4.1 Reliability and Validity
Three constructs of TSA were compressed into a single measurement of transaction cost variable, using principal components analysis method. Likewise, four constructs of REG were compressed into a single measurement of regulative institution variable. Cronbach’s α for three constructs of TSA is 0.518, and that for four constructs of REG is 0.603, indicating that the compressed measures are reliable, albeit marginally (Nunnally 1967, 1978). The results of scale purification of TSA and TSA are provided in Appendix B.
All measures of our research were based on the literature of distinguished journals, and interviews with and experts; therefore, theses appropriate contents suggest that our items relating to transactions costs and institutions for EMS test have strong content validity.
However, there is no scientific cut-off point for the construct reliability. Nunnally
recommended that the minimally acceptable reliability statistic such as Cronbach alpha to be in the range of 0.5 to 0.6 in his 1967 edition of Psychometric Theory (p.226). But he raised the recommended level to 0.7 in the 1978 edition, which is widely cited in the literature. A survey of the articles published in the marketing-related journals by Churchill and Peter (1984) found the mean value of reliability statistics to be 0.751 with a standard deviation of 0.156.
33
Peterson’s (1994) updated survey of marketing-and psychology- related studies found that about three-quarters of the reported Cronbach alpha’s are equal or greater than 0.7, but the rest is smaller than 0.7.Education levels of respondents, among other things, are found to affect the measurement reliability.
4.2 Data Analysis
Given that the dependent variable representing the control level of the entry mode is a discrete variable and ordinal in nature, an ordered probit model was employed for regression analysis. Three regression models were investigated to compare the explanatory power of transaction costs and the institutional environment with regard to the choices of entry mode, with the results being presented in Table 3. In Model 1, only the control variable and transaction cost variables were considered. In Model 2, institutional variables were combined together with the control variable. In Model 3, which was the fullest model, transaction cost and institutional variables were joined together with the control variable.
34
Table 3
Using Transaction Cost and Institutional Environment as Variables for Ordered Probit Analysis on Entry Mode
Model 1 Model 2 Model 3 Model
Variables Coefficient Std. Error Coefficient Std. Error Coefficient Std. Error Transaction
cost variables TSA
0.1221 0.063893 0.1046 0.0693 REG 0.1571* 0.0723 0.1504* 0.0726
PN -0.069 0.0576 -0.0831 0.0585 NET -0.1235** 0.0443 -0.1336** 0.0449 MIMIC 1.1872** 0.1810 1.1631** 0.1820 COGNITIVE 0.1488** 0.0555 0.1387* 0.0560 CD -0.079 0.0533 -0.0754 0.0534 Institutional
variables
AGE -0.0522** 0.0182 -0.058** 0.0189 -0.0564** 0.0190 Log likelihood -311.1174 -278.0027 -276.8615
Degree of freedom 2 7 8
P value 0.0019 0.0000 0.0000
Likelihood ratio test
Comparison model (Model 1) (Model 2) Log likelihood ratio 68.5118** 2.2823
Probability 0.0000 0.1309
Notes:
1. Dependent variables are wholly-owned subsidiary (WOS), exclusive dealership (ED), and non-exclusive dealership (NED), with WOS=2, ED=1, and NED=0.
2. TSA is the transaction variable constructed from TSA1~TSA3, and REG is the regulative institution variable constructed from REG1~REG4 (See Appendix A).
3. Coefficients (standard errors) are given for each independent variable; *p<0.05; **p<0.01
35
In all models, time of entry (AGE) exhibits a significant and negative effect on the choice of entry mode. This suggests that the later the time of investment, the higher the control mode that will be chosen. In Model 1, the three indicators representing transaction costs, when combined into a construct, exert a significantly positive effect on the entry mode choice. This indicates that when the transaction costs are higher, a higher control level in the entry mode will be preferred. The regression results support our hypothesis H1.
Model 2 indicates that the institutional environment does have some significant explanatory power on the entry mode choice. The regression results support our hypotheses H2, H4, H5, and H6; however, H3 and H7 are not supported by our data as professional norms and cultural distance have been shown to be inconsequential to the choice.
Model 3, which is the most comprehensive model, includes both transaction costs and institutional variables at the same time. When compared with Model 1, Model 3 shows that institutional variables exert a significant incremental explanatory power (with a log-likelihood ratio of 68.5118). When compared to Model 2, Model 3 shows that the incremental explanatory power of the transaction cost variables is not significant (with a log-likelihood ratio of 2.2823). The above empirical findings support our assertion that the institutional theory has more explanatory power than the transaction cost theory on entry mode choice in the China market.
36
4.3 Results summary
In what follows, we sum up the effects of the individual institutional variables. In terms of regulative institutions, higher institutional risks lead to lower levels of control in entry mode choice. This suggests that when the local government is more liberal, when the tax regime is more transparent and when the financial institutions are better developed, PFC is more likely to enter the region by establishing a wholly-owned subsidiary. The tighter the network ties, the lower the level of control that will be sought in terms of the entry mode.
Moreover, foreign enterprises are inclined to follow the market leaders in their choice of entry mode, which corresponds to the mimetic isomorphism of the institutional theory. In our case, PFC tended to adopt an entry mode similar to that of the market leader, which was Master Kong in most sales districts, as the leader had already accumulated experience and established reputation in the regional markets. Nevertheless, professional norms do not appear to have had a significant effect on the entry mode choice. This suggests that industrial associations in the Chinese markets have not yet established professional credibility to protect the interests of their members. This is probably because state-owned enterprises do not need the help of industrial associations as they have political recourses, and privately-owned enterprises only play a minor role in the “market economy” of China. Finally, in terms of cognitive pressure, the regression results indicate that the more easily that a newly-established distributor is received by the local markets, the higher the level of control in terms of entry mode that will
37
be adopted. However, cultural distance is shown to be an insignificant factor in entry mode decision. A possible explanation is that Taiwanese cultural backgrounds are so close to those of China that this is no issue. The other possibility is the measurement problem. Tihanyi et al.
(2005) has shown that whether cultural distance affects entry mode choice is sensitive to the way cultural distance is measured.