TABLE 6.
Self-sampling test of a dynamic threshold effect.
| Core explanatory variable | Model | Threshold value | Wald statistic | P-value | BS times | 95% confidence interval | |
| FDI | SYS-GMM | 0.326 | 16.008*** | 0.000 | 1000 | 0.058 | 0.442 |
| DIF-GMM | 0.085 | 55.380*** | 0.000 | 1000 | 0.058 | 0.442 | |
| OFDI | SYS-GMM | 0.357 | 36.560*** | 0.000 | 1000 | 0.058 | 0.442 |
| DIF-GMM | 0.430 | 222.818*** | 0.000 | 1000 | 0.058 | 0.442 | |
| Trade | SYS-GMM | 0.442 | 118.123*** | 0.000 | 1000 | 0.058 | 0.442 |
| DIF-GMM | 0.085 | 1.130*** | 0.000 | 1000 | 0.058 | 0.442 | |
***Significant at the level of 1. P-value and critical value are obtained by repeated sampling of the GMM threshold panel regression program for 1,000 times. Wald statistic is used to judge whether the threshold feature is obvious. The smaller the corresponding probability, the more obvious the threshold feature is.