The Armington Model

Multi-country computable general equilibrium (CGE) models used to analyse tariff and trade policy changes typically incorporate the Armington structure which differentiates commodities by their country of origin (national product differentiation), and assumes them to be imperfect substitutes for each other.

In contrast to the well-known Heckscher-Ohlin model, relatively little is known about ‘Armington models’ and their properties despite their wide acceptance among model builders and policymakers. This makes it difficult to interpret the trade and welfare results that might arise from trade liberalisation simulations that are based on Armington models.

Introducing the Armington structure changes fundamentally the properties of a trade model regardless of the values assumed for the elasticities of substitution between imported and domestically produced goods. In particular:

• there is no comparative advantage and hence no gains from trade due to product specialisation;
• the number of products is fixed and hence there are no gains from trade due to increased product variety; and
• large terms of trade effects tend to offset other gains from trade.

As a consequence of these properties, Armington models tend to understate the gains from tariff and trade policy liberalisation.

A numerical, 3-good, 3-country modification of the Global Trade Analysis Project (GTAP) model is used to illustrate these properties. Compared to a Heckscher-Ohlin model, a unilateral across-the-board cut in tariffs in an Armington model results in:

• a larger shift in consumption from domestically produced goods to imported goods;
• a larger decline in terms of trade; and
• a smaller resource reallocation across industries.

The paper indicates possible future directions for methodology and practice.