Assessing the discriminant validity of the Leadership practices inventory

Theories of transformational leadership attempt to describe leadership behaviours which are associated with above average performance by subordinates. Kouzes' and Posner's (1987) visionary leadership describes five key transformational leadership behaviours which can be assessed by the Leadership Practices Inventory (LPI: Kouzes & Posner, 1988). This study examined the construct validity of the LPI by using confirmatory factor analysis to test three alternate conceptual models. The sample consisted of 1400 subordinates who worked for an international finance company. It was concluded that LPI assessed an over-arching higher order transformational leadership. The implications of the findings for management development programmes were discussed.

Transformational leadership theories endeavour to explain how leaders motivate and commit subordinates to well-above-average organizational performance (e.g. Bass, 1985; Conger & Kanungo, 1988; Tichy & Devanna, 1986). It is argued that through the use of transformational leader behaviours, subordinates become inspired to transcend their own interests and become committed to achieving the leaders' vision for the organization. A key aspect of transformational leadership theories is that they attempt to operationalize effective leader behaviours. However, recent evidence indicates that measures designed to capture transformational leader behaviour lack discriminant validity (Bycio, Hackett, & Allen, 1995; Carless, 1998; Den Hartog, Van Muijen, & Koopman, 1997; Tracey & Hinkin, 1998).

Kouzes and Posner's (1987) visionary leadership theory has been used extensively by business organizations for management development purposes (Posner & Kouzes, 1993).(n1) In common with most management development programmes, an assessment tool, the Leadership Practices Inventory (LPI: Kouzes & Posner, 1988), is used to obtain feedback about the leadership practices of participants. The authors claim the LPI assesses five leader behaviours: (a) Challenging the Process, the extent the leader questions assumption, experiments and takes risks; (b) Inspiring a Shared Vision, the degree the leader describes an exciting view of the future; (c) Enabling Others to Act, the amount of cooperative and participative decision making used by the leader; (d) Modelling the Way, the extent the leader consistently practices his or her espoused values; and (e) Encouraging the Heart, the degree the leader gives positive feedback, publicly recognizes individual contributions and celebrates team achievements. However, there has been a notable lack of published evidence substantiating the original claims of the authors. Thus, the aim of this study was to examine the construct validity of the LPI. Confirmatory factor analysis was used to test three alternate conceptual models of transformational leadership.

Consistent with the notion that transformational leadership is multidimensional (Bass, 1985; Kouzes & Posner, 1987), the first model examined the proposition that transformational leadership can be defined by separate and distinct behaviours (Model 1). An alternate view, tested in the second model, is that transformational leader behaviours are so highly related that it is more appropriate to conceptualize them as a single factor (Model 2: Conger & Kanungo, 1988). Model 3 is a test of the assumption that leadership is a hierarchical concept explained by a number of independent behaviours that share in common a strong relationship with a higher-order construct. Recent evidence on an alternate measure of transformational leadership, the Multifactor Leadership Questionnaire (Avolio, Bass, & Jung, 1995) supports this view (Avolio, Bass, & Jung, 1999; Carless, 1998).

Leadership scholars have been criticized for their focus on the apex of the organizational hierarchy (e.g. Bryman, 1992). In contrast, the transformational theories of Kouzes and Posner (1987) and Bass (1985) have widespread applicability, in particular to, lower and middle-level management (Bass, 1997; Jermier, 1993; Lowe, Kroeck, & Sivasubramaniam, 1996). Thus, the target group of this study was managers at lower to middle-level management positions.

Data were obtained for research purposes from 1440 subordinates (response rate: 54%) who worked in Australia for an international banking organization. The majority of subordinates were female (69%) and their mean age was 31.2 years (SD = 7.92). The managers with whom they worked, were in lower to middle-level management positions.

The LPI (Kouzes & Posner, 1988) measures five leadership practices: Challenging the Process (alpha = .81), Inspiring a Shared Vision (alpha = .90), Enabling Others to Act (alpha = .89), Modelling the Way (alpha = .86), and Encouraging the Heart (alpha = .94). Six items are used to measure each of the five scales. The response format was a 5-point Likert scale ranging from 'Rarely or never' (1) to 'Very frequently, if not always' (5) was employed.

LISREL 8 (Joreskog & Sorbom, 1993) was used to analyse the 30 LPI items. A covariance matrix was used and the method of estimation was maximum likelihood. The goodness-of-fit statistics for the tested models are shown in Table 1. The chi-square difference test shows that the five factor model (Model 1) is a significantly better fit than the single factor model (Model 2): Chi2, sub diff (10) = 2641, p < .001. This conclusion is supported by the Relative Noncentrality Index (RNI), the goodness of fit (GFI), the adjusted goodness of fit (AGFI), the Non-Normed Fit Index (NNFI), Expected Cross-Validation Index (ECVI) and Akaike's Information Criterion (AIC). The root mean square error of approximation (RMSEA), the standardized Root Mean Square Residual (RMSR), the NNFI and the RNI are within the bounds of acceptable models (Bagozzi & Heatherton, 1994; Bentler & Bonnett, 1980; Browne & Cudeck, 1993; Byrne, 1998). In Model 1, the intercorrelations between the subscales were extremely high (average correlation = .89).

The third model examined was a higher order with five first-order factors. It can be seen that compared to a first-order five factor model (Model 1), the higher-order model (Model 3) is a slightly poorer fit. This is to be expected, as the 'first-order model represents an upper bound or optimum for all subsequent HCFA [hierchical confirmatory factor analysis], (Marsh & Hocevar, 1988, p. 108). Although in comparison to the first-order model, the higher-order model is a poorer fit, the NNFI and RNI indicate the higher-order model (Model 3) is an adequate fit to the data.

The higher-order model explained most of the variance in the sub-scales: 92% of Challenging the Process, 90% of Inspiring a Shared Vision, 81% of Enabling Others to Act, 94% of Modelling the Way and 76% of Encouraging the Heart. These findings provide evidence that a high proportion of the covariation among the first-order factors is explained by a higher-order construct. Following the recommendation of Marsh and Grayson (1994) the variance of the LPI was partitioned into three components: (a) common variance, that is, the variance which is shared with the second-order factor, (b) unique variance, that which is specific to the item, and (c) error variance, the variance which is unexplained and due to measurement error. Most of the variance associated with the items was accounted by the common factor (M common variance = .50) or was error variance (M error variance = .43). A small proportion of the variance was item-specific variance (M unique variance = .08).

In noting the problem of evaluating higher-order models, Marsh and his colleague (Marsh, 1987; Marsh & Hocevar, 1988) notes that the goodness of fit of higher-order models cannot exceed the corresponding first-order model and because the number of parameters are less than the first-order model the fit may be slightly poorer. Marsh and Grayson (1994) suggest the investigator must 'rely on subjective interpretation to determine whether a difference [between a first-order and higher-order model] is important' (p. 127). Hence, it is concluded that the LPI assesses an over-arching transformational leadership factor. This is based on the evidence that the second-order model (Model 3) is an adequate fit to the data and is consistent with Marsh's (1987) recommendation that a higher-order model is acceptable as long as it has a 'reasonable' fit to the data. In addition, most of the shared variance is explained by a higher-order factor and there is little unique variance in the sub-scales.

This study examined three alternate models of transformational leadership. The analyses indicated that two models provided an acceptable fit to the data: a single order five-factor model (Model 1) and a higher-order model (Model 3). Although comparison of the fit indices indicated that the single order multidimensional model was a better fit to the data, it was argued that the higher order model provided the better substantive interpretation. This view is consistent with recent findings on other measures of transformational leadership (Carless, 1998; Tracey & Hinkin, 1998).

The findings of this study suggest that while it is possible to distinguish conceptually among separate transformational leader behaviours, either these distinctions are not captured by the LPI or subordinates do not notice the differences. Another plausible explanation is that based on observations of transformational leader behaviour, subordinates attribute charisma to the person and this invokes an automatic categorization of a charismatic person (Gardner & Avolio, 1998). The attribution of charisma leads to automatic processing of leader behaviours (Gardner & Avolio, 1998), which implies that subordinates may not distinguish between separate leader behaviours. Indirect evidence supports this argument. Bycio and his colleagues (1995) suggest that perceptions of charisma primarily accounted for support of the augmentation hypothesis. That is, charismatic leadership accounted for most of the additional variance explained over and above that of transactional leadership in a range of dependent variables (e.g. satisfaction with the leader, leader effectiveness). Experimental research by Yorges, Weiss, and Strickland (1999) showed that the extent of self-sacrificing or benefiting leader behaviour influenced perceptions of leader charisma. On the other hand, research by Shamir, Zakay, Breinin, and Popper (1998) has shown that it is possible to distinguish empirically between separate charismatic leader behaviours, as well as charismatic effects. There is a need for theoretical and empirical work that extends the work of Shamir et al. to include a wider range of transformational leader behaviour and examines the relationship between leader behaviour and attributions of charisma.

The findings of this study have practical implications for management development programmes. The evidence that the LPI has weak discriminant validity, suggests there is little justification for giving feedback on specific transformational leader behaviours, nor could one defend promoting the development of specific transformational leader behaviours. Without evidence to show that distinct leadership behaviours are in fact measured, feedback may be misleading and detrimental. Caution must be used, however, when interpreting these findings; there is clearly a need for replication of these results.

In conclusion, the findings of this study suggest the LPI assesses on over-arching construct of transformational leadership. Although it is possible to identify meaningful differences among transformational leader behaviours, the high correlations between the constructs suggest the LPI has weak discriminant validity. Theories about transformational leadership have made an important contribution to our knowledge about leadership. However, measurement problems have constrained the potential contribution of empirical research. There is a need for further theoretical and empirical research on the assessment of transformational leader behaviour.

Requests for reprints should be addressed to Sally A. Carless, Psychology Department, Monash University, PO Box 197, Caulfield East, Australia 3145 (e-mail: Sally. Carless@sci.monash.edu.au).

(n1) The label transformational leadership is typically used as a generic term (Bryman, 1992) and includes visionary leadership (Kouzes & Posner, 1987) and charismatic leadership (Conger & Kanungo, 1988).

Avolio, B., Bass, B., & Jung, D. I. (1995). MLQ Multifactor leadership questionnaire: Technical report. Palo Alto, CA: Mind garden.

Avolio, B., Bass, B. M., & Jung, D. I. (1999). Re-examining the components of transformational and transactional leadership using the Multifactor Leadership Questionnaire. Journal of Occupational and Organizational Psychology, 72, 441-462.

Bagozzi, R. P., & Heatherton, T. F. (1994). A general approach to representing multifaceted personality constructs: Application to state self-esteem. Structural Equation Modelling, 1, 35-67.

Bass, B. (1985). Leadership and performance beyond expectations. New York: Free Press.

Bass, B. M. (1997). Does the transactional--transformational leadership paradigm transcend organizational and national boundaries. American Psychologist, 52, 130-139.

Bentler, P. M., & Bonnett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88, 588-606.

Browne, M. W., & Cudeck, R. (1993). Alternate ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136-162). Newbury Park, CA: Sage.

Bryman, A. (1992). Charisma and leadership in organisations. London: Sage.

Bycio, P., Hackett, R., & Allen, J. S. (1995). Further assessments of Bass's (1985) conceptualization of transactional and transformational leadership. Journal of Applied Psychology, 80, 468-478.

Byrne, B. M. (1998). Structural equation modelling with LISREL, PRELIS and SIMPLIS: Basic concepts, applications, and programming. Mahwah, NJ: Erlbaum.

Carless, S. A. (1998). Assessing the discriminant validity of the transformational leadership behaviour as measured by the MLQ. Journal of Occupational and Organizational Psychology, 71, 353-358.

Conger, J., & Kanungo, R. (1988). Behavioural dimensions of charismatic leadership. In J. Conger, R. Kanungo, & Associates (Eds.), Charismatic leadership: The elusive factor in organizational effectiveness (pp. 78-97). San Francisco: Jossey-Bass.

Den Hartog, D. N., Van Muijen, J. J., & Koopman, P. L. (1997). Transactional versus transformational leadership: An analysis of the MLQ. Journal of Occupational and Organizational Psychology, 70, 19-34.

Gardner, W. L., & Avolio, B. J. (1998). The charismatic relationship: A dramaturgical perspective. Academy of Management Journal, 23, 32-58.

Jermier, J. M. (1993). Introduction--Charismatic leadership: Neo-Weberian perspectives. Leadership Quarterly, 4, 217-223.

Joreskog, K., & Sorbom, D. (1993). LISREL 8: User's reference guide. Chicago, IL: Scientific Software.

Kouzes, J., & Posner, B. (1987). The leadership challenge: How to get extraordinary things done in organizations. San Francisco, CA: Jossey-Bass.

Kouzes, J., & Posner, B. (1988). The Leadership Practices Inventory. San Diego, CA: Pfeiffer.

Lowe, K. B., Knoeck, K. G., & Sivasubramanian, N. (1996). Effectiveness correlates of transformational and transactional leadership: A meta-analytic review of the MLQ literature. Leadership Quarterly, 7, 385-425.

Marsh, H. (1987). The hierarchical structure of self-concept and the application of hierarchical confirmatory factor analysis. Journal of Educational Measurement, 1, 17-39.

Marsh, H., & Grayson, D. (1994). Longitudinal confirmatory factor analysis: Common, time-specific, item-specific, and residual-error components of variance. Structural Equation Modelling, 1, 116-145.

Marsh, H., & Hocevar, D. (1988). A new, more powerful approach to multitrait-multimethod analyses: Application of second-order confirmatory factor analysis. Journal of Applied Psychology, 73, 107-117.

Posner, B., & Kouzes, J. (1993). Psychometric properties of the leadership practices inventory--updated. Educational and Psychological Measurement, 53, 191-199.

Shamir, B., Zakay, E., Breinin, E., & Popper, M. (1998). Correlates of charismatic leader behavior in military units: Subordinates' attitudes, unit characteristics, and superiors' appraisals of leader performance. Academy of Management Journal, 41, 387-409.

Tichy, N. M., & Devanna, M. A. (1986). Transformational leadership. New York: Wiley.

Tracey, J. B., & Hinkin, T. R. (1998). Transformational leadership or effective managerial practices? Group and Organizational Management, 23, 220-236.

Yorges, S. L., Weiss, H. M., & Strickland, O. J. (1999). The effect of leader outcomes on influence, attributions, and perceptions of charisma. Journal of Applied Psychology, 84, 428-436.

Received 28 October 1999; revised version received 2 June 2000