Sunteți pe pagina 1din 7

Journal of Applied Psychology 1989, Vol. 74, No.

4, 561-567

Copyright 1989 by the American Psychological Association, Inc. 002l-9010/89/$00.75

Development and Application of New Scales to Measure the French and Raven (1959) Bases of Social Power
Timothy R. Hinkin
Mclntire School of Commerce University of Virginia

Chester A. Schriesheim
School of Business Administration University of Miami

Presented are results from a multistep, three-sample study that designed measures of reward, coercive, legitimate, expert, and referent power that are conceptually consistent with respect to the source of power. Construct definitions were developed and items were generated and evaluated for content validity. Exploratory and confirmatory factor analyses and item analyses were conducted to develop perceptually distinct scales with acceptable internal consistency and stable factor structures. The independence of the scales was next examined, and discriminant validity was assessed. Finally, zeroorder and partial correlation concurrent validity analyses were conducted, and conclusions were drawn concerning both the new scales and the extant literature.

The acquisition and maintenance of power is, according to McClelland and Burnham (1976), one of the most socially motivating processes that occurs in organizations. As one might expect, power is a central concern of most employees (Gioia & Sims, 1983), and it has also been examined by scholars from a wide variety of perspectives (e.g., Grimes, 1978; McClelland, 1975; Pfeffer, 1981). Although a number of power typologies or frameworks exist, perhaps the most influential is that of French and Raven (1959). French and Raven (1959) distinguished between reward, coercive, legitimate, referent, and expert power. Their typology is presented in most major textbooks in the field and, according to Mintzberg (1983) and others, it is also the framework most frequently used in power research. A major criticism of the French and Raven typology, however, is that the power bases lack conceptual consistency regarding the source or origin of influence (Patchen, 1974; Yukl, 1981). For example, the source of reward power is the power holder's perceived ability to administer outcomes that are rewarding. However, the source of referent power is the power holder's perceived attractive characteristics. In addition, empirical research using the French and Raven typology has been limited almost exclusively to field studies that have primarily used single-item ranking scales (as opposed to rating or Likert-type measures). This research has been severely criticized (Podsakoff & Schriesheim, 1985), and a four-sample study by Schriesheim, Hinkin, and Podsakoff (1985) strongly demonstrated that the results of most existing

research on the French and Raven power bases must be questioned because of psychometric inadequacies of the measures that have been used. Although the concept of power has garnered much attention over the last several decades, key questions remain unanswered. Thus, it is important that the investigation of power in organizations continue. Because theoretical progress is impossible without adequate measurement technology (Korman, 1974; Schwab, 1980) and because many studies of power use psychometrically unsound measures (PodsakofF & Schriesheim, 1985), this study was conducted. This article first reports the results of a multisample study that refined the Schriesheim et al. (1985) scales to develop conceptually consistent and psychometrically sound measures derived from French and Raven (1959). Relationships between the new power base measures and subordinate outcomes are next reported and compared with prior research findings. Finally, conclusions and suggestions for future research in this domain are discussed.

Item Development Process


On the basis of findings of Podsakoff and Schriesheim (1985) and Schriesheim et al. (1985), the development of new power base measures was begun. A process recommended by Nunnally (1978), Churchill (1979), and others was used. The starting point involved the development and use of new, conceptually consistent, theoretical definitions of the various power bases and focused on developing scales that describe behaviors (rather than traits or characteristics). One goal of this research was to develop sound scales that were also parsimonious, consisting of as few items as possible while still adequately sampling the domains of interest.

Timothy R. Hinkin was supported during this research through funding provided to the University of Virginia's Mclntire School of Commerce by the Marriott Corporation Summer Research Fellowships. Chester A. Schriesheim was supported by the Summer Awards in Business and the Social Sciences Program, University of Miami, and the School of Business Administration Corporate Affiliate Program, University of Miami. Correspondence concerning this article, including requests for a longer and more detailed report of this research, should be addressed to Timothy R, Hinkin, Mclntire School of Commerce, University of Virginia, Charlottesville, Virginia 22904.

Theoretical Construct Definitions


To begin constructing scales, theoretical definitions are needed to permit the assessment of scale content validity and to provide a starting point for the generation of items (Nunnally, 1978). Here, a major concern was to create a consistent set of definitions that would remain compatible with the theory pro561

562

TIMOTHY R. HINKIN AND CHESTER A. SCHRIESHEIM and the sample was 59% men. The respondents worked an average of 22 hr per week and were employed in a wide variety of organizations. The second sample (B) consisted of 375 full-time employees of a large southern U.S. psychiatric hospital. The average age was 34 years, average tenure was 4.5 years, and 73% were women. Ninety-one percent of respondents had graduated from high school, and 40% had college degrees. The third sample (C) consisted of 220 part-time MBA students taking organizational behavior and business policy classes at a medium-sized southern U.S. university. The average age was 27 years, and the sample was 56% men. Average organizational tenure was 4 years, and all were currently working full-time employees, with 61 % having supervisory or managerial and 27% sales or clerical positions.

posed by French and Raven (1959). We felt that denning all of the power bases as involving the ability to administer tangible ("things") or intangible ("feelings") outcomes for another would do this and would also facilitate the generation of behaviorally phrased items that are consistent with prior theory and research on power (e.g., Dahl, 1957; Etzioni, 1961; Kanter, 1979). We first independently generated definitions for each base of power. We then jointly evaluated the proposed definitions and came to consensual agreement (for more detail, see "riinkin, 1985). The power bases involved in this research were thus defined as follows: Reward power is the ability to administer to another things he or she desires or to remove or decrease things he or she does not desire. Coercive power is the ability to administer to another things he or she does not desire or to remove or decrease things he or she does desire. Legitimate power is the ability to administer to another feelings of obligation or responsibility. Referent power is the ability to administer to another feelings of personal acceptance or approval. Expert power is the ability to administer to another information, knowledge, or expertise.

Procedure
Survey questionnaires were administered to the respondents in their natural (classroom or work) settings during normal hours. Participation was optional, and anyone not wishing to take part was given the opportunity to decline. The purpose of the research was explained in a brief oral presentation, and participants were given the opportunity to ask questions. Participants were encouraged to answer all of the questions in the survey honestly and were assured of complete anonymity. Of the original 254 Sample A respondents who participated, 3 had to be eliminated because of missing data. Twelve of the initial 387 in Sample B had to be eliminated, as did 6 of the original 226 Sample C respondents.

Item Generation and Content Validity Examination


We next generated lists of potential scale items on the basis of these definitions. The lists were then evaluated for conformity to the theoretical definitions and for redundancy, and agreement was reached on 53 items to be retained for further assessment. The 53 items were then submitted to a formal content validity examination by two independent panels of judges (Ns = 37 and 42), following Schriesheim's (1978) approach. The judges were presented with the construct definitions and asked to classify the randomly ordered items into one or more power categories. The 42 items that were assigned to the proper a priori category more than 60% of the time by both judge panels were retained. They appeared to be content valid, having first been generated by us independently and then having been selected by two independent panels of judges as measuring the appropriate theoretical constructs.

Measures
Power. The measures of power used in Sample A were the 42 items that survived the content judging process described earlier: 8 reward, 10 coercive, 11 legitimate, 7 referent, and 6 expert power items. Several items from the Schriesheim et al. (1985) scales were retained. The 20 items used in Samples B and C were selected on the basis of Sample A's results, with 4 items for each of the power bases. In all samples, the respondents were asked to indicate the extent of their agreement with the randomly ordered items. Satisfaction. Three measures of satisfaction were used, consisting of the five-item (each) Supervisor-Human Relations and SupervisorTechnical Ability subscales of the Minnesota Satisfaction Questionnaire (MSQ; Weiss, Dawis, England, & Lofquist, 1967) and a 12-item short global MSQ satisfaction measure that has been used in previous research (e.g., Schriesheim et al., 1985). These measures were used in the criterion-related validity analyses, quite simply because satisfaction is a traditional dependent variable in power research (cf. Bass, 1981; Yukl, 1981) and because satisfaction has been shown to be related to other important variables (cf. Locke, 1976). Coefficient alphas were .97, .84, and .84 for human relations, technical ability, and global satisfaction, respectively, in Sample A; .89, .90, and .90 in Sample B; and .91, .85, and .86 in Sample C. Commitment. Although empirical researchers of power do not often study commitment, it has been a central concern of theory in this domain (cf. Pfeffer, 1981), and scholars such as Yukl (1981) have specifically called for future research on power to routinely include commitment as a dependent variable. Thus, the 15-item Porter, Steers, Mowday, and Boulian (1974) Commitment scale was used to measure this construct. This instrument has been subjected to considerable psychometric examination, and data pertaining to its reliability and validity are generally quite positive (cf. Mowday, Steers, & Porter, 1979). Coefficient alpha reliabilities of .90, .87, and .90 were obtained in Samples A, B, and C, respectively.

Scale Development and Testing Process


Scale development immediately followed the item development phase we have outlined. The power items (and other measures) were administered to three independent samples. A series of exploratory and confirmatory factor analyses were conducted, leading to a final set of scales with acceptable content validity, discriminant validity, and internal consistency reliability. The Results section details the exact sequence of analyses used. Additional sets of analyses were then conducted to provide preliminary evaluative psychometric evidence on the validity of the new scales.

Samples
As noted, the data for this study were collected from three diverse samples. The first sample (subsequently labeled Sample A) consisted of 251 upper-level undergraduates enrolled in two different business courses at a large southern U.S. university. The average age was 22 years,

Results
The results from the analyses described previously are presented here. Those from the more complex and interrelated

NEW POWER SCALES

563

Table 1 Factor Analytic Results


Sample A Item
h2 R

Sample B

Sample C

02. R 27. R 33. R 38. R 04. C 18.C 21. C 22. C 07. L 30. L 39. L 42. L 16. E 19. E 31. E 40. E 03. F 06. F 08. F 12. F

.69 .84 .47 .78 .07 .09 .01 -.07 .07 .11 .12 .17 .10 .03 .06 .21 .19 .17 .07 .20

.03 .03 .01 .06 .50 .62 .88 .84 .14 .15 .08 .14

.05 .13 .02 .18 .15 .12 .03 .04 .64 .64 .86 .70 .11 .24 .26 .15 .22 .29 .23 .22

.11 .15 .13 -.03 .05 -.12 -.01 -.05 .25 .13 .17 .22 .71 .75 .62 .74 .25 .23 .15 .24

.03 .12 .22 .12 .11 -.09 -.07 -.11 .22 .27 .23 .19 .19 .24 .30 .11 .71 .72 .72 .70

.50 .75 .29 .66 .30 .44 .78 .73 .55 .54 .84 .62 .58 .68 .55 .62 .66 .68 .62 .64

.57 .74 .41 .64 .07 .00 .04 .07 .08 .06 .19 .19 .20 .15 .20 .28 .20 .14 .15 .19

.03 .11 .01 .07 .51 .84 .88 .85 .19 .09 .11 .04

.05 .17 .02 .23 .14 .06 -.04 .07 .57 .60 .79 .77 .29 .32 .41 .38 .27 .27 .23 .24

.04 .06 .17 .07 .04 .07 -.01 -.08 .15 .17 .11 .20 .55 .48 .51 .63 .12 .14 .19 .28

.03 .15 .11 .29 .08 -.02 -.09 -.12 .30 .23 .24 .24 .42 .44 .44 .33 .80 .77 .73 .76

.33 .62 .21 .56 .29 .72 .79 .75 .48 .46 .74 .72 .61 .55 .69 .74 .77 .71 .64 .73

.72 .90 .32 .73 .09 .06 -.01 .02 .09 .08 .10 .10 .01 .04 .18 .13 .28 .11 .01 .10

.07 .03 .02 .05 .57 .74 .91 .86 .32 .15 .14 .15 .00 .00 .07 -.03 .05 .27 -.05 -.04

.02 .11 .12 .11 .15 .10 .08 .14 .53 .71 .89 .65 .06 .15 .23 .10 .11 .22 .29 .26

.03 .02 .29 .17

.05 .10 .20 .14 .14


-.04

.53 .83 .24 .59 .37 .56 .83 .76 .58 .61 .86 .51 .71 .62 .59 .71 .50 .80 .55 .57

-.05 .06 .01 -.03


.24 .18 .09 .09 .83 .69 .60 .82 .16 .19 .17 .28

.03 .03 .36 .21 .20 .21 .13 .33 .38 .08 .62 .79 .66 .64

-.05 -.04 -.14 .06


.01 -.03 .11 -.06

-.02 -.01 -.13 -.11 -.04 -.07 -.06 .02

Note. Single letters in column headings and stub denote the proposed power base for each item: R = reward, C = coercive, L = legitimate, E = expert, and F = referent. The Appendix shows the specific wording of each item. Communalities are denoted by h2.

analyses for Sample A are discussed first. These analyses were performed prior to the collection of data from Samples B and C, and they were used to select items from the initial item pool for inclusion in the new scales (for administration to Samples BandC).

Initial Item Selection Study: Sample A


Initial exploratory factor analyses. The 42 power items administered to Sample A were factor analyzed by using the principal-axis factor method, with squared multiple correlations as initial communality estimates, and a varimax rotation. A second factor analysis was then conducted, retaining the 30 items with appropriate a priori factor loadings of .40 or greater. A close examination of the resulting factor structure and item content revealed that there would be little change in scale domain sampling adequacy, factor structure, or internal consistency reliability if the scales were further reduced. Two sets of items were then retained for further analysis, one composed of 25 items, and one of 20 items. The following analyses were conducted on both sets of items. All evidence suggests that the 20item scales were superior, and thus only these results are subsequently reported. These 20 items involve five scales of 4 items each. The instructions, response format, and final 20 items are presented in the Appendix. The five-factor solution for the best 20 items in Sample A explained 69.5% of the total item variance and is presented in the left-hand columns of Table 1. As shown in those columns, the item-loading patterns are excellent. Initial confirmatory factor analyses. LISREL maximum likelihood confirmatory factor analysis makes it possible to statisti-

cally assess the goodness of fit of a factor structure to a set of data (Joreskog & Sorbom, 1985). Thus, we conducted LISREL analyses on the 20 items of the power scales, initially specifying that each item load on only its appropriate factor (i.e., no crossloadings were allowed) and that the five factors were not orthogonal (i.e., intercorrelations were allowed among the factors). The resulting goodness-of-fit (.90) and adjusted goodness-offit (.86) indices, as well as the root-mean-square residual (.06), indicated an excellent fit to the data. All hypothesized factor loadings were .47 or greater, and all were significant at p < .001 or better. Allowing the estimation of initially nonhypothesized crossloadings, which significantly (p < .001) improved the overall goodness of fit of the factor structure to the data, produced only one small loading (.23) for an item that loaded .51 on its appropriate factor. Thus, because the goodness-of-fit index for this solution was unchanged (.90) and only slightly improved with respect to the adjusted index (.87) and root-mean-square residual (.05), the theoretical dimensionality of the 20 items and five scales appeared to be strongly supported.

Follow-Up Scale Construction Studies: Samples B and C


After the analyses were conducted, the 20 items shown in Table 1 (along with the other measures described) were administered in Samples B and C, and the following analyses conducted. Replicative exploratory factor analyses. The principal-axis factor analyses described earlier were also conducted on Samples B and C, with similar results; these are shown in the middle and right-hand columns, respectively, of Table 1. For Sample B, the five-factor solution explained 70.0% of the total item vari-

564

TIMOTHY R. HINKIN AND CHESTER A. SCHRIESHEIM bases in the three samples are presented in Table 2, along with their coefficient alpha reliabilities. As shown in Table 2, the scales were highly reliable in all three samples (cf. Nunnally, 1978). In Sample A, legitimate, expert, and referent power were highly intercorrelated as well as moderately correlated with reward power. Coercive power was not strongly related to any other power, its only significant relationship being with legitimate power. A very similar pattern of relationships was also obtained in Samples B and C, although coercive power demonstrated a significant yet very small positive relationship with referent power in Sample C. Thus, although some large and significant relationships existed between some of the powers, the overall correlations demonstrated a meaningful amount of perceptual independence, particularly given the high obtained scale reliabilities and earlier factor analytic results. Factor analyses of item discriminant validity. Because of the nature of the dependent variables typically used in research on power in organizations (cf. Podsakoff & Schriesheim, 1985) and because it is important that supposedly descriptive measures (the power scales) have minimal contamination with affective evaluation (e.g., satisfaction; cf. Ferratt, Dunham, & Pierce, 1981), we felt that an additional examination would be helpful in assessing discriminant validity. Using the data from all three samples (separately), we factor analyzed the 15 commitment items with the power items to determine whether respondents discriminated between their affective attitudes and the perceived bases of power. We repeated the analyses in all three samples for the 22 satisfaction items. The factor analytic results, as a whole, appeared supportive of the discriminant and, hence, construct validity of the new measures. None of the individual power items demonstrated inappropriate loadings across the three samples. The obtained item-loading patterns were thus supportive of the relative freedom of the new scales from substantial contamination by affective evaluation. Concurrent validity analyses. To examine the relationships between the power scales and the dependent variables that were measured in Samples A, B, and C, we computed zero- and fourth-order Pearson correlations (the fourth-order partial correlations involved removing the effects of the four other power scales from the independent-dependent variable relationships). The results of the zero-order correlations between the power scales and the measures of satisfaction and commitment are presented in Table 3. In Sample A, (a) expert and referent power demonstrated strong positive relationships with all of the satisfaction measures; (b) reward and legitimate power demonstrated weaker, yet significant, positive relationships; and (c) coercive power demonstrated moderately strong negative relationships. In addition, (d) expert and referent power had moderate positive relationships, whereas (e) coercive power demonstrated a moderate negative relationship with commitment. The results for Samples B and C are also presented in Table 3. The pattern of relationships was very similar to those of Sample A, except that in Sample C the magnitude of the negative correlations between coercive power and the satisfaction variables was much smaller and there were no significant relationships between coercive power and commitment. However, on an overall basis, the pattern of relationships was very similar across the three samples. To date, researchers have not examined the independent rela-

ance. All items loaded as predicted, and all but two of these loadings were greater than .50. (The four expert power items, however, had bothersome cross-loadings on the referent power factor.) The results for Sample C were quite similar to those for Samples A and B. The five-factor solution explained 70.2% of the total item variance, and all of the items had strong loadings on their appropriate factor, except for Reward Power Item 33. No items had even moderate cross-loadings on other factors. Replicative confirmatory factor analyses. To statistically test the goodness of fit of the hypothesized scale factor structure to the data of Samples B and C, we conducted LISREL (Joreskog & Sorbom, 1985) analyses mirroring those reported for Sample A. In Sample B, the goodness-of-fit (.92) and adjusted goodnessof-fit (.90) indices and root-mean-square residual (.05) for the initial factor model (with no cross-loadings specified) indicated excellent fit of the hypothesized factor structure with the data. (All loadings exceeded .42 and were significant atp < .001 or better.) Estimating two significant (p < .001) cross-loadings produced only slight changes in the goodness-of-fit (.93) and adjusted goodness-of-fit (.91) indices as well as in the rootmean-square residual (.04). Because these cross-loadings were only .22 and .21 (and occurred with items having appropriate loadings of .47 and .64, respectively), the a priori factor structure of the scales appeared well-supported in Sample B. The LISREL results for Sample C indicated an initial goodness-of-fit index of .84, with an adjusted index of .79, and a root-mean-square residual of .09. The lowest estimated loading was .39, and all loadings were significant at p < .001 or better. Allowing the estimation of seven significant (p < .001) crossloadings improved the goodness-of-fit (.89) and adjusted goodness-of-fit indices (.84) and root-mean-square residuals (.06) of the factor model. However, five of these cross-loadings were less than .28, and the two highest were only .32 and .39 (and these occurred for items with loadings on their appropriate factors of .45 and .61, respectively). Thus, although suggesting a slightly more complex structure, the LISREL results in Sample C also provided good support for the factor structure of the new power scales.

Item and Scale Examination and Assessment Process


The process of construct-validating any measure is one that never ends (Nunnally, 1978; Schwab, 1980). This suggests that no measure can be said to be validated in any final sense, only that more or less data can be said to support or not support a particular measure for a particular purpose. In recognition of this, the psychometric adequacy of the new power scales was initially explored through analyses of scale independence, item discriminant validity, and concurrent validity. For clarity and conciseness, the results from all three samples are presented together. Assessment of scale independence. Although the five bases of power are conceptually distinct, Bass (1981) and others have noted that there are theoretical reasons for expecting the powers to be correlated in real-world samples. Thus, the power scales were examined for independence, although significant intercorrelations were expected among at least some of the powers. The zero-order intercorrelations computed among the power

NEW POWER SCALES

565

Table 2 Scale Intercorrelations


Reward Coercive Legitimate Expert Referent
C A B C

A
Reward Coercive Legitimate Expert Referent (.80)

B
(.77)

C
(.77)

B
(-85)

C
(.86)

.06 .29 .27 .35

.11 .36 .43 .40

.14 .29 .29 .32

(.80)

.20
-.06 -.07

.16
-.08 -.06

.41 .06 .16

(.87)

(.85)

(.86)

.46 .53

.64 .57

.39 .55

(.85)

(.90)

.52

.72

(-83) .52

(.88)

(.87)

(.86)

Note. Coefficient alpha reliabilities are in parentheses. In Sample A ( J V = 251), correlations attain significance at . 10(p< .05) and at .15 (p< .01); in Sample B (./V = 375), correlations attain significance at .09 ( p < .05) and at. 14 ( p < .01); in Sample C(N= 220), correlations attain significance at .11 ( p < .05)andat .16(p< .01).

tionships of the five bases of power with criterion variables; this has created substantial difficulties in interpreting research results (Podsakoff & Schriesheim, 1985). For this reason, the fourth-order partial correlations for the above dependent variables are presented in Table 4. The results of these analyses present a picture quite different from those of the zero-order correlations. In Sample A, expert and referent power demonstrated moderate positive correlations with the satisfaction measures, whereas coercive power had moderate negative correlations. Reward and legitimate power had no significant relationships with any of the measures of satisfaction. Only coercive power had a strong relationship with commitment, although the relatively weak commitment-referent power correlation did attain statistical significance. A very similar pattern of relationships was revealed in Sample B, except that reward power and expert power had weak positive commitment correlations, and referent power had a weak negative correlation with commitment. Sample C, however, demonstrated a somewhat different pattern of results. The relationships between referent power and the criterion variables were weaker than those obtained in Samples A and B. Reward power also had a moderately strong positive relationship with commitment, whereas coercive power was not significantly correlated with commitment. Discussion The results of these correlation analyses provide an opportunity to compare the findings of this study with those reviewed

by PodsakofFand Schriesheim (1985). Looking at the zero-order correlations, one can see many discrepancies with the prior literature, although zero-order correlational analyses were used in most prior research. With regard to satisfaction with supervision, our results for coercive, expert, and referent power are consistent with those of past research. However, both reward power and legitimate power were consistently positively related to satisfaction with supervision in this study, whereas in past research, positive, negative, and no relationships had been found. Past researchers using global satisfaction as a dependent variable had also arrived at very inconsistent results. Reward, coercive, legitimate, and referent powers had demonstrated primarily nonsignificant relationships with global satisfaction. Such was not the case in our study, which had very consistent results across the three samples. In addition, no conclusions could be drawn from past research with respect to expert power, but our study clearly uncovered a strong positive relationship. The relationships between the power bases and organizational commitment were also more distinct than in past research, wherein the relationships with all but expert power had been nonsignificant. We feel that the discrepancy between these results and those of prior research is primarily due to effects caused by the single-item ranking scales typically used in prior research, as suggested by Podsakoff and Schriesheim (1985) and Schriesheim etal. (1985). Of perhaps more interest and greater substantive importance are the results of the partial correlation analyses. It has been demonstrated that the power bases are not empirically indepen-

Table 3 Power Scale Zero-Order Correlations With Satisfaction and Commitment


Satisfaction Global Power base Reward Coercive Legitimate Expert Referent Technical Human relations urganizauonai commitment

A .19
-.25

B .28
-.27

C .19
-.07

A .21
-.26

B .30
-.31

C .13
-.12

A .18
-.30

B .28
-.30

C .18
-.07

A .12
-.31

B .14
-.29

C .38 .01 .22 .33 .30

.28 .49 .52

.33 .50 .51

.24 .45 .34

.27 .54 .44

.46 .72 .63

.24 .52 .34

.22 .43 .52

.41 .65 .62

.24 .45 .34

.16 .24 .29

.12 .22 .11

Note. In Sample A (N = 251), correlations attain significance at .10 (p < .05) and at .15 (p < .01); in Sample B(N= 375), correlations attain significance at .09 (p < .05) and at .14 (p < .01); in Sample C (TV = 220), correlations attain significance at. 11 ( p < .05) and at. 16 ( p < .01).

566

TIMOTHY R. HINKIN AND CHESTER A. SCHRIESHEIM Table 4 Power Scale Fourth-Order Partial Correlations With Satisfaction and Commitment Satisfaction Global Power base Reward Coercive Legitimate Expert Referent A .01 -.23 -.01 .30 .32 B .09 -.29 .05 .14 .24 C .04 -.14 .07 .32 .11 A .05 -.26 .01 .39 .19 Technical B .02 -.38 .08 .43 .23 C .04 -.19 .08 .42 .10 Human relations A .01 -.28 -.05 .22 .36 B .02 -.34 .03 .32 .29 C .04 -.26 .03 .27 .20 Organizational commitment A .04 -.31 .07 .08 .14 B C

.11 .28 -.29 .06 .06 .03 .14 .17 -.11 .09

Note. In Sample A (N = 251), correlations attain significance at .10 (p < .05) and at .15 (p < .01); in Sample B (N = 375), correlations attain significance at .09 (p < .05) and at .14 (p < .01); in Sample C (N= 220), correlations attain significance at. 11 (p< .05)andat.l6(p< .01).

dent (Podsakoff & Schriesheim, 1985; see also Table 2), and most supervisory personnel would theoretically be expected to possess at least a small degree of each power base (Bass, 1981). Thus, the partial correlation analyses more accurately depict the "true" (independent) effects of each base of power. With regard to satisfaction with supervision, the relationships with expert and referent power are consistent with past research. However, in all cases, coercive power had a negative independent (partial) relationship with satisfaction with supervision; this is a much stronger finding than that found in prior research. In past research, the relationships between reward and legitimate power and satisfaction with supervision had varied greatly, but not so for the partial correlations of this study. Neither base of power was significant in any sample, suggesting that these types of power have no statistically independent relationships with supervision satisfaction. In regard to global satisfaction, the results of our study are consistent with past studies for reward power and legitimate power, wherein no relationships exist. However, the results for coercive, expert, and referent power are inconsistent with prior research, as these three power bases demonstrated strong independent relationships (where they had not been found previously). Although some significant relationships did emerge between the various bases of power and the commitment measure, the correlations were small and may have little practical significance. Perhaps supervisory power does not have strong relationships, positive or negative, with organizational commitment. To summarize, it is important to note that the substantive results of this study, although somewhat consistent with past research, differ in major ways. However, because of the diversity and independence of the samples used here, the results of this research can be viewed with some confidence. Another reason for viewing the current findings positively is that the current power scales, as well as the partial correlation analyses used, overcome two major shortcomings of past research (Podsakoff & Schriesheim, 1985).

serious reservations that Podsakoff and Schriesheim (1985) had about methodological inadequacies of the extant literature on the French and Raven (1959) bases of power seem justified. Despite the need for additional research, the results of this study are quite positive and encouraging with respect to the new measures. The results indicate that the new scales have built-in content validity (cf. Nunnally, 1978) and that they have demonstrated reasonable internal consistency reliability, factor structure, discriminant validity, and criterion-related validity. Future investigators would therefore be best advised to use the new scales in preference to ad hoc measures, particularly when the new scales are theoretically congruent with the nature and purpose of the investigation. The new scales could be particularly useful in examining relationships between leader behavior, such as reward and punishment behavior (cf. Podsakoff, Todor, Grover, & Huber, 1984), and perceived leader power, relationships that are seen as theoretically important (cf. Bass, 1981; Yukl, 1981) but that have not yet been examined.

References
Bass, B. (1981). Stogdill's handbook of leadership. New York: Free Press. Churchill, G. A. (1979). A paradigm for developing better measures of marketing constructs. Journal of Marketing Research, 16, 64-73. Dahl, R. A. (1957). The concept of power. Behavioral Science, 2, 201215. Etzioni, A. (1961). A comparative analysis of complex organizations. New York: Free Press. Ferratt, T. W., Dunham, R. B., & Pierce, J. L. (1981). Self-report measures of job characteristics and affective responses: An examination of discriminant validity. Academy of Management Journal, 24, 780794. French, J. R. P., & Raven, B. (1959). The bases of social power. In D. Cartwright & A. Zander, Group dynamics (pp. 150-167). New York: Harper & Row. Gioia, D. A., & Sims, H. P. (1983). Perceptions of managerial power as a consequence of managerial behavior and reputation. Journal of Management, 9, 7-26. Grimes, A. J. (1978). Authority, power, influence, and social control: a theoretical synthesis. Academy of Management Review, 3, 724-735. Hinkin, T. R. (1985). Development and application of new social power measures in superior-subordinate relationships. Unpublished doctoral dissertation, University of Florida, Gainesville.

Conclusion
This research makes two valuable contributions to the study of power. First, measures of power were developed that demonstrate adequate psychometric properties on at least a preliminary basis. Second, by using these scales, one can see that the

NEW POWER SCALES Joreskog, K. G., & Sorbom, D. (1985). LISREL vi. Mooresville, IN: Scientific Software. Kanter, R. M. (1979). Power failure in management circuits. Harvard Business Review, 57, 65-75. Korman, A. K. (1974). Contingency approaches to leadership. In J. G. Hunt & L. L. Larson (Eds.), Contingency approaches to leadership (pp. 189-195). Carbondale: Southern Illinois University Press. Locke, E. A. (1976). The nature and causes of job satisfaction. In M. D. Dunnette (Ed.), Handbook of industrial and organizational psychology (pp. 189-195). Chicago: Rand McNally. McClelland, D. C. (1975). Power: The inner experience. New York: Irvington. McClelland, D. C., & Burnham, D. H. (1976). Power is the great motivator. Harvard Business Review, 54, 100-110. Mintzberg, H. (1983). Power in and around organizations. Englewood Cliffs, NJ: Prentice-Hall. Mowday, R. T., Steers, R. M., & Porter, L. W. (1979). The measurement of organizational commitment. Journal of Vocational Behavior, 14, 224-247. Nunnally, J. C. (1978). Psychometric theory. New York: McGraw-Hill. Patchen, M. (1974). The locus and basis of influence on organizational decisions. Organizational Behavior and Human Performance, 11, 195-221. Pfeffer, J. (1981). Power in organizations. Marshfield, MA: Pitman. Podsakoff, P. M., & Schriesheim, C. A. (1985). Field studies of French

567

and Raven's bases of power: Critique, reanalysis, and suggestions for future research. Psychological Bulletin, 97, 387-411. Podsakoff, P. M., Todor, W. D., Grover, R. A., & Huber, V. L. (1984). Situational moderators of leader reward and punishment behavior Fact or fiction? Organizational Behavior and Human Performance, 34, 21-63. Porter, L. W, Steers, R. M., Mowday, R. T, & Boulian, P. V. (1974). Organizational commitment, job satisfaction, and turnover among psychiatric technicians. Journal of Applied Psychology, 59, 603-609. Schriesheim, C. A. (1978). Development, validation, and application of new leadership behavior and expectancy research instruments. Unpublished doctoral dissertation, Graduate School of Business, Ohio State University, Columbus. Schriesheim, C. A., Hinkin, T, & Podsakoff, P. M. (1985). Have measurement shortcomings produced erroneous results in field investigations of the five French and Raven power bases? An empirical examination. Proceedings of the Southern Management Association, 100102. Schwab, D. P. (1980). Construct validity in organization behavior. In B. M. Staw & L. L. Cummings (Eds.), Research in organizational behavior (Vol. 2, pp. 3-43). Greenwich, CT: JAI Press. Yukl, G. A. (1981). Leadership in organizations. Englewood Cliffs, NJ: Prentice-Hall. Weiss, D. J., Dawis, R. V., England, G. W, & Lofquist, L. H. (1967). Manual for the Minnesota Satisfaction Questionnaire. Minneapolis: University of Minnesota, Industrial Relations Center.

Appendix Final Scale Items'


Instructions: Below is a list of statements which may be used in describing behaviors that supervisors in work organizations can direct toward their subordinates. First carefully read each descriptive statement, thinking in terms of your supervisor. Then.decide to what extent you agree that your supervisor could do this to you. Mark the number which most closely represents how you feel. Use the following numbers for your answers: (5) = strongly agree (4) = agree (3) = neither agree nor disagree (2) = disagree (1) = strongly disagree My supervisor can . . . (Reward Power) 02. increase my pay level. 27. influence my getting a pay raise. 33. provide me with special benefits. 38. influence my getting a promotion. (Coercive Power) 04. give me undesirable job assignments. 18. make my work difficult for me. 21. make things unpleasant here. 22. make being at work distasteful. (Legitimate Power) 07. make me feel that I have commitments to meet. 30. make me feel like I should satisfy my job requirements. 39. give me the feeling I have responsibilities to fulfill. 42. make me recognize that I have tasks to accomplish. (Expert Power) 16. give me good technical suggestions. 19. share with me his/her considerable experience and/or training. 31. provide me with sound job-related advice. 40. provide me with needed technical knowledge. (Referent Power) 03. make me feel valued. 06. make me feel like he/she approves of me. 08. make me feel personally accepted. 12. make me feel important. ' The items shown are grouped for reader convenience only; the item number corresponds to the order actually used on the original survey questionnaire.

Received January 7,1988 Revision received October 28,1988 Accepted November 1, 1988

S-ar putea să vă placă și