Strategic IT: Best Practices for Executives
mudassir1219
casestudy2.1article.pdf
• Acodemy of Managemanl JoumaJ 1988. Vol. 31. No. 4. 737-770.
P O L m C S OF STRATEGIC DECISION MAKING IN HIGH-VELOCITY ENVIRONMENTS:
TOWARD A MIDRANGE THEORY
KATHLEEN M. EISENHARDT Stanford University
L. J. BOURGEOIS III University of Virginia
How do politics affect the strategic decision processes of top executives in the "high-velocity" microcomputer industry? We induced a midrange theory linking power, politics, and performance from a study of eight firms in the microcomputer industry. We found that politics—the observable, hut often covert, actions by which executives enhance their power to influence decisions—arise from power centralization. Auto- cratic CEOs engage in politics and generate political behavior among subordinates. We also found that politics are not organized into shifting and temporary alliances based on issues. Rather, they are organized into stable coalitions based on demographic characteristics such as age and office location. Finally, politics within top management teams are associated with poor firm performance.
Most strategic decision processes are ultimately political in that they involve decisions with uncertain outcomes, actors with conflicting views, and resolution through the exercise of power (Allison, 1971). However, not all strategic-decision-making processes evidence politics. Pettigrew's (1973) study of decision making within a British retail firm contained numerous examples of political activity including agenda control, withholding informa- tion, and behind-the-scenes coalition formation. Politics played a major role (Pettigrew, 1973). In contrast, according to Allison's (1971) description, in the Cuban missile crisis the actors relied on open and forthright discussion, with full information, in group meetings to influence decision making. Although there was substantial conflict, there was little evidence of the be- haviors which we have defined as politics.
The purpose of this study was to explore the politics of strategic deci- sion making. We defined politics as our informants did. Politics are the
We would like to thank our graduate assistants. Theresa Lent, Mike Boehm, Anita Callahan, Dave Ellison, and Paul McCright, for their invaluable help. We also benefited from the com- ments of Howard Aldrich, Deborah Ancona, Graham Astley, Jim Dean, [ane Dutton, George Huber, John Kimberly. |im March, Elaine Romanelli, Dick Scott, Bob Sutton. our reviewers, and the participants at the Institute for Management Science Workshop on Organizations, Industrial Engineering Department Seminar, and National Institute of Mental Health Golloquium at Stanford.
737
738 Academy of Management /ournaJ December
ohservable, but often covert, actions by which executives enhance their power to influence a decision.^ These actions include behind-the-scenes coalition formation, offline lobbying and cooptation attempts, withholding information, and controlling agendas (Pettigrew, 1973; Pfeffer, 1981). Politics contrast with the straightforward influence tactics of open and forthright discussion, with full sharing of information, in settings open to all decision makers.
The research setting was the high-velocity environment of the microcom- puter industry. By high velocity, we mean those environments in which there is rapid and discontinuous change in demand, competitors, technology, or regulation, so that information is often inaccurate, unavailable, or obso- lete {Bourgeois & Eisenhardt, 1988).^ During our study (1984-85), the micro- computer industry was undergoing substantial technological change such as the introduction of UNIX, 64K RAMs, and RISC computer architecture, as well as competitive change such as the entry of IBM, decline of Texas Instruments, and double-digit growth in demand (Bell, 1984). Thompson (1967) hypothesized that political processes would accelerate in such dy- namic conditions, yet our previous research (Bourgeois & Eisenhardt, 1988) indicated that politics were associated with poor performance in such environments.
We organized our research around three questions. First, why do politics emerge? Our interest was in why politics emerge among some executive groups, and not others. Do they arise from ambition, rivalry, or attempts to gain personal advantage (Schumpeter, 1934)? Or are they the natural out- growth of conflict between functional units (March, 1962), centralization (Hage, 1980), or power imbalances within a top management team (Bachrach & Lawler, 1960)?
Second, what is the shape of politics? Some authors have suggested tiiat politics are organized into warring factions, each engaging in a power strug- gle for supremacy (Butcher, 1988; Jay, 1967). If so, how do individuals choose allies, and how do coalitions evolve? Alternatively, others have suggested that politics are fluid (Stevenson, Pearce, & Porter, 1985). Individuals vary their political tactics from issue to issue, forming alliances based on payoffs and preferences. However, that view assumes people have the time to under- stand their own preferences, ascertain those of others, and engage in politics.
' There are many definitions of poiitics (Gandz & Murray, 1980). Although our definition is narrower tban some, tbis precision bas advantages: our definition (1) captures tbe meaning of politics common in organizations (Gandz & Murray, 1980) and among our respondents, yielding better empirical validity; (2) distinguishes politics from related concepts sucb as conflict and power; (3) makes no empirically unobservable assumptions about tbe intentions of actors: and (4) captures meaningful differences in strategic decision making behavior across executive groups.
^ The term, high velocity, refers to environments in wbicb there is dynamism (Dess & Beard, 1984) overlaid by sbarp and discontinuous cbange sucb as changing competitors, tecbnology. or government regulation (Sutton, Eisenhardt, & jucker, 1986). Using tbis definition, microcomputers and probably airlines are high-velocity industries. In contrast, altbough they score bigh on dynamism indices (Dess & Beard, 1984), cyclical industries sucb as machine tools are not.
1988 Eisenhardt and Bourgeois 739
Is such extensive cognitive processing realistic to assume, especially in high- velocity environments?
Third, how do politics affect firm performance? One view is that politics interfere with effective management (Jay, 1967). Empirical studies have indi- cated that this view is common in organizations (Gandz & Murray, 1980). However, other authors have taken the more balanced view that politics may be harmful in some situations and helpful in others (Pfeffer, 1981; Stevenson et al., 1985). However, the empirical evidence for all of these positions is limited (Pfeffer, 1981).
The conflicting answers to these questions and the limited research base on politics, especially regarding executive teams in fast-paced environments, led us to an inductive study. We found that politics arise from power centralization. Domination by powerful chief executive officers (CEOs), com- bined with the desire for control by top management teams, leads to politics. Conflict, although necessary, is not a sufficient condition for the emergence of poiitics. We also found that politics are not fluid. Rather, they become entrenched into stable patterns that are often based on characteristics such as age and office location. Finally, politics, because they restrict informa- tion flow and are time-consuming, are associated with poor firm performance. The empirical derivation of those ideas is the subject of this article.
RESEARCH METHOD
We studied the politics of top management teams in their natural setting, by means of what Yin (1984) termed multiple case design. Such a design allowed us to follow a replication logic (Yin, 1984). whereby multiple cases are treated as a series of experiments, each case serving to confirm or disconfirm the inferences drawn from previous ones.
Our study included 8 firms from the microcomputer industry in the San Francisco Bay Area, Each firm was functionally organized and privately held. We used theoretical sampling to determine our number of cases—that is, we stopped adding cases when our incremental learning diminished (Glaser & Strauss, 1967; Sutton & Callahan, 1987). We contacted 12 firms to obtain the 8 studied.
Data Sources
We interviewed every member of each top management team, including CEOs and their immediate subordinates. The teams typically included the heads of major functions such as sales, engineering, and finance. We relied on four data sources: (1) an initial interview with a firm's CEO, (2) semi- structured interviews with every member of a top management team, (3) a questionnaire completed by each member of the team, and (4) secondary source data.
CEO interview. The entry interview with the CEO of each firm had a semistructured format. We began the interviews by asking the CEOs to de- scribe the competitive strategy of their firm and its position within the in- dustry. We then asked them to describe the distinctive competencies of the
740 Academy of Monagement /ournal December
firm and key success factors in the industry. They then described their major competitors and rated the competitors' performance and their own firm's performance. After asking the CEOs to identify two or three recent or ongoing major decisions, we selected one decision to study in depth. We chose deci- sions that (1) involved strategic positioning, (2) had high stakes—outcomes the executives believed would have significant consequences for their firms, (3) were pervasive, involving as many functions of a firm as possible, and (4) represented the process by which a firm had made other major decisions. We then traced the making of the chosen decision from the perspective of every member of the top management team through semistructured interviews, described below. Table 1 lists each decision and gives descriptive statistics for each firm.
Team interviews. After our initial interview with a firm's CEO, we con- ducted semistructured interviews with every executive in the top manage- ment team. The interview consisted of 16 open-ended questions. Following the methods of inductive research, we supplemented these questions with ones that seemed fruitful to pursue during the interview. The interviews typically took 90 minutes, but occasionally took as long as three hours. We divided the interview into two parts. In the first part, we gained a general impression of the firm and the political climate surrounding the top manage- ment team. In the second part, we traced the decision that we had identified with the CEO.
Each interview began with our asking for a description of the firm's competitive strategy and its position within the industry. We then asked the
TABLE 1 Descriptions of the Eight Microcomputer Firms Studied
Number of Number of Firm Employees Informants Decision Stndied
6 Name cbange: Do we need a new name?
5 New product: Sbould we develop an IBM-compatible product?
8 New product: What should our next product be?
6 Alliance: Sbould we form a strategic alliance?
9 Strategy: Do we need a new strategic direction?
6 Strategy: Do we need a new strategic direction?
7 New product: What sbould our next product be?
7 Alliance: Sbould we form a strategic alliance or go "public"?
First
Alpha
Cowboy
Neutron
Omicron
Promise
Forefront
Zap
115
50
417
200
192
IBS
go
500
1988 Eisenhardt and Bourgeois 741
executive to describe (1) the firm's distinctive competencies, key industry success factors, and the functional strategy of his or her area within the firm; (2) colleagues, in whatever terms came to mind—descriptions of personality, skill levels, and functional background resulted; (3) the frequency and na- ture of interactions with each colleague; and (4) any routine decision-making sessions, in terms of conflicts, consensus, cordiality, and so forth. The last item provided us with a sense of the culture of the top management team.
In the second portion of the interview, we traced the history of the strategic decision identified in our entry interview with the CEO. We traced the decision from the perspective of every participant, using a standard set of interview questions. We asked questions, which concentrated on facts and events, using standard courtroom interrogation (e.g., "When did this first become an issue?"}.
Each interview was conducted in tandem—by two investigators—with one primarily responsible for the interview and the other for taking notes and filling in gaps in the questioning. Immediately after an interview, the investigators cross-checked facts and recorded their impressions. We fol- lowed several rules for within-case analysis (Yin, 1984). The "24 hour" rule required that detailed interview notes and impressions be completed within one day of the interview. A second rule was to include all data, regardless of its apparent importance at the interview. A third rule was to end our inter- view notes with our own ongoing impressions. We tried to sharpen those impressions by asking ourselves questions (e.g., "What did I learn?" "How does this interview compare to prior interviews?").
Questionnaires. We also obtained quantitative data on political patterns within each top management team from questionnaires introduced during interviews. The questions focused on variables from research on politics (e.g., Hinings, Hickson, Pennings, & Schneck, 1974). We measured goal conflicts, alliance formation, interpersonal disagreements, and power. From these data, we developed power maps and ailiance network diagrams for each executive team. The Appendix describes the questions, their admini- stration, and computation of the variables.
Secondary-source and other data. We examined available industry re- ports and internal documents and made informal observations. We obtained data on office location, demographics of team members, and each firm's financial performance. Finally, we also observed a day-long strategy-making session in one firm.
Data Analysis
Unlike hypothesis-testing research, inductive research lacks a generally accepted model for its central creative process. In the absence of a standard, we used the following approach: after collecting both qualitative and quanti- tative data from each firm, each author independently analyzed one of the data types. For eacb firm, one author calculated group-level scores of conflict, power, alliance formation, and so forth. That author then analyzed those
742 Academy of Management Journal December
data for patterns. The other author combined qualitative responses. We de- veloped profiles of each executive from the descriptions given by each member of the top management team, including traits mentioned by more than one executive in the profiles. For example, Don, the CEO of Alpha Computer, was described as "impatient" by three of his four colleagues. We thus included this trait in Don's profile, whereas other traits that only one person mentioned (e.g., "large ego") were dropped. This approach was also used to profile the decision climate.
"Decision stories" were developed by combining the accounts of each executive into a time line beginning with decision initiation. We included all events mentioned. In each firm, team members agreed on the critical issues of when the decision began, when it was made, and how it was made. For example, at Alpha the executives all agreed that the impetus for the decision was a board meeting, that the CEO made the decision alone, and that he did so just before the annual planning conference. Also, all Alpha executives, including the president, agreed that the decision was unpopular. Although they were few, conflicting reports were preserved. They usually concerned one person's assumptions about another's motives or opinions, not observable actions and events.
Once each of us had developed preliminary hypotheses from our respec- tive data sets, we exchanged analyses and searched for patterns in the data. That search was assisted by taking pairs of firms and listing similarities and differences between each pair. From this pairwise comparison, we induced tentative relationships between variables. We then went back to each case to see if the relationship was confirmed, and if it was, to use the case to develop a better understanding of the underlying dynamics. After many iterations between data and propositions, we used comparisons with existing litera- ture to sharpen our insights. What emerged were propositions linking power centralization, politics, and economic performance. As in deductive research, our propositions fit well with the evidence, but did not perfectly explain the cases (Sutton & Callahan, 1987).
THE ORIGINS OF POLITICS
Why do politics arise? Many authors have argued that the source of politics is conflict (Baldridge, 1971; March, 1962; Pfeffer, 1981). Absent conflict, there is no need for people to use politics to influence decision making. Many authors have also argued that politics arise when power is decentralized (Dean, Sharfman, & Ford, 1987; Hage, 1980; Pfeffer, 1981). When the power of individual actors is roughly equivalent, individuals band together to influence decision processes. Conversely, when power is highly centralized, conflict is submerged and the use of politics declines (Pfeffer, 1981).
The evidence from our data tells a different story. Although conflict was important, power imbalance was crucial. Specifically, the use of politics was closely linked to centralized, not decentralized, decision making. The more powerful a CEO, the greater the tendency among remaining executives to
1988 Eisenhardt and Bourgeois 743
consolidate power and engage in alliance and insurgency behaviors, while the CEO engaged in tactics for controlling and withholding information. In formal terms.
Proposition 1: The greater the centralization of power in a chief executive, the greater the use of politics within a top management team.
Table 2 summarizes the distribution of power within the top management teams of our firms. Consistent with the methods of others (Baldridge, 1971; Patchen, 1974; Pfeffer & Salancik, 1974), our measures of power emphasized influence on major decisions and used multiple indicators. We quantita- tively measured CEO power on a 0 to 10 scale rating influence on ten key decision areas. We also measured the power of the next most powerful executive and the number of functional areas in which a CEO, rather than the functional vice president (VP), was the principal decision maker on the team. We qualitatively assessed CEO power using descriptions of CEOs, descriptions of the decision climates and team interactions, and the stories of the decision-making processes by which specific strategic decisions had been reached.
Table 3 summarizes the use of politics in each team. We quantitatively measured politics using a questionnaire item on the frequency of ailiance behavior between pairs of executives and qualitatively assessed the inci- dence of politics by searching the interview data for behaviors by which executives tried to influence decision making indirectly, covertly, or, as Pfeffer (1981) described, unobtrusively. We distinguished between the use of evidence and full disclosure of information in open meetings, and the tactics of politics—observable, but often covert, actions by which executives en- hanced their power to influence decision making. Consistent with our respondents, we considered behaviors such as formation of insurgency groups, internal and external alliances, withholding information, agenda control, and private attempts to coopt or lobby key executives to be examples of politics. The political tactics we identified were similar to those identified by Pettigrew (1973) and Pfeffer (1981).
The results shown in Tables 2 and 3 indicate that there is wide variation in the use of poiitics across top management teams and that the variation is closely related to the centralization of power. In fact, the first four firms listed—First, Alpha, Cowboy, and Neutron—scored highest on both power- centralization and politics measures. We refer to those firms throughout this article as the politically active firms. By contrast, the second four firms, Omicron, Promise, Forefront, and Zap, scored lower on those measures.
For example, the chairman of First had the highest power score of our CEOs, 9.6 on a 0 to 10 scale rating influence on ten key decision areas. His distance on the power scale from the next most powerful person was 3.5, the second largest gap in the set of firms. The chairman was also the principal decision maker in every functional area. Anecdotes from our interviews cor- roborated the quantitative data. One executive told us: "When Geoff makes
744 Academy of Management Journal December
o BU
p Ed U
S a01 o u
go
£S
o B) S
b.s- 8s
.3 13
^ t- ^ !- g p_ - ^ * — -3 _ ; ^ CO « * to -5 to 5 — C a> > > > o n
2 B s o
I = 2 5 O (B
..-' o "^ 0 5
£ - •
b o
o to e io ^ - 2 - :t: .2 a. .2
— 0) -S
-2 £ s
S 5 c w ^ • - !
^ ..£ tl.
oors 2
i o o c/3 > D
gg- a; T3 C O
I
ag E S 3
, 5 Q. E-
n
11 3 a
_• "
O % to
CC G. 5
J
^
3 CO c
- ' O
•0 S uN .— 'E "^
DO C 0 DO >-.
0
K
o
1988 Eisenhardt and Bourgeois 745
V 3 .s "S ou M
a oa
H
vi si
on
Q
-o
n B 'E o
T 3
6 u
h
o U l
o u
L .
» 0
CL. IA
C E
O '
i n
b u u
J 'o.
fi
I f l
'S u s
«
1"5
u 0 u
c/3
a a
Q
F ir
m -— .5 o. >> VI to
a o 0)
O u O 1 1 a
s . •i3 CL, a> >
'•C tn o — 00 "E i- 3 5 ^
1-2 M (fl ,£) (U
2 u
ve .
'Zi 3
0 )
X <u
rf ul
1 o
t m
os t
>< 0) B tD
"o 01
O u<fl «
T3 B to 03 I-. ou n U
)'s p
o w
u to
a
X )
ff er
en c
x: a
B -2 "to
O
u >•
-a to
a
;i on
s
u -o II a CO
U th
o
<
u
or i
Ifl
E
si o
' o th
e d
l
c .^ c o CO
*2
., ,o If)
ba s
•s ao a
in g
st yi
on -m
ak i
(n u
- D
te go
ri zi
CO U
S U
S
E
E O
U
G CO a;
o
CQ
o I f l
o
O •a
2 3 (nao u
.s O
by C
E
0 )
-o CO
g (O
O
is i
u -a
ll
iv e
CO
a
a B O tn 1.. 0)
y on
e p
B O
w it
h ta
ti o
n
D (A
ou B
O
g ro
u p
.
to (fl a:
03
en ti
r id
e by
B t f l E
.2 u o;
T !
a
jr at
i
a o
E
0 (fl VI OJ U
ro p ro
y G 03
•£ 0 0)
i u
ha d
o; J2
ti
uc t
-oo o.
D m
pu te
r
(_> is E CO
l l O U l
a _s "H. p.
o"Q)
T 3 o
o
ua l
st y
] co
ns en
s
to
-a
t us
e es
id en
a. o>
JS H o
-2">- t f l
E
ri ta
ri 1
au th
o
c CO
B
E aj t f l > ^ I f l
746 Academy of Management /ournaJ December
s SI
— ea S 3 x 3
1-3 O B
15
HS "3 "^
H >
•£ c
•5 "^ -A — g o > « n:
5 :D >
X CO O.
.> o
Ifl 03
o
c a
05 J^ .Ŝ
0)
•fi o *- p. y
^-IS:
'c -a " S. tfl- >-
E .2 3 o ~ a U CD >
«.j 4 _ . CO - ^ ^ (f3 0) ^ OJ O 00 O
S o ' c § *i „> O B < .2 T! £S
CO
-G Q.
<
O .n o
U
a
tr o
3 V
2
1988 Eisenhardt and Bourgeois 747
c ou
CO
3 oa
o .2 •2 Qa o s^ o V
e u tg e
03 BO
M G
o • • =
S E 00 ~
0) QJ <U XI C 3 -̂̂ O c >̂
> • XI
—
V O au. >' tn
.1 g p a o o E -53
^ •£
X 00 OJ B
on 1^
CO 03 5 — - A H to DO S ^ u G 2 c
••5 - C o- .2
o p G G •a g
P-'
• " oo -t; F G S
t f l
to
-a '> -5 G
ti pl
m u!
E o
0)
c T3 ' >
1 c 0)
• £
B o
K o a. a p E/1
st ro
n g
:
^ ^
CO
3 >
c H)
in g!
tn CO
E o
OJ u G
id e
>
2i £
E o
•c oo. a 3 Cfl
m o
d es
t
-3C 01
It ip
li
p E B o i ~
01 CJ
B ^
X I
[e vi
L l O
B
is io
d ec
u •5
E o
•c o a o. 3 crt
st ro
n g
1 •a 3
-o iv
i -a G ta
li ng
l
CO
ro m
OJ u G 01
w
?o tn
a o
ci si
OJ
-a
E o •c 0 a a 3 t f l
m o
d es
t
II II II II — — W3 w
748 Academy of Management /ournal December
a decision, it's like God." Another said: "Geoff is the decision maker. He runs the whole show." A third said; "Most decisions around here are made at the top." One executive used understatement to sum up the power of the chair- man, remarking "This is not a democracy."
As indicated in Tahle 3. there was extensive use of politics at First. The company ranked at the top of the eight cases in the use of alliances, which we defined as hehind-the-scenes coalition building among proponents of a particular view. First's executives relied heavily on such handing together and even held "outlaw staff meetings" (the executives' term), in which the VPs of marketing, finance, and operations met regularly outside the formal chain of command. The meetings circumvented the chairman and "[kept] issues ironed out hetween us."
First's VPs also reported using cooptation, which we defined as private attempts to change the position of a key decision maker. The VPs of market- ing, finance, and manufacturing all told us that they tried to influence deci- sions through one-on-one meetings with the president, who sat hetween them and the chairman on the organization chart. Politics also surfaced in the strategic decision that we studied at First. Several of the VPs tried to coopt the president, first to change the decision and then to delay its implementation. For example, the VP of finance told us: "I went to Bob [the president], who said his hands were tied." The VP of marketing said: "I tried to slow down the implementation as much as possible. . . . I talked to the president ahout that." This same individual then went on to describe how he tried to form an alliance with external consultants to push his view: "I also had the consultants make a presentation to the chairman. They legitimized our view of the situation."
Overall, the atmosphere at First was one of frustration. The VP of sales descrihed the situation well: "The atmosphere is frustrating. Geoff goes off on a lot of tangents. He tends to have already made up his mind, and we spend time just trying to change his mind. I get frustrated because he doesn't really listen when you talk to him." We did not find evidence that First executives enjoyed politics. Rather, the executives felt that politics were necessary to counter the power of the chairman and to get things done. For example, the rationale for the outlaw staff meetings was descrihed as: "It takes the place of what we think should happen at weekly staff meetings." Another executive summed up the rationale for cooptation: "You HAVE to build a case through him [the president] and have it presented to the chairman." The VP of marketing described the reason for his use of external alliances: "The consultants legitimized our view of the situation. We wouldn't have been able to do it without them. We would have been labeled foot- draggers." Most felt like the executive who told us: "Sometimes I wish that Geoff would go away for six months!"
Alpha is another example of a firm with a powerful leader and an execu- tive team engaged in poiitics. The CEO (the president) was descrihed as a "parent" and "benevolent dictator." His power score was 9.6, tied for high- est in our cases. The next most powerful executive at Alpha scored only 5.8.
1988 Eisenhardf and Bourgeois 749
The strategic decision we studied corroborated those data. For example, the VP of sales said of the decision process: "The decision was a Don Rogers edict—^not a vote." The president agreed: "1 made the decision myself, de- spite the objections of everyone. I said 'the hell with it, let's go with the PC interface.' "
We also found strong evidence of politics at Alpha. For example, several of the VPs described others as "political," "secretive," and "close to the vest." This secretiveness was perceived as jockeying for power through the control of information (Pettigrew, 1973). As one VP told us: "Rich was trying to get Don's ear and since I had been close to Don for a long time, he would skip me." The VP of R&D descrihed the VP of sales in the following way: "He does all the games and the hidden agendas. He needs power." Another ex- ecutive told us of attempts to coopt the president by everyone. He explained: "We [the top management] are all trying to establish a pecking order among the VPs."
Frustration was evident at Alpha. For example, the president frequently withheld information. As one executive told us: "Don collects a lot of information, but he often doesn't share it. When he does share information, he prefers to give out partial information in one-on-one meetings." The VPs reacted with their own brand of politics: outlaw staff meetings. As one execu- tive said: "We started having outlaw meetings where we would meet with- out Don to share information. It appeared like we weren't getting a lot of information as a group—Don was distrihuting information piecemeal to each person." The outlaw meetings were effective. This individual continued: "One result of the meetings was that we gave Don a list of the things we needed. We got some things—we even got regular staff meetings."
By contrast, we found less centralized power along with less politics at Omicron, Promise, Forefront, and Zap. As Tahle 3 indicates, these top man- agement teams reported less alliance behavior than did the politically active teams. Our qualitative data did not reveal one-on-one attempts by executives to coopt key individuals. Neither did we see alliances with external individu- als such as consultants. Nor was there withholding of information by the CEO or by others. We saw no outlaw meetings or other attempts at insurrec- tion. In short, we saw little evidence of politics in the teams with decentral- ized power.
What we did see in those teams was evidence that their CEOs shared power and information. For example, the CEOs were described hy phrases like "consensus style," "people-oriented," and "team player." As Tahle 2 indicates, the power scores of the GEOs of these firms are lower than those of the CEOs of the politically active firms, and the distances between them and the other executives on the power scale are much smaller. Although the CEOs are still powerful executives, their functional VPs are usually the key deci- sion makers in their own areas (see Table 2}.
Our decision stories corroborate this. The Omicron decision to reassess strategic direction was conducted primarily in group meetings spread over several months, providing many opportunities for top managers to contribute.
750 Academy of Management Joumal December
The Forefront new-product decision was a consensus decision made in a team meeting. Promise also had a consensus decision made over several months and finalized at a multiple-day retreat attended hy all top managers. The Zap alliance decision was made by the CEO in consultation with the VPs of sales, finance, and engineering. The rest of the top managers were kept informed at weekly staff meetings and planned the implementation as it related to their functional areas.
Our cases suggest that executives in these politically less active teams regard politics as unnecessary. For example, one VP at Promise told us: "At most companies the early talk is to try to get allies, but not here—we don't have any superegos." The VP of manufacturing at Forefront said: "If you can present your ideas, people will change their opinions. We're not political at this point." These executives also regarded politics as unnecessarily time- consuming. For example, the VP of marketing at Zap told us: "1 don't have time to lobby, I hardly have time to make the meetings." The VP of sales said: "There isn't any premeeting politicking: there isn't any time. 1 just manage to get to the meetings." The VP of engineering at Zap summed up the prevailing view: "Lohhying isn't a good use of time."
The difference between our results, and those of other authors (Baldridge, 1971; Pfeffer, 1981) may be related to differences in setting: universities versus corporations. Other research has suggested that universities have un- usually decentralized power structures (March & Olsen, 1976). If this ex- treme decentralization exists, it may lead to the same frustration and inabil- ity to accomplish objectives without politics that we saw in our top management teams at the extreme of high centralization. Perhaps extreme decentralization so dilutes power that action is possible only through politics. Possibly then, politics emerge at the extremes of power imbalance—high cen- tralization and high decentralization—but do not emerge with moderate decentralization,
Our results also have similarities with the findings of studies of demo- cratic and autocratic leadership (e.g., Leavitt, 1951; White & Lippitt, 1960). Both our results and theirs suggest that autocratic leaders create frustration among their suhordinates. In those studies the frustration led to abuse and withdrawal among children (White & Lippitt, 1960) or to dissatisfaction among laboratory suhjects (Leavitt, 1951). Our executives, engaged in actual high- stakes decisions, manifested their frustrations through politics.
Conflict, Power, and Politics
Some authors (March, 1962; Mintzburg, 1983) have argued that the root of political behavior is conflict. Our data also indicate that conflict affects politics. For example, the decision at Neutron was less characterized hy politics than those at Alpha and First, at least in part because the Neutron decision was in itself less controversial (Tahles 2 and 4). By contrast, all executives at Alpha and First disagreed with the decision, and their use of politics was vigorous. However, our data also indicate that confiict does not necessarily result in politics. At Omicron, Zap, and Forefront, there was
1988 Eisenhardt and Bourgeois 751
extensive conflict. However, in each case, the conflict was resolved without politics. In more formal terms,
Proposition 2; Conflict is not a sufficient condition for the use of politics. Rather, confiict Jeads to poJitics onJy when power is centralized.
Table 4 summarizes the data grounding this proposition. We quantita- tively assessed conflict in terms of (1) disagreement on organizational goals, (2) disagreement on the importance of key strategic decisions, and (3) inter- personal disagreement among key executives. Qualitatively, we assessed the decision climate and tallied the agreement on the decision studied at each firm.
The data reveal that high conflict is associated with politics when a CEO is very powerful (e.g.. First), but not when a CEO shares power (e.g.. Zap). Forefront's executives are an example of a team experiencing relatively high conflict yet low politics. As Tahle 4 indicates. Forefront's conflict scores are moderate to high among the cases. Anecdotes from our interviews strengthen the quantitative data. One executive told us: "There is a lot cf disagreement. We air opinions and they're often heated. They're even abusive and insulting sometimes. . . . We argue about most things." Another VP said: "There is a lot of dehate. There is a lot of disagreement. . . . Art [the president) doesn't want yes people."
Despite this conflict, we saw little evidence of politics at Forefront. Rather, the executives seemed to operate using open argument. As the VP of manufacturing said: "You don't need to get the others behind you hefore the meeting. If you can explain your view [at a meeting], people will change
TABLE 4 Conflict on Top Management Teams
Finn
First Alpha Cowboy'̂ Neutron Omicron Promise Forefront Zap
Goal Conflict"
2.01 1.43 2.05 1.36 1.84 1.83 1.83 2.13
Policy Conflict''
1.78 1.87 2.39 2.14 1.58 1.99 2,06 2.58
Interpersonal Disagreement
4.1 3.3 3.7 3.3 3.2 3.1 3.3 3.8
Rank*̂
3 6 2 5 8 7 4 1
Decision Conflict''
Disagreement Disagreement Agreement Agreement Mixed Agreement Mixed Mixed
^ Sum of witbin-team variance on importance of eacb goal (see tbe Appendix). Sum of variances on importance of key decision areas (see the Appendix).
^ Rank was determined by computing tbe mean of a team's rank from the first three columns. Disagreement = final decision opposed by the entire team. Mixed = final decision opposed
by some on team. Agreement = final decision agreed to by the entire team. Altbough the Cowboy team scored bigb on the quantitative measures of conflict, tbere was bigb agreement on the decision tbat we studied in deptb.
752 Academy of Management /ournal December
their opinions. Forefront is not political at this point. But you must give your reasons or your ideas don't count." This theme of forthright discussion, instead of behind-the-scenes politicking, was echoed by the VP of finance: "There is some open disagreement—it's not covered up. We don't gloss over the issues, we hit them straight on." The firm's president said: "I prefer a 'stand up for what you think' approach."
The new-product decision we studied at Forefront corroborated the em- phasis on open discussion. New product introductions are particularly criti- cal in this industry because of the short product life cycles and the rapid changes in technology. Some executives favored an incremental product that could be rapidly introduced to counter the new product of a major competitor. However, other executives thought that such a product would cannibalize existing product revenues and dilute engineering resources. They preferred to introduce an innovative product, even though doing so would take more time. There was a series of meetings. The final decision was made through intense discussion at one such meeting. The VP of sales described the pro- cess: "The attitude is that if we disagree, we'll fight until someone changes his mind. It's painful." How was the conflict resolved? One VP described the decision as a push for consensus, followed by the CEO's decision: "When everybody sounded like they wanted it, Art [the president] said it sounds good. When he joins in, that's the consensus."
Our results at Zap and Promise echoed this approach to conflict resolution. What they describe is not simply consensus. Rather, one VP at Promise termed the approach "consensus with qualification." A team would try to reach consensus through an open airing of views, but if consensus failed, the CEO and often the relevant vice president would make the decision. Although the consensus with qualification process is political in the sense that the final decision is often a judgment by the most powerful individuals, the influence process employs open argument and full sharing of information, not politics.
Zap illustrates this process of conflict resolution. As one VP stated: "It's very open. . . . We're successful most of the time in building consensus. Otherwise, Randy [CEO] makes the decision." The decision we studied at Zap corrohorated their use of a consensus-with-qualification approach. The decision, concerning how to raise additional money, involved choosing whether the firm should go public or engage in a strategic alliance for an infusion of private funds. There was substantial conflict because of uncer- tainty about the stock market and the impact on recruiting key engineers. All the executives claimed that the entire team discussed the issue, with every- one participating and informed. One executive described the approach to the conflict as "open and forthright." Another said: "It's very open. . . . The meetings are fairly rigorous. We talk as a group, not committees—the meet- ings are intense and we usually walk out with a decision." The president made the final decision to form an alliance. Although some executives op- posed the decision, we saw no evidence of politics. Rather, as the VP of engineering told us: "1 think we should have gone [public]. But my position
1988 Eisenhardt and Bourgeois 753
is well known, and lobbying is not a good use of time." The VP of manu- facturing, who would have personally profited immensely in a public offering, told us: "I'm satisified just to bring the issue up." The VP of market- ing summarized conflict resolution at Zap: "We scream a lot, laugh, and then resolve the issue."
Why do our results fail to support the view that politics are driven by conflict [Cyert & March, 1963; Stevenson et al., 1985)? That view assumes that the primary way to resolve conflict over substantive policy issues is the use of politics. Such an assumption may be valid for legislative voting (Riker, 1962). laboratory games (Camson, 1961: Murnighan, 1978). and budget allo- cation decisions (Baldridge. 1971: Pfeffer & Salancik. 1974) from which politi- cal models have been developed. The structure of the conflict in each of those situations is a competitive game. There are winners and losers—what benefits one person harms another.
In contrast, in our cases the major decisions addressed survival in times of rapid change. The key decisions involved charting a strategy, changing to a new technology, and creating additional financing. These are not zero- sum, competitive situations as are many laboratory simulations [Murnighan, 1978) and budget allocation decisions (Pfeffer & Salancik, 1974). Rather, each decision was collaborative. It was in the interests of every executive to obtain the best possible decision because the decision was critical to the future, to even the survival, of the firm.
Where power was relatively decentralized, we found that the team main- tained a collaborative viewpoint. In effect, we found cooperative behavior focusing on group, rather than individual, goals. The comment of the VP of engineering at Zap was typical: "We emphasize a company view rather than a functional view most of the time." Although it is probably naive to assume that the executives were always able to do this, his statement nonetheless typifies the collaborative orientation. The executives argued, often strenuous- ly, for their views in an open forum and avoided politics.
Where power was centralized, we found competition among executives. For example. Alpha executives talked about "establishing a pecking order among tbe VPs." Dominant CEOs transformed a collaborative situation into a competitive one, and politics emerged as people competed for the time and attention of the CEO.
THE ORGANIZATION OF POLITICS
What is the shape of political hehavior? Political activity has often been described as organized into temporary and shifting alliances (Bachrach & Lawler, 1980: Gamson, 1961: March, 1962). The argument is that individuals form alliances around common points of view on a given issue in order to enhance their influence on the decision and the alliances disband when the issue is resolved. Further, Pfeffer (1981) argued that individuals try to make these alliances as large as possible in order to smooth implementation.
Our data, by contrast, suggest that coalition patterns are stable. Execu- tives do not shift allies as issues change, particularly in politically active
754 Academy of Management /ournal December
teams. Rather, they develop stable coalitions with one or possibly two other executives. They routinely seek out alliances with the same people. When usual allies disagree on an issue, they generally do not seek out more favorably disposed executives. Rather, they either drop the issue or pursue their interests alone.
Proposition 3: The greater the use 0/ politics within a top management team, the greater the likelihood of stable alli- ance patterns.
Table 5 summarizes the quantitative data that we used to ground our ideas on stable alliance patterns. The first column in Table 5 gives the vari- ance in alliance scores across the teams. Large variances indicate more stable alliance patterns: executives formed alliances frequently with some execu- tives and infrequently with others. Conversely, a low variance suggests shift- ing alliances, a pattern in which executives formed alliances with each other executive on their team with similar frequency. We also assessed the number of what we termed stable coalitions in each team. We considered an alliance to be a stable coalition when the mean frequency of alliance formation be- tween two executives was at least 7 on a 0 to 10 scale (see the Appendix). No executive was in more than one stable coalition. Finally, we calculated the percentage of top management executives who participated in stable coali- tions.
The data suggest that politically active top management teams are likely to be organized into stable alliance patterns. For example, the top manage- ment team at First was politically active and had a stable alliance pattern. All the executives at First could be identified with a stable coalition, and the variance score is the second highest among the teams. The VP of sales had a typical alliance profile for the top managers at First. He rated the frequency of his allying with the VP of marketing an 8 and the frequency of his allying
TABLE 5 Stability of Alliance Patterns
Firm
First Alpba Cowboy Neutron Omicron Promise Forefront Zap
Variance in Alliance
Scores
2.95 2.90 3.75 2.82 2.62 2.29 2.50 2.43
Number of Stable
Coalitions"
3 2 2 1 1 0 0 0
Percentage of Executives in
Stable Coalitions
100 80 85 30 33
0 0 0
° A stable coalition was defined as existing when the mean aUiance score between execu- tives was at least 7 on a 0 to 10 scale (see the Appendix).
1988 Eisenhard! and Bourgeois 755
with the rest of the team a 0 (0-10 scale, with 10 = constantly). His qualitative responses were consistent. Describing his reiationship with the VP of marketing, he said: "Often it's our building against everyone else. We are the bad guys."
In contrast, the teams at Zap, Forefront, and Promise were not only less likely to use alliances but were more fluid in their alliance patterns. Their variance scores were smaller, and none of their executives had formed stable coalitions. The qualitative data support those findings. For example, the VP of marketing at Zap told us: "It [the alliance pattern) really switches. Patterns just aren't that evident." She went on to say: "Actually I don't have a lot of time to lobby."
Why do stable aUiance patterns emerge? A possible reason, one consis- tent with the threat-rigidity hypothesis (Staw, Sandelands, & Dutton, 1981}, is that politically active teams have stressful cultures, leading executives to rely on habitual responses like stable coalitions. Our data indicate that stressful and even threatening cultures were prevalent in the firms with politically active teams. For example, the VPs of R&D and operations at Alpha were described as "burned out." Another executive spoke of being "sabotaged." At Neutron, meetings were described as "confrontive—Jim |CE01 beats us up when we don't meet our goals." At First, one executive described the decision climate as "violent." Another described the CEO as "a gun about to go off, but you never know in what direction he will fire." Several executives alluded to heing "caught in the crossfire." Another VP told us: "In fact, we used to joke about which one of us would get chewed out at a meeting. If he [the CEO] attacks, we don't get much done in the meeting. . . . It isn't smart to challenge Geoff in a group. He'll lash back." Perhaps then, stable alliance patterns are a rigidity response to the stressful and threatening environment of politically active teams.
A second reason for stahle alliance patterns may be that executives do not engage in the extensive cognitive processing necessary for ascertaining the preferences of others and planning political actions. Rather, consistent with bounded rationality arguments (Cyert & March, 1963), they make conve- nient assumptions about the opinions of others and fall into familiar patterns of politicking.
For example, as described earlier, all executives opposed the chairman's decision to change the name of First, with the VPs of finance and marketing being the most active in their opposition. However, they persisted in familiar hehaviors—that is, both tried to coopt the president and to engage their respec- tive coalition partners. However, the VP of sales, the coalition partner of the VP of marketing, was husy with sales channel problems, and the VP of manufacturing, the coalition partner of the VP of finance, simply decided that he was too busy to bother. The finance and marketing executives never did join forces. We asked why. The VP of marketing claimed: "This is more of a marketing issue—it didn't concern finance. I didn't try to get Tim [VP of finance] involved with me." The VP of finance responded; "Jon [VP of mar- keting] and I usually disagree. We have a very different view of the world."
756 Academy of Management fournaJ December
Neither executive was aware of the views of the other, and neither even considered the possibility that they might effectively ally.
The story at Alpha was similar. The VPs of R&D and sales particularly opposed the development of an IBM-compatible product that the president favored. However, as at First, these executives persisted in the routine use of politics. They tried to lobby the president and garner support within their usual coalitions. However, they did not attempt to cross coalition boundaries. We asked the VP of sales if he had considered allying with the VP of R&D on this issue. He replied: "It doesn't matter what he thinks—sales is necessary for the decision to go." The R&D executive said: "I didn't understand what Rich [VP of sales] was doing until the very end. We didn't talk much. I was too busy with what I was doing, and he probably wouldn't hear what I was saying anyway."
Finally, Cowboy's top management team consisted primarily of execu- tives who had worked together at another firm. When they arrived at Cowboy, they simply maintained the patterns established previously. The only Cow- boy executive who did not belong to a stable coalition had also not worked at the previous firm.
Why do our results fail to support the view (Bachrach & Lawler, 1980; Stevenson et al., 1985) that political activity is organized into temporary and shifting alliances? One reason may be environment. Rapid and discon- tinuous changes in technology, competition, and demand characterize the microcomputer industry. In such a fast-paced environment, executives sim- ply may not have the time to engage in the extensive cognitive processing necessary for forming fluid alliances. A second reason may be that the evi- dence for shifting alliances has been largely based on the results of laboratory studies of one-time-only coalition formation under conditions of perfect information about payoffs and preferences (Gamson, 1961; Murnighan, 1978). Such studies probably do not capture the information uncertainty and com- plexity of a real environment or the stressful culture of a continuously political team, both of which may lead executives to habitual behavior patterns.
Coalitions and Demographics of Team Members
The literature on politics has often argued that people will form alliances on the basis of agreement on issues (Gamson, 1961). Taking that view, we would expect executives to form alliances on the basis of issue-specific agreement. However, as discussed above, fluid alliances around issues did not form in the politically active teams. Rather, coalitions developed on the basis of demographic factors such as age, office location, similarity of titles, and prior experience together. Thus,
Proposition 4: When the use of politics is high, the basis of alliance is likely to he similarity of demographic attributes.
In Tahle 6, we have listed the stable coalitions, their demographic commonalities, and examples of the relationships between coalition members.
1988 Eisenhardl and Bourgeois 757
As indicated in the table, similarities in demographic attributes such as age, status, office location, and experience together occurred in all but one stable coalition. Age differences played a role in several of the stable coalitions. For example, the VPs of finance and manufacturing at First formed a coalition. They were both in their early 30s and were at least 15 years younger than the rest of the team.
Past history together is also an important factor. For example, as de- scribed above. Alpha had two coalitions: the VPs of R&D and operations,
TABLE 6 Demographics of Stable Coalitions^
Finn Commonalities Examples
First Chairman-president
VP. finance-VP, manufacturing
VP, sales-VP, marketing
Alpha VP, finance-VP, sales
VP, R&D-VP, operations
Cowboy VP, manufacturing-VP, human
relations-VP, finance
Director, software-director, marketing
Neutron VP, sales-VP, marketing
Omicron Chairman-VP. corporate
development-VP, strategic planning
Promise Forefront Zap
Adjoining offices Highest-status titles
School friends Youngest by 15 years Adjoining offices in
separate building
Adjoining offices Joined firm at tbe
same time Joined firm at tbe
same time
Same-status titles Prior experience
together Oldest Same-status titles Joined firm at the
same time Prior experience
together Youngest
Cofounders of Omicron Cofounders of prior firm School friends No stable coalitions No stable coalitions No stable coalitions
We talk about every thing and notbing, We're a good pair (president). We are good friends (VP, finance). We are tbe "bad guys" (VP. sales).
We talk about every thing (VP, sales).
Nice guy. Fits with me (VP, operations).
I like Jon, I under- stand wbat makes him tick (director, software).
Alter egos (VP, operations).
^ A stable coalition was defined as existing when tbe mean alliance score between executives was at least 7 on a 0 to 10 scale (see the Appendix).
758 Academy o/Management /ournaJ December
and VPs of finance and sales. In the first coalition, the two executives were among Alpha's original employees and had worked together for eight years. In contrast, the VPs of sales and finance joined the firm at the same time, about two years prior to our study. The same type of pattern existed at Cowboy. The directors of marketing and software, who formed a coalition, had worked together at a prior firm, and they joined Cowboy at the same time, about one year before the other executives. At Omicron, the VP of corporate development and the chairman had a long history together, includ- ing being school friends and the cofounders of both Omicron and a prior firm.
Office location is also an important factor in coalition formation. The VPs of sales and marketing at First had adjoining offices in a building about one mile from the work site of the rest of the executives. The chairman and the president had adjoining offices. Only the VP of manufacturing was lo- cated in their building and he was on its opposite side, on the manufacturing floor. The VPs of sales and finance at Alpha had adjoining offices while the VP of operations was at the other end of the building and the VP of R&D was located in another building.
Finally, status differences play a role in coalition formation. For example, at Cowboy, VPs formed one coalition and directors formed the other. At Omicron, the two cofounders engaged in a coalition. At First, the president and chairman were in one coalition, and VPs were in the others.
Our results are similar to those of studies of interpersonal attraction that have associated propinquity (Festinger, Shachter, & Back, 1950), similarity (Newcomb, 1961), and simply time together (Newcomb, 1956) with the for- mation of friendships. This similarity of results suggests that the same fac- tors that predict friendships also predict alliance patterns. People choose to ally with those with whom they interact frequently and with whom they feel comfortable. The qualitative data in Table 6 corroborate this view. For example, executives in many of the coalitions described their coalition rela- tionships with affection—"good friends," a "good pair," and a "good team" were among the phrases used. In contrast, the more Machiavellian rational view would suggest that alliances are based on similar opinions on issues. Our finding, that alliances are not particularly issue-based, is consistent with the views that most people are not comfortable with politics (Gandz & Murray, 1980), that they use politics only when they think that they must, and that when they do use politics they try to engage safe, familiar allies. Overall, our finding gives further credence to the existence of a social, rather than an ideological, basis of alliance formation.
Our fourth proposition also relates to the demographic approach to the study of top management teams (e.g., Hambrick & Mason, 1984; Wagner, Pfeffer, & O'Rielly, 1985). The evidence from our cases is consistent with the view that demographics can play an important role in the functioning of top management teams. However, our results differ from those of previous re- search in that the existence of such demographic patterning is not causal. Demographics did not trigger political action. For example, the Forefront,
1988 Eisenhardt and Bourgeois 759
Zap, and Promise top management teams did not have stable coalitions, despite having demographic patterns such as adjoining offices and prior history together. As at Cowboy, many of the Promise executives had worked together at another firm. However, these executives did not form stable coalitions. Two Forefront executives had gone to school together, but they were not stable allies. Several of the Zap executives were 15 to 20 years younger than some of the others. However, coalitions along age lines did not develop at the firm. In these firms, some executives had adjoining offices, while others were elsewhere, even in other buildings. Again, stable coali- tions did not develop. Therefore, our evidence suggests that when condi- tions are ripe for politics, demographics play a role. However, demographic patterns do not produce politics. Moreover, simply studying demographics would probably have led to overprediction of the extent of political behavior and to missing the true nature of politics within the teams.
Proposition 5: Demographic similarity is not a sujjicient condition for stable coaJition formation. Rather, demo- graphic similarity leads to stable alliance patterns only when power is centralized and the use of politics is high.
Effects of Time
Most observers of political process have paid little attention to changes over time (Stevenson et al., 1985). However, two cases gave us the opportu- nity to observe the effects of time and transitions on the politics of executive teams.
What we found was that stable patterns of political behavior develop slowly, but once formed, are slow to change. Like a river, politics can take an almost random early course, following the convenient paths offered by ad- joining offices, prior experience together, and so on. However, with time, political behavior becomes channeled into stable patterns, and like the course of a river, resistant to change. In formal terms:
Proposition 6: The formation of stable alJiance patterns iags changes in the use of poJilics.
Two cases in which there were major transitions within the top manage- ment teams provided grounds for this proposition. One case was Neutron, where we observed a top management team at the beginning of their time together. When we visited the firm, the president had been at Neutron for only six months. The VPs of sales, engineering, and manufacturing were all hired after his arrival. In fact, five of the six executives had worked for the firm for less than a year. Neutron's team was politically active, as measured by alliance scores and by anecdotal and decision-story evidence (Table 2). However, there was only one stable coalition within the team, and the vari- ance of the alliance scores was the lowest among the politically active teams (Table 5). Thus, despite evidence that the Neutron team was politically active, the political activity was not as organized into stable alliance patterns as it was at the politically active teams at Cowboy, Alpha, and First, which
760 Acodemy of Management /ournal December
had been in place for several years. We inferred that perhaps the team was so new that stable alliance patterns had just begun to emerge.
The second case was Omicron, where we observed a top management team that had recently undergone a change of CEO. At the time of our study, the current CEO had been in place for only a few months. His predecessor was viewed as a "domineering personality" with a "strong ego." Bill, the prior CEO, himself told us that he did not think that most people were even capable of thinking about strategic issues. In contrast, his successor was described as "easygoing," and "easy to work with" and as having "lots of respect for people." The power scores corroborate that he had decentralized responsibility (see Table 2). Consistent with Propositions 1 and 2, the qualita- tive evidence indicates that the increased decentralization reduced the team's use of politics. For example, the VP of manufacturing told us: "The politics at Omicron are becoming less and less. . . . I'm not aware of any politics lately." Another executive said: "A lot was getting decided by Bill and a few others over the weekends that ended up impacting us [the rest of top manage- ment]. Now we have more frank assessments in meetings and everyone attends." However, there was one stable coalition at Omicron, whereas there were none at the other politically inactive firms such as Zap, Forefront, and Promise, and the variance in alliance scores was the highest among the less political teams. We inferred that vestiges of past stable alliance patterns, although dissipating, still lingered.
In summary, although the four propositions in this section are tentative, they do make the case that politics, particularly in politically active teams, are organized into small and stable coalitions whose membership is based on demographic characteristics. Over time, a pattern of stable alliances emerges in politically active teams and lingers even when the conditions that trig- gered its rise are no longer present.
TOP-MANAGEMENT-TEAM POLITICS AND FIRM PERFORMANCE
Several authors (Daft, 1983; Pfeffer, 1981) have argued that politics can be beneficial. Moreover, those authors have argued that politics may be particularly advantageous in rapidly changing environments because they serve as an important mechanism for adaptation.
Although limited, our data indicate a different view. Among the firms we studied, the top management teams of the effective firms avoided politics, whereas the management teams of poor performers tended to use politics.^ in more formal terms:
Proposition 7: The greater the use of politics within (he top management team, the poorer the performance of a firm.
^ There are many factors related to performance, and power centralization may be related to performance other tban tbrough politics. However, our interest Is in politics and we found the politics-performance link compelling. So we bave focused our attention directly on politics, rather than on these other relationships.
1988 Eisenhardt and Bourgeois 761
Performance was assessed by (1) CEOs' numerical self-reports of com- pany effectiveness (0-10 scale), (2) that rating compared to ratings CEOs gave competitors, and (3) sales growth and profitability figures from before and after our study. Table 7 summarizes these data.
The variation in performance that emerges supports our proposition that firms with politically active top management teams perform less well. For example. Alpha's performance did not live up to its president's expectations. Sales were declining, and the firm was only marginally profitable. First was a moderate performer, with low growth and modest profits. And Neutror. and Cowboy were no longer in business. Neutron closed its doors as a result of bankruptcy, and Cowboy ran out of money because of delays in its new product line.
By contrast. Zap's performance was spectacular, with sales growing at 25 to 100 percent per quarter. Forefront was also a strong performer, with sales tripling and after-tax profits at 9 percent during the year of our study. Since the study, the firm has gone public. When we studied Omicron, the firm was in a turnaround situation, and in transition with a new president (cf. Bourgeois & Eisenhardt, 1987). We found evidence of reduced use of politics with the new president, and the firm has since rebounded, with 50 percent sales growth and 6 percent after-tax profitability. Although it is still too early to gauge Promise's performance on strictly financial grounds, the president gave them a high self-rating, they have received strong support from the venture capital community (over $15 million), and their order hacklog is strong.
Why are politics linked with poor firm performance? The threat-rigidity literature claims that poor performance is likely to trigger power centraliza- tion, which we have argued leads to politics. These linkages suggest that poor performance creates politics. The data from Neutron are consistent with this line of reasoning. The new CEO at Neutron was hired to "take charge" of the weak firm. Our data indicate that he centralized power, and the team hecame politically active.
However, our data also make a case for the opposite (but not mutually exclusive) causal direction: politics lead to poor performance. The qualitative data indicate several reasons why politics limit firm performance. One rea- son is that it is time-consuming to engage in politics. Politics distract execu- tives and dissipate their energy. The executives we studied are very husy people, and using politics means that they are taking time away from their functional responsibilities. For example, the VP of marketing at First spent extensive time lobbying the president and arranging presentations by exter- nal consultants in order to convince the chairman to delay his decision. He felt that he could not he effective by simply going to the chairman with his case. As he told us: "I would just have been labeled a foot-dragger." Not only was this political activity time-consuming, it also kept him from important marketing tasks. The firm had introduced a major new product line. The product was being distributed through multiple channels that competed with one another. Several distributors had begun to impinge on others'
762 Academy o/Management /ournal December
[ ^ a
OS " 5 H
=̂
s 3
x: z^
If "5 o
o *̂..< o a, oo 3 a.
a. in
lit ^ " 3 g a; "B
u
3 2
CO S,
o
o _3J
O CO O
,- z: o c .,- 'C 0̂
S .2 K Q
- E Q. 3 3 x:
D.
< 6
in -a ^ 3
1988 Eisenhardt and Bourgeois 763
territories, and price wars were developing. The politicking of the VP of marketing limited the time availabie for him to perform his regular job.
A second reason is that politics restrict information flow (Pettigrew, 1973). Such restriction is particularly problematic in high-velocity environ- ments, where strategic decisions must be made quickly and rationally (Bourgeois & Eisenhardt, 1987, 1988). In such environments, a premium is placed on timely and accurate information. Yet politics, by their very nature, often impede the flow of information [Pettigrew, 1973). For example, in the decisions we studied, both the president of Neutron and the chairman of First withheld information that a decision was occurring. In the case of First, the idea of changing the name of the firm arose over a year before the final decision. Although the VPs thought the issue had been dropped, the chair- man continued to consider the change privately. As one executive said; "The decision had gone underground." The chairman brought the issue up again only three weeks before the final decision was made, thereby limiting opposition. The president of Neutron also limited information about his considering an alliance with another firm. These actions constrained discus- sion of the issues, and Neutron eventually went bankrupt. Executives at Alpha, another politically active team, were described as "secretive" and "close to the vest" by their colleagues. One VP described how another hy- passed him by only sharing information with the president: "Rich IVP of sales] was trying to get Don's ear, and since I had been close to Don for a long time, he would skip me." The result was that the VPs had poor information from other functional areas. Particularly problematic was the new-product- development interface between sales and R&D. Rather than share informa- tion, the VPs of the two areas withheld information from one another as they jockeyed for position in the "pecking order."
The stability of alliances in politically active teams also distorted percep- tions about the opinions of others. For example, as described earlier, execu- tives at Alpha failed to develop natural, issue-based, alliances with execu- tives in different coalitions. The VP of R&D, who was in one coalition, thought that the VP of sales, in another coalition, favored development of an IBM-compatible product. Yet, the VP of sales expressed complete opposition to an IBM-compatible product to us. Overall, the executives in the politically active teams were often incorrect in their perceptions of others' opinions, thereby limiting their ability to form effective alliances. Descriptions of poor communication made by several executives from these teams corroborate this observation. For example, we were told: "He [the chairman] doesn't really listen to me," and "The group wasn't really hearing what I was saying." In contrast, the executives at Zap, Omicron, Forefront, and Promise had a much better understanding of the views of others. For example, the execu- tives at Zap could accurately describe the varying positions held by others on whether the firm should go public. Promise executives could also accu- rately articulate the positions of others on the issue of strategic redirection that we studied.
764 Academy of Management Journai December
Events at Omicron, where there was a change in CEO followed by a reversal in performance, make a strong case for the effects of politics on per- formance. Colleagues described the prior CEO as "domineering." In contrast, the new CEO was seen as "easygoing" and "people-oriented." Our quani- tative data confirm the power-sharing of the new CEO (Table 2). Consistent with Proposition 1, several of the VPs confided to us that the team was less political under the new CEO, and our quantitative data indicated a team in transition, becoming less prone to use politics with time. At the time of our study, the firm had stagnant sales and was losing money. In the year after our study, the firm, with its new power-sharing CEO and less political team, increased sales 50 percent and became profitable. In this case, diminished politics preceded a financial turnaround.
Overall, the best-performing firms were those in which the CEO shared power with the functional VPs and politics were minimal. For example. Zap was a star in the microcomputer industry. The president of Zap had relin- quished much of his power to strong functional heads. Zap executives, in return, saw no need to engage in politics. As one VP told us: "I've made my position known." Although this lack of politics was probably not the only reason for Zap's success (Eisenhardt, 1988), it seems an important factor. Similarly, Forefront was also a successful firm. Like Zap, Forefront had an empowered top management team and limited political activity among its top management. Again, there was little time to waste on alliances, outlaw staff meetings, and other forms of political behavior. As one executive told us: "We're not political."
Proposition 7 expresses a more negative view of politics than that of some authors (Baldridge, 1971; Pfeffer, 1981). Our results may differ because our definition of politics differs (Gandz & Murray, 1980). Ours is a focused definition grounded in the data from our informants. Another reason may be the stability of the alliance patterns we found in our cases. We simply did not find the shifting, fluid alliances others have described (Baldridge, 1971; Camson, 1964). The absence of shifting alliances undermines the adaptation argument (Daft, 1983; Pfeffer, 1981). We found that the stress of political cultures and the satisficing behavior of executives seem to lead to stable alliance patterns.
A final reason for the differences in results may be related to environment. The microcomputer industry is a high-velocity environment, characterized by rapid and discontinuous change (Eisenhardt & Bourgeois, 1989). At the time of our study (1984-85), the industry had existed for only seven years (Apple Computer was founded in 1977). Between 1980 and 1985, there were major changes in technology, including the advent of the UNIX and DOS operating systems, 32 bit microprocessors, and RISC architecture, and major changes in competitive positioning—IBM and DEC entered and Texas Instru- ments and Sinclair left. We have argued elsewhere that it is crucial both to move rapidly and to avoid mistakes in such an environment (Bourgeois & Eisenhardt, 1987, 1988). However, our evidence suggests that firm politics make it difficult to achieve either of those goals. Politics create rigid barriers
IQSB GJsenhardl und Bourgeois 765
to communication and slow down the pace of decision making. In contrast, the large and stable bureaucracies studied by others (Baldridge, 1971; Pfeffer & Salancik, 1974) may afford the time to engage in assessment of preferences and in fluid political patterns, and they may be more forgiving of poor judgments.
TOWARD A MIDRANGE THEORY OF POLITICS IN HIGH-VELOCITY ENVIRONMENTS
Our initial goal was to enhance understanding of how politics operate when top management teams make strategic decisions. The overall result of our work was a model of the politics of strategic decision making in high- velocity environments, depicted in Figure 1. Tlie rationing of power by autocratic CEOs, combined with the desire for control by frustrated top management teams, is an impetus for politics (Proposition 1). The CEOs preserve their power through the use of politics, and the other memhers of the teams attempt to gain power through politics of their own. In contrast, when the power of a CEO is decentralized and the other members of a team are empowered, most executives see little need to engage in politics. Rather, open, forthright conflict is the norm, with the CEO and possibly the relevant VP resolving stalemates. Thus, although conflict may be necessary, it is not a sufficient condition for the emergence of politics (Proposition 2).
We also found that politics are organized around stable coalitions (Proposition 3). Rather than being organized as shifting issue-based alliances, politics in our teams were organized around small and stable coalitions whose membership was not based on issues but on demographic characteris- tics such as age, joh title, and office location [Proposition 4). Demographic characteristics were not causal (Proposition 5), but rather seemed to provide a convenient alliance basis for executives caught up in the stress of political environments and the pace of high-velocity environments. Such stable pat- terns of alliances were particularly disruptive because communication within the executive teams was seriously impaired. We also found that the develop- ment of strategic coalitions lagged changes in power centralization and the use of politics (Proposition 6).
Finally, the use of politics is related to diminished firm performance (Proposition 7). The firms with politically active teams exhibited slow growth and low profitability. Our cases suggest that politics are time-consuming and information-restricting, creating communication barriers and inflexibility within a team. These negative effects are probably exacerbated in the fast- paced industry that we studied. Thus, the use of politics in high-velocity environments is associated with diminished firm performance.
As presently constituted, this model provides the rudiments of a midrange theory of political behavior among top managements, explaining relation- ships between variables in a particular setting: high-velocity environments. To the extent that our results are valid and can be supported by data from other
766 Academy of Management /ournal December
u
a Ho F a
>.
la r
fi 3 ^ .2 S
sa u
cu
M
1
H
.° o to a, ilc
u U
1^ .
• 9 = 1
a 0( Ia o
c o a 01
o
ED
V3 0)
o
a o
' n o a.o
3 c o.
1986 Eisenhardt and Bourgeois 767
research settings, we think that a theory of power and politics in top manage- ment teams is possible and that such a theory could inform a more general political theory of organizations.
REFERENCES
Allison, G. T. 1971. Essence of decision. Boston: Little, Brown & Co.
Astley, W. G. 1978. Sources af power in organizational life. Unpublished doctoral disserta- tion. University of Washington. Seattle.
Bachrach, S. B., & Lawler, E. J. 1980. Power and politics in organizations. San Francisco: Jossey-Bass.
Baldridge. J. V. 1971. Power and conflict in the university. New York: John Wiley.
Bell, G. 1984. The mini and micro industries, IEEE Transactions on Computing. October,
Bourgeois, L. J. Ul. 1980. Performance and consen-sus. Strategic Management Joumai, 1: 227-248.
Bourgeois, L. J. IU, & Eisenhardt, K. M. 1987. Strategic decision processes in Silicon Valley: The anatomy of a "living dead." Califomia Management Review, 30(1): 143-159.
Bourgeois, L. J. UI, & Eisenhardt, K. M. 1988. Strategic decision processes in high velocity environments: Four cases in the microcomputer industry. Management Science. 34: 816-835.
Butcher. L. 1988. Accidental millionaire: The rise and fall of Steve Jobs at Apple Computer. New York: Paragon House.
Cyert, R. M., & March. J. G. 1963. A behavioral theory of the firm. Englewood Cliffs, N. J.: Prentice-Hall.
Daft. R. L. 1983. Organization theory and design. St. Paul: West Publishing.
Dean, J., Sharfman, M., & Ford, C. 1987. Strategic organizational decision'making: A multiple- context model. Unpublished working paper, Pennsylvania State University, University Park.
DesB, G., & Beard, D. 1984. Dimensions of organizational task environments. Administrative Science Quarterly, 29: 52-73.
Eisenhardt, K, M. 1988. Making fast strategic decisions. Working paper, Stanford University.
Eisenhardt, K. M., & Bourgeois, L. J. IU. 1989. Charting strategic decisions in the microcomputer industry: Profile of an industry star. In M. Van Glinow & S. Mohrmann (Eds.). Managing complexity in high technology organizations, systems, and people: forthcoming. New York: Oxford University Press.
Festinger, L., Shachter, S., & Back. K. 1950. Social pressures in informal groups: A study of human factors in housing. New York: Harper.
Gamson, W. A, 1961. A theory of coalition formation. American Sociological Review, 26: 373-382.
Gandz, J., & Murray, V. V. 1980. The experience of workplace politics. Academy of Manage- ment Joumai, 23: 237-251.
Glaser. B.. & Strauss, A. 1967. The discovery of grounded theory: Strategies for qualitative research. London: Wiedenfeld and Nicholson.
Hage, I. 1980, Thearies of organizations: Form, process, and transformation. New York: Wiley.
Hambrick, D. C, & Mason, P. 1984. Upper echelons: The organization as a reflection of its top managers. Academy of Management Beview, 9: 193—206.
768 Academy o/Management /ournal December
Hinings. C , Hickson, D.. Pennings. [.. & Schneck. R. 1974. Structural conditions in interorgani- zational power. Administrative Science Quarterly. 19: 22-44.
Jay. A. 1967. Management and Machiavelli. New York: Holt. Rinehart & Winston.
Leavitt. H. J. 1951. Effects of certain communication patterns on group performance. Joumai of Abnormal and Social Psychology. 46: 38-50.
March, J. G. 1962. The business firm as a political coalition. Joumai of Politics. 24: 662-678.
March. J. G., & Olsen, }. P. 1976. Ambiguity and choice in organizations. Bergen, Norway; Universitetsforlaget.
Mintzberg. H. 1979. An emerging strategy of "direct" research. Administrative Science Quar- terly. 24: 580-589.
Mintzberg. H. 1983. Power in and around organizations. Englewood Cliffs, N. J.i Prentice-Hall.
Murnighan. J. K. 1978. Models of coalition behavior: Game theoretic, social psychological, and political perspectives. Psychological Bulletin, 85: 1130-1153.
Newcomb, T. M. 1956. The prediction of interpersonal attraction. American Psychologist, 11: 575-586.
Newcomb, T. M. 1961. The acquaintance process. New York: Holt, Rinehart & Winston.
Patchen, M. 1974. The locus and basis of influence on organizational decisions. Organizational Behavior and Human Performance, 11: 195-221.
Pettigrew, A. 1973. The politics of organizational decision making. London: Tavistock.
Pfeffer, J.. & Salancik. G. 1974. Organizational decision making as a political process: The case of the university budget. Administrative Science Quarterly, 19: 135-151.
Pfeffer, J. 1981. Power in organizations. Marshfieid, Mass.: Pitman Publishing.
Riker. W. H. 1962. The theory of political coalitions. New Haven, Conn.: Yale University Press.
Schumpeter. J. A. 1934. The theory of economic development. Cambridge. Mass.: Harvard University Press.
Staw, B. M., Sandelands, L. E., & Dutton, J. E. 1981. Threat-rigidity effects in organizational behavior: A multilevel analysis. Administrative Science Quarterly, 26: 501-524.
Stevenson, W., Pearce, J.. & Porter, L. 1985. The concept of "coalition" In organization theory and research. Academy of Management Beview. 10: 256-268.
Sutton, R. I., Eisenhardt, K. M., & Jucker, J. V. 1986. Managing organizational decline: Lessons from Atari. Organizational Dynamics, 14: 17-29.
Sutton. R. I.. & Callahan. A. L. 1987. The stigma of bankruptcy: Spoiled organizational image and its management. Academy of Management Journal, 30: 405-436.
Thompson, J. 1967. Organizations in action. New York: McGraw-Hill Book Co.
Wagner. G.. Pfeffer, J., & O'Reilly, C. 1984. Organizational demography and turnover in top management teams. Administrative Science Quarterly. 29: 74-92.
White, R.. & Lippitt, R, 1968. Leader behavior and member reaction in three "social climates." In D. Cartwright & A. Zander (Eds.), Group dynamics: 318-385. New York: Harper & Row.
Yin, R. K. 1984. Case study research: Design and methods. Beverly Hills. Calif.: Sage Publications.
1988 Eisenhordl and Bourgeois 769
APPENDIX
Several variables were measured with a questionnaire. The goal and policy-conflict ques- tions were derived from Bourgeois (1980}. The questions on interpersonal disagreement, alli- ance formation, and power were derived from Hinings and colleagues (1974). as modified by Astley (1978).
The questionnaire was introduced to each respondent during the course of an interview. We had the respondents go through two or three questions in order to ensure that they understood the intent of each question and the numerical rating system we were using. For example, the questions about disagreements and alliances were introduced with the commentary that "conventional wisdom holds that a decision climate of harmony and consensus is ideal, but the empirical evidence doesn't necessarily support this. In fact, some people feel that healthy debate and open conflict is more effective to decision making." The purpose of this introduction was to diminish social desirability bias.
Conflict
There were two questions on the importance of different goals and strategic decision areas. Conflict was assessed as the variance within each top management team.
Goal conflict. The goal question consisted of a list of ten organizational goals. The question read: "In the space provided, indicate how important each of these goals is to your firm" (0-10 scale). The goals included long-term profitability, growth, innovation, stock price, company prestige, and service to the community. We computed goal disagreement by summing the top management team's standard deviation on each of the goal items.
Policy conflict. Our questionnaire contained a matrix in which 12 key decision areas were listed down one side of the sheet and the firm's executive titles were listed across the top. The question was: "Here is a list of various decision areas which may be of strategic importance to your firm. Please indicate how important each of these decision areas is to the long-run health of your firm" (0 = not at all important, 10 = extremely important). The decision areas included marketing strategies and product pricing. R&D project selection, expansion of production capacity, major financing (e.g.. issuing stocks or bonds), and restructuring the organization. We computed policy disagreement by summing a team's standard deviation on each of the items.
Interpersonal disagreement. In order to obtain interpersonal disagreement scores for each top management team, we asked each executive to evaluate frequency of disagreement with each specific other member of the team. The question was, "How often, during the process of deliberating, debating and making policy decisions, have you found yourself in open disagree- ment with the suggestions or proposals of each of these individuals?" (0 = never.lO = constantly). We first computed the mean score for each executive and then the overall mean for the team.
Power
Using the same key-decision-area matrix described under "policy conflict," executives rated each manager on their influence on each decision: "Now. for the same list of decision areas (excepting those scored 0 or 1). indicate how much infiuence you think each manager has in making decisions concerning that decision area. If the manager has a very great deal of influence over the decision area, give a rating of 10, no influence would score 0, and so on." Power scores for each executive were computed by taking the mean scores assigned to the executive by every other respondent. We took three steps, computing (1) a mean power score for each person on each decision, (2) a mean score for each decision area (R&D, marketing, finance, operations, organization), and (3) an overall mean.
Alliance Formation
Executives were given a list of their colleagues on the team and asked to indicate the frequency (0 = never to 10 = constantly) with which they formed alliances with each colleague. The question was: "How often have you had to join forces with and form an alliance with each manager in order to influence or 'push through' a policy proposal, or to get your alternatives or points of view 'on the table' ?" We obtained each team's score in two steps, computing (1) the
770 Academy 0/ Management /oumal December
mean for each executive and [2] the mean of each of those means. We also constructed network diagrams of the alliance patterns of each team hy taking the mean alliance score hetween each pair of executives and used those means to determine the stahle coalitions. Finally, we com- puted the variance in the alliance score as a measure of alliance stahility.
Kathleen M. Eisenhardt received her Ph.D. degree in organizational hehavior from Stan- ford University. She is an assistant professor in the Department of Industrial Engineer- ing and Engineering Management, School of Engineering, Stanford University. Her research interests include strategic decision making and entrepreneurship in high- technology environments, and agency theory.
L. ]. Bourgeois III, whose Ph.D. degree is from the University of Washington, is currently an associate professor of business administration at the Darden School, Uni- versity of Virginia. His research interests are strategic decision making and top- management-team hehavior in dynamic husiness environments.
