GTAP Models

Current Model
Previous Versions
Dynamic Model
Other Models


Land Use and Cover

GTAP Models: Computable General Equilibrium Modeling and GTAP

General equilibrium, which dates back to Leon Walras (1834-1910), is one of the crowning intellectual achievements of economics. It recognizes that there are many markets and that they interact in complex ways so that loosely speaking, everything depends on everything else. Demand for any one good depends on the prices of all other goods and on income. Income, in turn, depends on wages, profits, and rents, which depend on technology, factor supplies and production, the last of which, in its turn, depends on sales (i.e., demand). Prices depend on wages and profits and vice versa.

To make such an insight useful, economists have to be able to simplify it sufficiently to derive predictions and conclusions. Theorists typically do this by slashing the dimensionality, say to just two goods, two factors and two countries, and often focusing on just a few parts of the system. An alternative approach is to keep the complex structure but to simplify the characterization of economic behavior and solve the whole system numerically rather than algebraically. This is the approach of Computable General Equilibrium (CGE) modeling. CGE models specify all their economic relationships in mathematical terms and put them together in a form that allows the model to predict the change in variables such as prices, output and economic welfare resulting from a change in economic policies, given information about technology (the inputs required to produce a unit of output), policies and consumer preferences. They do this by seeking prices at which supply equals demand in every market goods, factors, foreign exchange. One of the great strengths of CGE models is that they impose consistency of one's view of the world, e.g., that all exports are imported by another country, that the sum of sectors' employment does not exceed the labor force, or that all consumption be covered by production or imports. This consistency can often generate empirical insights that might otherwise be overlooked in complex policy analysis - such as the fact that import protection gives rise to an implicit tax on exports. A key component of GTAP is a CGE model known as the GTAP Model, which is briefly documented in the GTAP book (Hertel, 1997). Another component of this project is the GTAP Data Base which underlies the GTAP Model.

The mathematical relationships assumed in the GTAP Model are generally rather simple, and although many markets are recognized, they still have to be very aggregated-particularly for global economic analysis. The GTAP Data Base underlying the GTAP Model has 57 sectors (in version 6), so, for example, transport and communications services appear as a single industry. In principle all the relationships in a model could be estimated from detailed data on the economy over many years. In practice, however, their number and parameterization generally outweigh the data available. In the GTAP Model, only the most important relationships have been econometrically estimated. These include the international trade elasticities (Hertel et al., 2005), and the agricultural factor supply and demand elasticities (OECD, 2001). The remaining economic relationships are based on literature reviews, with a healthy dose of theory and intuition. An important limitation of CGE models is that very few of them are tested as a whole against historical experience-although GTAP is one such (Valenzuela et al., 2007, Liu et al., 2004).

The standard GTAP Model is amenable to modifications. Many of these modifications are documented in the GTAP Applications.

CGE modeling is a very powerful tool, allowing economists to explore numerically a huge range of issues on which econometric estimation would be impossible; in particular to forecast the effects of future policy changes. The models have their limitations, however. First, CGE simulations are not unconditional predictions but rather thought experiments about what the world would be like if the policy change had been operative in the assumed circumstances and year. The real world will doubtless have changed by the time we get there. Second, while CGE models are quantitative, they are not empirical in the sense of econometric modeling: they are basically theoretical, with limited possibilities for rigorous testing against experience. Third, conclusions about trade policy are very sensitive to the levels assumed for trade restrictions in the base data. One can readily do sensitivity analysis on the parameter values assumed for economic behavior, although less so on the data, because altering one element of the base data requires compensating changes elsewhere in order to keep the national accounts and social accounting matrix in balance. Of course, many of these criticisms apply to other types of economic modeling, and therefore, while imperfect, CGE models remain the preferred tool for analysis of global trade policy issues.

Further information on CGE modeling can be obtained Wikipedia and there are numerous books on CGE modeling.

Adapted from Hertel, T., Keeney, R., Ivanic M. and Alan Winters, L. Distributional effects of WTO agricultural reforms in rich and poor countries. Economic Policy, April 2007, pp. 289-337.