable and valid sourceof data. The apparent ability of these observers tocapture nuances in association among compclitorsover time suggests that this dala can considerablyenrich archival studies.

ACKNOWLEDGEMENTS

The authors wish to thank Irene Duhaimc. LuisGomez-Mejia. Richard G(x)ding and anonymousStrategic Management Journal reviewers for theircomments.

REFERENCES

Bamey. J. B. and R. E. Hoskisson. 'Strategicgroups: Untested assenions and research proptisals\Managerial and Decaion Economics. 11, I99().pp. 187-198.

Berlin. B. antl P. Kay. Bmic Color Terms: TheirUniverstiliiv and Evoluiion. University of CaliforniaPress. Berkeley. CA. 1969.

Boeker. W. Organizational strategy: An ecologicalperspective'. Academy of Management Journal. 54.1991. pp. 613-635.

Caves. R. E. and P. Ghemawat. 'Identifying mobility

barriers'. Strategic Management Journal, 13, 1992,pp. 1-12.

Cool. K. O. "Strategic group formation and strategicskills: A longitudinal analysis of the U.S. pharma-ceutical industry'. I%.V19K2'. Unpublished dtx'toraidissertation. Purdue t'nivcrsiiy. May \W5.

Cool. K. O. and D. I:. Schendel. 'Strategic groupformation and performance: The case of the U.S.pharmaceutical industry. 1963-1982". ManagementScience. 33(9), 1987. pp. 1102-1124.

Cool. K. O. and D. E. Schendel. "Performancedifferences among strategic group members', 5/re-regie Management Journal, 9. 1988. pp. 207-223.

Dess. G. G. and P. S. Davis. Torters (1980)generic strategies as determinants of strategic groupmembership and organizational performance'.Academy of Management Journal. 27. 1984.pp. 467-488.

DiMaggio. P. J. and W. W. Powell. The iron cagerevisited: Institutional isomorphism and collectiverationality in organizational tields", American Socio-logical Review. 48(2). 19H3. pp. 147-160.

Dunn. W. N.. A. G Cahill. M. J. Dukes andA. Ginsberg. "The policy grid: A cognitivemethodology for assessing policy dynamics". InW. N. Dunn (ed.). Policy Analysis: Perspectives,Concepts, and Methods. JAI Press, New York,1986. pp. .155-375.

Dunn, W. N. and A. Ginsberg. "A sociocognitivenetwork approach to organizational analysis'.Human Relations, 40(11). 1986. pp. 955-976.

Dutton. J. E. Separating the wheat from the chaff:The important dimensions of strategic issues'.Conference on Managerial Thinking in BusinessEnvironments. Boston University. 1987.

Dutton. J. E.. L. Fahey and V. K. Narayanan.'Toward understanding strategic issue diagnosis',Strategii Management Journal. 4,1983. pp. 307-324.

Eden. C , S. Jones and D. Sims. Thinking inOrganizations, Macmillan. London, 1979.

Eden, C . S. Jones and D. Sims. Messing Ahout inProblems. Pergamon. Elmsford. NY. 1983.

Everitt. B. Cluster Analy.m, second edition. Heine-mann Educational Bmiks. London. 1980.

Fiegenhaum. A. 'Dynamic aspects of strategic groupsand competitive strategy: Concepts and empiricalexamination in the insurance industry'. Unpublisheddoctoral dissertation. University of Illinois atUrbana-C!hampaign. 1987.

Fiegenbaum. A. and H. lliomas. 'Strategic groupsand performance: The U.S. insurance industry.I97i)-\^M. Strategic Management Journal. I I . 1990,pp. 197-215.

Fombrun. C J. and E. J. Zajac. "Struclural andperceptual influences on intraindu.stry siralilication'.Academy of Management Journal. 30. 1987.pp. 33-50.

Fransella, F. and D. Bannister. A Mamtalfor RepertoryGrid Technique. Academic Press. New York. 1977.

Ginsberg. A. "A cognitive methodology for assessingstrategic diversity". Conference on ManagerialThinking in Business Environments. Boston Univer-sity, 1987.


Green. P. E. 'A new approach to market segmentation'.Business Horizons. Fehruary 1977. pp. 61-73.

Hartigan. J. A. Clustering Algorithms, Wiley. NewYork. 1975.

Hatten. K. J. and M. L. flatten. "Strategic groups,asymmetrical mobility barriers and conteslability'.Strategic Management Journal. 8.1987, pp. 329-342.

Hatten, K. J.. D. E. Schendel and A. Cooper, 'Astrategic model of the U.S. brewing industry:1952-1971', Academy of Management Journal. 21,1978. pp. 592-610.

Huff. A. S. industry influences on strategy reformula-tion'. Strategic Management Journal, 3, 1982,pp. 119-131.

Muff. A. S. Mapping Strategic Thought, Wiley,Chichester, 1990.

Hunt, M. S. 'Competition in the major home applianceindustry, 1960-1970'. Unpublished doctoral disser-tation. Harvard University, 1972.

Jick, T. D. Mixing qualitative and quantitativemethods: Triangulation in action". AdministrativeScience Quarterly. 24. 1979, pp. 602-611.

Johnson. G. and H. Thomas. "The industry contextof strategy, structure and performance: The U.K.brewing industry'. Strategic Management Journal.8, 1987, pp. 343-361.

Johnson-Laird. P. N. and P. C. Wason. Thinking.Cambridge University Press, Cambridge, MA, 1977.

Kelly. G. A. The Psychology of Personal Comtructs.Vols. I and 2. Norton. New York, 1955.

Kuhn, T. The Structure of .Scientific Revolutions.University of Chicago Press, Chicago, IL. secondedition. 1970

Kumar. K. R.. H. Thomas and A. Fiegenhaum."Strategic groupings as competitive benchmarksfor formulating future competitive strategy: Amodelling approach'. Managerial and DecisionEconomics. II. 1990. pp. 99-109.

Lakoff. G. Women, Fire, and Dangerous Things.University of Chicago Press, Chicago. IL. 1987.

Uwless, M. W.. D. D. Bergh and W. D. Wilsted.'Performance variations among strategic group mem-bers: An examination of individual firm capability'.Joumal of Management. IS. 1989. pp. 649-661.

Lawless. M. W. and L. K. Finch. "Choice anddeterminism: A test of Hrebiniak and Joyce'sframework on strategy-environment fit*. StrategicManagement Journal. 10. 1989. pp. 351-365.

Lawless. M. A. and 1,. F. Tegarden, 'A testof performance similarity among strategic groupmembers in conforming and non-conforming indus-try structures*. Joumal of Management Studies. 28.I99I. pp. M.S-664.

Lewin, A. Y. and J. W. Minton. Determiningorganizational effectiveness: Another look, and anagenda for research". Management .Science, H,1986. pp. 514-538.

Lewis, P. and H. Thomas. 'TTie linkage betweenstrategy, strategic groups, and performance in theU.K. retail grocery industry". .Strategic ManagementJournal, 11. 1990,'pp. 3H.C397.

Mascarenhas. U. "Strategic group dynamics'. Academyof Management Journal. 32, 1989. pp. 333-352.

Mascarenhas, B. and D. A. Aaker. 'Mohiliiy barriersand strategic groups', .S'/m/f̂ ic Management Journal.10, 1489. pp. 475-*«5.

Mastin. R. O. and I. I. Mitroff. Challenging StrategicPlanning Assumptions. Wiley, New York. 1981.

McGee. i. and H. Thomas. 'Strategic groups: A usefullinkage between industry structure and strategicmanagement'. Strategic Management Journal. 7,19H6. pp. 141-160.

McKinsey and Company. Bankers' Merger and .Acqui-sition Choices: .Mapping a Course through Consoli-dation. Bank Administration Institute. RollingMeadows, IL. 1985.

McNamee, P. and M, McHugh. 'Mapping competitivegroups in the clothing industry (Part 2)', LongRange Planning, 22, 1989, pp. 89-97.

Miles, R. E. and C. C. Snow. Organization Strategy.Structure, and Process. McGraw-Hill, New York,1978.

Miller. G. A. 'The magic number seven plus or minustwo: Some limits tci our capacity for processinginformation'. Psychology Review, 50, 1956, pp. 81-97.

Mintzberg, H. D. 'Patterns in strategy formation',Management Science, 24, 1978, pp. 934-948.

Nohria, N. and C. Garcia-Pont. 'Global strategiclinkages and industry structure'. Strategic Manage-ment Journal. l2(Spccial Issue). 1991. pp. 105-124.

PassmorL'. S. W. "Strategic groups and the profitabilityof hanking*. Research paper No. 8501. FederalReserve Bank of New York. New York City, 1985.

Pfeffer, J. and G. R. Salancik. The External Controlof Organizations, Harper & Row, New York. 1978.

Porac. J. F. and H. Thomas. "Taxonomic mentalmoilels in competitor dehnition". Academy ofManagement Review, 15, 1990, pp. 224-240.

Porac, J. F., H. Thomas and C. Baden-Fuller.'Competitive groups as cognitive communities: Thecase of Scottish knitwear manufacturers". Journalof Management SttuUes. 15, 1989, pp. .397^*16.

Porac. J., H. Thomas and B. Emme. 'Understandingstrategists' mental models of competition." In G.N. Johnson, (ed.). Business Strategy and Retailing.Wiley, New York and Chichester, 1987, pp. 59-79.

Porter, M. E. Competitive Strategy. Free Press. NewYork, 1980.

Powell, W. W. and P. J. DiMaggio (eds.). TheNew Institutionalism of Organizational Analysis.University of Chicago Press. Chicago. IL. I99I.

Prahalad. C. K. and R. P. Bettis. 'The dominant logic:A new linkage hetween diversity and performance'.Strategic Management Journal. 7. 1986. pp. 48.^501.

Reger, R. K. 'Managerial thought structures andcompetitive positioning'. In A. S. Huff (ed.).Mapping Strategic Thought. Wiley. Chichester,1990a. pp. 71-88.

Reger, R. K. 'The repertory grid technique for elicitingthe content and structure of cognitive constructivesystems'. In A. S. Huff (ed.). Mapping StrategicThought. Wiley. Chichester. 1990b, pp. .301-309.

Reger. R. K. and T. B. Palmer. "Cognitive schemasof competition: Understandings of strategy in aturbulent industry'. Paper presented at the Academyof Management Meetings, August, 19911.


Reynolds. H. T. Analysis of Nominal Data, Sage,Beverly Hills, CA, 1977.

Romesburg. H. C. Cluster Analysis for Researchers.Lifetime Learning Puhlications, Belmont, CA.1984.

Rosch. E. 'Principles of categorization'. In E. Roschand B. Lloyd (eds). Cognition and Categorization.Erlhaum, Hillsdale. NJ, 1978. pp. 27-^8.

Rosch, R. 'Prototype classification and logical classifi-cation: The two systems'. In E. Scholnick (ed.).New Trends in Cognitive Repre.tentation: Challengesto Piagets Theory. Lawrence Erlbaum Associates,Hillsdale. NJ, 1981, pp. 73-«6.

SAS Institute, Inc. SAS Users Guide: Statistics. Version5 Edition. SAS Institute, Cary, NC, 1985.

Schroeder, D. M. 'A dynamic perspective on theimpact of process innovation upon competitivestrategies ' . Strategic Management Journal. I I , 1990,pp. 25-41.

Schwenk. C. R. "Cognitive simplification processes instrategic decision-making'. Strategic ManagementJournal. 5. 1984, pp. 111-128.

Simon. H. A. Administrative Behavior. Macmitlan,New York, 1945.

Spender, J.-C. 'Strategy making in business'. Unpuh-lished doctoral dissertation. Manchester BusinessSchool, University of Manchester, 1980.

Spender, J.-C. Industry Recipes, Basil Blackwell,Oxford, England, 1989.

Stubhart, C. I. 'The cognitive psychology of strategicplanning: Research opportunities'. National Acad-emy of Management Meetings, August 1986.

Stuhbart, C. I. "Cognitive science and strategicmanagement: Theoretical and methodological

issues'. National Academy of Management Meet-ings, August 1987.

Sudharshan. D., H. Thomas and A. Fiegenhaum.'Assessing mohility harriers in dynamic slrategiegroups analysis'. Journal of Management Studies.2S. 1991. pp. 429-438.

Tallman. S. B. 'Strategic management models andresource-hased strategies among MNEs In a hostcountry'. Strategic Management Journat, I2(SpeciaIIssue), 1991, pp. 69-82.

Thomas. H. and N. Venkatraman. 'Research onstrategic groups: Progress and prognosis'. Journalof Management Studies. 25, 1988, pp. 537-555.

Tushman, M. L. and E. Romanelli. 'Organizationalevolution: A metamorphosis model of convergenceand reorientation'. Research in OrganizationalBehavior, vol. 7. JAI Press, Greenwich, CT, 1985,pp. 171-222.

Ulrich, D. and B. McKelvey. 'General organizationalclassification: An empirical test using the UnitedStates and Japanese electronics industries'. Organi-zation Science. I. 1990, pp. 99-118.

Walton, E. "Managers' prototypes of financial terms'.Journal of Management Studies. 23, 1986pp. 679-^99.

Weick. K. E. The Social Psychology of Organizing,Addison-Wesley, Reading, MA, 1979.

Zadeh, L. 'Fuzzy sets'. Information and Control, 8,1965, pp. 338-353.

Zajac. E. and S. M. Shortell. "Changing genericstrategies: Likelihood, direction, and performanceimplications". Strategic Management Journal. 101989. pp. 413-430.

APPENDIX

This appendix lists examples of the dimensionselicited in the interviews. In all, 331 dimensionswere elicited, or approximately 13 per informant.The dimensions presented here have beengrouped together by categories agreed upon bythree raters. The numbers in parentheses indicatethe percentage of informants who provideddimensions in that category (only those categorieswhere at least 25 percent of informants gavedimensions are presented).

Every informant provided at least one idiosyn-cratic dimension (92 of the 331 dimensions or 28percent of the total). Some examples of dimen-sions provided by only one informant are 'ethnicclientele vs. broad based clientele.* "supptirtsliberalized branching laws vs. unit banks,' and'interested in buying savings and loans vs. notinterested in buying savings and loans.' Additionalinformation on procedures for categorizing the

dimensions and results of the dimension levelanalysis are available from the first author.

Geographic Scope (88% of informants)national market vs. local strategyinternational presence vs. no internationalpresence

Target Market: Lending (88%)money center, wholesale vs. retail operationmiddle market vs. retail, small business

Growth Strategies (79%)planned expansion through acquisition vs.internal developmentgrowth vs. retrenchment and survival weremajor concerns

Location (71%)downtown area vs. suburban operationloop area competitors vs. suburban banks


Management (63%)ability to implement strategic focus vs. nonstra-tegically focusedaverage management capabilities vs. high man-agement capabilities

HC Structure and Management (58%)unit bank vs. multibank holding company'federal approach': collection of banks vs.centralized

Trust (54%)strong trust department vs. no trust depart-menttrust department important (commercial andretail) vs. personal trust only

ProdtwtlMarket Scope (42%)broad range of financial services vs. focusedon a nichefull service banks vs. limited services

Successful Company (29%)good earnings performers vs. poor earningsperformancecontrol overhead vs, not careful with overhead

Ownership and Control (25%)public ownership vs. private ownershipindependently owned and managed by ownersvs. managers don t control

Asset Based Lending (25%)asset based lenders vs. unsecured lendingasset based lending important to strategy vs.asset based lending not important to strategy

strategic groups: a cognitive perspectiveeprints.maynoothuniversity.ie/5473/1/strategic... · by...

Documents