The import bill is an input cost, not a scoreboard

I want to state the position before the numbers, because the numbers are usually deployed the other way round. The $4.25 trillion of goods and services the United States imported in 2025 is not a debt, a loss, or a transfer of national wealth abroad. It is a bill for things Americans bought. A large part of that bill was run up by American firms buying things in order to make something else. A levy placed on that line is collected, first, from them.

The Bureau of Economic Analysis puts imports of goods and services at $3.42 trillion in 2021, $3.97 trillion in 2022, $3.86 trillion in 2023, $4.11 trillion in 2024 and $4.25 trillion in 2025. Exports on the matching series ran $2.57 trillion in 2021 and $3.32 trillion in 2025. Both lines grew across the window. Only one of them is routinely offered as an achievement, and the difference in treatment is not a finding about the economy. It is a habit of reading.

What the line actually records

The import total is a quantity bought, priced in current dollars. That second clause disposes of most of what the series is asked to prove. The one decline in the run, from $3.97 trillion in 2022 to $3.86 trillion in 2023, is consistent with fewer goods arriving and equally consistent with the same goods arriving cheaper, and nothing in an annual current-dollar total separates the two. I mention it because that dip is the observation most often recruited into arguments about production coming home, and it is the weakest number in the series to recruit. My argument does not need it. The argument is about what kind of quantity the total is, not about which way it last moved.

The point sharpens at the level where the decisions are actually taken. A machine-tool distributor, a chemical blender, a fabricator buying rolled steel — each of them sits inside that total, and each of them sits upstream of an American payroll. When a levy lands on the aggregate import line it lands on them before it lands on any foreign seller, and the sequence matters. The cost enters a domestic production run first. Whether it ever reaches the exporter depends on who is able to refuse to pay, which is an empirical question about particular markets and not a slogan about the national accounts.

The strongest case for the other reading

The serious argument on the other side is not about the total at all. It is about dependency. If a single supplier abroad controls a component that a defence programme, a power grid or a hospital system cannot function without, the price of that component is beside the point. Availability is the point. A country that has allowed such a position to form has taken on a risk that no efficiency gain compensates for, and it is entirely rational to pay more, at home, to unwind it.

That argument is correct and I would not want it softened. It is also, precisely, an argument about which imports rather than about how many dollars of them. It selects a handful of lines out of $4.25 trillion and says: these are different in kind. Every part of it is specific — the identification of the chokepoint, the instrument chosen, the stockpile or the qualified second source or the control on a particular technology. None of it follows from the size of the aggregate, and none of it becomes easier once the aggregate has been taxed.

Which is why a broad levy is a poor expression of a narrow concern. Set on the total, it taxes the fabricator and the chokepoint at the same rate, and the fabricator is no better able to switch supplier than the defence programme was. The policy that matches the security argument is narrow, named and unglamorous: find the dependency, price the insurance, buy it, and say what it cost. The policy that matches the scoreboard reading is broad, and its first incidence is domestic.

What this data cannot settle

I should be explicit about the limits of what is in front of me. The series behind this piece is an annual national total covering goods and services together. It does not separate commodities, it does not name partner countries, and it does not distinguish a finished consumer good from an industrial input. It carries no tariff rates and no estimate of what any tariff did. So I cannot tell you from this data how much of the $4.25 trillion is strategically exposed, and neither can anyone else working from a number this coarse.

What the data does settle is the direction the reasoning has to run. The total is a quantity bought at the prices of the year it was bought in. Any argument that moves from the size of that total to a verdict on national strength has skipped the step where you say what was inside it — and that step is the entire security case, not a footnote to it. The people making the security case in earnest already know this, which is why their papers name components and their opponents name totals.

So the test for the next announcement is not rhetorical and it is not hard. Ask which imports the measure addresses, and ask why those. If the answer names a component, a supplier and a risk, it is a security argument and deserves to be heard as one, including the part where somebody pays for the insurance. If the answer is the size of the import bill, what is being proposed is a tax on American production, and it should be defended as a tax, with the firms that will remit it named out loud.