Americans paid the tariffs. 2025 is running the test at scale

The 2018-19 tariffs on Chinese goods produced something unusual in economics: an empirical answer that arrived while the argument was still running. If foreign exporters were paying, the prices Americans paid at the border, before tariff, should have fallen as Chinese sellers cut prices to hold their market. They did not fall. Studies led by Amiti, Redding, and Weinstein, working through the actual customs microdata, found pass-through into US import prices of essentially 100%: foreign prices held, the duty was stacked on top, and the burden landed on American importers, firms, and consumers¹. A year of further data, covering the war's escalations, reached the same verdict¹.

That result would be trivia if the policy had stayed at 2019's scale. Instead it became the baseline for the largest American tariff experiment in ninety years. Through 2025 the effective average US tariff rate climbed to 16.8%, the highest since 1935³, as new duties rolled across allies and rivals alike. The incidence question, who actually pays, moved from seminar rooms to grocery bills, and the answer so far rhymes with 2019's.

How economists actually measured it

The 2018-19 finding is worth understanding mechanically, because its logic is what makes the 2025 numbers credible. Customs data records the price of every imported shipment before duty. Tariff incidence is then an observable: if exporters absorb the tariff, pre-duty prices fall by roughly the tariff amount; if Americans absorb it, pre-duty prices hold and the duty-inclusive price rises. Across thousands of products and two years of escalation, pre-duty prices barely moved¹. Follow-on work by Cavallo and co-authors traced the next link: at the border, complete pass-through; on the shelf, a mix of consumer price rises and importer margin compression, with retailers initially eating part of the cost and passing more through over time². Nobody found the foreign seller paying more than a sliver.

The one caveat the literature allows is market power in specific goods: where the US is a dominant buyer and supply is inflexible, some burden can shift abroad, and steel showed traces of it. As a description of the aggregate, though, the profession's verdict is unusually unanimous, spanning authors who disagree about nearly everything else in trade policy¹².

It is worth explaining why the intuition fails, because the belief that exporters pay is not stupid; it is a market-power assumption misapplied. A buyer who dominates a seller's business can force the seller to eat a tax; a buyer who is one customer among many cannot. America is a large market and, for most goods, still a minority of world demand, while Chinese producers of most tariffed goods had alternative buyers, thin margins already set by domestic competition, and, after 2018, a depreciation in the yuan that cushioned their receipts without any firm cutting its dollar price. The conditions under which foreigners pay a tariff exist in scattered niches. The policy was priced as if they were the general case.

The retail layer added its own lag, which confused the public reading for months at a time. Importers who stockpiled ahead of each tariff tranche sold old inventory at old prices; retailers spread increases across product lines to avoid headline items jumping; and some margins genuinely compressed, which is a real cost, just one borne by American firms rather than American shoppers. 'Prices have not exploded, so nobody is paying' misreads all three mechanisms. The bill still arrives, just smoothed out, and partly in forms that never show up on a receipt: thinner margins, skipped raises, delayed investment².

2025: the experiment at scale

The 2025 tariff wave dwarfs its predecessor: baseline duties on most partners, punitive rates on specific countries and sectors, and a cadence of announcements that made the effective rate a moving target the trackers updated monthly. The Budget Lab's November 2025 snapshot puts the short-run price level effect at 1.2%, about $1,700 of annual cost for the average household, and around $900 for households at the bottom of the income distribution, the regressive tilt coming from how much of a poorer household's budget is tariffed goods³. After consumers substitute away from the dearest imports, the effective rate settles near 14.4%³.

The revenue side is real money. The Congressional Budget Office scores the 2025 tariff increases, if they persist, as reducing primary deficits by about $2.5 trillion over eleven years⁴. That is the honest version of the pro-tariff case: not that foreigners pay, but that this is an import tax imposed through executive tariff authority rather than a conventional tax bill, collected at the border, yielding serious revenue. The case against is the same sentence with the emphasis moved: a large, regressive import tax, enacted without ordinary legislation, whose burden the customs data shows landing at home.

Retaliation adds a separate bill, a second policy step with its own cost. When China answered the 2018 round with duties on American farm goods, US soybean exports to China collapsed, Brazil took the market, and Washington spent tens of billions compensating farmers, a fiscal transfer that came out of the same taxpayers the import tariffs were already reaching. The 2025 wave has generated its own retaliation lists and its own compensations. A full accounting of who pays for a trade war therefore reads: domestic consumers through prices, domestic exporters through lost markets, and domestic taxpayers through the compensation paid to the second group. Three different costs, all landing in the same country.

What tariffs did and did not buy

For readers outside America the experiment reads as a price list for their own debates. Every government weighing import protection can now see, in unusually clean data, what the domestic bill looks like per point of tariff and per decile of income.

Paying the tariff is one question; getting something for it is another, and the second question needs the same discipline as the first. The 2018-19 round measurably shifted sourcing, the rerouting of trade through Vietnam is its direct creation, without shrinking the overall deficit or reviving the protected industries at scale. The 2025 round is younger, and its defenders point to reshoring announcements and to leverage in negotiations that produced deals from Hanoi to Brussels. Both can be true: tariffs reroute trade and extract concessions, and Americans finance the rerouting through prices. The error in the public debate was never claiming tariffs do things. It was claiming someone else pays for them.

The corporate adaptation layer also blunts the headline rates. Importers front-ran every announced increase, building inventory mountains that delayed the price impact; supply chains re-invoiced through third countries where rules allowed; and exemption lobbying became a growth industry, so the tariff actually collected per dollar of imports runs well below the statutory schedule. The gap between announced and effective rates is where most of the economy's tariff lawyers now live, and it is why serious trackers publish effective rates rather than reciting statutory ones³.

What to watch

The strategic case needs answering on its own ground, because 'who pays' does not end the argument for anyone whose goal was never consumer welfare. A country willing to tax its own consumption can use that tax to build industries, as Indonesia did when it banned raw nickel exports to force smelters into existence, and the CHIPS-style subsidies stacked alongside the 2025 tariffs read as exactly that trade. The test of the strategic case is not incidence but delivery: factories that exist, capabilities that would not have, at a price the public knowingly accepted. What the incidence literature forbids is only the free-lunch version, the claim that the building is funded by foreigners. Someone always pays, and the customs data says who.

The list of reasons a government might accept that bill knowingly is longer than the public debate suggests. Tariffs raise revenue without new legislation. They can shield industries a state considers strategic while domestic capacity is built, or keep alive capabilities, steelmaking, shipbuilding, chipmaking, that defence planners refuse to price at market value. They create bargaining leverage precisely because they hurt the partner's exporters as well as home consumers. And they can redirect supply chains away from a rival, a security objective rather than an economic one. Each of these can be a defensible aim; none of them is free; and a serious version of the argument states the cost alongside the objective instead of denying it.

Three time series settle the next round of this argument. Pre-duty import prices, the incidence instrument itself: any sustained fall would mean exporters finally sharing the burden, and through late 2025 the tracking shows little of it². The monthly effective rate against the litigation and negotiation calendar, because court rulings and trade deals move it more than announcements do³. And the CBO revenue line against actual collections⁴, because a tax whose yield depends on imports continuing is a tax that shrinks if it works as advertised. The 2018 experiment answered who pays; the 2025 one is answering what the payment buys. One prediction seems safe either way: a tax collected at the border, outside the ordinary budget process, is not a tool governments hand back quickly.