According to Baron and Kenny (1986), the model in Figure 5.10 assumes a three-variable system such that there are two causal paths feeding into the outcome variable: the direct impact of the independent variable (Path c) and the impact of the mediating variable (Path b). There is also a path from the independent variable to the mediating variable (Path a). If the coefficient (β- value) in the multiple regression analysis is greater for Path b than Path c, then this indicates that more of the resultant change in the dependent is explained by the mediating variable than by independent variable.
147 | P a g e Mediating Variable Organisational Culture a b Independent Variable Knowledge sharing capability c Dependent Variable Knowledge sharing success
Figure 5.10. The path diagram model (Adapted from Baron and Kenny, 1986, p. 1176)
On the other hand, should the independent variable coefficient for Path c be greater than the mediating variable coefficient for Path b, then the direct Path
c explains more of the resultant change in the dependent variable than the indirect Path b. When both Path aand b are controlled, a previously significant relationship between the independent and dependent variables is no longer significant. Where the coefficient for Path c is reduced to zero, perfect mediation is evident. A multiple or partial mediating effect can be concluded where the coefficient for Path b is greater than the coefficient for Path c.
5.4.4.1 Mediating Testing
The effects of the mediating variable on the relationship between the independent and dependent variables were tested using Baron and Kenny’s (1986) procedure and further developed by Chen and Huang (2009). This procedure involves first regressing the dependent variable (step 1) and then the mediating variable (step 2) on the independent variables, followed by a third step (step 3) where the dependent variable is regressed on the mediating variable. Finally, the procedure involves regressing the dependent variable on both the mediating and independent variables to examine whether the mediating variable reduces the effects of the independent variables to non-significance.
148 | P a g e The organisational entity may influence the perceived level of knowledge sharing success because different organisations may exhibit different organisational cultural values and resource deployment. Therefore, a control variable (i.e. organisation) was also included as a predictor in the regression analyses to measure the potential effects.
Step 1 demonstrates whether the independent variables have a significant relationship with the dependent variable and, then in step 2, with the mediating variable. Similarly, step 3 explains whether the mediating variable’s relationship with the dependent variable has a significant mediating effect. Mediation, where the impact of the mediating variable explains some or all of the independent variable’s effect on the dependent variable, cannot be present if either of these relationships are not significant. Thus, to test for a mediating effect, it must first be established that all paths are significant (Baron & Kenny, 1986; Chen & Huang, 2009; King & Marks Jr., 2008). Table 5.10 to 5.14 summarise these findings.
Table 5.10
Regression Analysis – Knowledge sharing success on control, independent and mediator variables
Variables β Std. Error Std. β t ρ
Organisation 0.503 0.571 0.059 0.945 0.346
Organisation 0.083 0.057 0.092 1.448 0.149
Knowledge sharing capability 1.424 0.154 0.502 9.237 0.000
Organisation 0.078 0.057 0.081 1.361 0.175 Formal collaboration 1.053 0.139 0.460 7.567 0.000 Trustworthiness 0.457 0.137 0.170 3.330 0.001 Expertise 1.175 0.428 0.157 2.743 0.007 Independence 0.248 0.195 0.069 1.272 0.204 Table 5.11
Regression Analysis – Organisational culture (Formal collaboration) on control and independent variable
Variables β Std. Error Std. β t ρ
Organisation -0.369 0.204 -0.093 -1.805 0.072 Knowledge sharing capability 0.645 0.063 0.529 10.237 0.000
149 | P a g e Table 5.12
Regression Analysis – Organisational culture (Trustworthiness) on control and independent variable
Variables β Std. Error Std. β t ρ
Organisation 0.270 0.209 0.078 1.296 0.196
Knowledge sharing capability -0.145 0.064 -0.136 -2.245 0.026
Table 5.13
Regression Analysis – Organisational culture (Expertise) on control and independent variable
Variables β Std. Error Std. β t ρ
Organisation 0.106 0.070 0.087 1.523 0.129
Knowledge sharing capability 0.138 0.021 0.364 6.411 0.000
Table 5.14
Regression Analysis – Organisational culture (Independence) on control and Independent variable
Variables β Std. Error Std. β t ρ
Organisation -0.182 0.152 -0.070 -1.200 0.231 Knowledge sharing capability 0.231 0.047 0.053 4.929 0.138 Step 4 combines these regressions to identify whether the impact of the independent variable on the dependent variable (perceptions of knowledge sharing success) remains significant in the presence of the mediator variable (perceived level of organisational culture) and can therefore be attributed (partially or fully) to its mediating effect. Table 5.15 summarises these findings.
Table 5.15
Regression Analysis – Knowledge sharing success on control, independent and mediator variables
Variables β Std. Error Std. β t ρ
Organisation 0.080 0.056 0.084 1.418 0.157
Knowledge sharing capability 0.059 0.021 0.047 2.844 0.364 Formal collaboration 0.761 0.018 0.448 2.217 0.028 Trustworthiness 0.052 0.016 0.167 3.129 0.002
Expertise 0.047 0.052 0.156 0.909 0.048
150 | P a g e To summarise, a variable is considered a significant mediator when the following conditions are met:
x The independent variable is significantly related to the dependent variable in the step 1;
x The independent variable is significantly related to the mediating variable in the step 2;
x The mediating variable is significantly related to the dependent variable in the step 3; and
x The β-value of the independent variable on the dependent variable is less in the step 4 than in the step 1.
No mediation is evident when the mediator variable does not have a significant effect on the dependent variable (step 3) or when there is an absence of significant relationships between the independent variable (knowledge sharing capability) and the mediating variable, (organisation culture) (step 2). Partial mediation is evident when independent variables have a significant effect on both mediating (organisation culture) and dependent variables (knowledge sharing success) in steps 1 and 4, which
demonstrates that the β-value of the independent variable on the dependent
variable is lesser in step 4 than in step 1. Perfect or dominant mediation occurs when the independent variable (knowledge sharing capability) has a significant effect on the mediating variable (organisation culture) in step 2, but not on the dependent variable (knowledge sharing success) in step 4.
In this study, as model 2 of Table 5.16 shows, perception of knowledge sharing capability has a positive impact on knowledge sharing success. To test the second condition, the results in Table 5.11 of models 6a to 6c demonstrate that perceptions of knowledge sharing capability have significant effects on perceived organisational cultural values of collaboration, independence and expertise. As Model 3 of Table 5.10 shows, the mediator, perceived organisational cultural values of formal collaboration, trustworthiness, and expertise have significant and positive effects on knowledge sharing success, indicating the direct effect of organisational culture on knowledge sharing success. Further, results show that the
151 | P a g e perceived organisational cultural values of formal collaboration, trustworthiness, and expertise significantly reduce the effects of perceptions of knowledge sharing capability on the outcome variable, knowledge sharing success to being non-significant, which is indicative of perfect mediation effects. The inclusion of the mediating variable also causes an increase in the proportion of variance explained (R2 = 0.254 to R2 = 0.361). These findings suggest that the inclusion of organisational culture variables attenuates the relationships between perceptions of knowledge sharing capability and knowledge sharing success. This is evidence that organisational culture (i.e. the perceived values of formal collaboration, trustworthiness, and expertise) plays a mediating role between the perceptions of knowledge sharing capability and knowledge sharing success.
Table 5.16
Results of regression analyses of knowledge sharing success (Mediation)
Variables Knowledge Sharing Success (KSS)
Model 1 Model 2 Model 3 Model 4
Control variable Organisation 0.059 0.092 0.093 0.084 Knowledge sharing capability 0.502* 0.047 Organisational culture Formal collaboration 0.460* 0.448* Trustworthiness 0.170* 0.167* Expertise 0.157* 0.156* Independence 0.069 0.070 R2 0.003 0.254 0.360 0.361 F 0.893 43.253 28.068 29.588
152 | P a g e Table 5.17
Results of regression analyses of organisation culture
Variables
Organisational culture Formal
collaboration Trustworthiness Expertise Independence Model 5a Model 6a Model 5b Model 6b Model 5c Model 6c Model 5d Model 6d Control variable Organisation -0.111 -0.093 0.083 0.078 0.075 0.087 -0.080 -0.115 Independent variable Knowledge sharing capability 0.529* -0.136* 0.364* 0.053 R2 0.012 0.291 0.007 0.025 0.006 0.138 0.006 0.017 F 3.301 54.688 1.850 3.460 1.499 21.411 1.708 2.153 To summarise, all conditions were met for demonstrating that perceived organisational cultural values of formal collaboration, trustworthiness and expertise perfectly mediated the link between perceptions of knowledge sharing capability and knowledge sharing success. However, no mediation by knowledge sharing capability was present with respect to perceived cultural value of independence.