The Global Trade Analysis Project (GTAP) is a research program initiated in 1992 to provide the economic research community with a global economic dataset for use in the quantitative analyses of international economic issues. The Project's objectives include the provision of a documented, publicly available, global, general equilibrium data base, and to conduct seminars on a regular basis to inform the research community about how to use the data in applied economic analysis. GTAP has lead to the establishment of a global network of researchers who share a common interest of multi-region trade analysis and related issues. The GTAP research program is coordinated by Thomas Hertel, Director of the Center for Global Trade Analysis at Purdue University. As Deputy Director of this Center, Robert McDougall oversees the data base work. Software development within the GTAP project has been assisted greatly by the efforts of Ken Pearson and other Australian researchers from Centre of Policy Studies, Monash University. (See Hertel , McDougall . A list of applications based on the GTAP framework can be found at the GTAP home page .)
The standard programming language for GTAP data and modeling work has been GEMPACK (Harrison and Pearson ). In the GEMPACK framework the model is solved as a system of nonlinear equations. The present paper describes a version of the GTAP model which has been implemented as a nonlinear complementarity problem in the GAMS programming language. Along with the core model I have developed several ancillary programs for dataset management. I call the package "GTAPinGAMS". These programs should be useful to economists who program in GAMS and wish to use GTAP in applied work. These programs include tools for translation of the GTAP files into GAMS readable form, GAMS programs for dataset aggregation, filtering and the imposition of alternative tax rates on trade or domestic transactions.
The GTAP version 4 database represents global production and trade for 45 country/regions, 50 commodities and 5 primary factors. The data characterize intermediate demand and bilateral trade in 1995, including tax rates on imports and exports. The core static model described in this paper does not have precisely the same structure as the GTAP model implemented in GEMPACK. There are several immediate differences between the standard GTAP model and the GTAPinGAMS model. First, the units of account are different by a factor of 10000. GTAP measures all transactions in millions of 1995$. GTAPinGAMS measure transactions in tens of billions of 1995$.2
Second, there is a potentially important difference concerns the structure of final demand. In the GEMPACK model final demand is represented by a constant-difference-elasticity (CDE) demand system whereas in the GAMS model final demand is Cobb-Douglas. Given differences in functional forms, even if benchmark value shares and reference prices are identical the two models may produce somewhat different estimates of policy changes due to differences in income and substitution elasticities.
A third set of differences concerns the representation of investment demand and global capital markets. The standard GTAP model assumes that a "global bank" allocates international capital flows in response to changes in regional rates of return. The GTAPinGAMS model makes the simplest possible assumptions regarding investment demand, international capital flows and the time path of adjustment: all of these variables are exogenously fixed at base year levels. 3
I chose to design the core model with simple Leontief, Cobb-Douglas and constant-elasticity-of-substitution (CES) functional forms so that it's structure could be as transparent as possible. These choices reflect my belief that any application of the GTAP data to a specific policy question should involve the development of a model tailored to the issues, and therefore the purpose of the core model is largely to illustrate how the benchmark data are organized.
This paper consists of three sections following this overview. Section 2 presents the core static model using Mathiesen's format for the Arrow-Debreu model. This section provides notation and equations describing technology, preferences and equilibrium conditions.
Section 3 describes how the GTAP model can be expressed in GAMS, either as an MPSGE model or as a system of algebraic equations. This material provides a short but complete overview of how the technology and preferences are calibrated along with GAMS code which performs this task.
Section 4 has a practical perspective with step-by-step instructions on how to install the GTAPinGAMS package. The intent of this material is to provide as short as possible a learning curve for economists who wish to perform a few calculations using the GTAP dataset. This section describes ancillary GAMS programs which have been developed for use with the GTAP 4 dataset. These include GAMS libinclude programs which read and write GTAP header-array files4; FILTER.GMS, a GAMS program which removes small trade flows and intermediate demands from a GTAP dataset to increase sparsity and provide improved computational performance in large scale models; IMPOSE.GMS, a GAMS program which permits arbitrary adjustment of benchmark tax rates with least-squares recalibration; and GTAPAGGR.GMS, a GAMS program which aggregates any GTAP dataset to a smaller number of goods, factors or regions.