The Effect of Pre-existing FTAs with Asymmetric Third Countries on the Formation of a New FTA

Cited 0 time in Web of Science Cited 0 time in Scopus


사회과학대학 경제학부
Issue Date
서울대학교 대학원
FTAInterdependenceFree Trade NetworkSpatial Dependence with Dyadic DataBig Data AnalysisLasso
학위논문 (박사)-- 서울대학교 대학원 : 경제학부 경제학전공, 2016. 8. 박지형.
Based on the theoretical model of 'Free Trade Network (2007)', this thesis defines the net welfare change caused by a new FTA derived under the condition that both parties do not have concluded any FTAs with other parties and under the condition that they have the pre-existing FTAs with asymmetric third countries. By substituting the real trade and economic data of 189 countries from 1993 to 2012 into the derived welfare change, this thesis can predict whether a new FTA is formed or not. To show how well it can be predicted, the substitutability, which is used for defining the welfare change but unobservable, and the best critical value of each country that matches `Real FTA Status' and `Predicted FTA Status' the most are estimated through the calibration. Then, this thesis can show that `Real FTA Status' corresponds to 90 percent of `Predicted FTA Status' which can be predicted by the welfare change caused by a new FTA and the critical value of each country that makes itself willing to sign a new FTA.
This thesis predicts how pre-existing FTAs with other asymmetric third countries, the industrialization level, and MFN tariff level of each country that participates in a new FTA affect the formation of a new FTA based on the aforementioned theoretical results, which has not been noted in the previous literatures. Those predictions are verified with the way of 'Spatial Dependence with Dyadic Data'. Specially, in this thesis, whether to sign a new FTA or not, which is predicted according to the critical value of each country estimated through the calibration, is used as a dependent variable to empirically test the unilateral incentives rather than `Real FTA Status' that represents the bilateral decisions agreed by both parties.
Finally, this thesis attempts to identify the most important issues that each country considers in signing a new FTA by selecting substantially influential variables through the Big Data Analysis. For this, I first generate the dataset that reflects the directly participating countries' economic characteristics, political characteristics, multilateral issues related to GATT/WTO, and the interdependence of preferential trading relationships as mentioned in the previous literatures. Then, I can get the different subset of variables for each individual country by applying the glmnet package and 10-fold cross validation method into R programming for `Lasso' and `Lasso' logit regression. Finally, I regress one year lagged variables selected in each `Lasso' and `Lasso' logit regression on `Real FTA Status' with each OLS and probit. The coefficients are very similar to each other in `Lasso' and OLS and `Lasso' logit and probit method. R² of each estimation result regressed with selected variables for the observations of each individual country is much higher than with the same selected variables for observations of all countries. In fact, the variables selected through the `Lasso' regression are powerful to explain what kinds of issues truly affect the willingness to sign a new FTA. The results seem to reflect each economical, geographical, or geopolitical tendency toward forming new FTAs, which is supported by the high R² regardless of the numbers of predictors.
This thesis contributes to using the theory-based calibration to empirically test the determinants of FTAs, empirically analyzing unilateral and bilateral incentives based on the generalized theory model of Furusawa and Konish (2007), which can explain that the effects of pre-existing FTA on the formation of a new FTA can be determined by the parameters such as the industrialization level, MFN tariff level, and economic size of each participating country, and lastly analyzing the determinants of FTAs for each individual country by selecting the relevant variables that substantially affect the formation of FTAs among the variables mentioned in the previous literatures by using Big Data Analysis.
Files in This Item:
Appears in Collections:
College of Social Sciences (사회과학대학)Dept. of Economics (경제학부)Theses (Ph.D. / Sc.D._경제학부)
  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.