Sunteți pe pagina 1din 14

Leadership

http://lea.sagepub.com A Contribution to the Validation of the Motivation to Lead Scale (MTL): A Research in the Italian Context
Andrea Bobbio and Anna Maria Manganelli Rattazzi Leadership 2006; 2; 117 DOI: 10.1177/1742715006057240 The online version of this article can be found at: http://lea.sagepub.com/cgi/content/abstract/2/1/117

Published by:
http://www.sagepublications.com

Additional services and information for Leadership can be found at: Email Alerts: http://lea.sagepub.com/cgi/alerts Subscriptions: http://lea.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.co.uk/journalsPermissions.nav Citations http://lea.sagepub.com/cgi/content/refs/2/1/117

Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

A Contribution to the Validation of the Motivation to Lead Scale (MTL): A Research in the Italian Context
Andrea Bobbio and Anna Maria Manganelli Rattazzi, Department of General Psychology, University of Padova, Italy

Abstract The study is a contribution to the validation of the Motivation to Lead Scale (MTL), recently proposed by Chan and Drasgow (2001) in order to reveal three types of motivation to be a leader. Participants are 624 university and post-degree specialization students, both male and female, who answered self-report structured questionnaires. The factorial structure of the MTL Scale was investigated through exploratory and conrmatory factor analysis and via multi-sample procedures. Following the analyses, the original scale was reduced into a 15-item form with satisfactory reliability. Other analyses showed the existence of correlations between the MTL Scale, the Social Desirability Scale and the McClelland Scales. Results indicate that the MTL Scale can be considered a useful research instrument in social, personality and organizational psychology. Correlations with the Social Desirability and McClelland Scales should be taken into consideration in order to improve and apply the MTL Scale. Keywords construct-validity; individual differences; leadership; motivation to lead

Introduction
The study of motivation has proved to be important to understand the nature of behaviour, of why it is activated, of why it takes a direction and continues to keep it. A number of researchers have tried to determine individuals basic needs that motivate behaviour. By focusing in particular on motivation to lead (to be and to become a leader), a number of fundamental reference points can be found in Murrays (1938), Atkinsons (1958), Whites (1959), Fielders (1967), Miners (1977) and McClelland and Boyatziss (1982) works, which actually opened the way to in-depth research in this eld. Recently, Chan and Drasgow (2001) proposed the Motivation to Lead (MTL) construct that is dened as an individual difference construct affecting the leaders or future leaders decision to take on both the role and responsibilities of leadership. The construct is articulated along three dimensions, called Affective-Identity
Copyright 2006 SAGE Publications (London, Thousand Oaks, CA and New Delhi) Vol 2(1): 117129 DOI: 10.1177/1742715006057240 www.sagepublications.com
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

2(1) Research Notes

MTL (AI), Non-calculative MTL (NC) and Social-Normative MTL (SN). A person may simply like leading others, hence the Affective-Identity MTL; whereas others might choose the role of group leader because they feel it is their duty or responsibility, hence the Social-Normative MTL. Some people may also decide to accept a leadership role only if they are not too calculative in the cost effectiveness of being the leader of a group, hence the Non-calculative MTL; leadership usually implies personal responsibility and/or costs as well as benets, the less a person calculates about the costs and benets of leading others, the less he or she wishes to avoid leadership roles. Chan and Drasgow (2001) developed the Motivation to Lead construct and a scale to measure it in order to study the relationship between individual differences and leadership behaviours and to contribute to the theoretical debate about the integration between leader development and leader performance. As to the last point, an interesting and well-documented discussion of the state-of-the-art on leadership development in organizational context can be found in Day (2000), where the author not only discusses the conceptual complexity but also the confusion regarding the distinction between leader development (i.e. based on an individualistic conceptualization of leadership) and leadership development (i.e. based on a relational model of leadership as an emergent property of social systems). A key assumption in the approach of Chan and Drasgow is that non-cognitive constructs, such as personality and sociocultural values, are related to leaders behaviour through the individuals MTL, which in turn affects his/her participation in leadership activities and roles. Through these activities, individuals acquire social abilities, useful experiences and the style of their leadership. The main hypothesis maintains that leadership ability and leadership style are learned and that MTL can change during the life course. All this is contextualized into an Integrative Theory of Leadership Behaviour, whose discussion can be found in the original work (Chan & Drasgow, 2001). According to the model, MTL antecedents cover distal constructs, such as personality traits, sociocultural values (i.e. vertical and horizontal individualism and vertical and horizontal collectivism; Singelis et al., 1985) and general cognitive abilities. Since leadership experience is one of the most dynamic and specic components of leadership, this construct is more proximal to MTL than the general characteristics described earlier. Finally, leadership self-efcacy is the nearest construct to MTL. As it can be affected by leadership experience, it may partly mediate the relationship between the other constructs of individual differences and MTL. Chan and Drasgows MTL Scale is a 27-item Likert-type scale. In their research conrmatory factor analyses have shown that the three-correlated-factor model was stable among three samples and showed better t indices than the single-factor model. The three MTL subscales have good internal consistency (all alpha coefcients between 0.65 and 0.91) and are all positively correlated. The main aim of our work is to analyse the psychometric properties of the Italian version of the MTL Scale. Another aim is to study the correlations of this new scale with McClelland needs (i.e. the need for achievement, afliation and power; McClelland & Boyatzis, 1982; see also Andersen, 1999) and social desirability (Paulhus, 1991). The idea is to supply some elements useful in verifying the convergent and discriminant validity of the MTL Scale and its degree of sensitivity to a

118
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

MTL in the Italian Context Bobbio & Manganelli Rattazzi

well-known response bias which falls under the wide category of impressionmanagement strategies.

Method
A total of 624 individuals 361 males and 263 females (mean age = 23.15, SD = 4.49) took part in the research: 487 were university students and 137 were postgraduate students (post-degree specialization and master courses). The data were collected in ve distinct groups over the same period of time, from April 2003 to June 2003 (see Table 1). The rst group of university students (n = 168) included for the most part engineering students; the second group (n = 319) comprised mostly psychology students. All questionnaires were made up of the Italian translation of the MTL Scale, a short form of the Marlowe and Crowne Social Desirability Scale (Crowne & Marlowe, 1960; Manganelli Rattazzi et al., 2000), the McClelland Scale (Tosi et al., 2002) and other instruments not addressed here. As to the Social Desirability Scale, we used a 12-item version with a six-point Likert-type response scale, where: 1 = totally in disagreement; 6 = absolutely in agreement. The McClelland Scale is composed of 30 items, 10 per factor (achievement, afliation, power) and the response scale was a 10-point Likert-type one ranging from 1 = totally in disagreement to 10 = absolutely in agreement. As stated earlier, the MTL Scale consists of 27 items, 9 for each dimension considered (see Table 2). The response scale was a Likert-type one, where 1 = totally in disagreement and 7 = absolutely in agreement.1

Analysis procedures
In order to verify the psychometric properties of the MTL Scale we considered the group of participants comprising all the university students (n = 487). This group was divided into two subgroups (n1 = 243 and n2 = 244). This division was done in order to examine model validity in an exploratory way in one sample (calibration sample) and to verify the model chosen in the other (validation sample; Jreskog & Srbom, 1996). Exploratory and conrmatory factor analyses were performed starting from the calibration sample. Goodness of t was checked using the following indices: 2 and the ratio between 2 value and degrees of freedom (2/d.f.), RMSEA, RMR, and CFI. The rst index suggests good t if it is not signicant (p > 0.05); however, it
Table 1 Samples involved in the research
Samples University students 1 University students 2 Specialization course Postgraduate students 1 Postgraduate students 2 Total Males 126 146 11 29 49 361 Females 42 173 22 8 18 263 Total 168 319 33 37 67 624 Mean Age 21.36 21.64 34.67 27.05 26.81 23.15 SD 1.7 2.55 8.42 2.28 2.33 4.49

119
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

2(1) Research Notes Table 2 Items of the MTL Scale


Factor 1: Affective-Identity MTL 1) Most of the time I prefer being a leader rather than a follower when working in a group. 4) I am the type of person who is not interested in leading others. 7) I am denitely not a leader by nature. 10) I am the type of person who likes to be in charge of others. 13) I believe I can contribute more to a group if I am a follower rather than a leader. 16) I usually want to be the leader in the groups that I work in. 19) I am the type who would actively support a leader but prefers not to be appointed as leader. 22) I have a tendency to take charge in most groups or teams that I worked in. 25) I am seldom reluctant to be the leader of a group. Factor 2: Non-calculative MTL 2) 5) 8) 11) 14) 17) 20) 23) 26) I am only interested in leading a group if there are clear advantages for me. I will never agree to lead if I cannot see any benets of accepting that role. I would only agree to be a group leader if I know I can benet from that role. I would agree to lead others even if there are no special rewards or benets with that role. I would want to know whats in it for me if I am going to agree to lead a group. I never expect to get more privileges if I agree to lead a group. If I agree to lead a group, I would never expect any advantages or special benets. I have more of my own problems to worry about than to be concerned about the rest of the group. Leading others is really more of a dirty job than rather than an honourable one.

Factor 3: Social-Normative MTL 3) 6) 9) 12) 15) 18) 21) 24) 27) I feel that I have a duty to lead others if I am asked. I agree to lead whenever I am asked or nominated by the other members. I was taught to believe in the value of leading others. It is appropriate for people to accept leadership roles or positions when they are asked. I have been taught that I should always volunteer to lead others if I can. It is not right to decline leadership roles. It is an honor and privilege to be asked to lead. People should volunteer to lead rather than wait for others to ask or vote for them. I would never agree to lead just because others voted for me.

Source: Chan and Drasgow (2001: 486, modied).

strongly depends on the number of cases. Further, the use of 2 is based on psychometric assumption too strong to be realistic (Hu & Bentler, 1998). For these reasons, alternative t indices, less sensitive to sample size and more appropriate to quantify the goodness of t along a continuum, were developed. As regards the second index, if the value of the 2/d.f. ratio falls between 1 and 3, it indicates an acceptable model, especially when n is high (Marsh et al., 1988). The RMSEA (Root of Mean Square Error of Approximation) measures the discrepancy between the hypothesized model and the data by degree of freedom; values equal to or less than 0.05 are considered satisfactory (Hu & Bentler, 1999). The RMR (Root Mean Square of Residual) represents the average residual value derived from the tting of the variancecovariance

120
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

MTL in the Italian Context Bobbio & Manganelli Rattazzi

matrix for the hypothesized model and the variancecovariance matrix of the sample data: recommended values are less than 0.08. If the variancecovariance matrix is analysed, the standardized version of the index must be used (SRMR; Byrne, 1998). The CFI (Comparative Fit Index) is based on the comparison between the t of the hypothesized model and the t of a basic model, represented by the null model in which all variables lack correlations and the only parameters estimated are error variances. This index is considered independent of sample size and one of the best indices of incremental t. Values close to 0.95 (or higher) are considered indicative of a good t (Hu & Bentler, 1999). Finally, a number of criteria were taken into account in judging the worth of individual parameters; these included statistical signicance as indicated by the LISREL t-values, goodness of t based on the standardized residual values and modication indexes (Byrne, 1991). The second step of the analysis involved the validation sample and aimed to support the selected model via the multi-sample procedure. Further multi-sample analyses allowed us to assess model invariance in the male and female sub-samples within the whole student group and to compare postgraduate and university student samples.

Results
An exploratory factor analysis was carried out on the calibration sample (n1 = 243) (principal components method, eigenvalue > 1, oblimin rotation).2 The three dimensions which emerged from the exploratory analysis (47.8 per cent of total variance explained) were consistent in meaning with the dimensions envisaged by the MTL Scale: Affective-Identity MTL, Non-calculative MTL and Social-Normative MTL. The three-correlated-factor model was then tested through a conrmatory factor analysis, applied to the covariance matrix and via maximum likelihood procedure using LISREL 8.30. First of all, a 27-item three-correlated-factor (Affective Identity AI, Noncalculative NC, Social Normative SN) model was tested. Goodness of t turned out to be totally unsatisfactory: 2 (321) = 1085.39, p 0.000, RMSEA = 0.099, SRMR = 0.11, CFI = 0.74 (2/d.f. = 3.38). Given rejection of the initially postulated model, the next step was to take an exploratory approach to the analyses in order to determine reasons for the ill-tting model (Byrne, 1991). By examining the parameter estimates some unsatisfactory results were found: (a) item 24 of the SN scale showed low, non-signicant loading ( = 0.23; t = 1.82); (b) items 11 (NC), 10 (AI), 22 (AI) 2 (NC), 21 (SN), 4 (AI), 8 (NC), 9 (SN), 15 (SN), 7 (AI), 23 (NC) had supplementary loadings on factors different from their own, as indicated by the modication indices on the lambda matrix (MI > 3.84); (c) several pairs of these items showed correlated measurement errors, as was clear from the too high thetadelta modication indices (MI > 3.84) (Jreskog & Srbom, 1996). Having discovered these defective items, it now seemed appropriate to work towards improving the MTL scale by identifying a core set of items that could most validly measure the three MTL factors (Anderson & Gerbing, 1988; Byrne, 1991). With this purpose, a second CFA model was specied in which item 24 was deleted. However, goodness of t indices remained unsatisfactory: 2 (296) = 1002.92, p 0.000,

121
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

2(1) Research Notes

RMSEA = 0.099, SRMR = 0.11, NNFI = 0.73, CFI = 0.75 (2/d.f. = 3.39). In an effort to further improve the measurement instrument, we removed one by one the items which exhibited cross-loadings onto a second factor. In specifying step by step a new model, the item-deletion procedure was deemed preferable in comparison to the correlation between error terms.3 The resulting and relatively more acceptable model in terms of signicant factor loadings and residuals was made up of three factors and the 15 items listed below: 1. for the Affective-Identity factor (AI): 1, 13, 16, 19, 22; 2. for the Non-calculative factor (NC): 5, 14, 17, 20, 26; 3. for the Social-Normative factor (SN): 3, 6, 12, 18, 27. Goodness of t was as follows: 2 (87) = 250.62, p 0.000, RMSEA = 0.08, SRMR = 0.06, CFI = 0.90 (2/d.f. = 2.88). The analyses described up to now were conducted in respect of what is called total disaggregation model (Bagozzi & Heatherton, 1994) where each single item from the scale is used to operationalize and measure its respective hypothesized dimension. The total disaggregation model provides the most detailed level of analysis of a scale because psychometric properties are provided for each individual item. In practice, however, the total disaggregation model can be unwieldy because of likely high levels of random error in typical items and the number of parameters that must be estimated. The further step that was considered necessary to reduce the number of parameters to be estimated and at the same time decrease the measurement error was item aggregation, carried out by calculating the item pair mean (partial disaggregation model; Bagozzi & Heatherton, 1994).4 The resulting model had three indicators per factor (AI, NC and SN), two of which are item pair means. The three-correlated-factor model with nine indicators showed a good t (only the 2 is signicant, while the 2/d.f. ratio is acceptable, see Table 3) and, most remarkably, a better t compared to the one of the single-factor model. Cronbachs alphas were respectively: AI = 0.81; NC = 0.74; SN = 0.61.5 The three-correlated-factor model was then tested on the validation sample (n2 = 244). The model ts the data well: 2 (24) = 54.89, p 0.001, RMSEA = 0.073, SRMR = 0.055, CFI = 0.96 (2/d.f. = 2.29). Cronbachs values were respectively: AI = 0.82; NC = 0.75; SN = 0.60. As we can see, in both the sub-samples reliability is acceptable for Affective-Identity and Non-calculative factors, and a little bit problematic for the Social-Normative factor (as we stated before, in the original study
Table 3 Goodness of t indices in the calibration sample
Model Single-factor model Three non-correlatedfactor model Three correlatedfactor-model 2 415.06 81.36 43.44 d.f. 27 27 24 p 0.000 0.000 0.001 2/d.f. 15.37 3.01 1.81 RMSEA 0.24 0.091 0.058 SRMR 0.19 0.13 0.045 CFI 0.49 0.88 0.97

122
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

MTL in the Italian Context Bobbio & Manganelli Rattazzi

the coefcients were all between 0.65 and 0.91 and the lowest one was always the SN, especially in the student samples; see Chan & Drasgow, 2001) The next step was a cross-validation done via the multi-sample procedure. This allows us to simultaneously analyse data from different samples, binding all or some parameters to be identical in the different groups. We analysed gradually more restrictive hypotheses: (a) invariance of the three-correlated-factor model (Model 1, Baseline); (b) invariance of factor loadings (Model 2); (c) invariance of factor loadings and correlations between latent factors (Model 3); (d) invariance of factor loadings and correlations between latent factors and error variances (Model 4) (Table 4). Testing for invariance requires the specication of a model in which parameters are constrained to be equal across samples and the comparison of this model with a less restrictive one in which parameters are free to assume any value. Comparison between models is carried out through the 2 difference test: a signicant 2 indicates non-invariance (Byrne, 1998). The specication of Model 1 allows us to check the goodness of t of the model itself through the simultaneous analysis of the data from the two groups and at the same time provides a criterion against which the two more restrictive models are compared. Model 1 (Baseline) shows a good t to the data from the two groups (Table 4). Model 2, which constrains factor loadings to be equal, and Model 3, which also constrains the correlations between factors to be equal, show a good t. The comparison between Model 2 and Model 1 and that between Model 3 and Model 2 are in favour of loading and correlation invariance. Besides loadings and correlations between factors, Model 4 constrains also error variances to be equal. The 2 difference between this model and Model 3 supports the total invariance of the model in the two samples. A second cross-validation was conducted by dividing the whole university students sample (n = 487) into a male group (n = 272) and a female group (n = 215).
Table 4 Testing invariance of factor structure, factor loadings, correlations between latent factors and error variances in the calibration and validation samples
Goodness of t indices Model 1 (Baseline): invariance of the factor structure 2 (48) = 98.33, p 0.000 RMSEA = 066.0 SRMR = 0.055 CFI = 0.96 2 (54) = 107.07, p 0.000 RMSEA = 0.064 SRMR = 0.059 CFI = 0.96 2 (60) = 109.64, p 0.000 RMSEA = 0.058 SRMR = 0.06 CFI = 0.96 2 (69) = 116.18, p 0.000 RMSEA = 0.053 SRMR = 0.062 CFI = 0.96 M2M1 2 (6) = 8.74 p > 0.05 M3M2 2 (6) = 2.57 p > 0.05 M4M3 2 (9) = 6.54 p > 0.05 Test of hypotheses

Model 2: invariance of factor loadings

Model 3: invariance of factor loadings and correlations between latent factors Model 4: invariance of factor loadings, correlations between latent factors and error variances

123
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

2(1) Research Notes

The model ts the data fairly well both in the male sub-sample: 2 (24) = 53.26, p 0.001, RMSEA = 0.067, SRMR = 0.052, CFI = 0.95 (2/d.f. = 2.22), and in the female sub-sample: 2 (24) = 62.35, p 0.001, RMSEA = 0.081, SRMR = 0.061, CFI = 0.93 (2/d.f. = 2.6). Then, following the multi-sample procedure, we tested the model invariance across the two samples. In this second analysis too, Model 1 shows sufcient t to the data of the two groups (Table 5); also Model 2, which constrains factor loadings to be equal, and Model 3, which constrains factor loadings and correlation between latent factors to be equal, show a fairly good t; the comparisons of Model 2 with Model 1 and of Model 3 with Model 2 are in favour of invariance of loadings. Besides loadings and factor correlations, Model 4 also constrains error variances to be the equal. The 2 difference between this model and Model 3 is not signicant and supports the total invariance of the model in the two sub-samples, even if the SRMR index gets progressively worse. A third cross-validation was conducted by examining the whole university student group (n = 487) and the postgraduate student group (n = 137). The model shows a fairly good data t both in the sub-sample comprising all the university students: 2 (24) = 68.88, p 0.000, RMSEA = 0.066, SRMR = 0.042, CFI = 0.96 (2/d.f. = 2.88), and in the postgraduate student sub-sample: 2 (24) = 34.63, p 0.08, RMSEA = 0.057, SRMR = 0.061, CFI = 0.95 (2/d.f. = 1.44). Then, following the multi-sample procedure, we tested the model invariance across the two groups. Also in this third analysis Model 1 shows good t to the data of the two groups (Table 6); also Models 2 and 3 show a fairly good t; the comparisons of Model 2 with Model 1 and of Model 3 with Model 2 are in favour of invariance of loadings. Model 4 forces also error variances to be equal. The 2 difference between this model and Model 3 is signicant and does not allow us to support the total
Table 5 Testing invariance of factor structure, factor loadings, correlations between latent factors and error variances in male and female samples
Goodness of t indices Model 1 (Baseline): invariance of the factor structure 2 (48) = 115.62, p 0.000 RMSEA = 076.0 SRMR = 0.061 CFI = 0.94 2 (54) = 125.55, p 0.000 RMSEA = 0.074 SRMR = 0.067 CFI = 0.94 2 (60) = 133.27, p 0.000 RMSEA = 0.071 SRMR = 0.078 CFI = 0.94 2 (69) = 146.39, p 0.000 RMSEA = 0.068 SRMR = 0.083 CFI = 0.93 M2M1 2 (6) = 9.93 p > 0.05 M3M2 2 (6) = 7.76 p > 0.05 M4M3 2 (9) = 13.12 p > 0.05 Test of hypotheses

Model 2: invariance of factor loadings

Model 3: invariance of factor loadings and correlations between latent factors Model 4: invariance of factor loadings, correlations between latent factors and error variances

124
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

MTL in the Italian Context Bobbio & Manganelli Rattazzi Table 6 Testing invariance of factor structure, factor loadings, correlations between latent factors and error variances in the university student and the postgraduate student samples
Goodness of t indices Model 1 (Baseline): invariance of the factor structure 2 (48) = 109.81, p 0.000 RMSEA = 064.0 SRMR = 0.061 CFI = 0.96 2 (54) = 115.18, p 0.000 RMSEA = 0.06 SRMR = 0.07 CFI = 0.96 2 (60) = 122.57, p 0.000 RMSEA = 0.058 SRMR = 0.086 CFI = 0.96 2 (69) = 168.60, p 0.000 RMSEA = 0.068 SRMR = 0.10 CFI = 0.94 M2M1 2 (6) = 5.37 p > 0.05 M3M2 2 (6) = 7.39 p > 0.05 M4M3 2 (9) = 46.03 p < 0.001 Test of hypotheses

Model 2: invariance of factor loadings

Model 3: invariance of factor loadings and correlations between latent factors Model 4: invariance of factor loadings, correlations between latent factors and error variances

invariance of the model in the two samples. Yet again, the SRMR parameter gets progressively worse. In the analyses that follow we will consider the whole sample of university students (n = 487). As said before, for the whole group the t indices are satisfactory. The phi matrix showed a negative correlation between Affective-Identity MTL and Non-calculative MTL (12 = .22, p < 0.05); and a positive correlation between Affective-Identity MTL and Social-Normative MTL (13 = 0.30, p < 0.05) and between Non-calculative MTL and Social-Normative MTL (23 = 0.42, p < 0.01). Unlike the original model, a negative correlation was found between personal preference for leadership roles (AI-MTL) and non-calculative attitude towards these positions. Consistent with the original model, a non-calculative attitude is correlated with a social-normative motivation and AI-MTL and SN-MTL showed a positive signicant correlation. Three composite scores were calculated. As regards AI, a high score indicates that people like to be leaders; for NC, here a high score will indicate that people weigh up the benets they can get from a leadership role, while a low score indicates that they do not (calculative direction); and for SN, a high score indicates that people think it is their duty to take a leadership role if they are in a position to do so. Considering the composite scores for the three dimensions of MTL, we carried out a multivariate variance analysis (MANOVA), with gender as between factor (males, n = 272; females, n = 215). A signicant multivariate effect was found: F(3,483) = 22.608, p < 0.0001. The analysis of the univariate effects highlighted differences between males and females in the Affective-Identity and Non-calculative factors: males, more than females, like to take on leadership roles (AI: 4.20 vs. 3.58),6 F(1,485) = 32.82, p < 0.0001; females are more willing to take on leadership roles even in the absence of specic benets or advantages (NC: 3.23 vs. 2.61), F(1,485)

125
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

2(1) Research Notes

= 41.15, p < 0.0001. However, in both genders a tendency is found to deny a calculative attitude. No difference was found concerning the SN scale: the scores, however, reveal a tendency to accept leadership roles out of a sense of duty (SN: 4.47 vs. 4.51), F(1,485) = 0.142, n.s.

Social desirability
The scale turned out to be sufciently reliable in both university student groups. Cronbachs alpha is 0.69 for student sample 1 (n = 168) and 0.66 for student sample 2 (n = 319). Comparisons of the mean scores of both males and females highlighted that women more than men tend to give socially desirable answers, t(166) = 2.89, p < 0.005; t(317) = 2.27, p < 0.03.

McClelland Scale
Analyses regarding the McClelland Scale were carried out considering only one of the university student groups (n = 319). The three sub-scales showed an acceptable level of internal consistency. Cronbachs coefcients are: need for achievement = 0.76; need for afliation = 0.68 and need for power = 0.79. The analyses carried out through the multivariate variance analysis (MANOVA), with gender as a between factor, highlighted a signicant multivariate effect: F(3,315) = 18.804, p < 0.0001. The analysis of the univariate effects highlighted expected differences between males and females related to all three factors. Males have higher scores in: need for achievement (6.39 vs 6.04), F(1,317) = 5.32, p < 0.03, and power (5.48 vs 4.60), F(1,317) = 32.61, p < 0.0001; while females in need for afliation (6.24 vs 5.80), F(1,317) = 9.87, p < 0.003.

Correlations between MTL, Social Desirability and McClelland Scale


The correlations between MTL sub-scales and Social Desirability were computed separately for the two student groups. Results highlighted a high signicant correlation between Calculative MTL and Social Desirability: r = 0.54, p < 0.01; r = 0.34, p < 0.01, respectively in student samples 1 and 2; correlations corrected for attenuation are: r = 0.77, p < 0.01 and r = 0.48, p < 0.01. Individuals who tend to give socially desirable answers tend also to declare that they are ready to take on leadership roles even without possible benets for them. Answers to the SocialNormative MTL scale also seem to be sensitive to social desirability effects: individuals who tend to present themselves under an excessively favourable light declare they would take on leadership roles out of a sense of duty (r = 0.28, p < 0.01; r = 0.15, p < 0.05, respectively in student samples 1 and 2; correlations corrected for attenuation are: r = 0.43, p < 0.01 and r = 0.28, p < 0.01). Concerning McClellands needs, all the sub-scales show signicant correlations with the Social Desirability Scale, even if less strongly than MTL Scale (all below r = 0.20, p < 0.01). In view of this outcome, the correlations between MTL Scale and McClelland Scales were calculated controlling social desirability effects (partial correlations). All three MTL factors show signicant correlations with the scales measuring Need for Achievement, Afliation and Power. The most interesting

126
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

MTL in the Italian Context Bobbio & Manganelli Rattazzi

results, that might support a partial overlapping between the two constructs, are: high correlation between Affective-Identity MTL and Need for Power (r = 0.67, p < 0.001) and between Affective-Identity MTL and Need for Achievement (r = 0.40, p < 0.01).

Conclusions
The data presented here must be considered as a rst step in the in-depth analysis of the MTL Scale proposed by Chan and Drasgow (2001) and in order to provide an adaptation to the Italian context. The MTL construct is without any doubt an original contribution to the eld of leadership and ts in with the line of research of Banduras (1986) social learning theory. Regarding the MTL scale, our rst results support the invariance of the threecorrelated-factor model in different groups in the Italian context. Furthermore, they also introduce the likelihood of reducing the number of scale items, thus originating a parsimonious instrument which is also quicker to administer. However, the model t indexes are acceptable rather than perfect (e.g. CFI between 0.93 and 0.97 and RMSEA between 0.06 and 0.076) and the 2 is always signicant. The reliability () also presents some problematic aspects especially regarding the Social Normative MTL sub-scale. All together, this information suggests caution in evaluating results. New analyses are needed after the reformulation of the most problematic items and it is also necessary to verify the model using samples made up of participants different from university students. In fact the most problematic one was the postgraduate student sample. Unlike the original model a different pattern of correlations between the MTL sub-scales (especially between the AI and NC scales) has been found and this raises interesting issues on the inuence of sociocultural variables. High correlations between the Social Desirability Scale and NC and SN factors (as well as with the McClelland Scales) have been found. This suggests caution in considering the scores resulting from the administration of the scale, especially concerning professional assessment or personnel selection situations, which are vulnerable to impression management modalities. Lastly, correlations between the MTL Scale and the McClelland Scales in particular with Need for Power and Need for Achievement leave some questions open as to discriminant validity which should be further evaluated in the light of an external leadership criterion. A hypothesis for future research is to examine how the MTL Scale interacts with situational factors and with an individuals choice to become a leader (or take the lead) in specic situations.

Notes
1. We changed the original response scale from 5 to 7 points. 2. The assumption of normality of variable distributions were checked with PRELIS 2.30: skewness and kurtosis were all between 1 and 1. 3. Concerning the issue of measurement error correlations, there is much debate in the literature of conrmatory analysis about their causes and possible solutions: for instance a high degree of similarity in item content and other effects due to the method used (for example the response scale) may trigger error correlation.

127
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

2(1) Research Notes 4. Item pairs were randomly formed. 5. The widely accepted and most common social science cut-off is that alpha should be 0.70 or higher while other authors are as lenient as 0.60. Reliability estimates in the range of 0.70 or 0.80 are good enough for most purposes in basic research even if in clinical settings high reliability (0.85 or more) is important (Miller, 1995; Nunnally, 1978). 6. The mean of the male group is greater than the central point of the scale (4): t(271) = 2.870, p < 0.005; while the mean of the female group is lower than the central point: t(214) = 5.077, p < 0.0001.

Acknowledgements
An early version of this paper was presented at the 2nd Studying Leadership International Conference, Lancaster University Management School, Lancaster, UK, 1516 December 2003. The authors would like thank the anonymous reviewers for their helpful comments on a previous draft of the article and are grateful to Eva Bolognese, Paola Boromello and Lucia Corrieri for supporting the data-collection efforts.

References
Andersen, J. A. (1999) Are Power Motivated Leaders More Effective? A Review of McClellands Theory, in M. A. Rahim, R. T. Golembiewski & K. D. Mackenzie (eds) Current Topics in Management, pp. 4159. Greenwich, CT: JAI Press. Anderson, J. C., & Gerbing, D. W. (1988) Structural Equation Modeling in Practice: A Review and Recommended Two-step Approach, Psychological Bulletin 3: 41123. Atkinson, J. W. (1958) Motives in Fantasy, Action and Society. Princeton, NJ: Van Nostrand. Bagozzi, R. P., & Heatherton, T. F. (1994) A General Approach for Representing Multifaceted Personality Constructs: Application to State Self-esteem, Structural Equation Modeling 1: 3567. Bandura A. (1986) Social Foundations of Thought and Action: A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice-Hall. Byrne, B. M. (1991) The Maslach Burnout Inventory: Validating Factorial Structure and Invariance Across Intermediate, Secondary and University Educators, Multivariate Behavioral Research 26: 47799. Byrne, B. M. (1998) Structural Equation Modelling with LISREL, PRELIS and SIMPLIS: Basic Concepts, Applications and Programming. Mahwah, NJ: Erlbaum. Chan, K., & Drasgow, F. (2001) Toward a Theory of Individual Differences and Leadership: Understanding the Motivation to Lead, Journal of Applied Psychology 86: 48198. Crowne, D. P., & Marlowe, D. (1960) A New Scale of Social Desirability Independent of Psychopathology, Journal of Consulting Psychology 24: 34954. Day, D. V. (2000) Leadership Development: A Review in Context, Leadership Quarterly 4: 581613. Fielder, F. E. (1967) Theory of Leadership Effectiveness. New York: McGraw-Hill. Hu, L., & Bentler, P. M. (1998) Fit Indexes in Covariance Structure Modeling: Sensitivity to Underparameterized Model Misspecication, Psychological Methods 4: 42453. Hu, L., & Bentler, P. M. (1999) Cutofff Criteria for Fit Indexes in Covariance Structure Analysis: Conventional Criteria Versus New Alternatives, Structural Equation Modeling 6: 155. Jreskog, K. J., & Srbom, D. (1996) LISREL 8: Users Reference Guide. Chicago: Scientic Software International.

128
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

Leadership

MTL in the Italian Context Bobbio & Manganelli Rattazzi Manganelli Rattazzi, A. M., Canova, L., & Marcorin, R. (2000) La desiderabilit sociale. Unanalisi di forme brevi della scala di Marlowe e Crowne, TPM Testing Psicometria Metodologia 7(1): 517. Marsh, H. V., Balla, J. R., & McDonald, R. P. (1988) Goodness Fit Indexes in Conrmatory Factor Analysis: The Effect of Sample Size, Psychological Bulletin 103: 391410. McClelland D. C., & Boyatzis R. E. (1982) Leadership Motive Pattern and Long-term Success in Management, Journal of Applied Psychology 7: 73743. Miller, M. B. (1995) Coefcient Alpha: A Basic Introduction from the Perspectives of Classical Test Theory and Structural Equation Modelling, Structural Equation Modeling 2(3): 25573. Miner, J. B. (1977) Motivation to Manage: A Ten-year Update on Studies on Management Education Research. Atlanta, GA: Organizational Measurement System Press. Murray, H. A. (1938) Exploration in Personality. Oxford: Oxford University Press. Nunnally, J. O. (1978) Psychometric Theory. New York: McGraw-Hill. Pauhlus, D. L. (1991) Measurement and Control of Response Bias, in J. P. Robinson, P. R. Shaver & L. S. Wrightsman (eds) Measurement of Personality and Social Psychological Attitudes, pp. 1759. San Diego: Academic Press. Singelis, T. M., Triandis, H. C., Bhawuk, D., & Gelfand, M. J. (1985) Horizontal and Vertical Dimensions of Individualism and Collectivism: A Theoretical and Measurement Renement, Cross-Cultural Research: The Journal of Comparative Social Science 29: 24075. Tosi, H. L., Pilati, M., Mero, N. P., & Rizzo, J. R. (2002) Comportamento organizzativo. Milan: Egea. White, R. W. (1959) Motivation Reconsidered: The Concept of Competence, Psychological Review 66: 297333.

Andrea Bobbio received a PhD in Personality and Social Psychology (2005, University of Padova, Italy), a Masters degree in Human Resources Management (2002, University of Castellanza, Italy) and a degree in Psychology (1999, University of Padova, Italy; specialization in Work and Organizational Psychology). His research interests are: social psychology; construction and analysis of measurement instruments in social and personality psychology; social identity theory; leadership and power dynamics in groups and organizations. [email: andrea.bobbio@unipd.it] Anna Maria Manganelli Rattazzi is Professor of Social Psychology, Department of General Psychology, University of Padova, Italy. Her current research interests are: research methodology in social psychology; development of measurement instruments in social psychology; theories of prejudice and measurement of modern forms of prejudice; economic psychology: attitudes towards money; money management strategies and their determinants.

129
Downloaded from http://lea.sagepub.com by Tomislav Bunjevac on May 4, 2009

S-ar putea să vă placă și