Foreign Direct Investment and Efficiency Benefits

Further econometric issues

When estimating production functions within a static framework as in the previous

section, the well-known problems of multicollinearity and endogeneity may arise. As

argued in section 3.1, these problems are alleviated by using the labor intensive form

                                                          

18

 A Wald test was performed to test for the hypothesis of equal spillover coefficients from small and

large foreign firms. The hypothesis was accepted at p=0.62.

19

 Despite the relatively large difference in the estimated coefficients a Wald test could not reject the

hypothesis of equal coefficients at p=0.32.

20

 All regressions were re-estimated without the industry dummies. The results remained unchanged.

---

15

of equation (3), nevertheless the obtained estimates may lack some of the required

properties, in particular unbiasedness and consistency. These problems are faced

adequately by the use of instrumental variable (IV) techniques when a time dimension

of the cross-sectional data is provided and, of course, in the case of panel data.

21

In an attempt to increase the robustness of our results and given the limitation

of our data set, we were able to trace back those firms from our sample that were

operating in 1992. Thus, using the 1992 values of the exogenous variables as

instruments, we re-estimated our regressions applying the IV procedure on the sample

of firms operating in both years and the results are shown in Tables 7 and 8. It turns

out that the estimated coefficients are close to the heteroscedasticity corrected OLS

estimates appearing in Tables 4 and 6. Actually, the spillover effects estimated with

IV keep their significant positive sign, hence supporting the robustness of our

estimations. No noticeable differences in the parameter estimates of the other

variables were observed.