Chapter 4: Research Results
4.6 Multi-Group Analysis for the Moderating Effect of Environmental Dynamism….110
To confirm the moderating effect of environmental dynamism on the relationship between EO and resource attributes (including value and rareness), this dissertation uses the methodology of a multi-group analysis within LISREL 8.54 and assesses two effects of environmental dynamism with the high and low levels on the structural model. Following the study by Brockman and Morgan (2006), the examination is conducted in a two-step approach.
First, the appropriate structural parameters are constrained to be equal among two groups, thereby generating a covariance matrix for each group and an overall chi-square value for the sets of sub-models as part of a system of structural models. Second, the parameter equality constraints, the EO-resource attributes (value or rareness) link, must be removed, resulting in the chi-square value with fewer degrees of freedom. The moderating effect is examined by estimating whether significant differences exist between the above two chi-square values.
When the second chi-square value is significantly less than the first chi-square value, the null hypothesis of parameter invariance is rejected, indicating a moderating effect.
Based on the studies of Brockman and Morgan (2006), an environmental dynamism index is formed by averaging the items. As mentioned in the above section, this study splits the sample into two groups on the basis of the scores for the environmental dynamism variable; thus, this sample is divided into two different groups, namely, high levels of environmental dynamism and low levels of environmental dynamism.
Subsequently, a multi-group analysis is conducted relying on both environmental dynamism groups when firm performance is competitive advantage, ROA, TQ, or
111
performance, respectively. These results are summarized in Table 4-6. The EO-value link and EO-rareness link show a statistically significant difference in chi-square values between the two groups for the competitive advantage, ROA, TQ, and performance, respectively. This means that environmental dynamism moderates the EO-value relationship and the EO-rareness relationship, respectively. According to the parameter estimates in Table 4-7, the EO-value path is significant in a high level of environmental dynamism but insignificant in a low level That is, the relationship between EO and value is stronger for firms facing a high level of environmental dynamism than for those facing a low level. The EO-rareness link has similar results; thus, Hypotheses 6a and 6b are supported.
In summary, the results of the multi-group analysis for the moderating effect are consistent with the results from subsection 4.5. Therefore, the influence of environmental dynamism on the EO-resource attributes link is supported again.
Table 4-6 Multi-Group Difference Test (Moderating Variable: Environmental Dynamism)
Hypothesis Firm 6a Satisfaction Influence of EO on value
8.515**
6b Satisfaction Influence of EO on rareness Note: * P < 0.10, ** P < 0.05, *** P < 0.01.
112
Table 4-7 Parameter Estimates, Low Versus High Environmental Dynamism
Hypothesis
Firm performance
Hypothesized Moderating
Path
Low estimate
Low t-value
High estimate
High t-value
6a CA EO-value link 0.15 1.41 0.41*** 3.16
6b CA EO-rareness link 0.15 1.44 0.55*** 5.11
6a ROA EO-value link 0.13 1.37 0.44*** 3.16
6b ROA EO-rareness link 0.13 1.41 0.68*** 4.98
6a TQ EO-value link 0.13 1.37 0.44*** 3.16
6b TQ EO-rareness link 0.13 1.41 0.68*** 4.98
6a Satisfaction EO-value link 0.15 1.42 0.40*** 3.45
6b Satisfaction EO-rareness link 0.14 1.33 0.53*** 4.73 Note: * P < 0.10, ** P < 0.05, *** P < 0.01.
113
4.7 Comparison of Alternative Models
To confirm whether our proposed model fits the data well, this study compares the efficacy of several alternative models (Aryee, Budhwar, & Chen, 2002). Based on the different indicators of firm performance, this study classifies into three hypothesized models to test the proposed hypotheses. The results are presented in Table 4-8. Hypothesized models l, 2, and 3 are the completely mediated models. Alternative models 1-1, 2-1, and 3-1 are the direct models and contain three paths: from EO to firm performance, from value to firm performance, and from rareness to firm performance. Alternative models 1-2, 2-2, and 3-2 are partially mediated models, with paths from EO to value, from value to firm performance, and from EO to firm performance. Alternative models 1-3, 2-3, and 3-3 are partially mediated models, containing three paths: from EO to rareness, from rareness to firm performance, and from EO to firm performance.
The results of Table 4-8 emphasize that our hypothesized models (Models 1, 2, and 3) fit the data better than the alternative models. The descriptions are as follows. First, the differences in the chi-square values (χ2) between hypothesized model 1 and the other models (Alternative models 1-1, 1-2, and 1-3) are 0.98, 0.94, and 2.15 respectively. These differences are all insignificant. Second, the differences in the chi-square values (χ2) between hypothesized model 2 and other models (Alternative models 2-1, 2-2, and 2-3) are 6.95, 1.20, and 1.64 respectively. Alternative models 2-2 and 2-3 are not better than hypothesized model 2, except for alternative model 2-1. Therefore, the results don’t support the optimization of hypothesized model 2. Finally, the differences in the chi-square values (χ2) between
114
hypothesized model 3 and other models (Alternative models 3-1, 3-2, and 3-3) are 3.19, 0.60, and 0.89 respectively. Overall, the hypothesized models 1, 2, and 3 are supported by the data.
Relying on former arguments, this study adds additional comparison tests to confirm whether the hypothesized models in the different two groups (high levels of environmental dynamism vs. low levels of environmental dynamism) still fit the data well; thus, these alternative models are compared again by this study. As shown in Tables 4-9 and 4-10, this study shows the similar results. All hypothesized models provide an adequate fit to the data, and alternative models are not significantly better than the hypothesized models.
In summary, on the whole, the hypothesized models 1 and 3 are more consistent with the data than any of four alternative models. Therefore, the results confirm the optimization of hypothesized models 1 and 3. That is, as dependent variables are competitive advantage and satisfaction, the completely mediated models are supported by data.
115
Table 4-8 Comparison of Competing Models with Full Sample (N=201)
Model Test Firm performance χ2 df χ2 / df △χ2 △df △χ2/△df GFI NFI CFI RMSEA
Hypothesized model 1 CA 93.61 60 1.56 0.93 0.98 0.99 0.055
Alternative model 1-1 CA 95.33 59 1.62 0.98 1 0.98 (insignificant) 0.93 0.98 0.99 0.055
Alternative model 1-2 CA 69.97 32 2.17 26.34 28 0.94(insignificant) 0.93 0.97 0.98 0.077
Alternative model 1-3 CA 18.86 24 0.79 77.45 36 2.15 (insignificant) 0.98 0.99 1.00 0.000
Hypothesized model 2 TQ, ROA 194.62 97 2.01 0.90 0.94 0.96 0.071
Alternative model 2-1 TQ, ROA 180.72 95 1.90 13.90 2 6.95 (significant) 0.90 0.94 0.97 0.067 Alternative model 2-2 TQ, ROA 150.12 60 2.50 44.50 37 1.20 (insignificant) 0.90 0.92 0.95 0.087 Alternative model 2-3 TQ, ROA 115.76 49 2.36 78.86 48 1.64 (insignificant) 0.91 0.93 0.96 0.083
Hypothesized model 3 Satisfaction 211.42 85 2.48 0.90 0.92 0.94 0.052
Alternative model 3-1 Satisfaction 122.32 83 1.47 89.10 2 44.5 (significant) 0.91 0.95 0.98 0.059 Alternative model 3-2 Satisfaction 107.67 51 2.11 103.75 35 2.96 (insignificant) 0.92 0.93 0.96 0.075 Alternative model 3-3 Satisfaction 90.43 41 2.20 120.99 45 2.68 (insignificant) 0.92 0.94 0.96 0.078 Notes:
△χ2 is the difference between the hypothesized model 1 and the competing models (Models 1-1 to 1-3), the hypothesized model 2 and the competing models (Models 2-1 to 2-3), and the hypothesized model 3 and the competing models (Models 3-1 to 3-3) respectively. If the value (△χ2/△df ) is smaller than 3.84, the model will not be adapted.
Models 1, 2, and 3 are the hypothesized models (completely mediated model).
Models 1-1, 2-1, and 3-1 are the direct models. The path is from EO, value, and rareness to firm performance.
Models 1-2, 2-2, and 3-2 are the partially mediated model. The path is from EO to value, from value to competitive advantage, and from EO to firm performance.
Models 1-3, 2-3, and 3-3 are also the partially mediated models. The path is from EO to rareness, from rareness to competitive advantage, and from EO to firm performance.
116
Table 4-9 Comparison of Competing Models with High Degrees of Environmental Dynamism (N=95)
Model Test Firm performance χ2 df χ2 / df △χ2 △df △χ2/△df GFI NFI CFI RMSEA
Hypothesized model 1 CA 103.22 60 1.72 0.86 0.95 0.98 0.089
Alternative model 1-1 CA 103.08 59 1.74 0.14 1 0.14(insignificant) 0.85 0.95 0.97 0.091
Alternative model 1-2 CA 51.91 32 1.62 59.81 28 2.13 (insignificant) 0.90 0.95 0.98 0.085
Alternative model 1-3 CA 37.76 24 1.57 39.88 37 2.05 (insignificant) 0.96 0.99 0.99 0.057
Hypothesized model 2 TQ, ROA 161.90 97 1.66 0.84 0.90 0.95 0.085
Alternative model 2-1 TQ, ROA 153.06 95 1.61 16.31 2 4.42 (significant) 0.83 0.91 0.95 0.082 Alternative model 2-2 TQ, ROA 100.48 60 1.67 61.42 37 1.66 (insignificant) 0.86 0.90 0.94 0.086 Alternative model 2-3 TQ, ROA 94.86 49 1.94 67.04 48 1.39 (insignificant) 0.86 0.91 0.95 0.100
Hypothesized model 3 Satisfaction 140.37 85 1.65 0.83 0.91 0.96 0.084
Alternative model 3-1 Satisfaction 141.29 83 1.87 2 -0.92 insignificant 0.83 0.91 0.96 0.086 Alternative model 3-2 Satisfaction 87.60 51 1.72 52.77 33 1.59 (insignificant) 0.87 0.90 0.95 0.090 Alternative model 3-3 Satisfaction 86.88 41 2.11 69.80 43 1.62 (insignificant) 0.86 0.91 0.95 0.109 Notes:
△χ2 is the difference between the hypothesized model 1 and the competing models (Models 1-1 to 1-3), the hypothesized model 2 and the competing models (Models 2-1 to 2-3), and the hypothesized model 3 and the competing models (Models 3-1 to 3-3) respectively. If the value (△χ2/△df ) is smaller than 3.84, the model will not be adapted.
Models 1, 2, and 3 are the hypothesized models (completely mediated model).
Models 1-1, 2-2, and 3-1 are the direct models. The path is from EO, value, and rareness to firm performance.
Models 1-2, 2-2, and 3-2 are the partially mediated model. The path is from EO to value, from value to competitive advantage, and from EO to firm performance.
Models 1-3, 2-3, and 3-3 are also the partially mediated models. The path is from EO to rareness, from rareness to competitive advantage, and from EO to firm performance.
117
Table 4-10 Comparison of Competing Models with Low Degrees of Environmental Dynamism (N=106)
Model Test Firm performance χ2 df χ2 / df △χ2 △df △χ2/△df GFI NFI CFI RMSEA
Hypothesized model 1 CA 62.90 60 1.04 0.92 0.96 1.00 0.025
Alternative model 1-1 CA 61.81 59 1.04 1.09 1 1.09 (insignificant) 0.92 0.96 1.00 0.025
Alternative model 1-2 CA 35.25 32 1.10 27.65 28 0.99 (insignificant) 0.96 0.97 1.00 0.036
Alternative model 1-3 CA 23.23 24 0.96 61.13 37 1.65 (insignificant) 0.96 0.97 1.00 0.000
Hypothesized model 2 TQ, ROA 161.90 97 1.68 0.86 0.92 0.96 0.085
Alternative model 2-1 TQ, ROA 144.27 95 1.69 17.63 2 8.82 (significant) 0.85 0.91 0.96 0.070 Alternative model 2-2 TQ, ROA 119.62 60 1.89 51.48 38 1.35 (insignificant) 0.88 0.91 0.95 0.084 Alternative model 2-3 TQ, ROA 92.93 49 1.99 68.97 48 1.44 (insignificant) 0.87 0.90 0.94 0.092
Hypothesized model 3 Satisfaction 104.15 85 1.52 0.88 0.91 0.98 0.048
Alternative model 3-1 Satisfaction 99.74 83 1.20 4.41 2 2.21 (insignificant) 0.88 0.91 0.98 0.052 Alternative model 3-2 Satisfaction 71.12 51 1.39 33.03 33 1.00 (insignificant) 0.90 0.90 0.96 0.063 Alternative model 3-3 Satisfaction 62.69 41 1.52 41.46 43 0.96 (insignificant) 0.90 0.90 0.95 0.071 Notes:
△χ2 is the difference between the hypothesized model 1 and the competing models (Models 1-1 to 1-3), the hypothesized model 2 and the competing models (Models 2-1 to 2-3), and the hypothesized model 3 and the competing models (Models 3-1 to 3-3) respectively. If the value (△χ2/△df ) is smaller than 3.84, the model will not be adapted.
Models 1, 2, and 3 are the hypothesized models (completely mediated model).
Models 1-1, 2-2, and 3-1 are the direct models. The path is from EO, value, and rareness to firm performance.
Models 1-2, 2-2, and 3-2 are the partially mediated models. The path is from EO to value, from value to competitive advantage, and from EO to firm performance.
Models 1-3, 2-3, and 3-3 are also the partially mediated models. The path is from EO to rareness, from rareness to competitive advantage, and from EO to firm performance.
118