Armington tariff simulator

Method

The policy surface implements an Armington (1969) CES partial-equilibrium model. It allocates each exporter's product exports across destinations in proportion to that exporter's total bilateral trade shares, converting BACI values from thousands of USD to plain USD.

For each importer, the solver forms baseline expenditure shares, applies the tariff-inclusive price ratio to each supplier, computes a CES import price index, and updates bilateral quantity and value. The ten largest countries by trade are represented separately, together with the country of interest and all policy countries; all others are grouped as ROW.

HS6 substitution elasticities come from Soderbery (2018). Stored sigmas are used in absolute value. A single HS6 uses its own sigma; HS4 or multi-code scenarios use an export-weighted mean for the country of interest. The fallback is sigma = 4.0 when no estimate is available.

Welfare uses a first-order Harberger approximation, most reliable for moderate tariff changes. Trade diversion is detected by comparing bilateral baseline and scenario flows.

The separate GET estimator applies an HS4 import-demand elasticity to a tariff change relative to the latest available MFN baseline.

Equations

Validation

• The policy endpoint reports a sigma-sensitivity band from 0.5, 0.75, 1.0, 1.25 and 1.5 times the selected sigma, with a floor of 1; the selected estimate is always included in the band.
• No out-of-sample or external benchmark validation is documented in the named repository sources.

Limitations

• Product-level bilateral flows are estimated from product exports and total bilateral trade shares; they are not observed bilateral product flows.
• This is a partial-equilibrium model. The first-order welfare approximation is most reliable for moderate tariff changes.
• The price result is a change in total import cost. The implementation has no quantity data from which to estimate a unit price.
• Most countries are aggregated into ROW, although the country of interest and policy countries are kept separately.
• The trade-flow inputs use the live BACI estate, while product elasticities remain on the pinned legacy estate. One result therefore combines vintages.
• The HS4 GET endpoint's dollar impact applies the elasticity to global exports of the HS4 line, not to one country's imports.
• When the elasticity table or a product estimate is unavailable, the full simulator uses sigma = 4.0.

References

• Armington (1969), A Theory of Demand for Products Distinguished by Place of Production.
• Soderbery (2018), Trade Elasticities, Heterogeneity, and Optimal Tariffs, Journal of International Economics 114.

Endpoints