• Tidak ada hasil yang ditemukan

The independent variables and the master style

6.5 Analysis of impacts of TPB components on leadership styles of deans

6.5.1 The independent variables and the master style

The master style represents the extent to which the deans perform the eight roles of CVF. As we indicated in section 6.3.1 the Kenyan deans perceive themselves as master managers who can very well balance the different leadership roles of CVF.

In this step, our interest is to establish the extent to which the independent variables (attitude, subjective norms and perceived behavioural controls) determine the master style of the Kenyan deans. We first undertook bivariate regressions whose outcome is presented in Table 6.11 to determine the extent to which each of the independent variables influence the master style of the deans.

The outcome indicates that attitudes explain 57% of the variance in the master style suggesting that there is a strong and positive relation between attitude and master style (F=79.19, p= 0.000). Since this outcome confirms that the attitude of the deans explains the master style to a large extent, it suggests therefore that the more positive the attitudes of the deans the more likely it is that they will deploy the master style. It should be noted that here the attitudes pertain to all the 8 roles making up the master style as presented in Table 6.4. Subjective norms or ‘social pressure’ on the deans by their significant referents explain 43% of why deans run their faculties as ‘masters’, which is also significant (F=42.11, p=0.000) but is less than the attitude value. Perceived behavioural controls, the third determinant, does not explain the master style (Adj. R2 = -0.02, F=0.01, p=0.93). This suggests therefore that whether deans do, or do not perceive any barriers or constraints would not directly affect their deployment of the master style.

150

Table 6.11: Bivariate regression analyses of independent variables on the master style.

Master style beta* t-value P Adj. R2 F

Attitude towards this style (n=59) 0.76 (0.74)

8.90 .000 0.57 79.19

Subjective norms towards this style (n=54)

0.66 (0.09)

6.49 .000 0.43 42.11

Perceived behavioural control (n=57) 0.01 (0.01)

0.08 0.93 -0.02 0.01

* The first coefficient is the standardized coefficient, between brackets the unstandardized b is presented.

We next executed a multiple regression based on the Theory of Reasoned Action combining the effects of both attitudes and subjective norms. The outcome (see Table 6.12) indicates that this model explains the master style (R2 =61%, F= 32.89, p=0.000) somewhat better than when each of the two variables was used singularly.

In the multiple regression, it still emerges that attitudes significantly and positively determine the master style of the deans as do subjective norms, though at different levels. We also note that when we compare the effects in table 6.11 and 6.12, the strength of the effect of the subjective norms and attitude decreases when both variables are included in one model. This suggests the existence of a relationship between attitudes and subjective norms.

Table 6.12: Multiple regression analysis of independent variables on the master style (n=53)

* The first coefficient is the standardized coefficient, between brackets the unstandardized b is presented.

Our further analysis confirmed that there is indeed a relative high correlation between the attitudes and the subjective norms (r = 0.65 – from the bivariate correlation matrix).8 This finding is also signaled by Ajzen and Fishbein (2010) who argue that while conceptually attitudes and social norms independently explain behaviours (in our case a leadership style), they empirically can be inter-related.

This might suggest that part of the effect of subjective norms runs via a person’s attitudes which can in turn influence one’s behaviour (here managerial style)9 . Subsequently, we investigated the possible effect of perceived behavioural control.

Here we move from the Theory of Reasoned Action (TRA) to the Theory of Planned Bahaviour (TPB). While the bivariate regressions (see Table 6.11) indicated that perceived behavioural control does not have a direct effect on the master style,

8 In all our regression analyses we checked for the presence of problems of multicollinearity. Here we did not find any evidence for such a problem. Subsequently we will only discuss this potential problem when it actually occurs.

9 Theoretically, this is the most plausible interpretation of the correlation. Both a person’s attitude and his or her behaviour is influences by norms of relevant others. But the correlation might also be in part the result of self projection where a person’s attitudes shape his or her perceptions of normative beliefs of relevant others.

Master style beta* t-value p

Attitude towards this style 0.54

(0.53)

4.98 0.000

Subjective norms towards this style 0.32 (0,04)

2.97 0.05

Adj. R2 = 0.61. F=32.89

152

the TPB implies that this factor might have a moderating effect. The assumption, as an alternate explanation of the effects of barriers on the master style, is that the existence of barriers could affect the relationship between attitudes and master style.

Thus, the expectation is that the more barriers the deans perceive, the less positive the effect of attitudes would be on the master style. When there are no barriers perceived it would be expected that attitude explain the master style.

To test this expectation we created a new variable: ‘attitude x pbc’, to be the moderator or interaction variable. In order to allow for a better interpretation of the coefficients we decided to center the variables in the analysis.10 To center the scores of the variable ‘attitude’ we took the mean of this variable and deducted it from the score on attitude (new variable ‘attitude_centered’). The same procedure was followed to calculate the new variable ‘pbc_centered’. The outcome of the regression analysis with the ‘new model’ with the moderator (see Figure 6.3) demonstrates that there is indeed an interaction effect; i.e. the degree to which deans perceive barriers affects the relationship between attitudes and master style.

As shown in Table 6.13, this model is significant (F=31.94, p < 0.00) and explains just more variance (R2=0.70) than the original model (R2=0.61).

10We should realize that the first three b-coefficients -- i.e. the ones that are relevant for understanding the interaction effect; viz. the coefficients for attitude (X1), perceived behavioural control (X2) and their product (X1*X2) in Table 6.13 – should be interpreted in conditional terms: The coefficient for X1 captures the effect of X1 on Y if and only if X2 is 0. Likewise the coefficient of X2 indicates the effect of X2 if X1 happens to be 0. And finally the coefficient of the product-term captures the extent to which X1 ’s effect on Y depends on X2, if neither of X1 nor X2 are 0 (see e.g.

Brambor et al. 2006: 72). By centering both X1 and X2 around their mean we make sure that the interpretation of the coefficients becomes easier because the values of 0 are set equal to the means of these respective variables, rather to an arbitrary value of 0 that may not even be a score that occurs in the sample.

Table 6.13: Multiple regression on master style with moderating variable (n=53)

Master style b* t-value p

Attitude towards this style (centered) 0.65 6.48 0.000 Perceived behavioural control (centered) -.11 -2.29 0.03 Moderator attitude x perceived behavioural control

(based on product of centered variables)

0.26 3.27 0.002

Subjective norms this style (not centered) 0.04 2.73 0.01 Adj. R2 = 0.70. F=31.94

* The table only contains the unstandardized b-coefficients

The results in table 6.13 in combination with the results presented in Figure 6.3 indicate that perceived behavioural control does have an impact on the relationship between attitudes and the master style. For all deans we see a clear statistically positive relationship between attitude and the style. This effect, as was theoretically expected, , is weakest when behavioural control is perceived to be relatively low (at -1 standard deviation), here we find a flatter slope (0.31) than at the mean level of PBC (0.65), and at a relatively high level (+1 standard deviation from the mean, where we find the steepest slope: b=1.00). The difference between the slopes at +1 and – 1 standard deviation from the mean is statistically significant.11

11 As can be noted in Figure 6.3, the confidence levels of the b’s at +1 and -1 from the mean do not overlap.

154

Figure 6.3: Conditional effects of attitudes on master style at different levels of perceived behavioural control 12

Minus 1 Standard Deviation PBC, b= 0.31 (95% confidence interval: 0.08 - 0.53) Mean PBC, b=0.65 (95% confidence interval: 0.44 - 0.85)

Plus 1 Standard Deviation PBC, b=1.00 (95% confidence interval: b= 0.81 – 1.18)

12 This figure and the conditional b-coefficients were produced by using Daniel Soper’s Interaction programme (Soper 2008): http://www.danielsoper.com/Interaction. This is also the case for the three similar figures that follow.

This outcome indicates that perceived behavioural control has its theoretically expected positive effect on the relation between attitudes and reported behavioural style. All in all these findings corroborate the expectations from the Theory of Planned Behavior. The initial results indicating the relevance of attitudes and subjective norms postulated by the Theory of Reasoned Action were confirmed.

But in addition it was shown that the effect of attitudes is contingent upon the dean’s perceived behavioural control. The more barriers the deans perceive, the less impact attitudes have on the master style.