V. The Estimation Results
2. Estimation results – Using disaggregate data with different levels of economic development
(1) When both exporting and importing countries are developed countries
The results of estimating the coefficients on gravity variables in various models are reported in Table 13. The estimated coefficients on the developed exporter’s and developed importer’s GDPs are all positive and significant statistically. According to the results of the PCS model, if a developed exporter’s GDP rises by 10%, it exports should rise by 8.73%, holding importer’s GDP constant. From the result, a 10% rise in developed importer’s GDP should be associated with a 3.97% rise in exports, all else constant. The PCS model predicts that exports tend to be capital-intensive as indicated by the significant negative sign on the developed exporter’s population coefficient.
And the positive coefficient developed importer’s population means that the
29
industry’s output is a necessity in consumption. As shown in Table 13, a developed exporter will send 76.5% less of exports to a market that is twice as distant as another otherwise-identical market, and 73.0% more to a country that uses the same language.
The fourth column of Table 13 reports the estimation result for the two-way fixed effects model. When both exporter and importer are developed countries, according to the results a 10% rise in exporter’s GDP should be associated with a 12.7% rise in exports, all else constant. If an importer’s GDP rises by 10%, exports should rise by 12.8%, holding exporter’s GDP constant. The estimated coefficients of population are about –1.34 and –0.68. This means that when the change of size of a developed exporting country is smaller by 10%, trade growth decreases by about 13.4%, and if the change of an importer’s size rises by 10%, the growth exports should fall by 6.8%.
Notice that controlling for the country-pair heterogeneity not only higher the estimated elasticity of trade with respect to GDPs but also reverses the signs on the developed importer’s population.
To avoid collinearity, the AFTA and MERCOSUR dummies are removed in the estimation. By the PCS model, the coefficient on the dummy variables for the EU and NAFTA are significantly negative and positive, and for the EURO is negative but insignificant statistically. All models predict that intra-NAFTA trade is higher than the average trade flow. The PCS model predicts a 103.1%, the three-way FE model with
30
time-invariant predicts a 123.8%, the FE 3-2 model predicts a 308.3%, and the FE 2 model predicts a 23.2% higher than an average bilateral trade.
As shown in the first column of Table 13, the within-EU effect is significantly negative, indicating that trade between developed exporters and importers is lower than the average trade flow by 14.3%. In contrast, the results for the FE 3-1, FE 3-2 and FE 2 models are presented in columns two, three and four of Table 13. The within-bloc effect of the EU is likely to be positive and significant statistically. For the EURO, the two-way fixed effects model indicates that the EURO countries trade among themselves 23.2% more than an average trading pair outside the bloc.
According the statistics shown in Table 13, the model (FE 2) has highest explanatory power (the adjusted R2), and also presents the lowest information criterion, thus is the favorite.
(2) When both exporting and importing countries are developing countries
The results of estimating the coefficients on gravity variables of the PCS model are presented in column one of Table 14. The estimated coefficients on the developing exporter’s and importer’s GDPs are 1.76 and 1.23, indicating that although faster growing countries do trade more and, the increase of trade is more than proportionate.
The PCS model predicts that exports tend to be capital-intensive as indicated by the
31
significant negative sign on the developing exporter’s population coefficient. The first column of Table 14 reports that a developing exporting country will send 122.3% less of exports to a market that is twice as distant as another otherwise-identical market, and 87.2% more to a country that uses the same language. The coefficient on Adjacency is positive and significant statistically.
To avoid collinearity, the EU, EURO and NAFTA dummies are removed in the estimation. According to the estimation results for the PCS model, the coefficient of AFTA and MERCOSUR are positive and significant statistically, indicating that the bilateral trade in the AFTA and MERCOSUR countries is more intense than the other countries. The PCS model predicts that intra-AFTA trade is 163.7% higher than an average trade flow, and intra-MERCOSUR trade is 112.4% higher than an average trade flow. From the third column of Table 14, the FE 3-1 and FE 3-2 models indicate that the within-bloc effects of the MERCOSUR and the AFTA are likely to be positive and significant statistically. But, they turn to be insignificant in the FE 2 model.
According to the statistics of the adjusted R2, AIC and SC criteria in the third and fourth columns of Table 14. To select available among the rest of models in Table 14, the FE 2 model with 0.192 AIC and 0.597 SC should then be preferred to the others. The FE 3-1 model remains the second-best.
32
(3) When either exporting or importing country is a developed country
Table 15 reports the results of the models for exporting and importing countries are developed and developing countries respectively. To avoid collinearity, the EU, EURO and MERCOSUR dummies are omitted. For NAFTA, the PCS model estimates a decrease in trade of 45.6% that is far from being statistically significant, whereas the FE 3-1 model with time-invariant estimates a statistically significant effect of –15.3%. The PCS model suggests a 338.7% increases in trade among AFTA members, whereas the FE 3-1 model suggests that the bloc led to only 48.7% increase in trade. The FE 2 model shows a statistically significant effect of 50% for the NAFTA and a statistically insignificant effect of –7.7% for the AFTA.
Table 16 reports the results of the models for exporting and importing countries are developing and developed countries respectively. To avoid collinearity the EU, EURO and MERCOSUR dummies are omitted. The PCS model shows a statistically insignificant effect of –5.5% for the NAFTA and a statistically significant effect of 332.3% for the AFTA. The FE 3-1 model also estimates the effects of the NAFTA and the AFTA as increases in intra-bloc trade of about 98.6% and 42.3%. The FE 2 model indicates that intra-NAFTA trade is about 114.1% higher than an average trade flow, and this model also shows a statistically insignificant effect of –29.5% for the AFTA.
33
3.