top management-team diversity and firm performance - examini

Top Management-Team Diversity and FirmPerformance: Examining the Role of Cognitions

Martin Kilduff • Reinhard Angelmar • Ajay MehraPennsylvania State University, 403 Beam, University Park, Pennsylvania 16802, [email protected]

INSEAD, Boulevard de Constance, 77305 Fontainebleau, France, [email protected] of Cincinnati, Cincinnati. Ohio, 45221, [email protected]

AbstractDemography research rarely examines the black box withinwhich the cognitive diversity of the top management team isassumed to affect firm performance. Using data from 35 sim-ulated firms run by a total of 159 managers attending executiveeducation programs, the current research tested several hypoth-eses concemed with (a) the relationship between demographicand cognitive team diversity and (b) the reciprocal effects ofdiversity and firm performance. Results showed that membersof high-performing teams tended to preserve multiple interpre-tations early in the team's life cycle, but that they moved towardgreater clarity near the end of the life cycle. These high-performing teams, therefore, exhibited both early interpretativeambiguity and late heedful interrelating. Cognitive diversity inteams affected and was affected by changes in firm perfor-mance. Surprisingly, there was no evidence of any effect ofdemographic diversity on measures of cognitive diversity.(Top Management Teams; Sensemaking; Cognitive Di-versity; Demographic Diversity; Organizational Perfor-mance; Ambiguity.)

At the fifth game of the National Basketball Association(NBA) championship series in Seattle, basketball legendJulius Erving remarked that the 1996 Chicago Bulls of-fered a glimpse of what the team of the future would looklike: The 1996 Chicago Bulls were one of the most na-tionally and ethnically diverse teams ever assembled inthe NBA. The Bulls' lineup consisted of the most skilledwhite and black players from three continents: Europe,Australia, and America.

The increasing diversity in the NBA is one instance ofa larger trend. The workforce throughout the developedworld is becoming more diverse, reflecting changingdemographics within nation-states (Johnston and Packer1987) and migrations of peoples across national borders(Hambrick et al. 1989, p. 33).

This growing demographic diversity may be accom-panied by a parallel increase in attitudinal or cognitivediversity, according to the literature on organizational de-mography. Within this literature, the assumption is com-monly made that demographic variation signals variationin underlying and invisible cognitive processes. From thisperspective, diversity may have important effects on teamand organizational performance.

Thus, the growing diversity of the workforce under-scores three questions of theoretical and practical impor-tance. First, how does demographic diversity affect cog-nitive diversity? Second, what are the effects of teamcognitive diversity on performance? Third, is there a re-ciprocal effect of firm performance on team cognitive di-versity? We explored these questions in a simulation oftop management-team decision making.

Literature ReviewThe victory of the Chicago Bulls in the NBA champi-onships notwithstanding, the relationship between teamdiversity and outcomes remains unclear, partly becausethere are numerous possible aspects of diversity that canbe investigated. Lawrence (1997, pp. 5-6) suggested thatdiversity can be studied across at least four different cate-gories of variables: visible demographic attributes (suchas gender): relational attributes (such as organizationaltenure); status attributes (such as marital status); and per-sonal attributes (such as personal beliefs and perceptions).In practice, research has tended to ignore the category ofpersonal beliefs and perceptions, a neglect this paper at-tempts to remedy.

Researchers have pointed to both the costs and benefitsof increased diversity in teams. Top management-teamdemographic diversity has been shown to predict turnoverrates (Jackson et al. 1991, Wiersema and Bird 1993); in-creased levels of work group diversity have been asso-ciated with lower psychological attachment to the orga-nization (Tsui et al. 1992); and less frequent

1047-7039/00/1101/0021/$05.001526-5455 electronic ISSN

ORGANIZATION SCIENCE, © 2000 INFORMSVol. 11, No. 1, January-February 2000, pp. 21-34

MARTIN KILDUFF, REINHARD ANGELMAR, AND AJAY MEHRA Top Management-Team Diversity

communication (Zenger and Lawrence 1989). But diver-sity has also been hailed as a competitive advantage be-cause minority views "can stimulate consideration ofnon-obvious altematives in task groups" (Cox and Blake1991, p. 50); and heterogenous teams have been shownto be more creative than homogenous teams (Hambricket al. 1996, Triandis et al. 1965).

The discrepancy in the research literature concemingthe effects of team diversity on organizational functioningrefiects two different approaches. The demographic ap-proach studies cognitive diversity in terms of proxy vari-ables such as gender, age, organizational tenure, ethnic-ity, and nationality (see Pfeffer 1983 for a review). Theemphasis is on directly measurable attributes of individ-uals. The cognitive approach studies cognitive diversitythrough direct questionnaire measures of attitudinal andnormative differences between individuals who may behomogenous on demographic indicators (e.g., Nemeth1986). Thus, cognitive diversity in this literature refers tovariability conceming relatively unobservable attributessuch as attitudes, values, and beliefs. Although both ap-proaches claim to be studying cognitive diversity, theymeasure very different attributes of individuals.

According to demography researchers, the reason togather demographic rather than cognitive data when ex-amining the consequences of diversity is that "mentalprocesses . . . are more difficult to access and reliablymeasure" (Pfeffer 1983, p. 351). Demographic variablesare argued to be more objective, to yield more parsimo-nious explanations of organizational phenomenon, and tobe more easily submitted to testing than cognitive vari-ables (Hambrick and Mason 1984;̂ Pfeffer 1983, p. 352;Wiersema and Bantel 1992, p. 94).

Although the demographic approach seems to offer ashortcut to studying the effects of diversity on organiza-tional outcomes, a careful reading of the demography lit-erature reveals a disjunction between theory and research.As noted by Lawrence (1997, p. 2), organizational de-mographers hypothesize "various subjective conceptsthat explain significant association between demographicpredictors and outcomes. . . . However, by invokingPfeffer's justification, researchers usually leave the con-cepts unmeasured and the hypotheses untested. As a re-sult, subjective concepts and their relationships within re-search models have become the 'black box' oforganizational demography." Demographic variables areoften treated as convenient proxies for cognitive ones inpredicting team outcomes.

The use of demographic variables as surrogates forcognitive ones is premised on the assumption that dem-ographic variables predict and explain variation in inter-vening cognitive variables (Hambrick and Mason 1984;

Pfeffer 1983, p. 350). Thus, for example, Michel andHambrick (1992) suggested that a common backgroundin terms of tenure in the top management team or func-tional specialization "contributes to the development ofcommon schemata among team members" (p. 18). Sim-ilarly, Tsui et al. (1992) argued that individuals use dem-ographic categories such as age and race to define psy-chological groups that reinforce self-identity; andWiersema and Bantel (1992, p. 112) argued that demo-graphic categories reflect individuals' cognitive bases.Team demographic diversity, measured over such visiblysalient variables as age, gender, and race, has been as-sumed to be an accurate refiection of how much the teamshares a common set of attitudes, values, and norms.

Researchers have tended to avoid specifying preciserelationships between diversity on particular demo-graphic variables and diversity on particular cognitivevariables in favor of general assertions that the more de-mographically diverse the team, the more members willtend to see the world differently, with this increased cog-nitive diversity affecting individual and firm performance(see the discussion in Lawrence 1997, p. 4). Researchresults on outcome variables have thrown this assumedgeneral correspondence between demographic and cog-nitive diversity into some doubt, however. For example,one study found that team heterogeneity had opposite ef-fects on performance depending on the measure of het-erogeneity selected: Whereas heterogeneity of experiencenegatively affected perfonnance, heterogeneity in yearsof education positively affected performance (Smith et al.1994). Another study reported that top management teamtenure heterogeneity had results directly contrary to thosehypothesized (Michel and Hambrick 1992). Similarly,Tsui et al. (1992) found, contrary to their predictions, thatthe larger the difference between the individual and thegroup on measures of organizational tenure and educa-tion, the higher the individual's intent to stay with thefirm. Wiersema and Bantel (1992, p. 14) reported that:"Very little support for the heterogeneity argumentemerged. . . . heterogeneity on age, organizational tenure,and team tenure were not significantly associated withstrategic change."

That these outcomes tumed out to be different frompredictions suggests that the link between demographicand cognitive diversity may be more complex than gen-erally assumed by demographers. As Smith et al. (1994,p. 432) summarized the results of their research: "Therelationships between team demography, team process,and organizational performance are not as straightforwardor as simple as scholars have previously believed," Ho-mogeneity along demographic markers does not neces-sarily engender homogeneity in attitudes, beliefs, or val-ues. Thus, one of the puzzling results in the

22 ORGANIZATION SCIENCEA'ol. 11, No. 1, January-February 2000

MARTIN KELDUFE, REINHARD ANGELMAR, AND AJAY MEHRA Top Management-Team Diversity

decision-making literature has been the wide array of in-dividual cognitive styles characteristic of people homog-enous on demographic indicators. Eor example, a studyof third-year elementary education majors (all femalesunder the age of 30) from the same university revealed astrikingly wide range of attitudes toward potential job in-terviews. The authors suggested that idiosyncratic pref-erences and personality traits could produce such indi-vidual cognitive differences (Rynes and Lawler 1983).Other research has found little support for the effects oftop management-team demographics on firm perfor-mance, but some effects of social process variables suchas communication frequency (Smith et al, 1994) may berelated to cognitive variables.

Theory and HypothesesIn the current research we examined the links betweendemographic diversity and cognitive diversity, and alsoexamined how these different kinds of diversity influ-enced team performance. We looked at three aspects ofdemographic diversity: functional specialization (e.g.,marketing, research and development, etc), national or-igin (i.e., Erench, German, etc), and age. Previous de-mography research has suggested that teams with highvariability on these kinds of markers may have differentschemata or ways of seeing the world (e.g., Michel andHambrick 1992). Eor example, executives from differentfunctional areas may view organizational problems fromdifferent perspectives (Dearborn and Simon 1958,Walker 1985, Waller et al. 1995: but see Walsh 1988 fora dissenting view), Eurthermore, managers from differentcountries tend to have different frameworks for approach-ing a wide range of issues (Hofstede 1980: Laurent 1983,1986). Similarly, research suggests that people of similarages tend to view the world in general in similar terms,reflecting shared experiences and socialization (cf. Tsuiet al. 1992, Wagner et al. 1984), with older cohorts tend-ing toward increased intracohort heterogeneity on a widevariety of characteristics (Dannefer 1987). Thus, basedon prior research and theory, we decided to test the hy-pothesis that demographic diversity on important indi-cators such as functional affiliation, national origin, andage would predict diversity over a wide range of attitudes,beliefs, and values relevant to organizational decisionmaking. We recognize that much of the prior research wehave summarized casts doubt on the predicted relation-ship between demographic variation and cognitive vari-ation. Nevertheless, we thought it important to at leastsubject this often unstated hypothesis to a formal test. Wealso recognize that there are other important demographicvariables, such as tenure, that are not studied in the cur-rent research.

HYPOTHESIS 1. The higher the demographic diversityof the top management team, the higher the cognitive di-versity exhibited in the team's decision-making pro-cesses.

Despite their differences over how best to examine theimpact of diversity on organizational outcomes, both theproponents of the demographic approach and those of thecognitive approach agree that cognitive diversity is re-lated to organizational outcomes. There is less agreement,however, on what the precise form of this relationshipmay be. We draw on a sensemaking perspective (Weick1979, 1995) to argue that cognitive diversity can be eithera blessing or a curse, depending upon the specific type ofcognition involved,

Eirst, from a sensemaking perspective (Weick 1979,1995), the preservation of multiple interpretations inteams is critical for registering complex environments.Weick reasons that for the detailed registering necessaryfor successfully coping with a complex, equivocal envi-ronment, the vatiety within the organization must matchthe variety outside it (1979, p, 190: cf, Ashby 1952). In-terpretive ambiguity within the top management teampreserves the requisite variety needed to sense and reg-ulate the variety facing the organization. Interpretativeambiguity is defined as follows: Lack of clarity within theteam conceming the degree to which team members sharecommon attributions conceming, for example, organiza-tional success and failure (cf, Martin 1992, pp, 133-134),Defined this way, interpretative ambiguity is a team-levelconstruct.

It is important to note here that the interpretative am-biguity valued by Weick should be distinguished fromdisorganization on the one hand (i.e., teams in which ev-eryone explicitly disagrees with everyone else) and un-animity or groupthink on the other (i,e,, teams in whicheveryone agrees with everyone else). Interpretative am-biguity resembles most closely that state of equivocalityin which both agreement and disagreement concemingthe environment are simultaneously possible, allowingthe same reality to be perceived by team members in dif-ferent but complementary ways. As Weick (1995, p. 120)points out, in organizations characterized by strategic am-biguity (a concept originated by Eisenberg 1984), "peopleare not pressed to articulate their individual understand-ing" of causal connections. Thus, people act effectivelytogether without the team as a whole ever clarifying howmuch interpretative ambiguity actually exists. Previousresearch from a sensemaking perspective has shown thatteams can act effectively despite an absence of sharedmeanings (Donnellon et al. 1986) or shared goals(Bourgeois 1980). Erom this perspective, efforts to clarify

ORGANIZATION SCIENCEA^ol, 11, No. 1, January-February 2000 23

MARTIN KBLDUFF, REINHARD ANGELMAR, AND AJAY MEHRA Top Management-Team Diversity

team disagreements may damage performance. As a ma-jor study of string quartets concluded: "[T]he more suc-cessful quartets . . . managed . . . inherent contradictionsimplicitly and did not try to resolve them" (MumighanandConlon 1991, p. 165).

HYPOTHESIS 2. The higher the interpretative ambi-guity within the top management team, the higher thefirm's subsequent performance.

A completely separate issue concems not how the teammembers interpret the causal connections in the extemalenvironment, but the degree to which team members per-ceive the same intemal environment. In some teams,members will tend to have differing perceptions of howsuch basic processes as decision making and task struc-turing are accomplished. An absence of agreement con-ceming basic team processes is evidence of lack of teamcohesion. This lack of cohesion can significantly impairperformance according to long-standing theory (Stodgill1959) and recent research (Smith et al. 1994).

Controlled conflict can, of course, be built into teamprocesses and can aid decision making by expanding per-spectives, but only to the extent that conflict remainsaffect-free and focused on how best to achieve objectives(Amason 1996). Executives in for-piofit firms report find-ing conflict unpleasant and leading to low-quality deci-sions (Schwenk 1990), suggesting that affect-free cogni-tive conflict in top management teams is rare. Conflictover decision-making processes and task structuring, likeother conflicts, can consume resources, such as time(March and Simon 1958, pp. 149-151), and thus ad-versely affect performance. From a sensemaking per-spective, in fast moving environments, speed of decisionmaking is critical in allowing the top management teamto influence events as they are developing (Eisenhardt1989). On balance, therefore, the research suggests a gen-erally negative effect of disagreement over basic teamprocesses on organizational performance.

HYPOTHESIS 3. The higher the top management teamdisagreement concerning basic team processes, the lowerthe firm performance.

In focusing on the effects of cognitive diversity on per-formance, we should not lose sight of the possibility of adynamic relationship between performance and cogni-tion. Erom an enactment perspective, feedback from theenvironment is expected to trigger changes in cognitionwhich in tum affect actions that help determine subse-quent performance (Weick 1979). The emphasis is on re-ciprocal relationships between cognitive change and per-formance change. Erom an enactment perspective, topmanagement teams help create environmental outcomes

(such as successful or unsuccessful product launches) thathave consequences for team cognition and functioning.Members of top management teams that see their fiimssucceeding in the competitive marketplace may becomeincreasingly homogenous in their adherence to idiosyn-cratic traditions, according to research on the develop-ment of strong-culture organizations (Kotter and Heskett1992). Eailing performance, on the other hand, may createa sense of crisis that results in a wider range of percep-tions and beliefs among top management (Kotter andHeskett 1992).

HYPOTHESIS 4. Increases in firm performance willdecrease top management team cognitive diversity.

Previous research has confirmed that team cognitivediversity does change over time, with team members' per-ceptions tending to coalesce around readily identifiablenorms (Jacobs and Campbell 1961, Weick and Gilfillan1971). How might this changing cognitive diversity affectperformance? The existing research concems the decayof arbitrary norms in laboratory experiments. We wereunable to find any discussion explicitly focused on theeffects of changing team cognition on changes in perfor-mance. Consequently, we decided to conduct exploratoryanalyses on this important topic.

Research SettingWe chose a setting in which it was possible to study thedemographic and cognitive diversity of top managementteams and to relate measures of diversity to performanceoutcomes: a business simulation involving teams of ex-perienced managers. Organizational simulation hasemerged as a valuable tool for studying complex orga-nizational processes (e.g., Cohen et al. 1972, Masuch andLaPotin 1989). The MARKSTRAT simulation (Larrecheand Gatignon 1977), in particular, is widely used forstudying decision making (e.g., Hogarth and Makridakis1981, Lant and Montgomery 1987, Walsh et al. 1988).Managers from different industries have reported thatMARKSTRAT reflects the competitive environment theyare familiar with, and that they take real actions in theirown organizations based on their experiences in MARK-STRAT (Kinnear and Klammer 1987). The realism ofthesimulation has been widely recognized (e.g., Dodgson1987, Remus 1978), and has been attributed to the game'sability to capture long-term market mechanisms(Gatignon 1987, p. 469). Eurthermore, interest by man-agers and other participants in the MARKSTRAT simu-lation is typically high (Hogarth and Makridakis 1981).

What is MARKSTRAT? Briefly, it is a game in whichteams of players compete with each other in a consumer



entertainment market. There are five teams, each teamrepresenting the overall strategic marketing orientation ofa company. Each team, in other words, represents the topmanagement of a company, and is responsible not onlyfor strategic marketing, but also for coordinating produc-tion, inventory control, and research and development.The top management team plans and implements the mar-keting strategy of the company in the context of compe-tition from four other companies in the same industry.Each player on each team typically has a MARKSTRATmanual (Larreche and Gatignon 1977) that describes thecharacteristics of the MARKSTRAT world, the ways itiwhich to design and implement strategy in this world, andthe operating instructions specific to this simulated en-vironment.

At the beginning of the simulation, each team is re-sponsible for marketing two brands, but part of the strat-egy involves modifying, withdrawing, and introducingbrands. For each decision period (equivalent to one year),teams must make decisions conceming the number andtypes of brands, how much research and development tofund, how much production to plan for, the advertisingbudget, recommended retail prices, number of salespeo-ple, and so on. These decisions are entered into a com-puter which simulates the competitive environment andproduces updated performance reports for each team,showing units sold for each branded good, gross market-ing contribution, market share, and other performance in-dicators.

In the current research, after a practice run designed toincrease familiarity with the simulation, each team re-ceived a computer printout detailing the initial positionof the organization. In the MARKSTRAT simulation,teams take over companies that are already established.The simulation was then run for six decision periods(each period is the equivalent of one organizational year),with each decision period lasting about two-and-a-halfhours. At the end of each period, teams made alterationsto the marketing mix (including pricing, sales force size,advertising budget, research and development budget,and production volume). These decisions were based onprior results given to participants as part of the game, andmarketing research studies that participants could pur-chase out of their game budgets (see Cook 1987 for moredetails of MARKSTRAT as a research tool).

MethodSampleThe sample consisted of all 159 managers (divided into35 teams) attending selected executive development pro-grams at a European management institute. The average

age was 41 (sd = 6), with the range extending from 25to 56. Fourteen countries were represented including 30or more people from France, Germany, and Switzerland.Because only three of the participants were women, gen-der diversity was not measured in this study. The man-agers occupied a variety of functions in European firms,including over 20 people in each of the following func-tions: marketing, research and development, manufactur-ing, and general management.

Unit of AnalysisThe unit of analysis was the simulated MARKSTRATteam to which managers were assigned. All variableswere calculated at the team level. There were 35 MARK-STRAT teams.

MeasuresQuestionnaires were distributed at two points in time:early in the game (at the end of the second decision pe-riod), and late in the game (at the end of the second fromlast decision period). These questionnaires were preparedin three languages (English, French, and German) corre-sponding to the language of instruction being used in theexecutive education programs of which the MARK-STRAT simulation formed a part. Following Hofstede" s(1980, p. 34) recommendations, translations from theoriginal English version ofthe questionnaire were accom-plished by a person fluent in all three languages and fa-miliar with the content matter of the document. Thismethod avoids the often clumsy phrasing that results fromback translation. The response rate for the questionnairewas 91.5% and 95% for the early and late administrationsrespectively.

Peifonnance. Because MARKSTRAT is a marketingsimulation, the focus of team effort is on improving mar-keting performance. The measures of performance, there-fore, relate to the results of decision making by the topmanagement of the marketing department, which is re-sponsible for the overall orientation ofthe company to itsmarkets (Larreche and Gatignon 1977, p. 5).

Two measures of performance were monitored by theteams throughout the game—net marketing contribution(in dollars) and market share (percent). The net marketingcontribution was the basic measure of profitability andwas defined as total revenues from sales minus expenses(cost of goods sold, research and development, inventoryholding costs, etc.). The market share was expressed as apercentage of the total market captured by the focal firmin competition with the other four firms in its industry.Recall that each team competed with four other teams formarket share.

In the analyses we look at these two measures of per-formance in different ways. To assess how well firms

ORGANIZATION SCIENCEA^OL 11, NO. 1, January-February 2000 25


were doing early in the simulation, we assessed Early NetMarketing Contribution Change as the net marketingcontribution at the start of the second period divided bythe net marketing contribution at the start of the game.This measure was greater than one for firms that im-proved their performance during the first period, and lessthan one for firms that suffered performance decreases asa result of decisions made during the first period. We alsoexamined Final Market Share and the Cumulative NetMarketing Contribution for each firm. These were themarket share and profitability results at the end of thesimulation. To measure performance increase and de-crease over the course of the game, we looked at MarketShare Change, measured as the percent change in marketshare for each firm from early in the game (end of secondperiod) until the end of the game.

Control Variables. In MARKSTRAT, teams takeover existing firms, and these firms have different startingprofiles. Five fimis make up each MARKSTRAT world,and of these five in the cuirent simulation, two firms hadrelatively disadvantaged starting positions whereas theother three firms had relatively advantaged starting po-sitions. Thus, we controlled for whether the team tookover an advantaged or disadvantaged firm by classifyingfirms into two groups: advantaged and disadvantaged.The managers attending these short executive programswere assigned to teams by the instructors and had notbeen in prior classes together.

We controlled for size because the size of teams variedbetween three and five people.

Cognitive Diversity. Six questionnaire items com-prised the raw data for team-level cognitive diversitymeasures. These single-item questionnaire measures werebased on the questions used by Zucker (1977) to measurecognitive variability in an institutionalization context.Like Zucker, we asked the managers to indicate (onseven-point scales) their perceptions of group processes.

First, we assessed perceptions of role specialization inthe team. As Berger and Luckmann (1967, p. 74) haveemphasized, "the construction of role typologies is a nec-essary correlate of the institutionalization of conduct."Before team members can function as an organization, itmay be necessary to agree among themselves on a divi-sion of labor. Managers were asked to indicate approxi-mately how specialized the members of the MARK-STRAT organization were, with the scale anchored withthe phrases "No person has a specialized role to play,"and "Each person has a specialized role to play."

The second item related to the distribution of power inthe organization. An important aspect of consensus in or-ganizations concerns the taken-for-grantedness of au-thority relations. The higher the consensus, the more dif-ficult it becomes to challenge the existing order of things

(cf. Garfinkel 1967). Managers were asked to estimatehow easy it would be to challenge the decision-makingpower of the dominant members. The scale was anchoredwith the phrases "Very easy to challenge decision-makingpower of dominant members" and "Very hard to chal-lenge decision-making power of dominant members."

Third, to measure agreement about the causes of per-formance, we asked team members the following ques-tion: "How much agreement was there in your MARK-STRAT organization about the causes of your marketshare results during the last decision period?" The scalewas anchored with the phrases "Everybody stuck to adifferent explanation," and "All members agreed." Thegreater the variation in members' responses to this ques-tion, the more interpretative ambiguity the team was as-sumed to exhibit with respect to attributions about pastperformance. If everybody in the team responded iden-tically, then this was taken as evidence of team consensusabout the level of agreement in the team.

Fourth, we assessed the degree to which each partici-pant perceived team decision making to be an habitual,routine, taken-for-granted activity. Berger and Luckmann(1967, p. 54) have pointed out that "Habitualizationmakes it unnecessary for each situation to be definedanew, step by step." To the extent that decision makingin the team was perceived by a participant to be part ofa high consensus, institutionalized process, then the par-ticipant should perceive the decision-making process asrelatively easy rather than difficult. Participants were re-quested to: "Think about how difficult or easy it is foryour MARKSTRAT organization to make decisions."The scale was anchored with the phrases "Very difficultto make decisions" and "Very easy to make decisions,"

The fifth item concerned each individual's agreementwith team decision making. The more consensus in theteam over decisions, the more each individual should feelnoiTnative pressure to agree with the team decision(Zucker 1977, p. 737). We asked participants to "Thinkabout your participation in the organization's decisionsin the last session. Did you feel you should agree withthe decisions reached by your MARKSTRAT organiza-tion?" The scale was anchored with the phrases "No—Ifelt that I could disagree," and "Yes—I felt that I shouldagree."

Finally, increasing consensus was expected to increasethe taken-for-grantedness of organizational procedures,thus making it more difficult to imagine ways to improveeffectiveness. Each manager was asked whether therewere some obvious ways in which the MARKSTRATorganization could be run more effectively, with the endpoints of the seven-point scale anchored with the phrases:

26 ORGANIZATION SCIENCEA^ol. 11, No. 1, January-February 2000

MARTIN KILDUEE, REINHARD ANGELMAR, AND AJAY MEHRA Top Management-Team Diversity

"Yes—there are many ways to increase effectiveness,"and "No—there are few ways to increase effectiveness."

The team-level coefficient of variation scores on thesesix variables represented our measures of cognitive di-versity.

Demographic Diversity. We measured three demo-graphic diversity variables—national heterogeneity,functional heterogeneity, and age heterogeneity—thatprior literature suggested as important.

AnalysisAll variables were measured at the team level of analysis.The performance variables were measures of how welleach team had performed. The diversity variables werecalculated as heterogeneity within each organizationalteam as follows.

Eor the two categorical demographic variables (nation-ality heterogeneity and functional heterogeneity), Blau's(1977) index of heterogeneity was computed (see Jacksonet al. 1991 for the formula and a recent example of itsuse). This index can vary from zero (indicating all teammembers are the same) to a high of one (indicating allteam members are different).

The other diversity measures (including age heteroge-neity and the cognitive diversity measures) were calcu-lated for each team as the coefficient of variation (i.e.,standard deviation divided by the mean). This is a scale-invariant measure judged to be superior in its psycho-metric properties to other measures such as the standarddeviation (Allison 1978).

The effects of diversity on performance were investi-gated using ordinary least squares multiple regression.The adjusted R-squared statistic is reported in all tablesin order to adjust the variance explained for the numberof parameters in the models. We also used multivariateanalysis of variance (MANOVA) to assess the effects ofindependent variables on multiple dependent variables.

ResultsTable 1 shows the effects of advantageous and disadvan-tageous starting position on the perfonnance and cogni-tive diversity measures. Note that the firms we labelledas advantaged did in fact have significantly higher per-formance than the firms we labelled as disadvantaged.The effects of an advantageous starting position signifi-cantly influenced early net marketing contribution change(t = - 5.734, p < 0.001), cumulative net marketing con-tribution (t = - 4.897, p < 0.001), and final market share(t = -3.456, p < 0.05). There was no significant dif-ference between advantaged and disadvantaged firms in

Table 1

Variables

Effects of Firm Starting Position

Starting Position

Advantageous Disadvantageous t

Performance Means (and Standard Deviations)

Eariy Net MarketingContribution Change

Final Market Share

Cumuiative NetMarketing Contribution

Market ShareChange

Note- n = 35 firms."p < 0.05; **p < 0,01;

2.04(0.27)

0.23(0.10)

339.09(122.68)

-0.01(0.088)

***p< 0.001,

1.55(0.23)

0.15(0,04)

159.76(73.77)

0.02(0.04)

-5.73***

-3.46**

-4,90***

1.31

how well they improved their market share over thecourse of the game.

Table 2 shows the descriptive statistics for the variablesin this study, including minimum and maximum values.Recall that the demographic and cognitive diversity vari-ables represent variation within each team. This tableshows that teams differed greatly both with respect tohow much market share was gained and lost, as well withrespect to final success or failure in the industry. Thediversity measures show considerable mean differencesin variation across teams.

Table 3 presents the correlations between the variables.Of interest is the significant negative correlation betweenteam size and national heterogeneity, reflecting the factthat some of the most heterogenous groups happened tocontain the least people. The six measures of cognitivediversity were relatively independent of each other: Ofthe correlations between these six variables, only one wassignificant at the 0.05 level (decision pressure and deci-sion difficulty, r = 0,36). None of the demographic di-versity measures were significantly correlated with thecognitive diversity measures.

The first hypothesis, which suggested that demographicdiversity would predict cognitive diversity was not sup-ported. There were no significant relationships betweenthe demographic diversity variables and cognitive diver-sity in any of the regressions we ran. Evidently, teamsthat were heterogenous on age, functional background, ornational origin were no more cognitively diverse early inthe game or late in the game than teams less heterogenouson these demographic indicators.

The second hypothesis suggested a positive effect of

ORGANIZATION SCIENCEA^ol. 11, No. 1, January-February 2000 27


Table 2 Descriptive Statistics

Variables

1. Market Share Change2. Early NMC Change^3. Final Market Share4. CNMC''5. Size6. Starting Position

Demographic Diversity7. National8. Functional9. Age

Cognitive Diversity10. Specialization11. Power13. Ambiguity14. Decision Difficulty15. Decision Pressure16. Effectiveness

Mean

-0.001.840.20

267.364.23

0.370.630.13

0.430.380.160.250.360.37

S.D.

0.070.350.09

137.400.73

0.280.120.07

0.260.160.130.130.210.19

Median

0.011.920.20

242.404

0.380.640.10

0.410.390.140.260.350.39

Minimum

-0.181.230.04

56.6031

00.320.03

00.100000

Maximum

0.232.930.50

606.0052

0.750.800.35

0.990.660.490.500.840.87

Note, n = 35 firms; ^Net marketing contribution change; ''Cumulative net marketing contribution.

Table 3 Intercorrelations Among Variables

Variable 10 11 12 13 14

1. Market Share Change2. Early NMC Change'̂ -0.213. Final Market Share 0.67*" 0.38*4. CNMC'' 0.37* 0.56*** 0.88***5. Size -0.00 -0.21 -0.04 -0.156. Starting Position -0.31 0.72*** 0.53** 0.67*** 0.06

Demographic Diversity7. National 0.34* -0.05 0.168. Functional 0.07 -0.30 -0.019. Age 0.30 -0.17 0.25

Cognitive Diversify10. Specialization 0.19 -0.12 0.16 0.13 -0.0511. Power -0.01 -0.18 0.01 0.03 -0.1012. Ambiguity 0.25 0.27 0.44** 0.50** -0.2213. Decision Difficulty 0.05 - 0 . 3 1 ^ -0.24 -0.33"^ 0.3714. Decision Pressure 0.19 -0.28 -0.12 -0.31+ 0.1615. Effectiveness 0.29' -0.42* 0.06 0.00 0.03

0.15 -0.40* -0.100.15 0.31 -0.050.16 -0.08 -0.07

0.08-0.03

0.26-0.32"*-0.39*-0.26

-0.240.21 -0.09

0.23 -0.05 0.190.13 0.09 0.24 0.110.18 -0.28 -0.07 -0.11 -0.02

-0.09 0.28 0.16 -0.13 0.070.13 0.13 0.20 -0.12 0.060.16 0.26 -0.04 0.03 0.31 •*

-0.09-0.10 0.36*0.05 0.33 + 0.11

Note, n = 35 firms; "Net marketing contribution change; "Cumulative net marketing contribution.*p < 0.05; **p < 0.01; ***p < 0.001.



interpretative ambiguity on subsequent firm performance.The relevant results are displayed in Table 4, where wesee that the overall models explain significant amounts ofvariance in performance, ranging from 21% to 54%. Theresults in Table 4 (model 3) show support for Hypothesis2: The greater the interpretative ambiguity early in thegame, the higher the final performance in terms of bothmarket share (beta = 0.272, p < 0.05) and profits (beta= 361.318, p < 0.05). Because these two performancedependent variables were highly correlated, we con-ducted a MANOVA test to see if there was an overalleffect of interpretative ambiguity on the two dependentvariables taken as a set. There was a marginally signifi-cant overall effect of interpretative ambiguity (F = 2.69;df = 2, 22; p < 0.10).

The results for interpretative ambiguity controlled for

both the significant effects of starting position and pos-sible effects of three types of demographic diversity onperformance. Looking at the demographic diversity vari-ables, functioning as controls in the regressions, we seethat there was a significant effect of age heterogeneity onperformance: the higher the variation in age, the higherthe performance of the firm (MANOVA test: F = 3.47;df = 2, 22; p < 0.05).

Hypothesis 3 suggested that team disagreement con-ceming basic team processes would reduce finn perfor-mance. The results in Table 4 show little support for thishypothesis. Over a range of five cognitive diversity vari-ables related to team processes (specialization, power, de-cision difficulty, decision pressure, effectiveness) onlythe decision difficulty variable was marginally significantin predicting performance: the greater the variation in

Table 4 Effects of Diversity on Final Organizational Performance

Variables

Size

Starting Position

Nationai Heterogeneity

Functional Heterogeneity

Age Heterogeneity

Cognitive Diversity VariablesSpecialization

Power

Ambiguity

Decision Difficulty

Decision Pressure

Effectiveness

Overail Model FAdj. R2

Root MSEDegrees of Freedom

1

0,01(0.02)0.10"

(0.03)0.08

(0.05)0,05

(0.118)0.39-^

(0.20)

3.30*0.250.08

5,29

Final Market Share

2

0.01(0.02)0.07"

(0.03)

0.03(0.06)0.04

(0.10)0.26'-

(0.13)-0.17(0.13)0.05

(0.08)0.08

(0.09)2.12 +0.210.08

8,26

3

0.02(0.02)0.08*

(0.03)0.05

(0.05)0.11

(0.122)0.53-

(0.21)

-0.01(0.05)

-0.02(0.10)0.27'

(0.13)-0 .23*(0.12)0.01

(0.07)0.10

(0.09)2.61*0.340.08

11,23

Cum.

1

0.74(27.02)195.45*"(36.22)97.81

(69.77)-54.13(151.46)420.26

(256.80)

6.55***0.45

101.965,29

Net Marketing

2

7.84(26.93)142.25'(42.49)

10.21(68.61)91.97

(128.77)387.83-

(161.73)-211.02(164.00)-16.29(93.37)113.85

(108.17)4.59*0.46

101.178,26

Contribution

3

24.69(25.60)157.10"(40.02)67.48

(67.87)19.76

(151.54)640.32*

(255.13)

-42.47(66.14)10.60

(121.77)361.32'

(155.77)-275.04*(154.86)-62.82(87.21)155.38

(106.11)4.70***0.54

92.7211,23

Note, n = 35 firms; the table shows unstandardized betas (with standard errors in parentheses).

*p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001.

ORGANIZATION SCIENCE/VOI. 11, NO. 1, January-February 2000 29


team members' perceptions of decision-making diffi-culty, the lower the performance of the firm in terms ofmarket share (beta = - 0.232, p < 0.10) and profits (beta= -0.275.043, p < 0.10). But this marginal effect wasnot evident when the two dependent variables were con-sidered as a set (MANOVA test: F = 1.76; df = 2, 22,ns).

Hypothesis 4 suggested that increases in firm perfor-mance would lead to decreases in top management-teamcognitive diversity. Table 5 provides two tests of this hy-pothesis, using an early measure of performance changeand a later measure of perfomiance change. Note that inthese analyses we omit the starting-position control vari-able because of its high multicoUinearity with one of theindependent variables, early net marketing contributionchange. Starting position had no significant effect on anyof the dependent variables and no overall effect on thedependent variables taken as a set. Its inclusion in theanalyses weakened model fit, but had no effect on thepattem of results that we report.

First, Table 5 shows the effect of changes in perfor-mance in the early stages of the game on cognitive di-versity. Contrary to the hypothesis, this effect is generallypositive. A MANOVA test on the six cognitive diversityvariables taken as a group shows a significantly positiveeffect of early improvement in profits (operationalized asEarly Net Marketing Contribution Change) on cognitivediversity (F = 4.60; df = 6, 26; p < 0.01). The oneexception to this generally positive relationship betweenperfonnance change and cognitive diversity change was

the interpretative ambiguity variable. As profits increasedin the early stages in the game, the interpretative ambi-guity of the top management team decreased (beta =-0.217, p < 0.10).

During the remainder of the simulation, however, thepredicted negative relationship between perfonnancechange and top management-team cognitive diversitychange was in evidence. As Table 5 shows, five of thesix cognitive diversity variables were negatively relatedto overall market share increase. A MANOVA test on theeffect of change in market share over the course of thesimulation on the six cognitive diversity dependent vari-ables taken as a set was significant (F = 3.06; df = 6,26; p < 0.05). However, only the interpretative ambiguitymeasure reached significance in the univariate tests: Thegreater the increase in market share, the more the teamtended to reduce interpretative ambiguity (beta =-1.412, p < 0.01).

Table 6 details the effects of cognitive diversity changeon the final performance of firms in the simulation. Thesewere exploratory analyses for which we formulated nohypotheses. The models explained significant amounts ofvariance, with adjusted r-squares ranging from 45% to58%. Immediately apparent is the continued importanceof starting position: Firms that started advantageouslyconcluded the game with significantly higher marketshare and revenues than firms that started the game indisadvantageous positions.

Looking at the cognitive diversity measures, we seethat increasing interpretative ambiguity was associated

Table 5 Effects of Performance Change on Team Cognitive Diversity Change

Size

Early Net MarketingContribution ChangeMarket Share Change

Overall Model FAdjusted R̂Root MSEDegrees of Freedom

Specialization

0.04(0.06)0.23-"

(0.12)-0.79(0.57)

2.33""0.100.24

3,31

Cognitive

Power

0.05(0.05)0.05

(0.10)0.62

(0.47)

.970.000.19

3,31

Diversity Change

Ambiguity

-0.02(0.04)

-0.22 +(0.09)

-1 .41**(0.41)

5.09"0.270.17

3,31

DecisionDifficulty

-0.02(0.04)0.13

(0.08)-0.02(0.37)

1.330.030.15

3,31

DecisionPressure

-0.01(0.07)0.18

(0.14)-0.59(0.65)

1.110.010.27

3,31

Effectiveness

0.03(0.05)0.20 •"

(0.10)-0.56(0.45)

2.38""0.110.19

3,31

MANOVA^(Wilks' Lamda)

0.49, F = 4.60**

0.59, F = 3.06*

6,26

Note, n = 35 firms. The table shows unstandardized betas (with standard errors in parentheses).^The MANOVA indicates whether the six dependent variables taken as a set are significantly affected by an independent variable.*p < 0.10, *p < 0.05, **p < 0.01.


MARTIN KILDUFE, REINHARD ANGELMAR, AND AJAY MEHRA Top Management-Team Diversity

Table 6 Effects of Cognitive

SizeStarting Position

Change in Cognitive DiversitySpecializationPowerAmbiguityDecision DifficultyDecision PressureEffectiveness

Overall Model FAdj. P?Root MSEDegrees of Freedom

Diversity Change on Final Performance

Final Market Share

-0.00-0,07*

-0.010,11

- 0 . 2 4 "0.14*

-0.03-0.10

4.54*0.450.07

8,26

(0.02)(0,03)

(0.06)(0.06)(0.07)(0.08)(0.05)(0.06)

Cumulative

-20.35148,51***

23.07113.75

-303,76**127,24

9.13-145.07*

6.86'*'0.58

89,078,26

Net Marketing Contribution

(21.62)(35.81)

(72.02)(81,81)(89.57)

(108.26)(61.59)(82.25)

Note, n = 35 firms; the table shows unstandardized betas (with standard errors in parentheses),""p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001.

with low final performance both in terms of market shareand profits, A MANOVA analysis, taking the two per-formance measures as a set, confirmed a significant effectfor interpretative ambiguity (E = 6,36: df = 2, 25: p <0.01). A closer investigation of this phenomenon showedthat all six firms with market share changes one standarddeviation below the mean showed increases in inteipre-tative ambiguity over the course of the simulation, Eorthose 19 firms that showed positive market share gains,13 of them showed a reduction in interpretative ambi-guity.

DiscussionThe results cast doubt on the thesis that teams composedof people of diverse nationality, functional background,or age are likely to be diverse in terms of their cognitiveprocesses. There was no evidence for an effect of dem-ographic diversity on measures of cognitive diversity inthese decision-making teams.

One possible reason why demographic diversity mayhave failed to affect measures of cognitive diversity re-lates to the competitive nature of the MARKSTRAT busi-ness simulation. To the extent that people within a teamare driven by the goal of maximizing performance rela-tive to other teams, the social group origins of team mem-bers may fade in importance. A team, such as the 1996championship Chicago Bulls basketball team, character-ized by national, ethnic, and age heterogeneity, may beheld together in part by the fierce interteam rivalry of

professional sports. Previous research and theory has sug-gested that superordinate goals can bring people from di-verse social groups together (Sherif et al. 1961: see thediscussion in Coser 1956). Thus team members may havefocused on their competitive goals rather than on theirnational, functional, or age differences. This suggests thatthe effects of demographic diversity on cognitive diver-sity may be minimized for teams facing extemal com-petition.

It is also possible that the effects of demographic di-versity on cognitive diversity may take more time to ap-pear than was available in this simulation. People mayinitially strive to harmonize their thinking with that oftheir new teammates. But after this initial honeymoonperiod, people may become more comfortable relying onthought pattems characteristic of particular age-cohorts,national cultures, or functional specialties. (See, however,a contrary argument in Pelled 1996, p. 623).

Another possibility is that demography does have ef-fects on cognitive diversity, but that these effects are spe-cific to certain types of demography and certain types ofcognitive diversity not studied in this research.

The hypotheses conceming the effects of cognitive di-versity on firm performance received mixed support.There was little evidence that team disagreement con-ceming basic team processes (such as work structuringand decision making) had any effect on firm perfonnance.

Interpretative ambiguity emerged as the key cognitivediversity measure that differentiated successful versus un-successful teams. Interpretative ambiguity among teammembers early in the game predicted overall performance

ORGANIZATION SCIENCEA'OI. 11, NO, 1, January-February 2000 31


of the firms. An exploratory analysis showed that teamsthat ended the simulation with high perfonnance tendedto reduce the degree of ambiguity over the course of thesimulation, even though they tended to start out with highinterpretative ambiguity. Exactly the opposite pattem wasobserved for low-perfoiTning teams. Thus, successful andunsuccessful teams showed quite different pattems of am-biguity management over the course of the simulation.

Gersick (1990, p. 103), in her qualitative analysis oftwo project groups, suggested that the successful teamtended in the early meetings to operate with implicit un-derstandings that were not openly discussed or chal-lenged. In our terms, this absence of explicit discussiontends to preserve high levels of interpretative ambiguity.Her observations of a less successful group showed themembers spending considerable time during the earlymeetings to explicitly outline their varying positions onimportant issues, thus, in our terms, reducing interpreta-tive ambiguity. Gersick suggested that all groups tend toundergo predictable transitions in the course of their lifecycles, but that groups' trajectories can vary. Our researchsuggests that the team's management of interpretativeambiguity over the course of the team's life cycle can bea factor affecting performance outcomes. Euture workshould examine in greater detail the time-related effectsof team diversity.

The only demographic diversity measure to affect over-all performance was age heterogeneity. Paradoxically, thegreater the diversity of team members' ages, the betterthe teams performed. This result is strikingly differentfrom previous research that has reported generally nega-tive effects of age diversity in groups (see the review inMilliken and Martins 1996), although there have beenfew if any direct tests of top management age diversityon firm performance. Recent research by Hambrick et al.(1996) found positive results of top management-teamheterogeneity on firm perfomiance, although they did notmeasure age heterogeneity. Their results from the realworld paralleled the results from this simulation in show-ing that top management-team heterogeneity had signifi-cantly positive effects both on growth in market share andgrowth in profits. According to Hambrick et al. (1996, p.665), the heterogeneous team has a broad potential be-havioral repertoire and is able to "conceive and launchactions on many fronts." Erom this perspective, demo-graphic heterogeneity may well complement rather thandetermine cognitive heterogeneity. Age diversity doesmatter, but not because it predicts cognitive diversity.Teams heterogeneous on demographic variables may bebetter able to build on the diverse experience base of theteam to validate diverse cognitions, and thus take advan-tage of innovative suggestions.

Not only did diversity affect performance; there wasalso a reciprocal effect of perfomiance change on changesin team cognitive diversity. In the latter part of the simu-lation, teams with improving performance tended to ex-hibit decreasing diversity across the range of cognitivevariables, suggesting a growing consensus among themembers of high-performing teams.

The current research is limited in looking at a relativelysmall number of highly competitive teams engaged in asimulation of real-world decision making. Eortunately,the realism of the MARKSTRAT world is high (Dodgson1987, Remus 1978), and the players in the current re-search were all practicing managers from the private sec-tor, familiar with decision processes in competitive firms.We were able to collect realistic performance data sepa-rately from our questionnaire measures, thereby eUmi-nating the problems of common method variance. Eutureresearch can extend the preliminary results in this studyby examining decision-making teams in the field ratherthan in the laboratory. One arena for which both demo-graphic and performance data is readily available is pro-fessional sports.

In using the results in this paper as a basis for furtherresearch, attention should be paid to the relatively smallnumber of teams relative to the Icirge number of regres-sors. The statistical tests, in particular, may be unstablegiven the low power of the study. We have tried to correctfor this problem by reporting adjusted r-squares, by test-ing hypotheses with more than one dependent variable,and by the extensive use of multivariate tests. Thus, thesignificant effects of interpretative ambiguity on perfor-mance show up whether the performance dependent vari-able is market share or profitability, and whether the testsare univariate or multivariate.

We also recognize that our reliance on one-item mea-sures of cognitive diversity violates arguments concem-ing the importance of reliable, multimethod measure-ments of individual dispositions and cognitions (e.g..Block, 1977). Because of this methodological shortcom-ing, this study's assessments of the effects of cognitivediversity are likely to comprise preliminary tests of thehypotheses.

One of the major tasks of management is to maintainwithin teams rich possibilities for sensemaking while atthe same time promoting coordinated work. In the currentresearch we saw that successful teams tended to allowinterpretative ambiguity to flourish in the early stages oftheir life cycle, but to exhibit more interpretative clarityin the later stages. This cycle of ambiguity and claritymay represent one dynamic solution to the twin problemsof impoverished sensemaking on the one hand and un-coordinated activity on the other. How can teams foster

32 ORGANIZATION SCIENCEA^OI. 11, No. 1, January-February 2000


both equivocality and mutual understanding? The answerfrom the present research is to take advantage of the nat-ural cycle of work: in the beginning let ambiguity flour-ish; in the end, strive for heedful interrelating.

AcknowledgmentsThe authors thank Barbara Lawrence for her consistently helpful edi-torial guidance. They also thank three anonymous reviewers for theirinsightful comments, and Dennis Gioia for his assistance with a pre-vious draft.

ReferencesAllison, P. D. 1978. Measures of inequality. Amer. Soc. Review 43

865-880.Amason, A. C. 1996. Distinguishing the effects of functional and dys-

functional conflict on strategic decision making: Resolving a par-adox for top management teams. Acad. Management J. 1 123-148.

Ashby, W. R. 1952. A Design for a Brain. Wiley, New York.

Berger, P. L., T. Luckmann. 1967. The Social Construction of Reality.

Doubleday, New York.Bettenhausen, K., J. K. Mumighan. 1985. The emergence of norms in

competitive decision-making groups. Admin. Sci. Quart. 30 350-372.

Blau, P. M. 1977. Inequality and Heterogeneity. Free Press. New York.Block, J. 1977. Advancing the science of personality: Paradigmatic

shift or improving the quality of research? D. Magnusson, N.Endler, eds. Personality at the Crossroads: Current Issues in In-

teractional P.sychology. Eribaum, Hillsdale, NJ. 37-63.Bourgeois, L. J., III. 1980. Performance and consensus. Strategic Man-

agement J. 1 227-248.Cohen, M. D., J. G. March, J. P. Olsen. 1972. A garbage can model of

organizational choice. Admin. Sci. Quart. 17 1-25.Cook, V. J., ed. 1987. Special section on Markstrat. J. Bus. Res. 15

(6).Coser, L. C. 1956. The Functions of Social Confiict. Free Press, New

York.Cox, T., Jr., S. Blake. 1991. Managing cultural diversity: Implications

for organizational competitiveness. Acad. Management Executive

5(3) 45-56.Dannefer, D. 1987. Aging as intracohort differentiation: Accentuation,

the Matthew effect, and the life course. Soc. Forum 2 211-236.Dearborn, D., H. A. Simon. 1958. Selective perception: A note on the

departmental identification of executives. Sociometry 35 3 8 ^ 8 .Dodgson, H. 1987. Management learning in Markstrat: The ICL ex-

perience. J. Bus. Res. 15 481^89 .

Donnellon, A. B. Gray, M. G. Bougon. 1986. Communication, mean-ing, and organized action. Admin. Sci. Quart. 31 43-55.

Eisenberg, E. M. 1984. Ambiguity as strategy in organizational com-munication. Comm. Monographs 51 227-242.

Eisenhardt, K. M. 1989. Making fast strategic decisions in high-velocity environments. Acad. Management J. 32 543-576.

Garfinkel, H. 1967. Studies in Ethnomethodology. Prentice-Hall, En-glewood Cliffs, NJ.

Gatignon, H. 1987. Strategic studies in Markstrat. J. Bus. Res. 15 469-

480.

Gersick, C. J. G. 1990. The students. J. R. Hackman, ed. Groups that

Work (and Those that Don't): Creating Conditions for Effective

Teamwork. Jossey-Bass, San Francisco, CA. 89-111.

Hambrick, D. C , T. S. Cho, M.-J. Chen. 1996. The influence of topmanagement team heterogeneity on firms' competitive moves.Admin. Sci. Quart. 41 659-684.

. L. B. Kom. J. W. Fredrickson, R. M. Ferry. 1989. Reinventing

the CEO. Kom/Ferry Intemational. New York.

. P. A. Mason. 1984. Upper echelon: The organization as a reflec-tion of its top managers. Acad. Management Rev. 9 193-206.

Hofstede, G. 1980. Culture's Consequences. Sage, Beverly Hills, CA.Hogarth, R. M., S. Makridakis. 1981. The value of decision making in

a complex environment: An experimental approach. Management

Sci. 27 93-107.

Jacobs, R. C , D. T. Campbell. 1961. The perpetuation of an arbitrarytradition through several generations ofa laboratory microculture.J. Abnontml and Soc. Ps}^ch. 62 649-658.

Jackson, S. E., J. F. Brett, V. I. Sessa, D. M. Cooper, J. A. Julin, K.Peyronnin. 1991. Some differences make a difference: Individualdissimilarity and group heterogeneity as correlates of recruitment,promotions, and turnover. / Appl. P.n'ch. 76 675-689.

Johnston, W. B., A. H. Packer. 1987. Workforce 2000. Hudson Insti-tute, Indianapolis, IN.

Kinnear, T. C , S.K. Klammer. 1987. Management perspectives onMarkstrat: The GE experience and beyond. / Bus. Res. 15 491-501.

Kotter, J. P., J. L. Heskett. 1992. Corporate Culture and Peifonimnce.

Free Press, New York.Larreche. J., H. Gatignon. 1977. MARKSTRAT: A Marketing Strategy

Game. The Scientific Press, Palo Alto, CA.Lant, T. K., D. B. Montgomery. 1987. Learning from strategic success

and failure. / Bus. Res. 15 Scientific Press, Palo Alto, CA. 503-517.

Laurent, A. 1983. The cultural diversity of Western conceptions ofmanagement. Intemat. Stud. Management and Organ. 13 75-96.

. 1986. The cross-cultural puzzle of international human resource

management. Human Resource Management 25 91-102.Lawrence. B. S. 1997. The black box of organizational demography.

Organ. Sci. 8 1-22.

March. J. G., H. A. Simon. 1958. Organizations. Wiley, New York.

Martin, J. 1992. Cultures in Organizations: Three Perspectives. OxfordUniversity Press, New York.

Masuch, M.. P. LaPotin. 1989. Beyond garbage cans: An AI model oforganizational choice. Admin. Sci. Quart. 34 38-67.

Michel, J., D. Hambrick. 1992. Diversification posture and top man-agement team characteristics. Acad. Management J. 35 9-37.

Milliken. F. J., L.L. Martins. 1996. Searching for common threads:Understanding the multiple effects of diversity in organizationalgroups. Acad. Management Review 21 402-433.

Mumighan, J. K., D. E. Conlon. 1991. The dynamics of intense workgroups: A study of British string quartets. Admin. Sci. Quart. 36165-186.

Nemeth, C. J. 1986. Differential contributions of majority and minorityinfluence. Psych. Review 93 23-32.

Pelled, C. H. 1996. Demographic diversity, confiict, and work groupoutcomes: An intervening process theory. Organ. Sci. 1 615-63L

ORGANIZATION SCIENCEA^OI. 11, NO. 1, January-February 2000 33


Pfeffer, J. 1983. Organizational demography. Res. Organ. Behavior 5299-357.

Remus, W. E. 1978. Testing Bowman's managerial coefficient theoryusing a competitive gaming environment. Management Sci. 24827-835.

Rynes, S., J. Lawler. 1983. A policy-capturing investigation of the roleof expectancies in decisions to pursue altematives. J. Appl. Psych.68 620-631.

Schwenk, C. R. 1990. Conflict in organizational decision making: Anexploratory study of its effects in for-profit and not-for-profit or-ganization.s. Management Sci. 36 436-448.

Sherif, M.. O. J. Harvey. B. J. White, W. R. Hood, C. W. Sherif. 1961.Intergroup Conflict and Cooperation: The Robber's Cave E.xper-iment. University Book Exchange, Normiin. OK.

Smith, K. G.. K. A. Smith, J. D. Olian, H. P. Sims. Jr.. D. P. O'Bannon.J. A. Scully. 1994. Top management team demography and pro-cess; The role of social integration and communication. Admin.Sci. Quart. 39 412-438.

Stodgill, R. M. 1959. Individual Behavior and Group Achievement.Oxford University Press, New York.

Triandis, H. C. E. R. Hall, R. B. Ewen. 1965. Member homogeneityand dyadic creativity. Human Relations 18 33-54.

Tsui, A. S., T. D. Egan, C. A. O'Reilly. III. 1992. Being different: Re-lational demography and organizational attachment. Admin. Sci.Quart. 37 549-579.

Wagner. W. G., J. Pfeffer, C.A. O'Reilly, III. 1984. Organizationaldemography and turnover in top management groups. Admin. Sci.Quart. 29 74-92.

Walker, G. 1985. Network position and cognition in a computer soft-ware fimi. Admin. Sci. Quart. 30 103-130.

Waller, M. J., G. P. Huber, W. H. Glick. 1995. Functional backgroundas a determinant of executives' selective perception. Acad. Man-agement J. 38 943-974.

Walsh, J. P. 1988. Selectivity and selective perception: An investiga-tion of managers' belief structures and information processing.Acad. Management J. 31 873-896., C. M. Henderson, J. Deighton. 1988. Negotiated belief structuresand decision performance: An empirical investigation. Organ. Be-havior and Human Decision Process. 42 194—216.

Weick, K.E. 1979. The Social Psychology of Organizing, 2nd ed.Addison-Wesley, Reading, MA.. 1993. Sensemaking in organizations: Small structures with largeconsequences. J. K. Mumighan, ed. Social Psychology in Orga-nizations: Advances in Theory and Research. Prentice Hall, En-glewood Cliffs, NJ. 10-37.. 1995. Sensemaking in Organizations. CA. Sage, Thousand Oaks., D. P. Gilfillan. 1971. Fate of arbitrary traditions in a laboratorymicroculture. 7. Personality and Soc. Psych. 17 179-191.

Wiersema, M. F., K. Bante. 1992. Top management team demographyand corporate strategic change. Acad. Management J. 35 91-121.. A. Bird. 1993. Organizational demography in Japanese firms:Group heterogeneity, individual dissimilarity, and top manage-ment team tumover. Acad. Management J. 36 996-1025.

Zenger, T. R., B. S. Lawrence. 1989. Organizational demography: Thedifferential effects of age and tenure distributions on technicalcommunication. Acad. Management J. 32 353-376.

Zucker, L. 1977. The role of institutionalization in cultural persistence.Amer. Soc. Rev. 42 726-743.

Accepted b\ Barbara Lawrence: received May 1995. This paper has been with the authors for three revisions.


top management-team diversity and firm performance - examini

Documents

net marketing contribution

oxford university press

organization sciencea ol

organization sciencea oi

basic team processes

cognitive diversity change

cognitive diversity variables

team cognitive diversity