5. EMPIRICAL STUDY
5.7 Multivariate testing
5.7.6 Comparison with competing models
High discriminant validity indicates that a construct is unique and captures some characteristics that other constructs do not.
The two models show that there is a very low level of cross-loading among the indicators of upstream competitiveness and downstream competitiveness (Table 5.6). Therefore, it is safe to conclude that the indicators only represent their own constructs.
5.7.3.3 Nomological validity: This is a test to examine whether the correlations between the constructs in the measurement theory make sense. As the model does not specify any correlations among the constructs and because it focuses on causalities, this test was not necessary.
5.7.3.4 Face validity is the extent to which the content of the indicators matches the construct definition. This judgement is based on logic and common sense. As the indicators and constructs used in this test are developed on the basis of theoretical literature and related empirical studies, they qualify for face validity.
run. All the goodness of fit indices showed a slight change from the original model. However, they still satisfied the conditions of good fit. It is important to note that all statistical coefficients pertain to the loading paths of the new factors are insignificant, and that the values of the coefficient estimates remained virtually unchanged (Table 5.7).
Table 5.7: Result of common method bias test
MODEL A MODEL B
Path relationships Coefficients Sig. Path relationships Coefficients Sig.
PriceQuality < MBF .024 .74 PriceQuality < MBF .041 .53
ProductionEfficiency < MBF -.135 .13 ProductionEfficiency < MBF .163 .06
ProductApplicability < MBF -.030 .69 ProductApplicability < MBF .072 .26
PhysicalDistant < MBF -.206 .23 PhysicalDistant < MBF .142 .17
PsychicDistant < MBF .014 .90 Nationalism < MBF -.380 .1
Nationalism < MBF .391 .12 PsychicDistant < MBF -.176 .02
StrategicResponseCompetitive < MBF .543 .22
StrategicFoothold < MBF .494 .08
Given the insignificance of the new parameters and the stability of the original parameters, there is no evidence to supports the hypothesis that the data suffers from common method bias.
It is therefore safe to conclude that the measurement method does not bias the results.
5.7.5 The mediation effect
The output from the AMOS tests indicates that the total effect of international strategies variables on export performance is the product of the effects of internationalisation strategies on particular upstream or downstream competitiveness and the effect of this competitiveness on firm performance. See Tables 9.7 and 9.8 in Appendix C, which provide details on total direct dissemination from AMOS. This result strengthens our argument that upstream and downstream international competitiveness are prerequisites for enhancing export performance.
5.7.6 Comparison with competing models
Based on other theoretical and empirical studies, two competing models were constructed. The primary objective was to ensure that the proposed model not only has an acceptable fit, but that it also performs better than some alternate models.
Figure 5.2: Competing model 1 – two export performance constructs as reflective indicators
Competing model 1: The export performance construct was split into two constructs (Diamantopoulos 1999, Zou et al. 1997) comprising two dimensions: economic performance and strategic performance. Therefore, overall export performance became a construct that reflects two subordinate constructs (Figure 5.2). The alternative model has goodness of fit indices that are much lower than the proposed model:
χ2= 375.1, CMIN = 2.89 CFI=0.809
RSMEA=0.094.
Therefore, the original model has a significantly better goodness of fit than competing model 1.
Export Facilitators Export Determinants
Export Performance Upstream
competitiveness Import of machinery and software
Downstream competitiveness Import of know-how
and user rights Import of material
Conduct of business trips Employment of sales staff with int’l experience
Employment of sales staff with foreign language proficiency
+ + + +
+ +
+ +
Controls (GVC insertion, size, ownership structure, capital
structure and industry)
+
+
Employment of foreigners with technical expertise
Use of Internet for day-to-day business
Collaboration with other firms
Use of governmental linkages
+
Use of formal business networks
+ +
Strategic Export Perf Economic Export Perf
+
Export Performance
Competing model 2: Competing model 2 was constructed with two mediation constructs:
upstream competitiveness and downstream competitiveness, which have recursive paths from each other. Previous studies (Korhonen 1999, Kuada and Sorensen 2000, Karlsen et al. 2003) indicate that there is an interdependent relationship between upstream and downstream competitiveness in which upstream competitiveness can lead to downstream competitiveness and vice versa. Therefore, the model was specified with the recursive paths from both of these constructs (Figure 5.3 with economic export performance as a dependent construct). For information on strategic export performance, see Figure 9.2 in Appendix C.
Figure 5.3: Competing model 2 − recursive paths between upstream competitiveness and downstream competitiveness
Competing model 2, with economic export performance as a dependent construct, has the following goodness of fit indices:
χ2= 448, CMIN= 3.34
Export Facilitators Export Determinants
Economic Export Performance Upstream
competitiveness Import of machinery and software
Downstream competitiveness Import of know-how
and user rights Import of material
Conduct of business trips Employment of sales staff with int’l experience
Employment of sales staff with foreign language proficiency
+ + + +
+ +
+ +
Controls (GVC insertion, size, ownership structure, capital
structure and industry)
+
+
Employment of foreigners with technical expertise
Use of Internet for day-to-day business
Collaboration with other firms
Use of governmental linkages
+
Use of formal business networks
+ +
EI Profit Growth
+
Export Performance
CFI=0.801 RSMEA=0.102
For competing model 2 with strategic export performance as a dependent construct, the following goodness of fit indices were displayed:
χ2= 450.54, CMIN= 3.78 CFI=0.70
RSMEA=0.11
For the economic and strategic export performance variables, all goodness of fit indices for competing model 2 are significantly lower than for the proposed model.
The Chi-Square difference test was then used to see if competing model 2 fits the data equally well as the proposed model. The difference between the Chi-square statistics of the two models are 223.8 and 240.9 respectively for economic and strategic export performance construct.
Thus, the Chi-square difference statistics is not statistically significant at 0.05. Therefore, we can conclude that the proposed model appears to have the best fit for this sample.