TOP MANAGEMENT TEAM DIVERSITY, GROUP PROCESS, AND STRATEGIC CONSENSUS

Strategic Management Journal
Strat. Mgmt. J., 20: 445–465 (1999)
TOP MANAGEMENT TEAM DIVERSITY, GROUP
PROCESS, AND STRATEGIC CONSENSUS
DON KNIGHT1*, CRAIG L. PEARCE2, KEN G. SMITH1, JUDY D. OLIAN1,
HENRY P. SIMS1, KEN A. SMITH3 AND PATRICK FLOOD4
1The Robert H. Smith School of Business, University of Maryland at College Park,
College Park, Maryland, U.S.A.
2Belk College of Business, University of North Carolina—Charlotte, Charlotte, North
Carolina, U.S.A.
3School of Management, Syracuse University, Syracuse, New York, U.S.A.
4University of Limerick, Plassey Technological Park, Limerick, Ireland
This study integrated concepts from upper echelons, group process and social cognition theories
to investigate how demographic diversity and group processes influence strategic consensus
within the top management team (TMT), where strategic consensus is defined as the degree to
which individual mental models of strategy overlap. Data from 76 high-technology firms in the
United States and Ireland were used to examine three alternative models. The results showed
that while demographic diversity alone did have effects on strategic consensus the overall fit
of the model was not strong. Adding two intervening group process variables, interpersonal
conflict and agreement-seeking, to the model greatly improved the overall relationship with
strategic consensus. For the most part, TMT diversity had negative effects on strategic consensus.
The model with superior fit showed both direct and indirect effects of diversity on strategic
consensus. Copyright © 1999 John Wiley & Sons, Ltd.
Diversity in groups and teams is often portrayed
Cross-functional teams, for example, are by defi-
as a positive force leading to effective functioning
nition designed with deliberate differences in
of the team. Diversity supposedly leads to greater
demographic diversity and technical specialization
variance in ideas, creativity, and innovation, thus
(Ancona and Caldwell, 1992).
generating better group performance (Cox, 1993;
In this research, we investigate the effects of
Jackson, May and Whitney, 1995). In the popular
demographic diversity within top management
press, diversity is almost always synonymous
teams (TMTs) on group process and strategic
with gender or ethnic diversity. Research on
consensus. Two related streams of research pro-
organizational work groups, however, has focused
vide the basis for this study. First, upper echelons
on other forms of diversity including differences
theory (Hambrick and Mason, 1984; Finkelstein
in age, education, firm tenure, and functional or
and Hambrick, 1996) links observable demo-
technical background (Jackson et al., 1995).
graphic characteristics of top executives to a va-
riety of organizational processes and outcomes.
Second, group process theory illustrates how
Key words:
top management teams; diversity; consen-
group interpersonal processes work to influence
sus; group process
various group and/or organizational outcomes
* Correspondence to: D. Knight, The Robert H. Smith School
of Business, University of Maryland, College Park, MD
such as firm performance (Shaw, 1981). By inte-
20742, U.S.A.
grating these two research streams we hope to
CCC 0143–2095/99/050445–21 $17.50
Received 27 September 1995
Copyright © 1999 John Wiley & Sons, Ltd.
Final revision received 13 July 1998

---

446
D. Knight et al.
provide a richer understanding of the combined
How do TMT demographic diversity and group
effects of TMT diversity and group processes.
processes relate to strategic consensus? The
TMT diversity and group process are used in
research question is broken into two parts. First,
this research to predict strategic consensus. The
how well do measures of demographic diversity
building
of
strategic
consensus
is
generally
alone explain strategic consensus and, second,
accepted as one of the first steps in the strategy
how does the inclusion of group process measures
formation process (Dess and Origer, 1987; Niel-
affect the model? We propose and test four alter-
son, 1981; Lyles, 1981; Bourgeois, 1980, 1985;
native models of how demographic diversity and
West and Schwenk, 1996). For example, Whyte
group process might work to influence the degree
(1989: 41) notes that the first task of all decision-
of TMT consensus about firm strategy. The sam-
making groups is to ‘produce consensus from the
ple for this research included top management
initial preferences of its members’. Similarly,
teams from 83 high-technology firms in the
Floyd and Wooldridge (1992: 27) contend that
United States and Ireland.
the successful implementation of strategy requires
that managers are ‘acting on a common set of
strategic priorities’. Studies of strategic consensus
THEORETICAL BACKGROUND
have either focused on its link to organizational
Strategic consensus as a mental model
performance (Bourgeois, 1980, 1985; Dess, 1983,
1987; West and Schwenk, 1996), or on the proc-
Hambrick and Mason (1984) argue that the
ess of consensus formation (Priem, Harrison and
psychological and cognitive characteristics under-
Muir, 1995; Schweiger, Sandberg, and Ragan,
lying observable demographic measures are criti-
1986; Cosier and Rechner, 1985). In this research,
cal to the group’s processes and subsequent
we focus on the process of consensus formation;
decisions. This is consistent with a growing body
that is, we treat TMT strategic consensus as a
of research on managerial cognition (see Walsh,
dependent variable.
1995, for a summary) which suggests that man-
We advance the strategic consensus literature
agers’ mental models will influence the decisions
by conceptually and empirically linking strategic
they make (Day and Lord, 1992).
consensus to managerial cognition (Walsh, 1995).
Mental models are similar to knowledge struc-
More specifically, we utilize a measure of each
tures (Walsh, 1995), schema (Fiske and Taylor,
team member’s interpretation of the business-
1984; Ireland et al., 1987; Sims and Gioia, 1986),
level strategy of his/her firm as a representation
and implicit theories (Brief and Downey, 1983).
of that team member’s mental model of the firm’s
With regard to managers, Mintzberg (1973: 183)
strategy. According to Barr, Stimpert, and Huff
observes that ‘it is the power of his mental
(1992: 16), mental models ‘consist of concepts
models that determines to a great extent the
and relationships an individual uses to understand
effectiveness
of
his
decisions’.
Kiesler
and
various situations or environments’. In this paper,
Sproull (1982: 557) assert that ‘managers operate
we use the term mental models to refer to clusters
on mental representations of the world and those
of
related,
though
not
necessarily
causally
representations are likely to be of historical
organized, concepts. When we aggregate individ-
environments rather than of current ones’. Prior
ual mental models of strategy to the group level,
research has shown mental models to be related
we assess the level of strategic consensus, which
to strategies (Day and Lord, 1992), strategic
is the extent to which the individual team mem-
actions, and performance (Thomas, Clark, and
bers’ mental models of strategy overlap. Linking
Gioia, 1993), and interpretations of and responses
the concept of strategic consensus to the underly-
to strategic issues (Dutton and Dukerich, 1991).
ing mental models of strategy allows us to explore
Scholars have also posited that mental models
the extent to which shared cognitions of strategy
can operate on the group level and have used
are influenced by TMT diversity and TMT group
terms such as shared cognition, team mental
process.
Such
relationships
have
long
been
model (Klimoski and Mohammed, 1994), collec-
assumed (Shaw, 1981; Hambrick and Mason,
tive cognitive map (Axelrod, 1976) or dominant
1984), but, to our knowledge, not widely tested,
logic (Prahalad and Bettis, 1986) to describe this
in the literature.
phenomenon. Here we use the term ‘strategic
The central research question in this study is:
consensus’ to represent the shared cognitions
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---

Diversity, Group Process and Consensus
447
among team members. This term mainly refers
orientation (Finkelstein and Hambrick, 1996),
to agreement or overlap among individual team
which consists of two elements: psychological
members’ mental models of strategy, but does
characteristics (including values, cognitive mod-
not necessarily imply a deliberative consensus-
els, and other personality factors) and observable
seeking process.
experiences. A fundamental principle of upper
We contend that each top-level manager will
echelons theory is that observable experiences
have a mental model or perception of the strategic
(i.e., demographic measures) are systematically
concepts and their interrelationships that his/her
related
to
the
psychological
and
cognitive
firm uses in an attempt to manage its environ-
elements of executive orientation. Upper echelons
ment. For example, each manager will have a
research employs the use of observable demo-
mental model of the role that various elements
graphic characteristics as proxy measures of
(such as innovation, costs, or service) play in the
executive orientation. Executive orientation works
firm’s overall strategy. Moreover, we argue that
through a perceptual or filtering process that
these mental models may vary among TMT mem-
results in what is called managerial perceptions
bers within a given firm. Accordingly, we esti-
(Hambrick and Mason, 1984) or construed reality
mate each manager’s mental model or perception
(Finklestein and Hambrick, 1996). Managerial
of his/her firm’s strategy from the responses
perceptions, in turn, influence strategic choices
he/she made to a broad set of 48 questions
and executive actions.
regarding the firm’s strategy. This is an indirect
Research using this theoretical framework has
measure of individual mental models. That is, we
linked the demographic characteristics of top
did not directly elicit the specific mental model
managers and/or the demographic diversity of the
of strategy from each subject, but we used a set
TMT to a variety of organizational outcomes
of terms common in the organizational strategy
including performance (Keck, 1991; Hambrick
literature and among executives to develop a
and D’Aveni, 1992; Michel and Hambrick, 1992;
quantitative measure of firm strategy.
O’Reilly and Flatt, 1989; Smith et al., 1994),
We use these individual measures to create
strategy (Finkelstein and Hambrick, 1990; Michel
a group-level measure that we called strategic
and Hambrick, 1992), strategic change (Grimm
consensus. We then investigate how TMT diver-
and Smith, 1991; Wiersema and Bantel, 1992),
sity and group processes relate to the strategic
management turnover (Wagner, Pfeffer and O’Re-
consensus of the TMT. Our contention is that
illy, 1984), and organizational innovation (Bantel
strategic consensus can result either because TMT
and Jackson, 1989; O’Reilly and Flatt, 1989;
members have similar backgrounds (e.g., low
Smith et al., 1993).
diversity), and/or because effective group proc-
Studies have also found that demographic
esses have been utilized to resolve differences in
diversity can influence group processes. In fact,
individual mental models of strategy. Although
diversity can influence group processes in contra-
there are other factors that might lead to strategic
dictory directions. For example, diversity has been
consensus (such as the fact that managers in a
shown to have negative effects on both group
firm are exposed to relatively similar stimuli from
cohesion (Katz, 1982; Lott and Lott, 1961; O’Re-
both inside and outside the firm), we focus in this
illy, Caldwell, and Barnett, 1989) and the fre-
research on the specific effects of demographic
quency or quantity of communication (Smith et
diversity and group processes.
al., 1994; Wagner, Pfeffer and O’Reilly, 1984).
In addition, diversity tends to lead to increased
conflict within the
group (Eisenhardt
and
Demographic diversity
Schoonhoven, 1990; Wagner et al., 1984) and to
Upper echelons theory builds on the idea of the
increased political activity (Pfeffer, 1981). How-
dominant coalition (Cyert and March, 1963) to
ever, diversity can also lead to enhanced creativ-
propose that executives influence organizational
ity and innovation by generating greater variance
performance through the decisions they make
in decision-making alternatives (Cox, 1993; Jack-
(Hambrick and Mason, 1984). Upper echelons
son et al., 1995). Consistent with prior TMT
theory
suggests
that
executives
will
make
research (e.g. Eisenhardt and Schoonhoven, 1990;
decisions that are consistent with their cognitive
Hambrick and D’Aveni, 1992; Keck, 1991;
base (Hambrick and Mason, 1984) or executive
Michel and Hambrick, 1992; Smith et al., 1994),
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---

448
D. Knight et al.
we used measures of diversity in functional back-
impact of interpersonal conflict on the level of
ground, age, education, and employment tenure
strategic consensus within the TMT.
as indicators of TMT diversity, and investigated
We define agreement-seeking behaviors as
how TMT diversity was related to group proc-
those that are intended to produce consensus or
esses and to the level of strategic consensus
agreement among TMT members regarding firm
within the TMT.
strategy. Gero (1985) found that group members
have
more
confidence
in
decisions
reached
through consensual decision
techniques (i.e.,
Group processes
agreement-seeking
techniques)
than
in
those
Previous
TMT
studies
have
examined
the
resulting from higher conflict techniques. Previous
relationships between organizational performance
studies (Schweiger et al., 1986; Schweiger, Sand-
and a variety of group processes. These include
berg, and Rechner, 1989) have shown that groups
comprehensiveness
in
the
strategic
decision-
using structured methods of task-oriented conflict
making process (Fredrickson, 1984; Fredrickson
(such as dialectical inquiry or devil’s advocacy)
and Iaquinto, 1989; Fredrickson and Mitchell,
during decision-making produced higher-quality
1984),
speed
in
decision-making
processes
recommendations than groups using agreement-
(Eisenhardt, 1989; Flood et al., 1997), and polit-
seeking processes. However, agreement-seeking
ical behavior within top management teams
processes led to greater member satisfaction with
(Eisenhardt and Bourgeois, 1988). Group proc-
the group and higher acceptance of their group’s
esses have also been shown to intervene in the
decisions. These findings indicate that the type
relationship between diversity and firm perform-
of group process used to manage or direct conflict
ance (Smith et al., 1994). The central arguments
can influence not only the quality of the decision
behind the study of group processes pertain either
but the affective responses of group members as
to group processes that provide greater efficiency
well. Because some groups may actively engage
(e.g., reducing costs or increasing speed in
in agreement-seeking behaviors (such as not mak-
decision-making) or greater effectiveness (e.g.,
ing final decisions until all members of the group
making better decisions). We contend that group
agree), the present study built on this research
processes are likely to influence the level of
stream and examined the influence of agreement-
strategic consensus within a TMT, and examine
seeking styles on strategic consensus within the
two important group processes from previous
TMT.
research: interpersonal conflict and agreement-
seeking.
TMT diversity, group processes and strategic
Consistent with prior research, we define inter-
consensus: Alternative models
personal
conflict
as
conflict
that relates
to
emotional
or
personal
relationships
between
Recall, the research question of this study is:
people (Amason, 1996) as opposed to conflict
How do TMT demographic diversity and group
that is task oriented in nature. While previous
process relate to strategic consensus? A review
studies have indicated that there can be different
of the literature suggests alternative explanations
kinds of conflict (Amason, 1996; Cosier and
or paths. We treat these alternative explanations
Rose, 1977; Jehn, 1992; Pelled, 1996; Pondy,
as three alternative or more finely elaborated
1969; Priem and Price, 1991), we chose to focus
models (as portrayed in Figure 1).
only on interpersonal conflict in this research.
Pfeffer (1983: 348) contends that ‘demography
Interpersonal conflict (called social or affective
is an important causal variable that affects a
conflict in some research) can have negative
number of intervening variables and processes,
consequences for an organization. For example,
and through them, a number of organizational
Amason (1996) found that affective conflict was
outcomes’. Pfeffer (1983) also questions whether
significantly
and
negatively
related
to
both
the study of the intervening processes will explain
decision quality and affective acceptance of the
any incremental variation in the dependent vari-
decision. Jehn (1995, 1997) found that member
ables beyond what could be explained by demo-
satisfaction with the group was negatively related
graphic measures alone. Thus, Pfeffer (1983)
to social conflict among group members. Accord-
would contend that the addition of group process
ingly, we built on this research and examined the
variables would not improve the explanation of
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---

Diversity, Group Process and Consensus
449
Figure 1.
Alternative models relating TMT diversity, group processes, and strategic consensus
strategic consensus. Also, demographic variables
Pfeffer (1983) suggests that individuals from dif-
may directly influence strategic consensus in
ferent age cohorts will have significantly different
addition to any influence expressed through group
values and perspectives. Hambrick, Cho, and
processes. Consistent with Pfeffer (1983), the first
Chen suggest that demographic heterogeneity will
part of the research question is how well meas-
lead to ‘dispersion in the group’s perspective’
ures of demographic diversity alone explain stra-
(1996: 664). This dispersion in perspectives (or
tegic consensus, and we first examine whether
construed reality as Finkelstein and Hambrick,
there is a direct relationship between demographic
1996, term it), which may arise from different
diversity and strategic consensus without con-
life or unique organizational experiences, make it
sidering the effects of group processes. We label
less likely that members of a TMT will have the
this model the ‘direct effects’ model in Figure 1.
same mental models of firm strategy.
Finkelstein and Hambrick (1996) and Priem
More specifically, if cognitions or mental mod-
(1990) suggest that demographic diversity will
els of the individual manager are based on his/her
be negatively related to strategic consensus.
past experiences and values as Kiesler and Sproull
Demographic diversity within a TMT reflects dif-
(1982) suggest, then differences in experience
ferences in experiences among its members,
should result in differences in mental models.
which should result in differences in individuals’
Thus, if past experiences are represented by a
mental models. Differences in mental models
surrogate measure—demography—we expect dif-
should, in turn, be reflected in differences in the
ferences in demography to be related to differ-
way that individual TMT members characterize
ences in managers’ cognitions. For example, a
or understand firm strategy. For example, Ireland
marketing vice-president might have a different
et al., (1987) posit that people of similar age are
mental model of organizational strategy than a
likely to have similar values and beliefs because
finance vice-president (e.g., ‘First on the market’
their life experiences are similar. Similarly,
vs. ‘Profits through tight cost control’). Therefore,
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---

450
D. Knight et al.
the literature suggests a direct effects model
when dealing with firm strategy since it involves
whereby diversity in functional position, age, edu-
a high degree of uncertainty and ambiguity
cation, and employment tenure will be negatively
(Eisenhardt, Kahwajy, and Bourgeois, 1997).
related to the level of strategic consensus.
Accordingly, we expect that interpersonal conflict
In contrast to Pfeffer, other research (Gist,
among TMT members will be negatively related
Locke, and Taylor, 1987; Smith et al., 1994)
to strategic consensus. That is, when interpersonal
suggests that the study of intervening group proc-
conflict is high within the TMT the differences
esses is important. With regard to strategic con-
in perceptions or mental models of what the
sensus, Langfield-Smith (1992) proposes that
firm’s current strategy is will be greater and
social (i.e.,
group)
processes
will
have
an
strategic consensus as we measure it will be
important effect on the development of shared
lower.
cognitive maps. Ford and Baucus (1987) suggest
Another important group process identified in
that individual cognitive interpretations will be
previous research is the manner in which groups
shaped by the personal contexts in which the
make decisions. As noted above, prior studies
individual operates. Even if Pfeffer (1983) is
(Schweiger et al., 1986, 1989) found that groups
correct, the study of potentially intervening group
using
agreement-seeking
behaviors
achieved
processes is still important as their effects on the
higher levels of consensus than groups using
organization may be easier to control or alter
processes that incorporated structured systems of
than the effects that arise from the group’s demo-
task-oriented conflict. Consistent with this litera-
graphic characteristics. In addition, group proc-
ture, we expected that the use of agreement-
esses may, in fact, be useful in overcoming some
seeking behaviors by TMTs (i.e., behaviors
of the potentially negative consequences of demo-
intended
to
foster
agreement
among
team
graphic diversity. Therefore, we also examine a
members) would be positively related to strategic
set of intervening models to assess whether
consensus (i.e., there would be greater similarity
inclusion of two group process measures—
among team members’ interpretations or mental
interpersonal conflict and agreement-seeking—
models of strategy).
improves the explanation of strategic consensus,
Since prior research offers little direction on
and, if so, which intervening model is most
the precise relationships among the variables, we
appropriate to the data.
analyze two possible intervening models. Model
If group processes influence the cognitions of
2, designated as the ‘partially mediated model’,
TMT members, then it is reasonable to believe
assumes that demographic diversity will have
that interpersonal conflict within the TMT may
both direct and indirect effects on the level of
diminish shared cognitions about firm strategy. In
strategic consensus. Model 3, which is labeled
fact, as Jehn (1995: 257) notes, conflict has been
the ‘fully mediated model’, assumes that TMT
defined as ‘perceptions by the parties involved
demographic diversity will have no direct effect
that
they
hold
discrepant
views’.
Minkes
on strategic consensus but will directly affect the
(1994: 80) defined conflict as ‘imperfect compati-
intervening group processes which, in turn, will
bility of views which necessarily follows from
affect strategic consensus.
the variety of human beings’. This suggests that
With regard to the intervening models, the
differences in mental models or construed realities
question of relationships between the TMT demo-
cannot exist without conflict. Research on group
graphic measures and the intervening group proc-
cohesiveness (Shaw, 1981), which might be inter-
esses, and relationships between the two group
preted as the inverse of interpersonal conflict,
process
measures
(interpersonal
conflict
and
suggests that interpersonal conflict may reduce
agreement-seeking) must be addressed. In general,
strategic consensus. Moreover, while interpersonal
we expect demographic diversity will be posi-
conflict may provoke disagreements about what
tively related to interpersonal conflict within the
firm strategy should be, it is also likely to result
team. This is consistent with prior research such
in different perceptions or interpretations about
as that of Zenger and Lawrence (1989), who
what current firm strategy actually is. Conflict
found that heterogeneity can lead to conflict.
that is personal or emotional in nature may result
Similarly, with regard to agreement-seeking, we
in or even create disagreements about a wide
propose that demographic diversity will be nega-
range of issues. This may be particularly true
tively related to the use of agreement-seeking
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---

Diversity, Group Process and Consensus
451
behaviors within the TMT. This is also consistent
The U.S. sample contained 56 companies from
with prior research. For example, heterogeneity
114 originally contacted. An almanac profiling
has been found to impede other group processes,
high-tech or technology-intensive firms that were
such as communication between team members
part of a technology and research and develop-
(O’Reilly et al., 1989, Smith et al., 1994), which
ment consortium was the original source used to
cause us to believe that it will have a negative
identify the firms. Forty-seven firms were dropped
impact
on
the
use
of
agreement-seeking
from the sample after initial contacts for a variety
behaviors.
of reasons, including unwillingness to participate,
Finally, the question of a relationship, if any,
mistaken
identification
(e.g.,
were
low-
and the direction of that relationship between
technology), mergers, diversification, or because
interpersonal conflict and agreement-seeking must
the firm had gone out of business. Interviews
be addressed. It is reasonable to expect that inter-
were conducted with the CEOs of the remaining
personal conflict and agreement-seeking would be
67 firms (an initial response rate of 59%).
related. However, there remains the question of
The personal interviews served two purposes.
exactly how these two group processes are
First, it allowed the researcher to explain more
related. In this research, we proceed with the
fully the goals of the study and to obtain the
expectation
that
interpersonal
conflict
would
CEO’s approval and endorsement of the study.
likely precede and affect the degree to which
The study design called for the CEO to identify
TMT members are willing to adopt agreement-
each of the team members and for each team
seeking decision processes within the team. This
member to complete a questionnaire. Second, as
is consistent with other researchers, such as Jehn
part of the interview, the CEO initiated a memo
(1997: 531), who suggests that ‘relationship con-
to each top management team member, which
flicts interfere with task-related effort’. Accord-
requested participation in the study and served
ingly, we expect that interpersonal conflict will
to endorse the study, increasing the likelihood
be negatively related to the use of agreement-
of participation.
seeking behaviors within a TMT, since agree-
Usable responses were received from 53 of the
ment-seeking is a task-related undertaking. We
67 CEOs interviewed, for a participation rate of
also expect that agreement-seeking behavior will
79 percent of the final population (47% of those
intervene in the relationship between interpersonal
originally selected). The size of companies in the
conflict and strategic consensus.
sample, measured in gross sales, ranged from
$200,000 to
$162 million.
Mean
sales were
$29 million,
with
a
standard
deviation
of
METHOD
$32.5 million and a median of $17.8 million. The
mean size of firms in number of employees was
Sample
357, with a standard deviation of 395, while the
The sample for this study consisted of the TMTs
median number of employees was 225.
of 83 high-technology firms located in the mid-
The companies that did not participate in the
Atlantic region of the United States and a group
study were involved in similar kinds of businesses
of
subsidiaries
of
U.S.
multinational
firms
to the firms included in the final sample. A one-
operating in Ireland. Data used in this research
way analysis of variance on gross sales indicated
were developed from CEO interviews and from
that responding firms were not significantly differ-
comprehensive questionnaires that were com-
ent from nonresponding firms in terms of size
pleted by the members of the top management
(F = 1.85; p
0.18; N = 114).
team at each company. The companies included
Hambrick and Mason’s (1984) upper echelons
in the sample were involved in a variety of
theory suggested that researchers can identify
technology-related
businesses
including
infor-
members of a top management team simply by
mation technology, research (biotechnology and
equating executive titles with membership in the
aerospace), hazardous waste management, and
team, and some recent studies of top management
defense. The sample included both large and
teams have used this approach (Norburn and
small firms, both publicly and privately owned,
Birley, 1988; Tushman, Virany, and Romanelli,
and, in the Irish sample, subsidiaries of U.S.
1989; O’Reilly and Flatt, 1989; Keck, 1991).
multinational companies.
However, to more closely approximate Cyert and
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---

452
D. Knight et al.
March’s (1963) notion of the dominant coalition,
Group process variables
we asked each CEO to identify the members of
his or her ‘real’ top management team. Of those
Two group process measures were assessed in
identified, 78 percent were also officers of the
this research: interpersonal conflict and agreement
corporation.
seeking. Interpersonal conflict was defined as con-
All members of the top management team,
flict that was emotional or person oriented in
including the CEO, were asked to complete ques-
nature rather than task or issue oriented (Amason,
tionnaires. From the 286 questionnaires requested
1996). It was measured with a five-item, Likert-
from team members, a total of 230 usable ones
type scale where one represented ‘definitely not
were returned. Eighty percent of the team mem-
true’ and five represented ‘definitely true’. The
bers who were asked to complete the question-
questions included items such as, ‘The members
naire did so, and the average number of question-
of the TMG get along together very well’ and
naires returned per firm was 4.5.
‘Relationships between members of the TMG are
In the Irish sample, 60 percent of the com-
best described as “win–lose;” if he/she wins, I
panies contacted agreed to participate, yielding a
lose’. Cronbach’s alpha for the interpersonal con-
final sample of 26 companies. From these 26
flict scale was 0.84. (See Appendix 1 for the list
companies,
98
usable
questionnaires
were
of items included in this measure.)
returned (approximately 3.77 per firm). Combin-
Agreement-seeking was defined as the degree
ing the samples resulted in an overall response
to which TMT members worked together in order
rate of 45 percent across both subsamples.
to achieve consensus or reach agreement on stra-
The questionnaires consisted of a group of
tegic issues. It was also measured using a five-
demographic and process-oriented questions, and
item, Likert-type scale where one represented
a series of questions designed to assess various
‘definitely not true’ and five represented ‘defi-
aspects of each executive’s mental model of firm
nitely true’. The questions included items such
strategy. The respondents were members of the
as, ‘When making decisions, the TMG works
company’s TMT, which included the CEO. The
hard to reach a decision’, and ‘TMG decisions
unit of analysis was the TMT.
are not final until all members agree that the
decision is acceptable to them’. Cronbach’s alpha
for this six-item scale was 0.79. (See Appendix
Demographic diversity variables
1 for the list of items included in this measure.)
Four measures of TMT diversity, drawn from
To assess the appropriateness of aggregating
prior studies in upper echelons theory, were used
individual responses to the team level, we
in this study. To measure functional diversity,
employed two techniques recommended by other
team members were asked to indicate the func-
researchers. First, the James, Demaree, and Wolf
tional category that most closely represented their
(1984) r(WG(J)) procedure was used. This pro-
background; functional diversity was calculated
cedure produces a measure of agreement among
in terms of Blau’s (1977) heterogeneity index:
respondents and ‘it provides the justification for
(1 − i2), where i is the proportion of the group
aggregation’ (Koslowski and Hattrup, 1992: 162).
in the ith category. A high score on this index
The r(WG(J)) for the interpersonal conflict scale
indicates variability in the functional responsi-
was 0.97 and the r(WG(J)) for the agreement-
bilities among team members or functional diver-
seeking scale was 0.96, both indicating high lev-
sity, while a low score represents greater func-
els of agreement within teams, thus suggesting
tional homogeneity.
that aggregation is appropriate. Second, using the
Age diversity was computed as the coefficient
technique recommended by Danserau, Alutto, and
of variation in age of the members of the team.
Yammarino (1984) we performed a within-and-
Educational diversity was computed as the coef-
between analysis (WABA). Generally, the ratio of
ficient of variation of the number of years of
the calculated eta-squared between to eta-squared
postsecondary education across team members.
within should be greater than 1 in order to justify
Employment tenure diversity was the coefficient
aggregating individual level responses. The
of variation on the number of months each team
ratio for the interpersonal conflict scale was
member had been employed by their current com-
1.046 and for the agreement-seeking scale was
pany.
1.009. Therefore, our data met the criteria for
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---

Diversity, Group Process and Consensus
453
aggregation using two commonly accepted tech-
ations were summed within each team. Since we
niques.
had named the variable ‘strategic consensus’,
which reflects agreement, we reversed the direc-
tion of this score by multiplying by −1. Thus,
Strategic consensus
for the final metric, high scores implied consensus
Strategic consensus was an outcome measure of
(i.e., high similarity among TMT members’ men-
similarity among TMT members’ interpretations
tal models), while low scores implied lack of
about the firm’s strategic orientation. In essence,
consensus. This measure, that we called ‘strategic
the measure attempted to determine the degree to
consensus’, was an indirect estimate of the degree
which the top management team shared a com-
to which top management shared a mental model
mon mental model with regard to the current
with regard to the strategy of the organization. It
strategy of the organization.
is important to note that the measure was not an
The strategic consensus variable was created
objective measure of the actual strategy, but
from
responses
to
48
Likert-type
questions
rather was a surrogate measure of the extent to
regarding firm strategy asked of each participant.
which the top management team possessed a
Our conception of strategic consensus was very
shared viewpoint—a common ‘mental model’—
broad and informed by the strategic management
about the strategy.
literature (e.g., Miles and Snow, 1978; Porter,
1980) and TMT interviews. We used a very broad
Data analysis
conceptualization to capture the great variety of
viewpoints in the firm’s strategy that are rep-
Structural
equation
modeling
using
LISREL
resented in both the conceptual and empirical
(version 8.12a) was used to determine whether
strategic management literature and in the view-
the pattern of relationships observed among the
points of executives. From strategic management
variables was consistent with any of the four
scholars, such as Porter (1980), we included the
alternative models proposed above. The alterna-
firm’s emphasis on costs, price, market differen-
tive models examined in this research represented
tiation, and product/service differentiation. From
a set of fully nested models. Since we were
Miles and Snow (1978), we added the firm’s
interested in determining if one of the four alter-
emphasis on risk, innovation, and proactiveness.
native models was superior to the others, struc-
Research of MacMillan, McCaffrey, and Van
tural equation modeling was particularly appropri-
Wijk (1985) and Chen, Smith, and Grimm (1992)
ate for this research. Unlike some other analytical
provided dimensions of competitive timing of
techniques, structural equation modeling allows
firm’s strategic actions. From Eisenhardt (1989)
for assessment of the fit of the data to the
we included the speed of the firm’s strategic
hypothesized model(s), and, especially important
behavior. Interviews with executives participating
for us, a comparison of fit among alternative
in the study provided additional insights into
models can be made.
strategic variables. Examples of these items
Joreskog and Sorbom (1993: 26) suggested
included the extent to which the firm emphasized
using the following formula to compute the mini-
total quality, product specialization, customiza-
mum sample size for estimation of the asymptotic
tion, branding, after sales support, and an accentu-
covariance matrices:
ation on strategic change. Each item was an-
swered using a five-item Likert response format
k (k − 1)
with anchors ranging from one to five, where one
2
indicated ‘definitely not true’ and five indicated
‘definitely true.’ (See Appendix 2 for the list of
where k equals the number of variables. The
items included in this measure.)
maximum number of variables in this research
The computation of the final variable was gen-
was 8, resulting in a recommended minimum
erally consistent with the previous work of Bour-
sample size of 38—substantially smaller than our
geois (1980) and Dess (1987) and was done as
final sample size of 76 cases.
follows. First, within each team, a standard devi-
An
additional
underlying
assumption
of
ation was calculated from the responses for each
maximum likelihood estimation is that all vari-
of the 48 items. Next, these 48 standard devi-
ables are normally distributed (Hayduk, 1987;
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---

454
D. Knight et al.
Table 1.
Means, standard deviations, and intercorrelations (N = 76)
Variable
Mean
S.D.
1
2
3
4
5
6
7
8
1. Interpersonal conflict
2.401 0.533
1.00
2. Agreement seeking
3.246 0.538
−0.62** 1.00
3. Strategic consensus
−36.397 6.463 −0.11
0.026*
1.00
4. Location
0.737 0.443
−0.03
0.00
−0.07
1.00
5. Functional diversitya
0.175 0.133
0.26* −0.12
−0.024* −0.04
1.00
6. Age diversity
0.157 0.068
0.046 −0.0185†−0.13
−0.17† 0.19† 1.00
7. Educational diversityb
−0.469 0.252
0.03
−0.06
−0.21* −0.19* 0.03 0.12 1.00
8. Employment tenure diversity
0.600 0.307
0.13
−0.09
0.02
0.32** 0.18† 0.09 0.01 1.00
†p
0.10; *p
0.05; **p
0.01
aThis variable was transformed using a power transformation. The descriptive statistics are pre-transformation.
bThis variable was transformed using a natural log transformation. The descriptive statistics are pre-transformation.
Tabachnick and Fidell, 1996). To address this,
which was included to test for systematic differ-
we screened all variables and found that two
ences between the U.S. and Irish samples, was
variables did not satisfy this condition. In accord-
significant, indicating that the level of consensus
ance with standard procedures (Hayduk, 1987;
was lower for U.S. firms. The squared multiple
Tabachnick and Fidell, 1996) we did transfor-
correlation (analogous to an R2) for this model
mations of these two variables. Education diver-
was 0.17.1 However, the fit statistics for this
sity was transformed using a logarithmic function
direct effects model (summarized in Table 2)
and functional diversity was transformed using a
indicated that the data did not fit the hypothesized
power transformation. Following the transfor-
model well. Therefore, although several of the
mations, both met all conditions for normality.
relationships were significant and in the expected
Finally, in order to test for systematic differ-
direction, the ‘direct effects’ model did not rep-
ences between the two subsamples, location was
resent the data well.
included as a control variable, and was entered
We next addressed the question of whether
as a dummy variable with U.S. companies coded
including intervening group process variables
as one and Irish companies coded as zero.
would increase explanatory power. The second
group of models posited that the effects of demo-
graphic diversity on the level of strategic consen-
RESULTS
sus would be mediated, either partially or fully,
by interpersonal conflict and agreement-seeking.
Descriptive statistics for the variables in this
Accordingly, we examined a set of fully nested
study and the correlation matrix of transformed
models with respect to the effects of TMT diver-
variables are provided in Table 1.
sity and group processes on strategic consensus.
The first research question was how well demo-
LISREL facilitated the examination of these alter-
graphic diversity alone explained the level of
native tests of the research question by allowing
strategic consensus within the TMT. Four meas-
comparisons
of
fit
between
the
competing
ures of demographic diversity were used with
hypothesized models. Three mediated models are
location included as a control variable, and the
shown in Figures 3–5. For clarity of inter-
results for the direct effects model are presented
pretation, only the significant paths were indicated
in Figure 2. Although all relevant paths were
although all relevant paths were tested.
tested, only those paths that were significant were
Figure 3 represents a partially mediated model,
included in the figure. Of the four measures of
in which both direct and indirect effects of all
demographic diversity, functional diversity, edu-
the variables on strategic consensus were allowed.
cational diversity, and employment tenure diver-
sity were significant, and with the exception of
1 For all models, we tested alternative calculations of the
employment tenure diversity the direction of the
measures (i.e., standard deviation instead of coefficient of
relationships was negative, as expected. Location,
variation) and the results were similar.
Copyright © 1999 John Wiley & Sons, Ltd.
Strat. Mgmt. J., 20: 445–465 (1999)

---