• No results found

5 Data Analyses

5.5 Secondary Research

This section will outline the results of the A-Meieriet case. The complete case text is displayed in appendix A. A univariate analysis of variance (ANOVA) was performed in order to evaluate whether there were significant differences between the groups (control, negative – and positive priming).

5.5.1 Hypothesis Testing

H7: Effective SCM reduces assignment of blame

To measure whether SCM has an effect on blame, a univariate analysis was performed. The result (m = 3.41, Std.Dev = 1.19, df = 4, F = 9.94, p<.000) confirms H7. A simple linear regression was run to further investigate the relationship. The output (∆ R2

= .17, t = 6.13, F = 37.52, B = .48, SE.B = .08, β = .41) confirms H7, that effective SCM reduces stakeholders‟

assignment of blame by increasing trust in the organization.

H8: A positive CSR reputation moderates the relationship between SCM and assignment of blame. In order to assess H8, a regression with interactions effects were performed, where the independent variable consisted of SCM and CSR.

Table 20 Regression with Interaction Effects

Linear Regression ∆ R2 t F B SE B β

Positive Priming .11 2.6 6.73 .18 .07 .35 **

Negative Priming - .00 .96 .93 .23 .24 .19

Note: *** p<.001, ** p<.01, ∆ R2: Adjusted R Square, t: T-value of t-test, F: F-value of F-test, B: Unstandardized Coefficient Standard error Beta, SE B: Unstandardized Coefficient Beta, β: Standardized Coefficient Beta.

The result for the positive CSR condition is significant, hence H8 is accepted. In the negative CSR condition the result is not significant.

H9: A positive CSR reputation reduces assignment of assignment blame.

A univariate analysis was performed, and the Scheffe values may be viewed in appendix G.

Table 21 Univariate Analysis Total Effects of CSR on Blame

Ind. Var: CSR, Dep.Var: Q60 Mean Std. Dev.

Control Group 3.84 .83

Negative priming 2.51 1.27

Positive priming 2.75 .97

Note: Q: Question numer.

Table 21 indicates that the positive priming group attributes more trust, thus reduced blame to A-Meieriet compared to the negative priming group, m(P) = 2.75 while m(N) = 2.51.

However, the control group attributes more trust, thus reduced blame m(C) = 3.84 compared to the positive priming group. The results indicate that significant (p=.000) differences

61

between the control and the negative priming group, as well as between the negative and positive priming group (p=.000). Test of the total effects (c) show that the locus, stability and controllability significantly enhance the effect CSR has on blame. The significant results indicate that positive CSR reputation reduces blame, thus H9 must be accepted.

H10a): A positive CSR reputation will cause the locus of the crisis to be perceived as external rather than internal.

H10b): A positive CSR reputation will cause the crisis event to be perceived as temporary rather than stable.

H10c): A positive CSR reputation will cause the crisis event will be perceived as controllable rather than uncontrollable.

Table 22 summarizes the descriptive statistics from the univariate analysis of whether CSR affects the attribution process. The Post hoc and Scheffe can be viewed in appendix G.

Table 22 Descriptive Statistics Univariate Analysis

Descriptive Statistics Q53 Locus 1 Q54 Locus 2 Q55 Locus 3 Q56 Stability

Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.

Control group 59.51 23.915 29.41 18.461 11.40 11.637 4.63 2.619

Negative priming 71.64 23.150 20.88 16.710 7.48 10.700 5.98 2.554

Positive priming 59.40 24.256 28.60 17.472 12.00 13.609 5.31 2.500

Q57 Controllability

Mean Std. Dev.

Control group 6.63 2.195

Negative priming 5.40 2.570

Positive priming 6.65 2.259

Note: Q: Question number

Table 22 indicate that the positive priming group attribute less responsibility to A-Meieriet than subjects in the negative priming condition, as depicted by the authors (m(P)= 59.4 while m(N)= 71.64). Appendix G reveals a significance (p=.012) between the control group and the negative priming group, and (p=.013) between the negative and positive priming group. The positive priming group assigns a larger extent of responsibility to the supplier than the negative priming group, (m(P) = 28.6 while m(N) = 20.88). This is significant (p=.019) between the control and the negative priming group as well as (p=.042) between the negative and positive priming group. The results also indicate that the positive priming group assign a larger extent of responsibility to the customer, than the negative priming group, (m(P) = 12.0 while m(N) = 7.48). However, the results are not significant. Based on these findings, the positive priming group perceive the locus of the crisis to be more external compared to the negative priming group who perceive the locus to be more internal. Thus, H10a) is accepted.

62

Respondents in the positive priming group perceive the product harm crisis to more temporary than those in the negative priming condition (and the control group), m(P) = 5.31 while m(N)

= 5.98). The results are only significant (p=.014) between the control and the negative priming group. Consequently, H10b) is rejected.

The positive priming condition assigns a higher degree of controllability of A-Meieriet to prevent similar types of crises from occurring in the future than the negative priming group, and a greater amount of controllability within the crisis compared to the negative priming subjects, as shown by m(P) = 6.65 while m(N) = 5.40). The results are significant (p=.013) among the positive and negative priming conditions, and (p=.015) between the control and the negative priming group. Therefore, H10c) is accepted.

5.5.2 Bootstrap Indirect Effects & Simple Mediation

H11a): Locus mediates the relationship between CSR and assignment of blame.

H11b): Stability mediates the relationship between CSR and assignment of blame.

H11c): Control mediates the relationship between CSR and assignment of blame.

The syntax macro equation may be viewed in appendix I.

Table 23 Simple Mediation Effects Results (Preacher & Hayes, 2008)

Coeff. a Coeff. b Coeff. c Coeff. c`

Locus1 -13.80* -.01* 1.25*** 1.08***

Locus2 9.51* .01 1.25*** 1.14***

Locus3 4.29 .02* 1.25*** 1.15***

Stability -.77 -.03 1.20*** 1.17***

Controllability 1.32*** .01 1.24*** 1.22***

Note: Coeff: Coefficient, *** p<.001,** p<..01, * p>0.5

The simple mediation (INDIRECT) macro for SPSS indicates that Locus1 mediates the effect of CSR on blame (Field, 2009), which is illustrated in figure 20. A-Meieriets responsibility within the PHC mediates the relationship between CSR and blame.

*** p<.001, ** p<.01, * P<.05

Figure 20 Simple Mediation – Impact of CSR on Locus1 via Blame

The results indicate however, that Locus2 and Locus3 do not represent mediators between CSR and assignment of blame. Thus, locus may not be viewed as a mediator between CSR

63

and blame. Stability and Controllability have no mediation effects on the relationship between CSR and blame.

Note: SE: Standard Error, LL: Lower Level, CI: Confidence Interval, UL: Upper level.

The bootstrap analysis showed that there was a significant indirect effect of Locus2 & 3 on attribution of blame, as indicated by the exclusion of zero in the bootstrapped confidence interval (Locus2: 95% CI = {.01, .27, Locus3: 95 % CI= {.01, .26}). The standardized indirect effect is significantly different from zero at the .001 level (p=.001). The bootstrap further demonstrates that H11b & c includes zero in the CI (Stability: 95 % CI = {-.03, .17} and controllability {-.11, .19}). For stability, the path of c and c` is significant, but not path a and b. Whereas for controllability c and c` is also significant, and path a and b is not significant.

Thus, H11a, b & c must be rejected (though Locus1 did represent a mediator, because the other constituents of locus did not). Note in figure 21 that Locus2 & 3 is rejected as mediators, thus H11a is only partially accepted. However, Locus is considered to represent an indirect effect.

Figure 21 Revised Secondary Research Model from Bootstrap Mediating Analysis