Cognitive Inertia in a Turbulent Market: the Case of UK Residential Estate Agents

COGNITIVE INERTIA IN A TURBULENT MARKET: THE CASE OF UKRESIDENTIAL ESTATE AGENTS*

GERARD P. HODGKINSON

Leeds University Business School

ABSTRACT

Recently, a number of researchers seeking to develop cognitive approaches forunderstanding processes associated with the development and implementation ofcompetitive positioning strategies have drawn attention to the notion of`cognitive inertia'. Once established, there is a danger that actors may becomeoverly dependent on their mental models of competitive space, to the extent thatthey fail to notice changes in the material conditions of their business environ-ment, until these changes have become so widespread, or signi®cant in otherways, that their organization's capacity for successful adaptation has beenseriously undermined. While there have been several anecdotal accounts ofcognitive inertia reported in the literature, and a number of researchers haveexplored processes of cognitive change in organizations more generally, to datethere have been virtually no studies published which have investigated thisphenomenon speci®cally within the domain of competitor assessment. This paperreports one such investigation, a longitudinal ®eld study which explored theextent to which actors' mental models of competitive space in the UK residentialestate agency (real estate) industry, an industry characterized by high volatility inrecent years, remained stable or changed over time. An initial sample of 208respondents from 58 ®rms completed detailed questionnaires at the onset of therecent recession in the UK property market. The questionnaires were designedto elicit the respondents' perceptions of their own organization and variouscompetitors on a number of key dimensions. A sub-sample of 114 respondentsfrom 41 ®rms returned a further set of completed questionnaires, 12 to 18months later, when the recession had become deeply established. The ®ndingsare entirely consistent with the cognitive inertia hypothesis. The results indicatethat the respondents' individual and collective cognitions remained highly stable,despite a signi®cant down-turn in the property market from T1 to T2.

INTRODUCTION

Recently there has been an upsurge of interest in the cognitive analysis ofcompetitive positioning strategies in industries and markets (e.g. Calori et al.,1992, 1994; Dess and Davis, 1984; Gripsrud and Gronhaug, 1985; Hodgkinson

Journal of Management Studies 34:6 November 19970022-2380

# Blackwell Publishers Ltd 1997. Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UKand 350 Main Street, Malden, MA 02148, USA.

Address for reprints: Gerard P. Hodgkinson, Leeds University Business School, University of Leeds,11±15 Blenheim Terrace, University of Leeds, Leeds LS2 9JT, UK.

and Johnson, 1994; Porac and Thomas, 1994; Porac et al., 1987, 1989; Reger1990a; Reger and Hu�, 1993; Walton, 1986). Central to this stream of work isthe premise that actors, both individually and collectively, form simpli®ed repre-sentations, or `mental models', of their competitive worlds and that these de®nethe boundaries of the competitive arena and bases of competition. As Porac andThomas observe:

From a cognitive perspective, decision makers act on a mental model of theenvironment. Thus any explanation for strategic responses to competitivepressures must ultimately take into consideration the mental models of compe-titive strategists . . .

. . . before competitive strategies can be formulated, decision makers must havean image of who their rivals are and on what dimensions they will compete.Given the diverse range of organizational forms and decision makers' limitedcapacity to process complex inter-organizational cues, the task of de®ning `thecompetition' is both important and problematic. (Porac and Thomas, 1990,pp. 224±5)

Drawing on this and related work, a number of theorists have advanced what wemight loosely term `social constructionist' accounts of competitive rivalry inindustries and markets (e.g. Bogner and Thomas, 1993; Easton et al., 1993;Levenhagen et al., 1993; Porac et al., 1989, Porac et al., 1995; Whitley, 1992).According to this emerging body of theory, over time, strategists from rival ®rmsdevelop highly similar (or `shared') mental models of the competitive arena dueto the fact that they share similar technical and material problems and frequentlyexchange information during the conduct of business transactions. This processof social exchange in turn leads to the development of a shared understanding ±throughout the community of ®rms within the marketplace ± of how to compete.In the words of Hu� (1982) and Spender (1989), `industry recipes' develop,informing competitors on what bases they are to compete with one another.This emerging body of theory has important implications for understanding

the dynamic processes underpinning organizational adaptation in turbulentindustries and markets. Clearly a key danger as material transactions stabilize tothe extent that `market domains' (Levenhagen et al., 1993) become rei®ed andindustry recipes begin to appear, is that actors may become overly dependent onthe shared mental model that has come to prevail, to the extent that theirperceptions of what it takes to compete e�ectively become out of step with thechanging material conditions of the marketplace (Baden-Fuller and Stopford,1992; Porac and Thomas, 1990; Reger and Palmer, 1996; Senge, 1990). Incircumstances where environmental contingencies shift and new forms of compe-titive strategy emerge to challenge an organization's once protected position,concomitant changes in the way in which strategists view competition wouldappear to be a fundamental prerequisite for successful adaptation. However, asPorac and Thomas (1990) observe, the literature abounds with anecdotalevidence which suggests that all too often strategists are unable to re-conceptua-lize the market identity of their businesses in this way, due to `cognitive inertia',with drastic consequences for the ®rms concerned. It is an empirical analysis ofthis phenomenon that forms the focus of the present paper.

GERARD P. HODGKINSON922

# Blackwell Publishers Ltd 1997

The notion of cognitive inertia implies that, to a certain extent, ®rms experien-cing a down-turn in their business may actually perpetuate this state of a�airsdue to the inability of strategists to revise their mental models of competitivespace su�ciently quickly to adapt successfully to the changing environment. Inorder to test this hypothesis there is clearly a need for longitudinal studies whichmonitor actors' mental models of competitive space, market conditions andstrategic behaviour and organizational performance over time. Only then can weascertain what role, if any, these mental models actually play in facilitating and/or inhibiting strategic change in dynamic environments.Unfortunately, such studies have not been forthcoming. While a number of

scholars (e.g. Barr et al., 1992; Bartunek, 1984; Fahey and Narayanan, 1989;Fletcher and Hu�, 1990; Hu� and Schwenck, 1990; Laughlin, 1991; Narayananand Fahey, 1990) have explored processes of cognitive change in organizationsmore generally, in a variety of domains related to the subject matter of thispaper, only two studies can be found speci®cally concerned with the analysis ofcompetitive positioning strategy from a cognitive viewpoint, which haveemployed some form of longitudinal design.In a study of the forest products industry, Gronhaug and Falkenberg (1989)

compared senior managers' perceptions of their own and competitors' strategiesduring periods of growth and recession (`boom and bust'). However, this studyfocused on just seven respondents from four organizations and employed a retro-spective design, i.e. the informants were required to report their cognitions forthe periods of interest on the basis of recall. The organizations, selected from theFortune 500 list of the largest US industrial ®rms, were pursuing strategies whichemphasized wood products over pulp and paper. Multiple sources of data wereused, including semi-structured interviews with top management, questionnaires,company reports and articles from relevant business periodicals. Competitivepositioning strategy was operationalized in terms of Miles and Snow's (1978)four-fold typology.Following Miles and Snow's (1978) original procedure, the participants were

each presented with a basic description of the four strategic types. They wererequested to classify their own ®rms and the three competitors with respect totype, retrospectively, over the two time periods covered by the study. Each ofthe four ®rms were also classi®ed by the researchers for both time periods inorder to provide an additional basis of comparison. According to Gronhaug andFalkenberg, great discrepancies were observed in relation to the self-evaluationsand competitors' evaluations of the ®rms' strategies, with no changes in strategiesreported by the ®rms themselves, despite the fact that such changes wereobserved by their competitors. Although this study is based on an extremelylimited sample, it clearly illustrates the potential value of longitudinal research inthis ®eld of inquiry and points overwhelmingly to the need for further studiesusing prospective designs with a greater sophistication of analytical methods andlarger sample sizes.More recently, Reger and Palmer (1996) have attempted to compare longitud-

inally, over a nine-year period, executives' cognitive maps of competitive posi-tioning in the US ®nancial services industry. The ®ndings of this researchprovide broad support for the validity of the cognitive inertia hypothesis. Consid-erable upheavals were experienced by the American ®nancial services industry


COGNITIVE INERTIA IN A TURBULENT MARKET 923

throughout the period encompassed by this research, following extensive deregu-lation and other changes. However, according to Reger and Palmer, due toinertia, strategists' mental maps failed to keep pace with this changing environ-ment.[1]

As with the Gronhaug and Falkenberg study, however, there are several meth-odological weaknesses associated with the design of this research that render the®ndings inconclusive. In order to explore the extent to which mental modelsaltered over the nine-year period encompassed by this research, Reger andPalmer compared retrospectively the ®ndings from three separate studies by meansof a qualitative content analysis.While the data were gathered over di�ering time periods (1981, 1986 and

1989), as Reger and Palmer themselves acknowledge, there are several locationaland industry segment disparities across the three studies. These potentiallyconfound a number of outcomes which Reger and Palmer would otherwise havebeen able to attribute with much greater certainty to increasing environmentalturbulence. Furthermore, there are several di�erences in the data analysismethods employed within each study which might also account for some of theoutcomes.As in the case of the Gronhaug and Falkenberg study, these ®ndings highlight

a pressing need for further studies using prospective designs, with a greater sophisti-cation of analytical methods and larger samples ± both in terms of the numberof individual research participants and the number of participating organizations.This paper reports one such investigation.

METHOD

Industry ContextThe present research objectives require an industry that exhibits considerableenvironmental volatility, i.e. an industry that has a marked tendency to su�ersigni®cant environmental jolts from time to time. Should it transpire ± in thistype of industry ± that mental models are found to remain highly stable over asigni®cant period of time, despite noteworthy changes in market conditions, thiswould be very strong supporting evidence for the cognitive inertia hypothesis.One industry in particular which appears to meet this criterion, and would

seem, therefore, to be highly appropriate for pursuing the present researchagenda, is the UK residential estate agency, or real estate, industry. Very fewindustries can have experienced the degree of turbulence experienced by the UKresidential estate agency industry in recent years. Within the past decade thisindustry has changed from being one in which large numbers of small ®rms ±the vast majority being single o�ce concerns ± dominate local markets, to one inwhich an increasingly smaller number of larger ®nancial services institutions havecome to dominate the national scene. According to Dietrich and Holmes (1990)the major reason for these changes has been the realization on the part of thelarge ®nancial institutions (i.e. banks, building societies and insurance companies)of the increased opportunities a�orded for marketing their ®nancial services byentry into the estate agency sector, in a trend towards one-stop house purchasingpackages.



Given that the market for estate agency services has traditionally been a localone, in order to assess the true impact of these structural changes on competitionDietrich and Holmes have argued that it is necessary to observe changes in theextent of market concentration, following the entrance of the national chains,within localized geographical areas. Empirical research by Dietrich and Holmes(1990, 1991) into the dynamics of the market structure of the estate agencyindustry in the Tyneside region of north eastern England illustrates the way inwhich the changes that have taken place nationally have had a profoundin¯uence on local competition. According to Dietrich and Holmes, in the early1980s there were some 79 ®rms active in the Tyneside region, with the fourlargest ®rms accounting for 15.7 per cent of the total industry. However,following the entrance into the local market of Black Horse, GA, and theHalifax and Prudential, together with two additional relatively large local ®rms,the Northern Rock Building Society and Parkes Estates, the percentage of thetotal industry market share accounted for by the four largest ®rms increased bya staggering 68 per cent. These six companies, which did not exist within thelocality of this study in 1987, accounted for 43 per cent of the total industry sizejust one year later. More precisely, the four ®rm concentration ratio whichmeasures the percentage of an industry's size accounted for by the largest four®rms in that industry rose by 86 per cent from 15.7 per cent to 36.1 per cent.The rapid and high pro®le entrance of the major national chains into estate

agency has had a highly signi®cant impact on the way in which businessescompete within this industry. One of the major consequences of the changeswhich have taken place nationally is that the markets for estate agency servicesare becoming increasingly concentrated at the local level. As we shall see later,these structural changes have clearly had a major and lasting impact on the wayin which actors construe competitive space within this sector.[2]

The performance of residential estate agency businesses is inextricably linkedto the state of the domestic housing market, a market which has proven to behighly volatile over recent decades. However, since the summer of 1988, whenthe UK housing market experienced its last peak in activity (following thedecision of the government to abolish multiple mortgage income tax relief), themarket has experienced a dramatic recession, to the extent that in many regions,most notably London and the south east of England, the market has all butcollapsed.The decision to abandon multiple mortgage tax relief, with e�ect from August

1988, served to fuel dramatic increases in house prices as potential purchasersrushed to beat the deadline. The e�ect was that many home owners, especiallyin London and the south-east of England, undertook ®nancial commitmentswhich rendered them extremely vulnerable to the e�ects of subsequent increasesin interest rates.[3]

House prices began to fall rapidly in the wake of rising interest rates during1990, followed by rising unemployment in 1991. As sales fell to barely half oftheir normal annual total, estate agency ®rms began to contract drastically, tothe extent that more than 20 per cent of the 20 leading estate agents' o�cesclosed (Key Note, 1992). One leading chain, the Prudential, has withdrawn fromthe sector altogether ± with trading and purchase losses estimated to besomewhere in the region of £300 million.



In summary, within the past decade this industry has witnessed several ratherdramatic environmental jolts, not least of which are the entrance into theindustry of the major banks, building societies and insurance companies in themid-1980s, and the rapid rise in house prices in the summer of 1988, followingwhich the housing market has severely stagnated and, in some areas of thecountry, declined. Consequently, a number of estate agency ®rms havecontracted on a major scale or, alternatively, withdrawn from the sector alto-gether. In short, this industry has experienced boom, stagnation and change overa relatively short time period. These features make it an ideal context in whichto explore the validity of the cognitive inertia hypothesis.

Research DesignAs noted earlier, the vast majority of previous cognitive studies of competitivepositioning strategy have tended to employ cross-sectional designs in which theresearch participants' perceptions of competitive space are assessed at a singlepoint in time. As we have seen, in the small number of cases where longitudinalanalyses have been attempted, the studies concerned have employed retrospectiveresearch designs, thereby limiting our ability to draw valid inferences from thedata. While the reluctance on the part of previous researchers to engage inprospective longitudinal ®eldwork is understandable, this type of research isessential if our understanding of the role of mental models in strategy-makingprocesses is to advance signi®cantly beyond present levels. The present study,therefore, utilized a two-wave panel design. The ®eldwork at T1 was carried outbetween mid-July 1989 and October 1989, while the ®eldwork at T2 ran over acomparable period from September 1990 to mid-December 1990.

SampleThe sample was drawn from the north-east Midlands region of the UK,spanning the area encompassed by West Yorkshire, South Yorkshire, the EastMidlands and Humberside. The participating organizations were recruited byworking systematically through all entries under the heading of `Estate Agents' inthe relevant volumes of the Yellow Pages telephone directory (1989 edition)covering this geographical area.The research participants were recruited by means of a telephone call to the

most senior representative available within the participating organization (invari-ably this was a member of the senior management team). The ®nal sample atT1 comprised a total of 208 research participants from 58 participating organiza-tions.Unfortunately, as is common in longitudinal research, many respondents

who participated in the study at T1, declined to continue their involvement inthe study (or could not be traced) at T2. A total of 114 participants from 41organizations took part in the second phase of the study, a sample attritionrate of 45.19 per cent for individuals and 29.31 per cent for organizations,respectively.It is not possible to identify the reasons for non-response at T2, but the most

likely explanation is the rapid deterioration in market conditions experiencedduring the course of the study. Several attempts were made to try and follow upthose who failed to return completed questionnaires at T2, but many ®rms were



reluctant to discuss their reasons for dropping out of the study. Others, however,gave as a reason that large numbers of sta� had been made redundant in theintervening period. This suggests that the most likely explanation for the highattrition rate is that the non-returners were either no longer with their organiza-tions at T2, or were too busy in the wake of their ®rms' redundancyprogrammes to commit further time to the study.

Research Instruments and Data Collection ProcedureBefore proceeding, it is important to note that the ®ndings reported here formbut one part of a much wider investigation into the nature, role and impact ofcompetitor cognition in relation to strategic behaviour and organizationalperformance. The wider issues which this study addressed are multi-level innature, spanning the functional group, organizational and, as reported here,individual and industry, levels of analysis. The research objectives to be accom-plished in this study necessitated an approach to data collection that wassuitable for use with relatively large samples, both in terms of the number ofparticipating organizations and the number of individual research participants,and which would yield data in a form which could subsequently be analysedat a variety of levels. Consequently, the research method of choice was thequestionnaire survey.The primary advantage of questionnaire-based approaches is in terms of the

relative ease with which they can be administered to large numbers of researchparticipants in comparison to other methods of data collection. The sheervolume of supplementary data (beyond the scope of this paper) to be collected inthe present investigation, in addition to the respondents' perceptions of theircompetitive worlds, ruled out other, more labour intensive, methods if a su�-ciently large sample of participants was to be included in the study in order totest the various hypotheses with any degree of rigour.Furthermore, a common methodological limitation associated with many

previous cognitive studies of competitive positioning strategy, is that the variousknowledge elicitation techniques employed in order to gather the data necessitateextensive interactions between the investigator(s) and the research participants(for a review see Hodgkinson, 1997). During the course of these interactionsthere is ample opportunity for a range of factors associated with the dynamics ofthe interview (chie¯y the length of the interview, and the behaviour of the inter-viewer and interviewee) to in¯uence the outcomes of the study (c.f. Calori et al.,1994). In other words, the application of these techniques may lead researchparticipants to generate idiosyncratic responses which may be a function of thedemand characteristics of the data collection task, rather than a re¯ection ofunderlying substantive di�erences in cognition across di�ering units of analysis(individuals, functional sub-groups, organizations, etc.) and/or di�ering timeperiods. As noted by Hodgkinson (1997), the extent to which the respondent iswilling to generate further, more detailed responses (additional attributes, furthercategories/sub-categories of competitor, etc.) on a given occasion may be deter-mined by such factors as how much time he or she has available to continuewith the task, the extent to which they are motivated to elaborate their previousresponses, and the extent to which the interviewer presses for further elaboration,either overtly, or by means of relatively subtle non-verbal cues. A second reason



for adopting a structured questionnaire-based approach in this study, therefore,was to ensure a greater degree of rigour and control during the ®eldwork.

The competitor analysis questionnaire. Following the lead of Walton (1986), Reger(1987) and Thomas and Venkatraman (1988), a modi®ed repertory gridapproach to cognitive assessment was employed in the present study. Theprimary strength of the repertory grid technique lies in its inherent ¯exibilityfrom the point of view of both data collection and analysis.Competitor cognition was assessed by means of a questionnaire speci®cally

constructed for this study by the author. The competitor analysis questionnaire(CAQ) comprises a series of 21 bipolar attribute rating scales, or `constructs'(Kelly, 1955), which the research participants used in order to evaluate their ownbusiness organization and some 19 competitors. The competitors were elicited bymeans of a standard list of categories, or `elements' in Kelly's (1955) terms, thusensuring that the research task was meaningful to the research participants, yetcontrolled, systematic comparisons of actors' mental models could be made atvarious levels of analysis. In this way, the CAQ is intended to combine keystrengths associated with both ideographic and nomothetic approaches toknowledge elicitation, while minimizing their associated weaknesses (c.f. Danielset al., 1995; Daniels et al., 1993).[4]

The constructs incorporated in the CAQ were derived through a pilot studyin which eight volunteers from separate estate agency organizations wereasked to list the attributes which immediately came to mind when di�erentiat-ing various competitors. This yielded a total of 21 distinguishable attributes.These attributes were formed into a series of seven-point semantic di�erentialscales (Osgood et al., 1957) ± for example: size of branch network (verysmall/very large); quality of sta� (very poor/very good); marketing pro®le(very low/very high); scale of charges (very low-priced/very expensive);operating practices (very poor/very good); and range of services (very narrow/very extensive). (The complete list of attributes appears in table II, later in thepaper.)In order to elicit the elements, the pilot study participants were also asked to

list the various types of estate agent which came to mind when thinking aboutcompetition within their industry. In addition, a search was made of advertise-ments placed by estate agency businesses in the Yellow Pages telephone directoryand various local property newspapers in order to ensure that no important cate-gories of estate agent had been omitted. This yielded a total of 13 types of estateagent, to which seven additional categories (the respondent's own business,together with categories denoting major, secondary and inferior competitors, andcategories denoting high, medium and low performing ®rms) were added by theauthor (see table I). These additional categories were incorporated in an e�ort togain insights into the factors which di�erentiate the successful and unsuccessfulplayers in the minds of the research participants, i.e. as a means of revealing thebases of `the industry recipe'.Further pilot interviews were conducted with ®ve fresh volunteers in order to

decide which, if any, of the elements and/or constructs could be omitted fromthe main study on the grounds of redundancy. However, there was unanimousagreement that the complete set of elements and constructs should be retained in



its entirety. Hence the 20 categories and 21 bipolar rating scales were assembledinto a questionnaire booklet.The research participants were instructed to draw up a list of 19 competitors

in response to the categories. These were then entered on a prepared strip ofpaper which could readily be inserted into each page of the booklet in turn.Each page of the CAQ is devoted to a single attribute. Hence respondentsevaluated their own organization and the 19 competitors on a particularattribute rating scale before proceeding to additional scales.

Market buoyancy scale. In order to assess the participants' awareness of the state ofthe property market which prevailed at the time of the study, market buoyancywas assessed by means of a single item six-point Likert scale devised by theauthor. Respondents were required to indicate the extent to which theyperceived the local property market to be buoyant at the time of data collection.Responses ranged from `very depressed' (low score) through to `very buoyant'(high score).As noted above, a two-wave panel design was employed in the present study

in order to facilitate the comparison of mental models of competitive space overtime. Both instruments described above were administered to the research parti-cipants on both occasions. In addition to these instruments, the participants alsocompleted a wide range of other measures, which are not discussed here sincetheir contents refer to issues beyond the scope of this paper.

Data AnalysisPrevious cognitive studies of competitive positioning strategy, typically usingcross-sectional research designs, have tended to focus on the industry (e.g. Poracet al., 1987, 1989) or the individual research participant (e.g. Daniels et al.,

Table I. Categories used to elicit lists of contrasting organizations

1. My business2. My major competitor3. A solicitor agent4. An estate agent owned by a building society5. A traditional estate agent6. An estate agent owned by an insurance company7. An estate agent o�ering a professional service8. An estate agent with a poor reputation9. An estate agent with chartered surveyor status10. An estate agent specializing in exclusive property11. An estate agent specializing in commercial and industrial property12. An estate agent specializing in residential property13. A secondary competitor14. An estate agent with a good reputation15. A diversi®ed estate agent16. An independent estate agent17. An inferior competitor18. A very successful estate agent19. A moderately successful estate agent20. An unsuccessful estate agent



1994; Hodgkinson and Johnson, 1994) as the focal unit of analysis. The presentresearch objectives, by contrast, necessitated an approach to cognitive assessmentthat would permit a multilevel comparison of actors' mental models of competi-tive space over time. However, as Hu� and Fletcher (1990) have observed,arguably one of the most challenging and complex set of issues facing researchersconcerned with the investigation of cognitive aspects of strategic management isrelated to the problem of how actors' mental models should be compared withone another, a problem which intensi®es with increased numbers of researchparticipants and levels of analysis. In the case of longitudinal studies, concernedwith the cognitive assessment of multiple actors, the time dimension gives rise toeven greater complexity. Fortunately, however, there is a collection of analyticalprocedures available which, on the basis of their prima facie appearance, wouldseem to be ideally suited for meeting the methodological challenges posed by theobjectives of the present study.

Three-way scaling for the longitudinal analysis of cognitive structures. The main feature ofthese techniques, known collectively as `three-way scaling' (Arabie et al., 1987),or equivalently, `weighted multidimensional scaling' (WMDS) (Schi�man et al.,1981), which makes them particularly attractive in the context of this investiga-tion, is that they are highly versatile, enabling the researcher to derive keyinsights into the nature of actors' shared cognitive structures and explore indivi-dual and/or subgroup di�erences in cognition, cross-sectionally and/or overtime. The analytical procedures adopted in previous cognitive studies of competi-tive assessment, such as basic two-way scaling techniques and principal compo-nents analysis (as employed by Walton (1986) and Reger (1990a, 1990b) in theirstudies of competitor cognition in the US ®nancial services industries, forexample), by contrast, are considerably less sophisticated. While simpler techni-ques such as principal components analysis and two-way scaling enable theresearcher to capture key features of particular individuals' mental models, or thecollective mental model of a given group of research participants as a whole,unlike their three-way counterparts, there is no means for exploring the subtleinterplay between individual and collective cognition.As with simpler two-way procedures, the aim of three-way scaling techniques is

to represent actors' perceptions of a collection of stimuli (in this case the variousestate agency categories) in the form of a spatial map comprising as few dimen-sions as possible, without unduly distorting the information contained in theinput dataset. In a three-way scaling analysis, the input dataset comprises a seriesof (stimuli 6 stimuli) proximity matrices (typically Euclidean distances). Thesematrices (one for each potential source of variation in cognition) capture theextent to which the stimuli are similar to one another. (The greater the distancebetween a given pair of stimuli, the more dissimilar they are to one another.Conversely, the smaller the distance, the greater their similarity.)The basic underlying assumption of these techniques is that a given group of

actors share a common set of underlying dimensions in their mental models of aparticular domain. However, at the individual (or subgroup) level, actors di�er interms of the extent to which a given dimension is of relevance in their `privatecognitions', including the possibility of zero relevance. Three-way scaling proce-dures yield an aggregate perceptual map, known as the group space, and a series



of weights re¯ecting di�erences in cognition at the individual or subgroup-level.These source weights (Arabie et al., 1987) vary between zero and unity as a re¯ec-tion of the extent to which a particular dimension is salient in the judgements ofa given individual or subgroup, or alternatively, the extent to which some parti-cular condition imposed by the researcher, acts as a source of variation incognition (for technical details see Arabie et al., 1987; Coxon, 1982; Kruskal andWish, 1978; Schi�man et al., 1981). The group space enables researchers tocapture aspects of collective cognition (for example at the industry level, or atthe organizational level), while the source weights capture individual or subgroupdi�erences. Using the source weights, it is possible to di�erentially `shrink' and/or `stretch' the axes of the group space in order to derive a `private space' foreach individual participant or subgroup. The greater the extent to which suchtransformations of the group space are necessary, in order to accommodate agiven source's input data, the less the group space is representative of thatsource's `private cognitions'. In essence, these weights inform researchers of theextent to which particular individuals, or subgroups, deviate from the collectivecognitive model in their `private views'.The extent to which the group space adequately represents the judgements of

the stimuli associated with particular sources is captured by two goodness of ®tindices (stress and RSQ). As a general rule, the lower the stress level, the greaterthe goodness of ®t. (A stress value = 0.00 would denote the fact that the derivedmodel captures perfectly all the information contained in the input dataset.) Thesum of the source weights associated with a given source varies in monotonicproportion to the percentage of variance accounted for (RSQ) in the inputdataset associated with that particular source by the group space. The greaterthe RSQ value, the greater the goodness of ®t. Poor average diagnostics (i.e.high stress, low RSQ) relating to the various proximity matrices suggest errorfuldata and/or the need for a group space of higher dimensionality in order tobetter represent the information contained in the input dataset as a whole, whilepoor diagnostics associated with particular input proximity matrices indicateerrorful data and/or the need for separate models of higher dimensionality inorder to represent adequately the judgements of the stimuli associated with theindividual sources in question.This ability to capture aggregate-level cognitions through the group space,

while exploring individual and/or subgroup variations in cognition using thesource weights, potentially renders three-way scaling an extremely insightfultechnique for addressing the empirical concerns of this paper. Using the groupspace con®guration derived from the initial dataset as a ®xed point of reference,to the extent that the source weights associated with the individual researchparticipants were found to remain stable from T1 to T2, this would provide verypowerful evidence indeed for the cognitive inertia hypothesis. Such a ®ndingwould indicate that the importance respondents placed on the dimensions asso-ciated with the industry-level mental model of competitive space, in their`private' assessment of competitors, had remained constant over time.The extent to which the group space con®guration, depicting the dimensional

structure of the collective mental model of the industry as a whole, remainsstable or changes over time can also be investigated by deriving an additionalcon®guration using the T2 dataset. This supplementary analysis provides an



additional test of the cognitive inertia hypothesis, i.e. to the extent that previoustheorizing is correct, we would expect to ®nd no signi®cant di�erences in termsof the dimensional structure of the group spaces associated with the T1 and T2datasets.

The three-way scaling analyses. In total, 206 of the 208 questionnaires from the 58organizations at T1 and all 114 questionnaires returned from the 41 organiza-tions at T2 were usable for the purposes of the present tests. The data wereanalysed by adapting the procedure outlined by Kruskal and Wish (1978,pp. 70±3) for the analysis of pro®le proximities (see also Forgas, 1981; Forgas etal., 1980). The bipolar attribute ratings of the various estate agency categories atT1 were converted into matrices of Euclidean distances (one matrix perattribute) by averaging over the individual respondents' judgements using aformula derived from the generalized distance score proposed by Osgood andSuci (1952), Cronbach and Gleser (1953) and Wish et al. (1976) for formingpro®le distances between stimuli, namely:

_________1 N

sjk(s) =H( ± ~ (Xij(s)7Xik(s))2)N i =1

where Xij(s) and Xik(s) are respondent i's ratings of the estate agency categories jand k on scale s, and N is the total number of research participants. In short, theresult of this formula is a dissimilarity matrix which represents all the partici-pants' judgements of the various stimuli on a given scale.Using this formula, a total of 21 pro®le proximities matrices were formed via

the proximities procedure on SPSS version 4 (SPSS, 1990) which in turn servedas input to the ALSCAL multidimensional scaling routine developed by Takaneet al. (1977) which is also available on SPSS. At this stage, the complete T1dataset was employed rather than restricting the analysis to the subsample whoreturned completed questionnaires on both occasions, in order to makemaximum use of the available data and hence increase the generalizability of the®ndings.The program was run for analyses from six down to two dimensions, using the

model = indscal, level = ordinal (untie), level = ordinal (tied) and level = intervaloptions. Similar con®gurations were obtained across each level of measurement.However, an analysis of the diagnostic measures (stress and RSQ) suggested thelevel = ordinal (untie) option to be the most appropriate. Taking into accountboth the diagnostic criteria (stress = 0.212; RSQ = 0.846) and substantive inter-pretability of the output, a two-dimensional solution was strongly indicated.Following this, the participants' judgements were aggregated over the scales,

using a procedure similar to that outlined above for deriving the attribute pro®leproximities matrices, in order to derive a series of pro®le proximity matricesrepresenting particular individuals' judgements across all the scales. In turn, thesematrices served as input to a subsequent analysis in which the two-dimensionalsolution considered optimum was held constant, in order to derive the individualsource weights. Similarly, a set of source weights was also calculated for each ofthe research participants at T2 by initially deriving an individual pro®le



proximity matrix (aggregated across the various attribute rating scales) for eachrespondent, then inputting these matrices into the ALSCAL program with theT1 group space held constant.As noted earlier, source weights, re¯ecting the saliences of the various group

space dimensions in the judgements of particular sources, vary between zero andunity in monotonic proportion to the percentage of variance in a particularsource's `private' cognitive structure explained by the group space. The closer agiven weight is to unity, the greater the salience of that particular dimension andvice versa.Unfortunately, it is not appropriate to apply ordinary linear statistical proce-

dures to these source weights, in their raw form, in the search for systematicvariation in cognition.[5] When logarithmically transformed, however, the ratiosof these weights (taking each pair of dimensions in turn) provide a convenientindex of di�erential cognition which re¯ects the extent to which particulardimensions are relatively salient in the judgements of the individual sources.These transformed ratios are amenable to conventional statistical analysis (for ajusti®cation of this procedure see Coxon and Jones, 1978, 1979; andMacCallum, 1977) and, therefore, provide a suitable basis for addressing themain substantive concern of the present paper. To the extent that previous theo-rizing is correct, we would expect to ®nd no signi®cant changes in the (logarith-mically transformed) ratios of these weights from T1 to T2.

RESULTS

The Industry-level Cognitive StructureAs explained in the previous section, the multidimensional scaling analysisindicated a two-dimensional solution to be the optimum, taking into accountboth the diagnostic information and substantive issues of interpretability. Thissolution, which accounted for 84.6 per cent of the variance in the input data,was considered su�cient to represent the participants' judgements both statisti-cally and conceptually. The addition of further dimensions did not increasesubstantially the variance accounted for (a third dimension added a mere 1.7 percent and as many as six dimensions a mere 5.5 per cent to the explainedvariance). Moreover, no convincing interpretation was obtained for additionaldimensions.Interpretation of the group space was accomplished by examining the source

weights calculated by the computer program for each of the 21 attributematrices (see table II). These weights (the `dimension weights'), ranging betweenzero and unity, are analogous to factor loadings in a conventional principalcomponents or factor analysis, in the sense that the greater the magnitude of agiven dimension weight, the greater the relevance of the associated attribute forconceptualizing that particular dimension.The results in table II strongly indicate that the ®rst dimension represents the

overall market dominance, or `power', of various types of estate agent, in termsof their sheer visibility in the marketplace, as re¯ected by the relatively highdimension weights for the following attributes: size of branch network (0.94),geographical coverage (0.93), amount of advertising (0.90), marketing pro®le



(0.94), market share (0.83), ®nancial resources (0.93), links with ®nancial servicescompanies (0.93), and strategic in¯uence/power (0.90). The second dimension,by contrast, seems to represent the general quality of service associated with eachtype of estate agent, as re¯ected by the relatively high dimension weights for thefollowing attributes: service to vendors (0.96), service to purchasers (0.94), qualityof sta� (0.96), degree of personal attention (0.94), and operating practices (0.96).It is noteworthy that several of the attributes appear to be of signi®cance for theinterpretation of both dimensions. Given the pattern of dimension weights, it isclear that sta� training, quality of advertising, pro®tability, location of businesspremises, range of services, scale of charges, degree of local knowledge, andtypical range of properties on sale, are of some importance in relation to bothmarket power and quality.The aggregate perceptions of the respondents with respect to each estate

agency category are plotted along the two dimensions of the group space in®gure 1. The results suggest that within this particular industry there is a generalconsensus that successful estate agents are estate agents who have establisheddominance in the marketplace and o�er a high quality of service to vendors andpurchasers alike. Such organizations tend to gravitate towards the upper right-hand quadrant of the ®gure. Unsuccessful estate agents, by contrast, tend to belocated in the bottom left-hand quadrant of the ®gure, re¯ecting the generallyheld view that they are considerably less powerful than their successful counter-parts and o�er a poorer quality of service.

Table II. Dimension weights for the ratings of the attributes aggregated across the full sample ofrespondents at T1 (n= 206)

Dimension

Attribute 1 2

Service to vendors 0.10 0.96Quality of sta� 0.12 0.96Service to purchasers 0.18 0.94Training of sta� 0.54 0.78Operating practices 0.13 0.96Quality of advertising 0.55 0.72Pro®tability 0.45 0.76Location of business premises 0.58 0.67Size of branch network 0.94 0.00Range of services 0.69 0.56Geographical coverage 0.93 0.01Scale of charges 0.64 0.50Degree of personal attention 0.12 0.94Market share 0.83 0.29Marketing pro®le 0.94 0.15Degree of local knowledge 0.37 0.78Strategic in¯uence/power 0.90 0.29Amount of advertising 0.90 0.22Financial resources 0.93 0.08Links with ®nancial services companies 0.93 0.10Typical range of properties on sale 0.75 0.39



Longitudinal Analyses of Cognitive Stability and ChangeAs noted earlier, the source weight ratios associated with a three-way scalinganalysis provide a convenient index of di�erential cognition which will enable usto explore empirically the extent to which the individual research participants'`private' mental models remained stable or changed over the time period encom-passed by this study. Following Coxon and Jones (1978, 1979) the ratios of eachpair of source weights associated with the individual research participants werelogarithmically transformed using the following formula:

Log (dim 2/dim 1)

where dim 1 and dim 2 are the raw source weights associated with the marketpower and quality dimensions respectively.These transformed ratios re¯ect the extent to which market power and quality

are relatively salient in the judgements of the various participants. The greaterthe magnitude of this ratio in the positive direction for a given individual, thegreater the salience of the quality dimension relative to the market power

Figure 1. The two-dimensional representation of the 20 estate agency categories for the full sampleof respondents at T1 (n = 206; stress = 0.212; RSQ = 0.846)



dimension. Conversely, the greater the magnitude of this ratio in the negativedirection for a given individual, the greater the salience of the market powerdimension relative to the quality dimension. To the extent that these logarithmi-cally transformed source weights (and the associated group space con®gurationsfrom which they have been derived) are found to remain stable over time,despite noteworthy changes in market conditions, supporting evidence for one ofthe most central, though previously unsubstantiated, claims of recent theoristswill have been uncovered, i.e. the assertion that mental models of competitivespace play a key role in actually perpetuating the conditions of market decline,through cognitive inertia.We turn now to consider the evidence relating to this issue. We shall begin by

considering the extent to which responses to the market buoyancy scaleremained stable or changed across the two time periods. This will enable us togain insights into the local market conditions which prevailed when the researchparticipants' mental models were assessed. Having considered these ®ndings, wewill then go on to compare over time the logarithmically transformed sourceweight ratios and group space con®gurations associated with the three-way multi-dimensional scaling analyses.

Longitudinal comparison of market conditions. In order to assess the extent to whichlocal property market conditions remained stable or changed during the periodof the study, a matched-pairs t-test was computed on the responses to the marketbuoyancy scale. In addition, the test±retest correlation coe�cient was also calcu-lated. The former procedure informs us of the extent to which the mean levelsof a variable have changed or remained stable from one time period to another,whereas the latter provides an indication of the extent of stability and change inindividual scores over time. Thus, a situation in which a particular variable wascharacterized by a signi®cant t-value, but a low and non-signi®cant test±retestcorrelation, would imply that the overall direction of the change from T1 to T2which produced the signi®cant t-value was not typical of the pattern for thesample as a whole. Rather, this would indicate considerable diversity among indi-viduals with some scores increasing over time, others decreasing and still othersbeing largely unchanged.The means for responses to the six-point market buoyancy scale were 3.09 (SD

= 0.85) at T1, and 2.33 (SD = 0.78) at T2, respectively. These data indicate ashift from a general perception that the local property market was `moderatelydepressed' at T1 to a perception that the market was `depressed' at T2.The result of the matched pairs t-test was highly signi®cant (t = 8.43, df =

113, p 5 0.001), indicating that there has undoubtedly been a decline in themarket between T1 and T2. The test-retest correlation, though moderate in size,was also found to be highly signi®cant (r = 0.32, df = 112, p 5 0.001),providing a further indication that this trend is rather marked for the sample asa whole.These ®ndings bear out our earlier discussion concerning the recent history of

this industry. As we have seen, the period immediately leading up to the initialwave of data collection was characterized by a marked boom in the propertymarket across the country as a whole. Since that time, however, estate agentshave endured a major long-term recession which arguably has persisted even to



the present day. Responses to the market buoyancy scale at T1 indicate that thisrecession was just beginning to emerge during the initial period of data collec-tion. By the time of the second wave of data collection, however, the recessionwas well and truly developed, as evidenced by the highly signi®cant decrease inscores on this scale at T2.[6] Having established that the participants perceived asigni®cant decline in the state of the property market from T1 to T2, it nowremains to be seen to what extent the research participants' mental models ofcompetitive space changed or remained stable during this same period.

Longitudinal comparison of the ALSCAL source weight ratios. In order to establish theextent of stability and change in actors' `private' mental models of competitivespace, the logarithmically transformed ratios of the source weights associatedwith each of the individual research participants were compared over time. Asdiscussed earlier, these ratios (which were derived from the two-dimensionalgroup space reported in ®gure 1), represent the relative salience of eachdimension (i.e. market power and quality) for each individual. Recall that thegreater the magnitude of this variable in the direction of the positive pole, thegreater the salience of the quality dimension relative to the market powerdimension. Conversely, the greater the magnitude of this variable in thedirection of the negative pole, the greater the salience of the market powerdimension relative to the quality dimension.The mean values of the logarithmically transformed ratios of the source

weights (i.e. the geometric means of the ratios) were 70.18 (SD = 0.25) at T1,and 70.19 (SD = 0.26) at T2, respectively. These results indicate clearly thatthere has been no meaningful change whatsoever in the mean source weightratios over time. This ®nding was con®rmed by the results of a matched-pairst-test which yielded a non-signi®cant di�erence (t = 0.17, df = 113, p 4 0.05).Conversely, the test±retest correlation was found to be highly signi®cant (r =0.57, df = 112, p 5 0.001), suggesting considerable stability in the relativesaliences of the market power and quality dimensions, from T1 to T2, despite asizeable down-turn in the market.[7][8]

Longitudinal comparison of the group space con®gurations. The previous analysesdemonstrated that there are high levels of stability in competitor cognition inthis industry, as assessed by the logarithmically transformed source weightratios. The relative saliences of the market power and quality dimensions didnot di�er from T1 to T2 and the test±retest correlation was found to behighly signi®cant.Clearly to the extent that the group space from which these source weights

have been derived is stable from one time period to another, the results of theseanalyses are informative. Unfortunately, however, this particular method ofassessing cognitive stability does not take account of the possibility that meaning-ful changes may have occurred in terms of the dimensional structure of theindustry-level mental model from T1 to T2 (cf. Reger and Palmer, 1996). Suchchanges would imply that rival ®rms within the estate agency industry wereactively attempting to adjust the bases on which they compete with one anotherin response to the signi®cant down-turn in the property market which occurredduring this period. Clearly, this would call into question fundamentally the roleof industry-level mental models in market decline.



If the dimensional structure of the industry-level mental model was found tovary concomitantly with the changes in market conditions experienced in thisindustry, within the relatively short time frame surrounding this study, this wouldimply that competitor cognition is little more than a covariate. If, on the otherhand, the two-dimensional cognitive structure obtained at T1 was to be repli-cated at T2, this would constitute powerful evidence indeed for the assertion thatchanges in industry-level mental models of competitive space lag behind changesin market conditions. In other words, this ®nding would o�er additionalempirical support for the cognitive inertia argument advanced earlier ± i.e. thatindustry-level mental models of competitive space, and associated `recipes', mayperpetuate the conditions of industry decline, by causing organizations tocontinue competing on the basis of strategies that have been outmoded by thechanged demands of the marketplace.In order to explore the extent to which the basic two-dimensional industry-

level mental model observed at T1 was replicable over time, separate MDScon®gurations were derived for the returners using the competitor analysis ques-tionnaire data from both time periods, i.e. separate con®gurations were derivedfor T1 and T2. As in the previous run, analyses were performed from six downto two dimensions, in order to ascertain the optimum dimensionality for eachsubgroup in turn. However, in each case a two-dimensional solution was selectedon the basis of the goodness of ®t criteria, together with a consideration of theinterpretability of the output. As in the previous run, the general patterning ofthe dimension weights associated with these analyses seems to re¯ect marketpower and quality, suggesting that there are no substantive di�erences in termsof the dimensional structure of the group spaces and that the industry-levelmental model is highly stable over time. These results provide a further indica-tion that the highly signi®cant down-turn in the property market from T1 to T2has had no material impact on the way in which estate agents generally construetheir competitive worlds.[9][10]

DISCUSSION

In this paper the extent to which mental models of competitive space are stableor change over time has been explored empirically. Previous studies have largelyneglected to address this fundamentally important issue or, alternatively, haveemployed inadequate research designs. This has been due to an understandablereluctance on the part of scholars to engage in studies which necessitate datacollection at more than one point in time.However, the present ®ndings demonstrate powerfully the bene®ts to be gained

from such prospective, longitudinal studies, using relatively sophisticated methods ofdata collection and analysis. The research design and methods adopted in thepresent study have greatly strengthened the contribution of the ®ndings to ourunderstanding of the inhibitory role cognition plays as organizations seek toadapt to changing material conditions in the marketplace.As noted earlier, previous cognitive studies of competitive positioning strategy

have tended to be restricted to a single level of analysis such as the industry-levelor the level of the individual research participant. The present research objec-tives, by contrast, necessitated an approach to cognitive assessment that would



permit a multilevel exploration of actors' mental models of competitive spaceover time. The repertory grid-based questionnaire approach to knowledge elicita-tion adopted in this study, together with the use of a three-way scaling techniquefor the analysis of the data, has proven highly e�ective for this purpose.The results clearly indicated that despite a signi®cant down-turn in the housing

market from T1 to T2, neither the private nor the collective mental models ofcompetitive space held by the research participants changed to any meaningfulextent. On the contrary, minimal changes were observed in terms of the sourceweight ratios associated with the individual research participants and the group-space representations depicting the collective cognitions of the sample as awhole. These ®ndings provide a convincing demonstration of the deleterious roleof cognition in ailing industries and markets.Judging from the group space con®gurations derived in this study, it would

appear that the entrance of the major institutional estate agents is a highlyvisible stimulus, representing a competitive threat of substantial proportions,which in turn has greatly narrowed actors' attention and in so doing has cometo dominate the thinking of the larger and smaller ®rms alike. However, as wehave seen, while these larger players may well have been considered relativelysuccessful during the 4±5 year period leading up to the ®eldwork associated withthis study, these ®rms actually fared rather badly over both periods of datacollection and beyond ± to the extent that, as noted earlier, one of these ®rms,the Prudential (with over 800 branches during the peak of the 1988 boom) hassubsequently withdrawn from the industry altogether. When placed in thisproper historical context, the results emerging from the multidimensional scalinganalyses suggest that from the outset of this study the research participants wereacting on a mental model which was already out of step with the changingcircumstances confronting the industry.There are several limitations associated with this research which should be

acknowledged before concluding this discussion. This study has tracked cognitionwithin just one industry over a relatively short time period. Consequently, thestudy is limited in terms of the extent to which we can generalize the ®ndings toother contexts. While no evidence emerged from the investigation to suggestactors' mental models changed within the 12±18 month period encompassed bythe ®eldwork, this is not to say that such changes might not have occurred hadthe study been conducted over a longer time frame.Within the period encompassed by the present study, the mean level of the

participants' perceptions of the local property market shifted only from `moder-ately depressed' at T1, to `depressed' at T2. While highly signi®cant statistically,this change may have been of insu�cient magnitude to be of cognitive signi®-cance. It could be that mental models of competitive space are updated mostsigni®cantly when there is a fundamental qualitative change in market conditions(for example from `moderately buoyant' to `moderately depressed'), but not for aquantitatively equal shift with no qualitative change, as observed presently. Futurestudies should extend the range of industries investigated and vary the timeintervals between data collection periods, in order to increase the generalizabilityof the ®ndings and advance knowledge of the e�ects of cognitive inertia overlonger periods, under qualitatively di�ering market conditions.Nevertheless, the present results are consistent with Reger and Palmer's (1996)



®ndings (based on data gathered over a nine-year time frame) and previous theo-rizing in this area (Hu� et al., 1992; Porac and Thomas, 1990), adding furtherempirical substance to the claim that within volatile business environments,changes in mental models of competitive space signi®cantly lag behind changesin the material conditions of the marketplace. The implications for practice arethat actors should periodically engage in processes of individual and collectivere¯ection in order to reconsider anew the extent to which their assumptions andbeliefs about the external environment provide a viable basis upon which tobuild e�ective strategies for competitive success (see also Bowman and Johnson,1992; Hu�, 1990; Mitro�, 1988). Should actors fail to engage in such processes,to the extent that they become impervious to the fact that their organizations'competitive strategies have been outmoded by the changed demands of theenvironment, as Senge (1990) observes, this state of a�airs may ultimately giverise to the total decline of particular businesses or indeed entire industries.

NOTES

*I am grateful to the participants of the sub-theme seven working group, `Thinking inOrganizations', of the 12th Colloquium of the European Group for OrganizationalStudies, and three anonymous referees of the JMS, for their constructive comments andsuggestions on an earlier version of this paper.

[1] Reger and Palmer posit a continuum of information processing modes, beyond thescope of the present paper, drawing on Schneider and Shi�rin's (1977) distinctionbetween controlled and automatic information processing, in an attempt to developa more comprehensive account of the role of cognition in strategic decisionmaking.

[2] Unfortunately, however, the large-scale entrance of the ®nancial services institutionsinto the estate agency sector has also, on the face of it, turned out to be nothingshort of an unmitigated disaster, with annual losses running to seven and eight®gure numbers. During the ®rst half of 1989, for example, Prudential PropertyServices (the largest player at that time with 800 branches) reported losses of£24.7 million, while Royal Life, GA, Black Horse and Cornerstone reported lossesof £14.5 million, £9.0 million, £5.9 million and £9.5 million, respectively(Dietrich and Holmes, 1991).

[3] At a more technical level, as reported by the Bank of England in its QuarterlyBulletin for November 1991, household income gearing (gross interest payments asa proportion of disposable income) rose from 5 per cent to 13 per cent in thedecade from 1980 to 1990, while housing related capital gearing (mortgage debt asa percentage of the value of owner-occupied housing stock) increased from 15 percent to 25 per cent over the same period.

[4] It should be noted that the adaptation of the repertory grid technique in this way,i.e. so as to incorporate the use of standardized elements and constructs, is notwithout its critics. This approach represents a signi®cant departure from Kelly's(1955) personal construct theory, the underlying theory upon which the repertorygrid technique was originally based. (Personal construct theory asserts that indivi-duals construe the world di�erently from one another, i.e. they each have a set ofconstructs which vary from one individual to another.) Consequently, as originallyformulated, repertory grid was intended purely as an ideographic technique and, inits purest form, both the elements and constructs are elicited separately for eachindividual research participant. However, this frequently yields data which are



di�cult to compare meaningfully from one individual to another and for thisreason a number of researchers have developed alternative approaches to theconstruction and analysis of repertory grids (see, for example, Fransella andBannister, 1977; Reger, 1990b; Slater, 1976, 1977; Smith and Stewart, 1977).However, critics remaining loyal to the Kellyian tradition have argued that manyof these developments have, inappropriately, taken repertory grid away from itstheoretical origins and are therefore unwelcome. The use of grids where elementsand constructs are totally provided by the researcher is one such development.According to Phillips (1989, p. 196): `This kind of grid would be more usefullysupplanted by a semantic di�erential.' In the present case, however, a compromisestrategy was adopted whereby all of the constructs and the majority of theelements were elicited from a subsample of practising estate agents, rather thanresearcher imposed, in an attempt to ensure that the research task was within theparticipants' `range of convenience' (Kelly, 1955).

[5] The reason for this anomaly is that the source weights are more appropriatelydepicted as vectors (the lengths of which vary in monotonic proportion to thevariance accounted for in the private cognitions of a given source by the groupspace), as distinct from points in a multidimensional space. As such, it is the relativedirectionality of these vectors that is most useful for exploring di�erential cognition,rather than their absolute magnitude, or length (Coxon, 1982).

[6] While this analysis has indicated highly signi®cant di�erences in terms of the parti-cipants' responses to the market buoyancy scale across the two time periodscovered by the study, there is a strong possibility that the research participantswho dropped out of the study after the ®rst phase, di�ered signi®cantly in theirinitial perceptions of the state of the property market from those who subsequentlyreturned at T2. If this should turn out to be the case, the meaning of the resultsreported above would have to be re-considered. Such a ®nding would render theresults of the longitudinal analysis inconclusive. The apparently high levels ofchange observed would be potentially confounded due to bias resulting fromsample attrition.In order to explore the extent to which sample attrition poses a threat to the

validity of the ®ndings, the initial responses of those research participants whoreturned completed questionnaires on the second occasion were compared with theresponses of those participants who subsequently dropped out of the study, usingthe unrelated t-test. The means were 3.09 (SD = 0.85) and 3.16 (SD = 1.08) forthe returners and non-returners, respectively. Fortunately, the results of the t-testcon®rm that this negligible di�erence is not statistically signi®cant (t = 70.54,df = 204, p 4 0.05).

[7] Strictly speaking, hypothesis testing is inappropriate, due to the fact that the sourceweights associated with three-way scaling procedures lack independence. For anygiven analysis, the source weights, even following logarithmic transformation inratio form, are dependent on the characteristics of the group space from whichthey have been derived. Consequently, any statistical procedures applied to sourceweights, should be used for descriptive purposes only (Coxon, 1982).

[8] Once again there is a strong possibility that the research participants who droppedout of the study after the ®rst phase di�ered signi®cantly in their initial responsesfrom those who subsequently returned at T2, thus posing a threat to the validity ofthe substantive ®ndings. Fortunately, however, the unrelated t-test performed onthe mean logarithmically transformed source weight ratios associated with thereturners (mean = 70.18, SD = 0.25) and non-returners (mean = 70.16, SD =0.23) derived at T1, revealed a non-signi®cant di�erence (t = 70.68, df = 204,p 4 0.05).



[9] In addition, a third con®guration was derived for the non-returners (using the T1dataset), in order to compare the extent to which the group level mental modelassociated with this particular subgroup di�ered from the group level mental modelassociated with the returners assessed at T1. Any substantial di�erences betweenthese con®gurations would imply that the T2 dataset was unduly biased owing tosample attrition. Fortunately, however, this additional con®guration was found tobe markedly similar to that derived for the returners at T1 and that derived fromthe T1 dataset as a whole, suggesting, once again, that sample attrition has had aminimal impact on the ®ndings.

[10] Details of these supplementary analyses are available from the author on request.

REFERENCES

ARABIE, P., CARROLL, J. D. and DE SARBO, W. S. (1987). Three-Way Scaling and Clustering.Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-065. London: Sage.

BADEN-FULLER, C. and STOPFORD, J. M. (1992). Rejuvenating the Mature Business. London:Routledge.

BARR, P. S., STIMPERT, J. L. and HUFF, A. S. (1992). `Cognitive change, strategic action,and organizational renewal'. Strategic Management Journal, 13, 15±36.

BARTUNEK, J. M. (1984). `Changing interpretive schemes and organizational restructuring:the example of a religious order'. Administrative Science Quarterly, 29, 355±72.

BOGNER, W. C. and THOMAS, H. (1993). `The role of competitive groups in strategyformulation: a dynamic integration of two competing models'. Journal of ManagementStudies, 30, 51±67.

BOWMAN, C. and JOHNSON, G. (1992). `Surfacing competitive strategies'. European Manage-ment Journal, 10, 210±19.

CALORI, R., JOHNSON, G. and SARNIN, P. (1992). `French and British top managers'understanding of the structure and dynamics of their industries: a cognitive analysisand comparison'. British Journal of Management, 3, 61±78.

CALORI, R., JOHNSON, G. and SARNIN, P. (1994). `CEOs' cognitive maps and the scope ofthe organization'. Strategic Management Journal, 15, 437±57.

COXON, A. P. M. (1982). The User's Guide to Multidimensional Scaling. London: Heinemann.COXON, A. P. M. and JONES, C. L. (1978). Images of Occupational Prestige: A Study in Social

Cognition. London: Macmillan.COXON, A. P. M. and JONES, C. L. (1979). Measurement and Meanings: Techniques and Methods

of Studying Occupational Cognition. London: Macmillan.CRONBACH, L. J. and GLESER, G. C. (1953). `Assessing the similarity between pro®les'.

Psychological Bulletin, 50, 456±73.DANIELS, K., DE CHERNATONY, L. and JOHNSON, G. (1995). `Validating a method formapping managers' mental models of competitive industry structures'. Human Relations,48, 975±91.

DANIELS, K., JOHNSON, G. and DE CHERNATONY, L. (1994). `Di�erences in managerialcognitions of competition'. British Journal of Management, 5 (Special Issue), S21±S29.

DANIELS, K., MARKOCZY, L., GOLDBERG, J. and DE CHERNATONY, L. (1993). `Comparingcognitive maps'. Paper presented to the International Workshop on Managerial andOrganizational Cognition, Brussels, Belgium.

DESS, G. G. and DAVIS, P. S. (1984). `Porter's (1980) generic strategies as determinants ofstrategic group membership and organizational performance'. Academy of ManagementJournal, 27, 467±88.



DIETRICH, M. and HOLMES, P. (1990). `The market structure of the estate agents industryin the 1980s: an empirical investigation'. Applied Economics, 22, 629±38.

DIETRICH, M. and HOLMES, P. (1991). `Financial institutions and the estate agentsindustry in the 1980s'. The Services Industries Journal, 11, 481±90.

EASTON, G., BURRELL, G., ROTHSCHILD, R. and SHEARMAN, C. (1993). Managers and Compe-tition. Oxford: Blackwell.

FAHEY, L. and NARAYANAN, V. K. (1989). `Linking changes in revealed causal mapsand environmental change: an empirical study'. Journal of Management Studies, 26,361±78.

FLETCHER, K. E. and HUFF, A. S. (1990). `Strategic argument mapping: a study ofstrategy reformulation at AT&T'. In Hu�, A. S. (Ed.), Mapping Strategic Thought. Chiche-ster: John Wiley, 165±93.

FORGAS, J. P. (1981). `Social episodes and group milieu: a study in social cognition'.British Journal of Social Psychology, 20, 77±87.

FORGAS, J. P., BROWN, L. B. and MENYHART, J. (1980). `Dimensions of aggression: theperception of aggressive episodes'. British Journal of Social and Clinical Psychology, 19,215±27.

FRANSELLA, F. and BANNISTER, D. (1977). A Manual for Repertory Grid Technique. New York:Academic Press.

GRIPSRUD, G. and GRONHAUG, K. (1985). `Structure and strategy in grocery retailing: asociometric approach'. Journal of Industrial Economics, XXXIII, 339±47.

GRONHAUG, K. and FALKENBERG, J. S. (1989). Èxploring strategy perceptions in changingenvironments'. Journal of Management Studies, 26, 349±59.

HODGKINSON, G. P. (1997). `The cognitive analysis of competitive structures: A reviewand critique'. Human Relations, 50, 625±54.

HODGKINSON, G. P. and JOHNSON, G. (1994). Èxploring the mental models of competi-tive strategists: the case for a processual approach'. Journal of Management Studies, 31,525±51.

HUFF, A. S. (1982). Ìndustry in¯uences on strategy formulation'. Strategic ManagementJournal, 3, 119±30.

HUFF, A. S. (1990). Mapping Strategic Thought. Chichester: John Wiley.HUFF, A. S. and FLETCHER, K. E. (1990). `Conclusion: key mapping decisions'. In Hu�,A. S. (Ed.), Mapping Strategic Thought. Chichester: John Wiley, 403±12.

HUFF, A. S. and SCHWENCK, C. (1990). `Bias and sensemaking in good times and bad'. InHu�, A. S. (Ed.), Mapping Strategic Thought. Chichester: John Wiley, 89±108.

HUFF, J. O., HUFF, A. S. and THOMAS, H. (1992). `Strategic renewal and the interactionof cumulative stress and inertia'. Strategic Management Journal, 13, 55±75.

KELLY, G. A. (1955). The Psychology of Personal Constructs (2 volumes). New York: Norton.KEY NOTE (1992). Key Note Report: An Industry Sector Overview ± Estate Agents. London: KeyNote Publications.

KRUSKAL, J. B. and WISH, M. (1978). Multidimensional Scaling. Sage University Paper Serieson Quantitative Applications in the Social Sciences, 07-011. London: Sage.

LAUGHLIN, R. C. (1991). Ènvironmental disturbances and organizational transitions andtransformations: some alternative models'. Organization Studies, 12, 209±32.

LEVENHAGEN, M., PORAC, J. F. and THOMAS, H. (1993). Èmergent industry leadershipand the selling of technological visions: a social constructionist view'. In Hendry, J.and Johnson, G. with Newton, J. (Eds), Strategic Thinking: Leadership and the Management ofChange. Chichester: John Wiley.

MACCALLUM, R. C. (1977). È�ects of conditionality on INDSCAL and ALSCALweights'. Psychometrika, 42, 297±305.

MILES, R. E. and SNOW, C. C. (1978). Organizational Strategy, Structure and Process. NewYork: McGraw-Hill.



MITROFF, I. (1988). Break-Away Thinking: How to Challenge Your Business Assumptions (and WhyYou Should). Chichester: John Wiley.

NARAYANAN, V. K. and FAHEY, L. (1990). Èvolution of revealed causal maps duringdecline: a case study of Admiral'. In Hu�, A. S. (Ed.), Mapping Strategic Thought. Chiche-ster: John Wiley, 109±33.

OSGOOD, C. E. and SUCI, G. J. (1952). À measure of relation determined by both meandi�erences and pro®le information'. Psychological Bulletin, 49, 251±62.

OSGOOD, C. E., SUCI, G. J. and TANNENBAUM, P. H. (1957). The Measurement of Meaning.Urbana: University of Illinois Press.

PHILLIPS, E. M. (1989). Ùse and abuse of the repertory grid: a PCP approach'. ThePsychologist, 2, 194±8.

PORAC, J. F. and THOMAS, H. (1990). `Taxonomic mental models in competitor de®ni-tion'. Academy of Management Review, 15, 224±40.

PORAC, J. F. and THOMAS, H. (1994). `Cognitive categorization and subjective rivalryamong retailers in a small city'. Journal of Applied Psychology, 79, 54±66.

PORAC, J. F., THOMAS, H. and BADEN-FULLER, C. (1989). `Competitive groups as cognitivecommunities: the case of Scottish knitwear manufacturers'. Journal of Management Studies,26, 397±416.

PORAC, J. F., THOMAS, H. and EMME, B. (1987). `Knowing the competition: the mentalmodels of retailing strategists'. In Johnson, G. (Ed.), Business Strategy and Retailing.Chichester: John Wiley.

PORAC, J. F., THOMAS, H., WILSON, F., PATON, D. and KANFER, A. (1995). `Rivalry andthe industry model of Scottish knitwear producers'. Administrative Science Quarterly, 40,203±27.

REGER, R. K. (1987). `Competitive positioning in the Chicago banking market: mappingthe mind of the strategist'. Unpublished doctoral dissertation, University of Illinois atUrbana-Champaign, USA.

REGER, R. K. (1990a). `Managerial thought structures and competitive positioning'. InHu�, A. S. (Ed.), Mapping Strategic Thought. Chichester: John Wiley, 71±88.

REGER, R. K. (1990b). `The repertory grid for eliciting the content and structure ofcognitive constructive systems'. In Hu�, A. S. (Ed.), Mapping Strategic Thought. Chiche-ster: John Wiley, 301±9.

REGER, R. K. and HUFF, A. S. (1993). `Strategic groups: a cognitive perspective'. StrategicManagement Journal, 14, 103±24.

REGER, R. K. and PALMER, T. B. (1996). `Managerial categorization of competitors: usingold maps to navigate new environments'. Organization Science, 7, 22±39.

SCHIFFMAN, S. S., REYNOLDS, M. L. and YOUNG, F. W. (1981). Introduction to Multidimen-sional Scaling. New York: Academic Press.

SCHNEIDER, W. and SHIFFRIN, R. M. (1977). `Controlled and automatic human informa-tion processing: I. Detection, search, and attention'. Psychological Review, 84, 1±66.

SENGE, P. (1990). The Fifth Discipline: The Art and Practice of the Learning Organization. NewYork: Doubleday.

SLATER, P. (Ed.) (1976). The Measurement of Intrapersonal Space by Grid Technique: Vol. I ±Explorations of Intrapersonal Space. Chichester: John Wiley.

SLATER, P. (Ed.) (1977). The Measurement of Intrapersonal Space by Grid Technique: Vol. II ±Dimensions of Intrapersonal Space. Chichester: John Wiley.

SMITH, M. and STEWART, B. J. M. (1977). `Repertory grids: a ¯exible tool for establishingthe contents and structure of a manager's thoughts'. In Ashton, D. (Ed.), ManagementBibliographies and Reviews, 3, 209±29.

SPENDER, J. C. (1989). Industry Recipes: The Nature and Sources of Managerial Judgement. Oxford:Basil Blackwell.

SPSS (1990). SPSS Reference Guide Chicago: SPSS Inc.



TAKANE, Y., YOUNG, F. W. and DE LEEUW, J. (1977). `Nonmetric individual di�erencesmultidimensional scaling: an alternating least-squares method with optimal scalingfeatures'. Psychometrika, 42, 7±67.

THOMAS, H. and VENKATRAMAN, N. (1988). `Research on strategic groups: progress andprognosis'. Journal of Management Studies, 25, 537±55.

WALTON, E. J. (1986). `Managers' prototypes of ®nancial ®rms'. Journal of ManagementStudies, 23, 679±98.

WHITLEY, R. (1992). `The social construction of organizations and markets: the compara-tive analysis of business recipes'. In Reed, M. and Hughes, M. (Eds), Rethinking Organi-zation: New Directions in Organization Theory and Analysis. London: Sage, 120±43.

WISH, M., DEUTSCH, M., KAPLAN, S. S. (1976). `Perceived dimensions of interpersonalrelations'. Journal of Personality and Social Psychology, 33, 409±20.



Cognitive Inertia in a Turbulent Market: the Case of UK Residential Estate Agents

Documents

Transcript of Cognitive Inertia in a Turbulent Market: the Case of UK Residential Estate Agents