How the numbers on this site are produced

The model: a Ricardian, multi-sector trade model

Every scenario on this site is solved with a general-equilibrium trade model in the tradition of Caliendo & Parro (2015) — itself a multi-sector extension of the Eaton–Kortum Ricardian trade framework, in which sectors also trade with each other as intermediate inputs. In plain terms:

• Each country's technology in each sector is described by a distribution of production efficiencies, so countries specialize according to comparative advantage — an outcome of the model, not an assumption fed into it.
• Producing anything requires labor plus inputs bought from other sectors, including sectors abroad. How much of each input a sector uses, and where it sources them from, follows the calibrated real-world structure of production captured in the underlying data — so a shock that hits one sector can raise costs anywhere that sector's output is used as an input, not only in the sector itself.
• How strongly trade between two countries in a given sector responds to a change in relative prices — including a tariff or trade-cost change — is governed by that sector's trade elasticity (see "Sectoral trade elasticities" below).

Given a scenario — a set of tariff or trade-cost changes — the model searches for the wages and prices, in every country and sector, at which every market clears and each country's overall trade balance stays fixed at its pre-shock level — its baseline deficit or surplus is preserved, not forced to zero. Everything reported on this site is the difference between that new equilibrium and the pre-shock baseline, not a forecast and not a path over time (see "What this model assumes", below).

What "welfare" means here

"Welfare" is the model's estimate of the change in real income for a representative household in a given country — how much better or worse off people there are, once the new prices and wages the model computes are taken into account, compared with the pre-shock baseline. It is reported as a percentage change: a value of +1.3, for instance, means real income is estimated to rise by 1.3%.

When a single global figure is shown, it is not a simple average across countries — it is an expenditure-weighted average, where each country's welfare change is weighted by its share of world spending at baseline. A change in a large economy therefore moves that global figure more than the same percentage change in a small one.

What a non-tariff-barrier change does

Alongside tariffs, this site lets you model non-tariff barriers — regulatory, logistical, or other frictions that make cross-border trade in a sector more costly without collecting revenue for anyone. The model represents a non-tariff-barrier change as an iceberg-type trade cost: like a block of ice that partly melts in transit, a fraction of what is shipped is effectively lost along the way, so more has to be produced and dispatched to deliver the same amount to the destination market. That is different from a tariff, which the model treats as a border charge rather than a resource cost. Raising a non-tariff barrier by, say, 25% is therefore not economically equivalent to raising a tariff by 25%, even though both are entered here as a single percentage change, because they act on the economy through different channels.

The data behind the numbers

The model is calibrated to a snapshot of global production and trade for 2023, based on the GTAP 12 database, aggregated to a workable size: 163 countries and territories, and 7 broad sectors.

• Agriculture
• Mining & fuels
• Manufacturing
• Utilities & construction
• Trade, transport & hospitality
• Business & financial services
• Public & other services

Every result you see — a country's welfare change, its sector-by-sector output moves, its exports and imports — is computed against this single calibrated baseline.

What this model assumes — and doesn't

Like any model, this one is a deliberate simplification, and it is most useful for what it is built for: comparing two snapshots of the world economy, one with a scenario applied and one without. In the standard specification used here:

• Full employment. Everyone who wants to work is working, in every country, before and after the scenario; the model reallocates workers across sectors within a country, but no one becomes unemployed as a direct result of a scenario.
• Trade balances held fixed. Each country's overall trade balance — exports minus imports, in total — is held at its baseline level; a scenario changes what a country trades and with whom, not whether it runs an overall surplus or deficit.
• Comparative statics, not a forecast. The model compares two equilibria — before and after — rather than tracing out a year-by-year adjustment path, and it has no notion of calendar time, so it says nothing about how long an adjustment might take.
• No investment, cross-border capital reallocation, or exchange-rate and monetary channels. Firms don't relocate production abroad in response to a scenario, and there is no exchange-rate or monetary-policy mechanism — every price change here is a relative price within the model's own equilibrium, not a currency movement.

These are standard, deliberate features of this class of model — following Caliendo & Parro (2015) — not limitations specific to this site. They matter for how to read the numbers: the results speak to the medium-run reallocation trade brings, not to unemployment spikes, capital flight, or currency moves that a real shock of this kind might also cause in the near term.

Sectoral trade elasticities

Numerical precision

Because there is no closed-form solution for a model like this, the equilibrium is found numerically: the model repeatedly updates its guess at wages and prices until they stop changing by more than a very small amount — its convergence tolerance, set to 1×10^-6 here, a difference of one part in a million between successive guesses — using a damped search so it settles down instead of overshooting on demanding scenarios. A scenario is only reported as solved once it clears that bar; short of it, the site reports that a scenario did not reach a stable solution rather than showing a number it doesn't trust.

How to cite this

@article{caliendo2015estimates,
  author  = {Caliendo, Lorenzo and Parro, Fernando},
  title   = {Estimates of the Trade and Welfare Effects of {NAFTA}},
  journal = {The Review of Economic Studies},
  year    = {2015},
  volume  = {82},
  number  = {1},
  pages   = {1--44}
}

@misc{kite_model_placeholder,
  author       = {{KITE Project}},
  title        = {{KITE}: a general-equilibrium trade model implementation},
  year         = {2024},
  howpublished = {\url{https://kite-model.org}},
  note         = {Placeholder citation --- replace with the definitive KITE reference}
}