Assessing the impact of data disaggregation level and non-tariff barriers in regional trade agreements utilizing the Global Trade Analysis Project Framework

摘要

Computable general equilibrium (CGE) models have been extensively used by economists for trade policy analysis due to their ability to quantify the impact of a shock on an entire economy. Providing economy-wide numerical results, and including linkages and interactions among main economic variables, agents, sectors, and regions make CGE models preferable in addressing a wide range of economic problems. Among various comparative static, multi-sector and multi-region general equilibrium models, Global Trade Analysis Project (GTAP) is one of the most extensively used. However, despite the widespread use of CGE models in trade policy analysis, there are still debates among researchers about the right choice of the model to apply. The discussions are frequently about the data aggregation level. The degree of data disaggregation within the CGE models has direct impact on policy simulation results stemming from the aggregation bias. Against this background, one of the focal points of this dissertation is the impact of aggregation bias occurring in GTAP simulations and the reasons behind this bias.Another focal point of this dissertation is the estimation of the ad-valorem equivalents (AVEs) of non-tariff barriers (NTBs) on food and agricultural sector through gravity approach and their subsequent implementa-tion into the GTAP framework for thorough analysis of regional trade agreements (RTAs). With the increas-ing number of economic integration agreements and multilateral trade negotiations of the World Trade Or-ganization, the importance of import tariffs has declined, while that of NTBs has risen, since NTBs are hard-er to address due to their complex structure. However, the welfare gains through the reduction of restrictive NTBs due to RTAs are not negligible. We either use the border effect approach or the free trade agreement (FTA) approach to identify NTBs in the trade between respective countries. NTBs are originally not consid-ered in the standard GTAP framework. However, they can be implemented into the GTAP model in several ways (i.e., as export taxes, import tariffs or as efficiency losses) depending on the policies with which they are related. Due to our focus on the agro-food sector in our articles and the predominance of technical NTBs on this sector, we mainly account for the efficiency-decreasing effect of NTBs. Hence, we model a majority of them using the efficiency approach. For the remaining part of trade costs we utilize the import-tariff ap-proach.In this context, the objective of this cumulative dissertation is threefold: (1) to reveal the impact of data ag-gregation level in trade policy analysis with the GTAP framework, (2) to expose the importance of NTBs in the evaluation of RTAs, (3) to demonstrate the effect of data aggregation level in gravity estimates of NTBs and its subsequent impact on trade policy simulations. Hence, this dissertation consists of four articles which are published or submitted to journals. In our first article entitled "Model Structure or Data Aggregation Level: Which Leads to Greater Bias of Results?", we focus on two fundamental characteristics of CGE models, i.e., the model structure and the data aggregation level. Our results demonstrate that there are substantial differences in results due to the use of GE or PE model structure or data disaggregation level. However, the deviations in results caused by sectoral breakdown are much more pronounced than those stemmed from the model structure. While the economy-wide setting of GE models causes differences across the results of GE and PE models, tariff averaging and false competition ground the reason for deviations in results due to data aggregation level.Following our theoretical work in the first article, in our second article, "Moving toward the EU or the Mid-dle East? An Assessment of Alternative Turkish Foreign Policies Utilizing the GTAP Framework", we focus on more applied analysis. In this article, we analyze Turkeys two different policy options by considering the simultaneous elimination of NTBs and import tariffs in the case of Turkeys membership either to the Euro-pean Union (EU) or Greater Arab Free Trade Area (GAFTA). For both experiments, gains from NTB re-moval outweigh the gains due to the elimination of import tariffs. Hence, based on our simulation results, we are able to confirm the importance of NTBs in the evaluation of RTAs. After indicating the importance of aggregation bias in our first article and confirming the impact of NTBs in the evaluation of RTAs in the second, in our third article, "The Effect of Aggregation Bias: An NTB-Modelling Analysis of Turkeys Agro-Food Trade with the EU", we expound the magnitude of aggregation bias in the calculation of AVEs of NTBs. Our estimations demonstrate that using aggregated gravity model to estimate the AVEs of NTBs results in overestimation of trade costs. Hence, the transfer of overestimated trade costs to the GTAP model also leads to overestimation in the simulation results of the EUs extension to include Turkey. Our last article, "Keep Calm and Disaggregate: The Importance of Agro-Food Sector Disaggregation in CGE Analysis of TTIP", is designed as a follow-up to our first article; however, it also includes the key find-ings from the second and third articles. We create five different versions of the GTAP database, which are aggregated at different sector levels. Thereafter, we simulate the Transatlantic Trade and Investment Partner-ship (TTIP) between the EU and the United States (US). In addition to what we constructed in our first arti-cle, in this article we also consider the reduction NTBs for each version of the GTAP database. Hence, in addition to averaging of tariffs and false competition, estimation of AVEs of NTBs at different data aggrega-tion levels also has an impact on deviations in simulation results across five versions of the GTAP database. As we have presented in our articles, the use of higher data disaggregation level commonly results in greater welfare and trade effects, but cases also exit in which more aggregated version of the GTAP database leads to larger changes in simulation results. The atheoretic method of trade-weighted tariff aggregation given in the GTAP database is the trigger of lower trade and welfare effects. By calculating of the Mercantalistic Trade Restrictiveness Index (MTRI) for bilateral import tariffs, and comparing them with the initial trade-weighted tariffs in the GTAP database, we are able to verify the underestimation effect of "tariff averaging". In contrast, "false competition" causes overestimation of trade and welfare effects when higher level of data aggregation is used in the simulations. False competition arises in such situations when competition for a particular subsector does not initially exist between two exporting countries, but this subsector can be aggre-gated with others in which competition actually exists. Hence, this situation leads to wrongly applied weights, and results in false substitution effects, which causes overestimation of results. The estimation of AVEs of NTBs at higher data aggregation levels also reduces the variation across sectors, and commonly leads to higher trade and welfare results. However, the contribution of tariffs to the deviation of results across versions is generally higher than the contribution of NTBs. Hence, based on our simulation results, we exhibit that aggregation of tariffs is more important than the NTBs. This dissertation concludes that neither the impact of aggregation bias nor the importance of NTBs in the evaluation of RTAs on trade policy analysis is negligible. There are considerable differences across simula-tion results depending on the data aggregation level used. The differences in results occur both in the estima-tion of trade costs of NTBs and also in the policy simulation results on the GTAP level. Hence, the selection of data aggregation level can be critical for thorough analysis of trade agreements, especially for the detailed examination of policy changes at the product level. Aggregation bias cannot be entirely overcome in econo-metric estimates or in CGE analysis; however, the extent of its possible effect can be born in mind. Depend-ing on the aim of the policy analysis, the appropriate level of data disaggregation should be chosen.