Measurement and Empirical Validity
STUDY 3 STUDY 3 3
leadership scales. These results provide evidence of convergent and discriminant validity for the Conger-Kanungo subscales.
This refinement of the Conger-Kanungo scale resulted in a revised 20-item measure of charismatic leadership with five behavioral dimensions measured by five distinct subscales. The alpha reliability of the total scale is .88, justifying its use as an overall measure of charismatic leadership as proposed in the Conger-Kanungo model.
Although the data in Table 3.12 provide evidence of the convergent and discriminant validity of the subscales, there is some evidence of the convergent and discriminant validity of the overall scale. The overall Conger-Kanungo scale has a correlation of .69 with the Bass charisma scale. Table 3.12 reveals that this degree of convergence is attributable to the positive relationship between four of the Conger-Kanungo subscales and the BASS scale. On the other hand, the dimension of "unconventional behavior" differentiates the Conger-Kanungo scale from the Bass scale, as is revealed by the slight but significant negative relationship between the subscale UB and the BASS scale.
Although the confirmatory factor analysis (CFA) supported a five-factor solution with adequate fit, it must be borne in mind that the CFA was performed on the same data collected in Study 2 as was the original principal components analysis. It is necessary, therefore, to replicate this result using other independent samples. This was done in Study 3.
Method
The study involved administering a questionnaire containing the Conger- Kanungo scale and the above leadership measure to a sample of organizational employees. To minimize possible method bias, the questionnaire was divided into two parts. Part 1 contained the 20-item Conger-Kanungo scale and demo- graphic variables. Part 2 included select items from the MPS. With the coopera- tion of the human resources department of a large multinational corporation based in the United States, the questionnaire was administered to employees attending a company training program. Participants were asked to describe their superior in terms of the questionnaire items. Participation was voluntary, and respondents were instructed to allow a minimum of 24 hours between the completion of the two parts.
The mean age of the total sample of 103 respondents (66% male) was 39.6 years. Ninety-seven percent had at least a college degree, and 87% were married.
The respondents were predominantly from the middle (69%) and senior (30%) levels and were drawn from all functional departments in the organization.
Salary levels ranged from $45,000 to $60,000 to more than $85,000. The average number of years worked was 10.8, and the mean job tenure was 2.8 years.
The Conger-Kanungo scale was first subjected to a confirmatory factor analysis (CFA) to reconfirm the five-factor structure. The subscales were then related to the MPS dimensions through a structural latent variable model to test specific convergent and discriminant validity hypotheses.
Analysis and Results
The confirmatory factor analysis (CFA) on the 20-item Conger-Kanungo scale was conducted using the LISREL 7 program. Given the somewhat small sample size (103) and the relatively large number of items (20), there is a possibility that the number of estimated parameters in the model might be too large for the available sample size. The effects of a poor sample size-to-estimator ratio can be minimized by using testlets of composite items. It may be recalled that the subscale "strategic vision and articulation" has seven items. In line with Anderson and Gerbing's (1988) suggestion, three indicators were constructed to represent this subscale by combining items through random assignment (see Shore, Barksdale, and Shore [1995] for an example of this technique). For the subscale "sensitivity to the environment" with four items, one composite indi- cator was constructed by combining two randomly selected items. The remain-
ing two items were retained as indicators, unaltered. The three other subscales have three items each, and these were retained as the respective indicators. Thus, all five subscales were represented by three indicators apiece.
The CFA resulted in a model with χ2(80) = 103.27, ρ = .041. Comparison to an absolute null model with χ2(105) = 698.77 yielded a Normed Fit Index (NFI) of .85 and a Non-Normed Fit Index (NNFI) of .95. These results lend support for the five-factor structure discussed earlier. The alpha reliabilities of the five subscales were .86 (strategic vision and articulation), .72 (sensitivity to the environment), .75 (sensitivity to member needs), .85 (personal risk), and .81 (unconventional behavior). The alpha reliability of the overall 20-item Conger- Kanungo scale was .87.
Convergent and Discriminant Validity
Evidence of convergent and discriminant validity can be assessed by relating the Conger-Kanungo subscales to the various MPS subscales. It is expected that the Conger-Kanungo subscales should be differentially related to the MPS subscales. For example, the subscale "strategic vision and articulation" (SVA) should correlate highly with the MPS subscale "inspiring" (INS) while having a low correlation with the MPS subscale "monitoring" (MON). This hypothesis is based on the expectation that the intent of leader behaviors classified as
"strategic vision and articulation" is to inspire employees, whereas monitoring is a more mundane managerial function. On the other hand, the Conger-Kanungo subscale "sensitivity to the environment" (SE) should correlate highly with MPS subscales associated with activities involving the assessment of the internal environment of the organization such as "monitoring" (MON), "problem-solving"
(PRB), and "recognizing" (RCG). With regard to the other Conger-Kanungo subscales, one can expect the subscale "sensitivity to member needs" (SMN) to correlate highly with the MPS subscale "supporting" (SUP). There is no a priori reason to expect significant relationships between the subscales "personal risk"
(PR) or "unconventional behavior" (UB) and any of the MPS subscales. These two Conger-Kanungo subscales constitute the distinguishing features of the Conger-Kanungo scale.
The above hypotheses were tested by a structural latent variable model using LISREL 7, with the five subscales treated as independent exogenous variables and the MPS subscales (INS, MON, PRB, RCG, and SUP) treated as dependent, endogenous variables. The alpha reliabilities for these latter scales are .87 (INS), .75 (MON), .81 (PRB), .87 (RCG), and .87 (SUP). Table 3.13 shows the LISREL (gamma) estimates.
TABLE 3.13 LISREL (Gamma) Estimates of the Relationships Between Conger-Kanungo Subscales and MPS Subscales
MPS Subscales SVA SE SMN PR UB
INS .400*** .167 -.041 .063 .067
MON -.064 .503*** -.112 -.045 .033
PRB .111 .515** -.175 -.058 .083
RCG .132 .477** -.079 .001 .016
SUP -.143 -.080 .706*** -.100 .171*
SOURCE: Adapted from "Measuring Charisma: Dimensionality and Validity of the Conger-Kanungo Scale of Charismatic Leadership" (p. 297), Conger, Kanungo, Menon, and Mathur. © 1997 by the Canadian Journal of Administrative Sciences. Reprinted with permission.
NOTE: Sec Table 3.11 for definitions of subscales.
*p < .05. **p < .01. ***p < .001.
As hypothesized, the Conger-Kanungo subscale "strategic vision and articu- lation" (SVA) is significantly and strongly related to the MPS subscale "inspir- ing" (INS). Furthermore, it is not significantly related to MPS subscales such as
"monitoring" (MON) or "problem solving" (PRB). On the other hand, the Conger-Kanungo subscale "sensitivity to the environment" (SE) is strongly related to the MPS subscales "monitoring" (MON), "problem solving" (PRB), and "recognizing" (RCG), as hypothesized. In line with expectations, the Conger-Kanungo subscale "sensitivity to member needs" (SMN) is strongly and significantly related to the MPS subscale "supporting" (SUP). Finally, as hy- pothesized, the Conger-Kanungo subscales "personal risk" (PR) and "uncon- ventional behavior" (UB) are not significantly related to any of the MPS subscales, with the exception of "supporting" (SUP), which shows a mild positive relationship with "unconventional behavior" (UB).
Overall, the results provide support for the five-factor structure of the Conger-Kanungo scale using an independent sample. The subscales as well as the overall Conger-Kanungo scale have good to excellent alpha reliabilities that range from .72 to .87. Convergent and discriminant validity were demonstrated by the distinct pattern of relationships between the Conger-Kanungo subscales and the subscales of the MPS.
STUDY 4: A TEST OF DISCRIMINATORY POWER