Introduction

The theory of international trade is a general equilibrium affair: offer curves, Edgeworth–Bowley box diagrams, and Travis boxes are frequently used tools. For a long time, however, applied trade policy analysis was restricted to partial equilibrium techniques. This was primarily due to constraints on data and computation for general equilibrium calibration and simulation. Nonlinear simulation software is now readily available, and general equilibrium techniques are often applied to trade policy problems. Surveys of these applications can be found in Shoven and Whalley (1984) and de Melo (1988).

Computable general equilibrium (CGE) models are calibrated to what are known as benchmark equilibrium datasets. The calibration process computes intercept and share parameters for the model's mathematical functions, given assumed or estimated values of behavioral elasticities, to reproduce the observed data as an equilibrium solution of the model. While much attention is devoted to the assumed or estimated behavioral elasticities, the calibrated intercept and share parameters are at least as important. Regardless of the quality of elasticity estimates, comparative static results have little empirical significance if they are not evaluated with respect to observed economic conditions.

A consistent and convenient means of compiling a benchmark equilibrium dataset is the social accounting matrix (SAM). This data framework has been extensively applied to developing countries under promotion by the World Bank and has been more recently applied to developed countries. A SAM for 1984 underlies the CGE model of the United States developed by de Melo and Tarr (1992). The CGE model in use at the U.S. International Trade Commission is currently based on a SAM updated biannually (Reinert and Roland-Hoist, 1991).