Professional Documents
Culture Documents
BERTRAMH. RAVEN^
University of California, Los Angeles
Social power can be conceived as the resources one person has available so
that he or she can influence another person to do what that person would not
have done otherwise. As used here, power sources or bases differ from influ-
ence techniques as the former refer to potential and the latter to the actual use of
the power. A manager may have the potential to use many resources, but in
practice uses only certain ones to influence subordinates. Empirically, the dis-
tinction implies that compliance is a function not only of powers actually used
but also of the power sources assumed to be available to the supervisor.
Over the years, the study of power sources has become a central concept in
organizational psychology, as well as in other analyses of social interaction. A
number of instrumentswere based on the original five sources presented by French
'We express our appreciation to Aroldo Rodrigues for his suggestions and assistance in
developing the scale items.
2Correspondence concerning this article should be addressed to Bertram H. Raven,
Department of Psychology, UCLA, Los Angeles, CA 90095-1563, or to Joseph Schwarzwald,
Department of Psychology, Bar-Ilan University, Ramat Gan 52900, Israel.
307
31n informational influence (and, in its potential, informational power), the word information
is used in its broadest sense to include any different perspective or knowledge which an influencing
agent might impart to a target. The absent-mindedparent who turns on the headlights in his car after
his child points out that it has become quite dark outside has been influenced by the child’s
informational power. Once the child has pointed it out, the parent realizes that he really must have
his headlights on.
POWER MODEL 309
1985; Schriesheim, Hinkin, & Podsakoff, 1991). Particularly pressing has been
the absence of measures which assess power profiles across situations.
Concerns have also been raised in regard to conceptual overlaps between
categories, inconsistent operational definitions, and, at times, low agreement
between observations and prediction. For example, allusions to expertise may
involve elements of referent and informational power. This problem is height-
ened by the fact that differing power bases are often used in combinations
(e.g., expert power and informational power), such that influencing agents tend
to reveal profiles of powers using unique constellations of bases rather than
simple clear preferences for individual bases (Bui, Raven, & Schwarzwald,
1994; Hinkin & Schriesheim, 1989; Kipnis, 1990; Schriesheim & Hinkin,
1990).
Finally, several other investigators using different methodological ap-
proaches have argued that the number of sources may be greater than the
five or six in the original formulation (Kipnis & Schmidt, 1983). Based on
their factor analysis, Yukl and Tracy (1992) reported a nine-factor taxonomy
(rational persuasion, inspirational appeal, consultation, negotiation, exchange,
personal appeal, collision, legitimating, and pressure). In another power study,
Yukl and Falbe (1991) identified two power sources, persuasiveness and cha-
risma, that did not appear in the original French and Raven taxonomy. How-
ever, their definition of persuasiveness (being convinced by a person’s logical
argument) overlaps with Raven’s (1992) definition of information. Charisma
may be seen as a projected character trait which is probably closely related to
the use of referent power, but may also contribute to legitimate and expert
power.
To answer some of these concerns, a number of investigators have at-
tempted to develop new scales and measures to assess French and Raven’s
bases of power (e.g., Hinkin & Schriesheim, 1989; Imai, 1989; Rahim, 1986;
Schriesheim & Hinkin, 1990). This, then, is our attempt to develop social
power measures which we hope will answer a number of past criticisms and
also take into account recent theoretical developments.
Recently, Raven (1992, 1993) presented an interpersonal powerhteraction
model. While still maintaining the basic conception of six bases of power, evi-
dence was cited from theory and research to point to the need to make differen-
tiations among some of the bases of power. Coercion and reward power, which
were usually conceptualized in terms of “impersonal” threats (e.g., threat of
being dismissed) or promises of reward (e.g., promise of a promotion), are now
distinguished from personal coercion and personal reward. The latter are indi-
cated when the agent threatens to disapprove or dislike the target for noncom-
pliance (coercion) or promises to approve or like the target (reward) if the
target does comply.
310 RAVEN ET AL.
scales, have alluded to the fact that individual power bases can be conceptual-
ized at different levels of abstraction. Bass (1960) and Yukl and Falbe (1991)
differentiated between position and personal types, and Kipnis (1984) as well
as Bui et al. (1994) differentiated among strong, weak, and rational bases. By
conceptualizing power sources at a higher level of abstraction, it may be pos-
sible to provide a better understanding of the observed relationships in the re-
cent taxonomy suggested by Raven (1992, 1993). As such, one of the goals of
the study was to determine whether the new 11 power bases (i.e., the surface
factors) identified in the model could be encompassed by a smaller number of
source factors.
The purpose of the two studies conducted here was to operationalize the 11
power bases by developing a new scale, and to determine whether the items
conceptualized for each tactic represent homogeneous and independent fac-
tors. In addition, we proposed to examine the empirical structure along com-
mon underlying dimensions as a measure of the interrelationship among the
power bases. This may provide a more parsimonious, yet comprehensive
model for understanding the available power bases. In the second study, sev-
eral features of the new scale were tested on another sample in a different cul-
ture and in an actual work setting.
Study 1
Method
Instrument
In developing items for the new scale, the IPI, we drew on several sources
including previous scales in the literature (e.g., Hinkin & Schriesheim, 1989)
and definitions for specific power bases as delineated in Raven’s (1992) model.
These items were piloted with open-ended interviews where respondents were
asked to complete trial questionnaires. For each of the 11 power bases pro-
posed in Raven’s model, four items were eventually constructed, yielding a to-
tal of 44 items.
Subjects were asked to respond to one of two forms of the IPI: subordinate
or supervisor. In the subordinate form, the instructions read:
POWER MODEL 313
Results
Table 1
Combined Individual
Factor Content Item Loadings a a
Table 2
Harsh bases
Legitimate power of reciprocity .76 .2 1
Impersonal coercive power .76 -.33
Legitimate power of equity .75 .09
Impersonal reward power .70 -.15
Personal coercive power .69 .10
Legitimate power of position .35 .34
Soft bases
Expert power -.20 .80
Referent power .11 .79
Informational power -.23 .68
Legitimate power of dependence .17 .66
Personal reward power .47 .60
Along the harsh dimension, one can find personal reward and personal coer-
cion at one end, and impersonal reward and impersonal coercion at the other
end. In between are legitimate equity and reciprocity. Along the soft dimen-
sion, one can find legitimate dependence and reference at one end and expertise
and information at the other end. The apex of these two dimensions occurs at
the legitimate position source which has characteristics of both soft and harsh
bases.
I m m
Figure I. Diagram of smallest space analysis of the 33 items of the Interpersonal Power
Inventory. LGR = legitimate power of reciprocity; COI = impersonal coercive power;
LGE = legitimate power of equity; RWI = impersonal reward power; COP = personal
coercive power; LGP = legitimate power of position; EXP = expert power; REF = refer-
ent power; INF = informationalpower; LGD = legitimatepower of dependence; RWP =
personal reward power.
showed that females did not differ from males (Wilks’s A = 0.94), x2( 11) =
1 8 . 7 1 , ~> .07.
In the second analysis, power factor profile by position perspective (super-
visor N = 134, subordinate N = 183) was performed. Here, results were signifi-
cant (Wilks’s A = .84), x2(7) = 53.15, p < .01. The two groups differ on
expertise, F(1, 315) = 1 8 . 8 4 , ~< .01; information, F(1, 315) = 1 0 . 5 7 , ~
< .01;
and reference, F(1, 315) = 15.78, p < .01. The difference between groups
emerged only in the soft bases for which supervisor mean compliance was
higher than subordinate in each case (Table 3). For the other eight power
bases, no significant differences were found between supervisors and subor-
dinates, indicating that, in general, the two groups overlap on the profile of
318 RAVEN ET AL.
Table 3
Means and Standard Deviations of Power Base Usage and Rank Preference
by Job Status
Super- Subordi-
visor nate Total
aDenotes a significant difference between position within job status for each sample.
power tactic compliance. The similarity in compliance scores between the two
groups was also examined with the Spearman rank order correlation coeff-
cient. The high value (.90) indicated that although some of the means may have
differed, the position of the bases within each group was approximately the
same.
Study 2
As the IPI was developed on a student sample, some of whom had only lim-
ited work experience, we felt a need to test the scale on a worker sample in a
field setting. For this purpose, a follow-up study was conducted to examine two
issues: the generalizability of the power tactic structure and the relationship
between compliance profiles and job satisfaction. The decision to use job satis-
faction stems from its popularity as a measure of organizational outcome that
has shown consistently positive relationships in the power literature (Podsakoff
& Schriesheim, 1985). Specifically, the following issues were examined:
POWER MODEL 319
(a) the consistency of the factor structure across cultures, (b) the similarity in
degree of compliance to power bases, and (c) the relationship between compli-
ance and job satisfaction.
Method
Participants
Instruments
Hebrew version of the IPI (subordinateform). The final version of the ques-
tionnaire consisting of 33 items identified in the American sample was de-
scribed previously in Study 1. For Study 2, the scale was translated into
Hebrew and retranslated into English and was found to match the American
version in content and purpose.
Minnesota Job Satisfaction Questionnaire (MSQ). The short version of the
scale (Arvey, Bouchard, Siegal, & Abraham, 1989) was used here. Respon-
dents are asked to indicate the degree of satisfaction with each of 20 statements
on a 5-point scale ranging from 1 (very dissatisfied) to 5 (very satisfied). Exam-
ples are: “The chance to do different things from time to time,” “The freedom
to use my own judgment,” and the “Praise I get for doing a good job.” The coef-
ficient alpha was .87.
Procedure
After obtaining permission from the heads of several departments in the hos-
pital, the questionnaires were administered to the subjects who were promised
confidentialityand anonymity. The order of questionnaires was counterbalanced
such that half received the IPI first and the other half received the MSQ first.
Results
Several features of the IPI scale were compared between studies. For the
Israeli sample, as with the American sample, the internal consistency of items
320 RAVEN ET AL.
within each power tactic was assessed with the Israeli sample. The alpha coef-
ficients were satisfactory, ranging from .63 (for legitimate dependence) to .88
(for impersonal reward and coercion). It appears that the relationship among
items obtained in the original study was repeated for the Israeli workers.
The factor analysis on the individual items was not repeated, as the size of
the sample here was too small for meaningful inferences (Nunnally, 1978). We
did, however, conduct factor analysis on the 11 power bases. As with the
American sample, a two-factor solution was obtained (Table 4). The first fac-
tor, explaining 39.8% of the variance, consisted of the harsh bases. The second
factor, explaining 20.3% of the variance, included the soft bases. Except for the
legitimate position tactic, which, in the Israeli sample, was included among the
soft tactics and personal reward that was included as a harsh tactic, the other
bases fell in the same general categories in both samples. Closer inspection of
the American data revealed that the legitimate position loadings were actually
quite similar on the harsh and soft bases (.35 vs. .34, respectively). The SSA
outcomes of the American sample also supported this duality, as findings
showed that legitimate position was placed at the apex, indicating characteris-
tics of both dimensions.
Degree of Compliance
Table 5 presents the means and standard deviations for the likelihood that
compliance could be attributed to each of the 11 power bases, ordered by
size of the means. Comparison of the values between the American sample
(Table 3) and the Israeli sample indicated high similarity. Indeed, the rank or-
der correlation between the two samples was .93. For both cultures, partici-
pants tended to report that some power bases (such as information, legitimate
position, and expertise) were more likely to be the reason for compliance, while
others (legitimate reciprocity, equity, and impersonal coercion) were less
likely.
Finally, the 11 power bases, as well as the harsh and soft categories, were
used as predictors of job satisfaction in separate discriminant analyses. The
MSQ score, split at the median (50 above median, 50 below, 1 case missing),
served as the grouping variable. The findings for the 11 bases were significant
(Wilks’s A = .77), x2( 11) = 2 4 . 2 6 , <
~ .01, implying that the compliance profile
distinguished between workers with high and low satisfaction.
Discriminant analysis employing only two predictors, means on harsh and
soft bases, and the general MSQ score as the grouping variables also produced
POWER MODEL 321
Table 4
Harsh bases
Personal coercive power .85 .07 .62
Impersonal reward power .84 -.12 .80
Legitimate power of reciprocity .83 .ll .78
Personal reward power .80 .29 .72
Impersonal coercive power .79 -.26 .83
Legitimate power of equity .68 .06 .80
Soft bases
Expert power .oo .77 .71
Referent power .34 .73 .68
Informational power ..24 .67 .82
Legitimate power of dependence .05 .62 .62
Legitimate power of position .03 .58 .68
significant findings (Wilks’s A = .93), x2(2) = 7.03, p < .05. As with the pre-
vious analysis using 11 bases, the univariate analyses showed that only the
soft tactic category was significant, F(1, 99) = 7 . 5 1 , ~< .01. It appears that
greater compliance to soft bases was related to higher degrees of job satisfac-
tion, whereas the degree of compliance to harsh bases did not relate to job sat-
isfaction.
General Discussion
The findings showed that the IPI based on Raven’s (1992, 1993) recent
elaboration was a reliable and valid instrument for assessing the bases of
power attributed to supervisors in influencing subordinates. Satisfactory inter-
nal consistency was obtained for the items comprising each of the 11 social
power bases as seen in the two different sets of respondents-American stu-
dents and Israeli hospital workers. Factor analysis of items indicated, however,
that some of the 11 bases tended to cluster together such that seven factors
emerged: personal sanctions (personal reward and personal coercion), imper-
sonal sanctions (impersonal reward and impersonal coercion), credibility (expert
322 RAVEN ET AL.
Table 5
Power base A4 SD
and personal reward with the soft; for the Israeli workers, the grouping was re-
versed.
Of special interest are the power factors from the item analysis of the
American sample which we had expected to be independent but which were
found to hang together. The fact that individual power bases often overlap with
each other has been reported previously and should not come as a surprise. It
should be emphasized, however, that the overlap was always within category.
For example, expertise and information, which were found to be related, were
both soft factors. This overlap is not unique to our study. Similarly, legitimate
equity and reciprocity, as well as impersonal reward and impersonal coercion,
which overlapped, were harsh bases. Others have also noted that power bases
are not necessarily independent of each other. Yukl and Falbe (1991) argued
that reward and coercion should be combined into a single factor. Podsakoff
and Schriesheim (1 985) argued that an agent who possessed referent power
may also be perceived as possessing expert power; that is, a halo effect of ex-
pertise is attributed to those whom we like. Finally, a person who has the power
to reward may also be perceived as having the power to punish (Bass, 198 1).
The overlap between information and expertise is compatible with the defi-
nitions for each source as proposed by Raven (1 992). With both bases, the
agent influences the target by saying there are good reasons why the target
should change his or her behaviors or beliefs. Yet, they differ. With informa-
tional influence, the agent actually gives the reasons, by directly providing
convincing data on the basis of which the target can be expected to change.
With expert bases, the agent, in effect, says, “On the basis of my substantial
knowledge and experience, I know good reasons why it would be better for you
to change, but you will have to trust me, since (for one reason or another) I am
not making these reasons explicit to YOU.”Our respondents apparently felt that
a supervisor who is likely to influence subordinates by presenting good reasons
tends to be equally successful if he or she relies on expertise without explicitly
stating reasons. It may also be that when the supervisor states good reasons, the
target tends to infer expertise.
The tactic which we have labeled legitimate equity also includes two over-
lapping hypothesized power bases, legitimate reciprocity and legitimate eq-
uity. With respect to both of these, the key underlying principle is “YOUshould
feel obliged to do as I ask, because you owe it to me. It is only fair or equitable.”
We had believed there were two different forms of “fairness”: (a) “You owe it
to me because I have done something for you” (reciprocity); and (b) “. . . be-
cause you have done something not nice to me” or “. . . I have really put a lot of
myself into this . . .” (equity). The reciprocity tactic refers to the “give-and-
take” norm (Cialdini, 1988), whereas equity can be considered a “compensa-
tory” norm (Walster et al., 1978). Apparently, our respondents did not make
324 RAVEN ET AL.
such a distinction between these two forms of equity. It is interesting that our
respondents did make clear distinctions between legitimacy based on position
and legitimacy based on dependence.
In interpreting these data, it is important to keep in mind the wording of the
questionnaire, which represents a hypothetical situation in which compliance
in fact did occur and then asks for a rating of the likelihood of each of these
power bases being responsible for that compliance. We thus ask the respon-
dents to make an attribution regarding the extent to which the bases of power,
in effect, are the cause of compliance. The original French and Raven (1959)
statement as well as subsequent formulations (e.g., Raven, 1992, 1993) empha-
size that power and strategies are also distinguishable in other ways: longevity
of change, importance of surveillance, subsequent liking or disliking for the in-
fluencing agent, and attribution of internal versus external causality of change.
There is good reason to believe that, even though coercion and reward are
highly correlated as factors leading to influence, they would be more independ-
ent ifwe were to ask how likely it is that the subordinate will like the supervisor
after the change. Informational and expert power, though correlated in these
data, might be more independent of one another if our questions dealt with how
long the behavior change would continue or to what extent the subordinate
would see the changes as internally versus externally caused (Brown & Raven,
1994; Litman-Adizes, Fontaine, & Raven, 1978).
One would also predict that the factor structure will differ in other interper-
sonal influence and power situations. For example, greater differentiation is
expected between impersonal reward and impersonal coercion in relationships
between husbands and wives (Raven et al., 1975), and greater differentiation
between expert and information in doctor-patient relationships (Brown &
Raven, 1994). The relative effectiveness of expert and informational power is
also related to the importance of the changed behavior or belief for the target.
Petty and Cacioppo (1986) found that while expert and informational power
may not differ in their impact for issues which have little practical significance
for the target, informational power has greater impact for matters that have im-
portant implications for the target-if the issue is really important, the target is
more likely to require clear facts and reason for the change, and is less likely to
put his or her trust in the expertise of the influencing agent.
The manner in which the items are phrased was meant to allow a measure of
social power which was more in keeping with psychological definitions pro-
posed by Cartwright (1959), French and Raven (1959), Heider (1958), and
others. Of course, there might also be objective definitions which others have
advocated: Supervisors have expert power if they have had a wealth of experi-
ence or training in the area in which influence is attempted. They have reward
power if they can promote their subordinates or give them raises. These then
POWER MODEL 325
would appear to be objective measures. Or we could move a bit toward the psy-
chological direction by asking subordinates’ opinions as to whether supervi-
sors have such resources, rating the extent to which the respondents agree with
statements such as “My supervisor can fire me”; “My supervisor can promote
me”.,LL My supervisor has had much experience in this area”; and so forth. With
such methods we might get good measures, but would they be measuring social
power? Not necessarily. If interviewed, some respondents might answer, “My
supervisor may have had a lot of training and experience, but I really don’t trust
his judgment and would not say he is an expert”; “My supervisor might be able
to fire me, but I don’t think he would since he really needs me, and besides, I
don’t really need this job anyway.” Thus, it is not uncommon that an influenc-
ing agent’s possession of ostensible power resources will fail to result in effec-
tive power and influence. It is for that reason that we used what has been called
an attributional approach for defining power: “If your supervisor asked you to
do something you preferred not to do and you did it, was it because you felt that
otherwise he would hurt or punish you in some way? How likely would that be
the reason (why you did as he asked)?’
Though we feel that this attributional approach more nearly taps social
power, it is not entirely without its own problems. As with most questionnaires
and survey instruments used in our field, one must always be concerned about
the extent to which responses may be biased by self-serving needs (e.g., “I
really did it because otherwise he may punish me, but I cannot admit this to the
person conducting the study, or even to myself’). If so, is there a solution that
would be more satisfactory than the objective, pseudo-objective, or our attribu-
tional approach? None immediately comes to mind. An anonymous reviewer
recommended that
This is a very intriguing suggestion indeed, but one which would also not be
free of problems, including the extreme difficulty in carrying out such a study
with any number of participants. In any case, one would have to have some rea-
sonable measures in order to carry out such a study.
In the meantime, we can get some satisfaction from the fact that we found
considerable consistency in our data. Though we have not solved all of the
problems in this area of research, we feel we have made considerable strides.
326 RAVEN ET AL.
References
Arvey, R. D., Bouchard, T. J., Jr., Siegal, N. L., & Abraham, L. M. (1989). Job
satisfaction: Environmental and genetic components. Journal of Applied
Psychology, 74, 187-192.
Barnes, R. D., Ickes, W., & Kidd, R. F. (1979). Effects of perceived intention-
ality and stability of another’s dependency on helping behavior. Personal-
ity and Social Psychology Bulletin, 5, 367-372.
Bass, B. M. (1960). Leadership, psychology, and organizational behavior.
New York, NY: Harper.
Bass, B. M. (1981). Stodgill’s handbook of leadership (rev. ed.). New York,
NY: Free Press.
Berkowitz, L., & Daniels, L. R. (1963). Responsibility and dependency. Jour-
nal ofAbnormal and Social Psychology, 66,429-436.
Briggs, T. H. (1927). Praise and censure as incentives. School and Society, 26,
596-598.
Brown, J. H., & Raven, B. H. (1994). Power and compliance in doctor/patient
relationships. Journal of Health Psychology, 6, 3-22.
Bui, K. T., Raven, B. H., & Schwarzwald, J. (1994). Influence strategies in dat-
ing relationships: The effects of relationship satisfaction, gender, and per-
spective. Journal of Social Behavior and Personality, 9,429-442.
Carlsmith, J. M., & Gross, A. E. (1969). Some effects of guilt on compliance.
Journal of Personality and Social Psychology, 11,232-239.
Cartwright, D. (1959). Power: A neglected variable in social psychology. In
D. Cartwright (Ed.), Studies in socialpower (pp. 1-14). Ann Arbor, MI:
Institute for Social Research.
Cattell, R. B. (1957). Personality and motivation structure and measurement.
Yonkers, NY: World.
Cialdini, R. B. (1988). Znfluence: Science and practice (2nd ed.). Glenview,
IL: Scott, Foresman.
Danet, B. (1971). The language of persuasion in bureaucracy: “Modern” and
“traditional” appeals to the Israeli customs authorities. American Socio-
logical Review, 36, 847-859.
Erchul, E. P., & Raven, B. H. (1997). Social power in school consultation: A
contemporary view of French and Raven’s bases of power model. Journal
of School Psychology, 35, 137-171.
Etzioni, A. (196 1). Comparative analysis of complex organizations. New
York, NY: Free Press.
French, J. R. P., Jr., & Raven, B. H. (1959). The bases of social power. In D.
Cartwright (Ed.), Studies in social power (pp. 150- 167). Ann Arbor, MI:
Institute for Social Research.
POWER MODEL 327
Appendix
Reward Impersonal
Coercive Impersonal
Expert Power
Referent Power
Informational Power
4. Once it was pointed out, I could see why the change was necessary.
* 17.My supervisor had carefully explained the basis for the request.
POWER MODEL 331
24. My supervisor gave me good reasons for changing how I did the job.
42. I could then understand why the recommended change was for the
better.
Legitimacy/Position
Legitimacy/Reciprocity
*7. My supervisor had done some nice things for me in the past and so I
did this in return.
12. For past considerations I had received, I felt obliged to comply.
32. My supervisor had previously done some good things that I had
requested.
43. My supervisor had let me have my way earlier so I felt obliged to
comply now.
Legitimacy/Dependence
LegitimacyIEquity
1 1. By doing so, I could make up for some problems I may have caused
in the past.
2 1. Complying helped make up for things I had not done so well
previously.
30. I had made some mistakes and therefore felt that I owed this to
himlher.
*36. I had not always done what he/she wished, so this time I felt I should.
332 RAVEN ET AL.
Personal Reward
Personal Coercion