We test the sensitivity of our results in several ways. First, we examine other performance measures, such as return on equity (Luo and Chung, 2013).
The results from both are similar, shown in Model 8 of Table 5. Second, because the data contain observations of foreign SMEs across several years, there may be unobserved characteristics or propensities that influenced the foreign SMEs in their performance. We therefore include two year dummies to control for the three years. The results are similar and presented in Model 9 of Table 5. Third, we also estimate our regression using lagged-structure models by regressing performance in year t+1 on all independent variables in year t to correct for potential endogeneity. The results do not change, though the remaining sample size consists of 7949 observations, which are reported in Model 10 of Table 5.
Fourth, while the number of lawyer represents a sufficient proxy for the development of legal institutions, several studies have suggested measuring the impact of patents granted to establishments (Delios and Beamish, 1999; Guler and Guillén, 2010). Therefore, an alternative measure used is the growth rate of the ratio of patent applications granted to gross capital within a province, which
Table 5
Regression results for performance in robustness testsa
Model 8
Firm sizeb .013(.030) .003(.003) -.005(.003)
Networking capital .000(.001) .001(.001)*** .001(.001)***
Industry dummyc .067(.084) .009(.004)*** .003(.005)
Potential market demand -.020(.020) .003(.001)*** .005(.003)**
Labor supply -.481(.509) -.001(.001) -.001(.001)
Growth rate of labor cost -.112(.077)* -.095(.023) -.077(.038) Independent variables
Local market orientation -.004(.001) .005(.008) -.007(.013) Legal institutions -.020(.017) -.004(.001) -.007(.004) Financial institutions .003(.018) .001 (.002) -.002 (.004) Specialized agglomeration economies .004(.030) .002(.004) .008(.008) Diversified agglomeration economiesb .015(.038) .009(.002)*** .007(.005)*
Interaction
Local market orientation * Legal
institutions .023(.016)* .006(.003)*** .005(.004)
Local market orientation * Financial
institutions .005(.003)* .009(.003)*** .010(.004)***
Local market orientation * Specialized
agglomeration economies .006(.004)* .009(.005)** .128(.009)*
Local market orientation * Diversified
agglomeration economies .009(.006)* .008(.005)* .012(.008)*
Year 2005 .020(.009)***
Year 2006 .030(.008)***
_cons .118(.029)*** .033(.005)*** .043(.009)***
Wald chi-square 84.57*** 175.64*** 79.02***
Df 16 18 16
Observations 14028 14042 7949
Number of firms 6147 6147 4544
R2 0.028 0.033 0.019
a: The method of General Linear Square (GLS) Random-Effects models is used to estimate the parameters of models, in which *** indicates significant at .01, ** indicates significant at .05, and * indicates significant at .10. Robust standard errors are reported in parentheses.
b: Logarithm.
c: Coded 0/1.
is reported by the China Statistical Yearbook. Additional regressions suggest that these four hypotheses are still firmly supported by using such an alternative measure, as shown in Model 11 of Table 6. Fifth, aiming to construct another measure relating to financial institutions, we calculate the number of financial practitioners within each province based on the China Statistical Yearbook.
Using such an index leads to similar results provided in Model 12 of Table 6.
Sixth, following the method outlined by Sobel (1982), we also examine the potential for a mediating (versus a moderating) effect of legal institutions, financial institutions, and specialized and diversified agglomeration economies where local market orientation influences the degree of the aforementioned four institutional variables, which in turn impacts firm performance. Three conditions are necessary for the presence of a mediation effect (Baron and Kenny, 1986). In the beginning, the independent variable must significantly influence the dependent variable(s); its corresponding equation is Y=cX. Next, the independent variable must significantly influence the mediator(s); its regression equation is M=aX. Finally, the mediator(s) must significantly affect the dependent variable(s) after the influence of the independent variable is controlled;
its related equation is Y=c2X+bM. If the Sobel test Z-value3 is significant (z>1.96), then the mediation effect is likely present. Non-significant z-values (p>.05) indicate that legal institutions and financial institutions, as well as specialized and diversified agglomeration economies, respectively, are not mediators. Thus, our data negate the possibility of a mediating relationship.
Seventh, a potential concern for the empirical analysis is that four institutional factors—legal institutions, financial institutions, and specialized and diversified agglomeration economies—may be endogenous. Since performance and the aforementioned institutional factors are continuous, we use a two-stage least squares (2SLS) regression analysis to overcome the endogeneity problem.
In the first stage, we choose two instrument variables (IVs)—regional population and new registrations of medium-sized trucks—from the China Statistical Yearbook and find that these two IVs are positively and strongly correlated with
3 𝑧𝑧 = 𝑎𝑎∗𝑏𝑏
�𝑎𝑎2∗𝑆𝑆𝑏𝑏2+𝑏𝑏2∗𝑆𝑆𝑎𝑎2
Table 6
Regression results for performance using different proxiesa Model 11 Model 12 Control variables
Firm ageb .025(.096) -.006(.006)
Firm sizeb -.053(.069) .004(.003)*
Networking capital .001(.001) .001(.001)***
Industry dummyc .012(.090) .009(.004)***
Potential market demand -.022(.022) .002(.001)***
Labor supply .001(.001) .001(.001)***
Growth rate of labor cost .151(.723) -.042(.020)
Independent variables
Local market orientation .104(.069)* -.005(.006)
Legal institutions .001(.001)
Legal institutions: patents .246(.160)*
Financial institutions -.026 (.027)
Financial institutions: financial practitioners -.002 (.001) Specialized agglomeration economies .085(.070) .002(.004) Diversified agglomeration economiesb .104(.098) .011(.002)***
Interaction
Local market orientation * Legal institutions .008(.003)***
Local market orientation * Legal institutions:
patents .384(.292)*
Local market orientation * Financial institutions -.065(.048)*
Local market orientation * Financial institutions:
financial practitioners .001(.001)***
Local market orientation * Specialized
agglomeration economies .029(.020)* .009(.005)**
Local market orientation * Diversified
agglomeration economies .134(.176) .011(.006)**
cons .143(.037)*** .048(.002)***
Wald chi-square 22.16*** 159.40***
Df 16 16
Observations 14022 14042
Number of firms 6147 6147
R2 0.004 0.027
a: The method of General Linear Square (GLS) Random-Effects models is used to estimate the parameters of models, in which *** indicates significant at .01, ** indicates significant at .05, and * indicates significant at .10. Robust standard errors are reported in parentheses.
b: Logarithm.
c: Coded 0/1.
the presence of legal institutions and financial institutions (p < .00), and that they are also jointly statistically significant (F = 356.89, p< .00 in the legal institutions model, and F = 677.59, p< .00 in the financial institutions model, respectively). We choose another two IVs—length of highway transport routes and persons employed in SOEs—in the China Statistical Yearbook and test the relevance of the four IVs. Our findings show that the length of highway transport routes is positively associated with specialized and diversified agglomeration economies, while the number of persons employed in SOEs is negatively correlated. Furthermore, these two IVs are jointly statistically significant (F = 62.11, p < .00 in the specialized agglomeration economic model, and F = 722.11, p < .00 in the diversified agglomeration economic model, respectively). However, these four IVs are uncorrelated with foreign SMEs’
performance. Additionally, the Sargan statistic (overidentification tests of all instruments) fails to reject the null hypothesis that the instruments are not correlated with the main equation errors (χ2 = .01, p > .05 in the legal institutions model, and χ2 = 1.48, p > .05 in the financial institutions model, and χ2 = .77, p
> .05 in the specialized agglomeration economic model, and χ2 = .89, p > .05 in the diversified agglomeration economic model, respectively). We are therefore confident in the relevance and exogeneity of the IVs. Note that these instruments are not perfect, but are useful for eliminating reverse causality problems. In the second stage, the results show that the coefficients on the interaction of local market orientation and the instrumented legal institutions and financial institutions, as well as the interaction of local market orientation and the instrumented specialized and diversified agglomeration economies, are statistically significant and positive in the second-stage regressions, indicating that our findings remain valid even after controlling for the potential endogeneity problem.