A tariff is borne at home, and the published data closes two of the three answers that say otherwise

The sentence that a tariff is paid by the country it is aimed at has lasted a long time on the strength of sounding like a description of a payment. It is not one. A tariff is a tax assessed on goods as they enter, and the party assessed is the importer of record, which in the ordinary case is a firm inside the border. My position is that the burden stays close to where the payment starts, and that the published data, which cannot measure a burden directly, closes two of the three answers usually given to the contrary.

Where the payment is made has never been the dispute. Who ends up poorer is. Three answers are usually given. The exporter cuts the dollar price he charges and absorbs the duty in his own margin. The exchange rate moves and absorbs it for him. Or none of this need matter, because the purchase comes home and the country receives a factory in exchange for the tax. The first of those cannot be tested with anything held on this site, and I return to it at the end. The second and third can be tested, and both fail.

The buying moved sideways

U.S. goods imports from China, stated as a monthly average, read 36 billion dollars in 2020, 42 in 2021, 45 in 2022, 36 in 2023, 37 in 2024 and 26 in 2025. From the 2022 peak to the last reading the series lost 19 billion dollars a month, which is 42.2 per cent of where it stood. Taken alone, that is the picture a tariff is meant to produce.

Set it beside the total. Imports of goods and services were 3.42 trillion dollars in 2021, 3.97 in 2022, 3.86 in 2023, 4.11 in 2024 and 4.25 in 2025. Across the same three years in which the bilateral series lost more than two fifths of itself, the total gained 0.28 trillion dollars, or 7.1 per cent. In 2025 alone the China series fell 11 billion dollars a month while the total rose 0.14 trillion. The two are stated on different bases, one a monthly average and one a yearly figure, and nothing here divides, adds or ratios them; what is compared is the direction each moved.

A tariff that redirects a purchase to a third supplier has changed the address on the invoice. It has not, by that act, moved the tax onto anyone abroad, and it has not reduced what the country buys from abroad. A displaced supplier loses a sale, which is a different thing from bearing a tax. The total is the line that would have to fall for the purchase to have come home rather than moved. It rose in three of the four steps the series records, and 2025 is its highest reading.

The production that did not arrive

Had the purchases come home rather than moved sideways, two lines would carry it. Manufacturing output, indexed with 2017 at 100, reads 93 in 2020, 97.6 in 2021, 98 in 2022, 97.3 in 2023, 96.4 in 2024 and 97.3 in 2025. Every reading is below the base year, and 2025 sits 0.7 index points under 2022. Manufacturing employment, in millions of workers, reads 12.1, 12.3, 12.8, 12.9, 12.8 and 12.6 across the same six years. Its peak is 2023 and the last reading is 0.3 million below it.

Neither line proves anything about a tariff, and I will not claim otherwise. They are aggregates moved by demand, energy costs, the exchange rate and a dozen other things at once, and no series on this site separates one cause from another. What they do is deny a particular claim its evidence. A policy defended on the ground that it brings production back has to be able to point at production. An output index that has not regained a base set in 2017, and a workforce count lower than it was two years ago, are not that.

The exchange rate did not do the work

The second answer requires nobody to decide anything. If the dollar buys more of the exporter's currency, he can hold his home-currency price while the American buyer pays fewer dollars, and the duty is met out of the difference. That needs a currency move of roughly the size of the duty. The broad dollar index, with January 2006 set to 100, reads 117.8 in 2020, 113.1 in 2021, 120.7 in 2022, 120.3 in 2023, 122.9 in 2024 and 122.7 in 2025. Between 2022 and 2025 it gained 2.0 index points, which is 1.7 per cent of its 2022 level. Between the last two readings it fell 0.2.

Two index points over three years will not pay a duty of any size worth legislating for. The measure has limits and they are worth naming. It is a basket taken against a group of partner currencies rather than any single one, each point is an average of that year's observations rather than a rate on a date, and no bilateral rate for the currency most often named in this argument is held here. Those limits make the figure less precise than a reader might want. They do not turn 1.7 per cent into a number large enough to carry a tax.

What the data cannot settle

That leaves the first answer, which I set aside earlier. Whether a foreign exporter cut his dollar price cannot be established from anything on this site. There is no series here for duties collected, no tariff rate, no import price index and no bilateral exchange rate, and nothing that would split a consumer price index into tariff and non-tariff parts. So the size of any pass-through is not computable here, in either direction. That is the honest limit and it is a wide one.

One further thing the bilateral series cannot see. Goods that leave China, are finished or relabelled elsewhere and then enter the United States are counted against the country they last shipped from. Some unknown share of the 19 billion dollars a month is that, and this data holds no way of sizing it. The possibility does not rescue the claim under attack. It makes it worse, because a purchase re-routed is a purchase still made abroad, now carrying whatever the re-routing costs on top.

The strongest reply

The best answer to all of this is that the altitude is wrong. Annual averages of national aggregates cannot see a single tariff line, and a firm that moved a production line because of one would be invisible here even if the policy worked exactly as its authors intended. I accept that without qualification. It is a real limitation and it applies to every figure above.

Accepting it fixes the size of the disagreement rather than ending it. The claim I am against is the national one, made in the national register: that the tax is paid abroad, and that the country is made whole by production coming back. That claim is pitched at the altitude of exactly these series. At that altitude the import total did not fall, the output index did not rise, the employment count did not rise, and the currency did not move far enough to pay for anything. A defence that retreats to the individual firm has given up the sentence it started with.

This page carries no chart. Incidence is not a series, and the two lines doing most of the work here are stated on bases that cannot share an axis, so a picture of either alone would make a narrower claim than the one I am making. What I will sign is short. A tariff is collected from a domestic firm, and nothing published here shows the cost leaving that firm's books. Whether some part of it was absorbed abroad by a seller cutting his price is a question this data cannot answer, and anyone who states the answer is using numbers they have not shown you.