Estimation results – Using disaggregate data with different IIT levels

(1) IIT between 0.75 and 1

Table 17 reports the estimation results for higher IIT bilateral trade flows. In the column one of Table 17, the PCS model predicts that bilateral trade among countries with a higher IIT level are positively related to the two countries’ GDPs, and negatively related to the two countries’ populations and the distance between the two countries. According to the result, a 10% rise in exporter’s GDP should be associated with 9.54% rise in exports. If an importer’s GDP rises by 10%, exports should be rise by 9.85%. If an exporter’s population rises by 10%, exports should decrease by 1.43%, while an importer’s population rises by 10%, its imports should decrease 0.67%. The PCS model implies that exports tend to be capital-intensive as indicated by the negative signs and significant statistically on the exporter’s population coefficient, and the negative coefficient importer’s population means that the industry’s output is a luxury in consumption. The first column of Table 17 reports that a exporting country will send 87.9% less of exports to a market that is twice as distant as another otherwise-identical market, 9.7% more to a country that uses the same language and 64.9% more to the two countries share a common border.

The PCS model predicts a 32% lower of bilateral trade if exporting and importing countries with a higher IIT trade flow are the EU members. The coefficient

34

on EURO is negative and insignificant statistically. The coefficient on NAFTA, AFTA and MERCOSUR are positive and significant statistically, indicating the bilateral trade in the NAFTA, AFTA and MERCOSUR countries with a high IIT trade flow is more intense than the other countries. In contrast, the FE 3-2 model indicates that intra-EU trade is lower than the average trade flow. And, the effect is statistically significant.

According to the statistics of the adjusted R², AIC and SC criteria in the second and third columns of Table 17, inclusion of the time-invariant variables slightly improves the estimation result. To select the best one among the rest of models in Table 17 the FE 2 model with –2.034 AIC and –1.089 SC should then be preferred to the others.

(2) IIT between 0 and 0.25

The estimation results of bilateral trade flows with a lower IIT levels are reported in Table 18. The PCS model predicts that bilateral trade flows with a lower IIT levels are positively related to the two countries’ GDP, and negatively related to the two countries’ population and the distance between the two countries. As usual, the two countries trade more with each other if they have common language and share a common border. According to the result, if an exporter’s GDP rises by 10%, it exports

35

should rise by 15.5%, holding importer’s GDP constant. From the result, a 10% rise in importer’s GDP should be associated with a 7.67% rise in exports, all else constant.

The PCS model predicts that exports tend to be capital-intensive as indicated by the negative signs and significant statistically on the exporter’s population coefficient.

To avoid collinearity, the NAFTA and AFTA dummies are omitted in the estimations of models PCS, FE 3-1 and FE 3-2. The estimation result of the PCS model shows the coefficients of EU and EURO are insignificant statistically. The PCS model predicts a 338.3% higher of a lower IIT bilateral trade if exporting and importing countries are the MERCOSUR members. Both models FE 3-1 and FE 3-2 predict that the EU effect is insignificantly positive. As the estimation models predict that the EURO effect is negative. In contrast, MERCOSUR is expected to have a positive effect on a lower IIT trade flow.

36

VI. Conclusions

This dissertation attempts to study the determinations of trade flows using traditional and modified models of gravity type. A pooled cross-section (PCS) model is used to serve as the benchmark. Three-way fixed effects models with and without time-invariant variables, a two-way fixed effect model are estimated for the comparison. By doing this, this dissertation investigates whether the bilateral trade flows are affected by trading agreements, discusses how different level of economic development affects bilateral trade flows, and measures how different level of intra-industry trade (IIT) has impact on bilateral trade. Based on the results, the study concludes the conventional variables such as the exporter’s and importer’s incomes have positive effects on bilateral trade, and the income elasticity of bilateral trade is higher when the modified gravity specification is used. However, the coefficients of the exporter’s and importer’s population are positive and statistically significant, which is contrary to the conventional wisdom (see Table 12).

This study analyzes the determination of bilateral trade flows for different countries with various levels of economic development. Each model used in the study predicts positive and statistically significant coefficients on the developed exporter’s and developed importer’s GDPs, as well as for the coefficients on the developing exporter’s and developing importer’s GDPs (see Tables 13-14). Based on the results,

37

the estimated elasticities of trade with respect to GDPs are higher when both exporting and importing countries are developing countries, compared to the case where both exporting and importing countries are developed countries. When the PCS model is used, the estimates on distance, language and adjacency are also higher in the case where both exporting and importing countries are developing countries.

This study also investigates the difference of various IIT bilateral trade flows, i.e.

classified into four ranges by using the intra-industry index (IIT) for the estimation.

Each model used in the study predicts positive and statistically significant coefficients on the developed exporter’s and developed importer’s GDPs, as well as for the coefficients on the developing exporter’s and developing importer’s GDPs (see Tables 17-18). Based on the results, in the case of higher bilateral IIT trade flows, the estimated elasticities of trade with respect to exporter’s and importer’s GDPs are quite similar. But, the impacts of exporter’s and importer’s GDPs on trade flows show uneven in the case of lower bilateral IIT trade flows. Also, when the PCS model is used, the estimates on distance, language and adjacency are smaller in the case of higher bilateral IIT trade flows.

To appropriately provide the comparison of the estimation results of conventional and modified models, this study considers the following regional economic arrangements: EU, EURO, NAFTA, AFTA and MERCOSUR. As we are noticed that

38

multilateral trade barriers in recent decades under GATT and WTO are declining, and specific preference between trading partners is not considered in the standard gravity models, failure to take account of falling transportation costs can cause the REI effects to be misleading. Based on my results, the REI effects from estimating the PCS model vs. the FE models are quite different.

The results suggest that the trade pattern both within-NAFTA and within-MERCOSUR exhibit a sign of trade creation. The two-way FE model, a favorite model suggested by the information criteria, predicts that within-EU and within-EURO effects are positive, while within-AFTA effect is negative, but neither of them is statistically insignificant. The same signs on the REI dummies can be found in the case in which the models are estimated with different levels of countries’

economic development. When different IIT trade flows are considered, the within-bloc effect of NAFTA, AFTA and MERCOSUR is strictly positive and statistically significant. However, the intra-bloc effect of EU or EURO is ambiguous under different specifications. The model controlling for trading pair (Model FE2) predicts that within-bloc effect are positive and statistically significant, except for the EU.

Although this study already includes more countries, in particular in the study of the effect of economic integration on trade flows than earlier literature, this study does

39

not consider if trade flows differ across different industry, due to restriction on data sources to calculate the IIT indexes. Since changes in trade policy and the cost structure of transportation may become an important indicator of trade patterns nowadays, thus a parallel analysis of long run trends in these factors is needed. As noted in the literature, failure to obtain appropriate proxies for transportation costs and trade policy would cause some biases. In this study, to solve the estimation biases I have proposed a two-way fixed-effects gravity model in which the country-specific intercepts capture the effects of trade policy and transport costs. A further area of investigation would be to include the variables that measure tariff and non-tariff barriers in the estimation. In condition, to distinguish how of economic development and intra-industry bilateral trade flows, further search may include dummy variables to indicate if a trade flow is higher IIT, and if both trading countries are developed countries, or developing countries..

40