Each instrument of economic statecraft has its own published series and the doctrine meant to rank them has none

Where this is written from should be settled before anything in it is weighed. I am an analyst, not a practitioner. I have not sat in the room where an instrument was chosen, drafted a licence condition, or negotiated terms with a ministry. What I have is the published record: the statistical series named at the foot of this page as they are stored on this site, retrieved on 19 September 2026, and the public documents any reader can pull down without asking anyone's permission. That vantage is narrow and it is checkable, and the second of those is the reason to write from it at all.

The narrowness runs in a particular direction and the direction matters. From here it is possible to see what was spent, what was produced, what was borrowed and at what price, each measured by a named agency on a stated base and published on a schedule. A decision is not visible from here. No statistical series records which instrument was considered and rejected, which department objected, what was promised to whom, or what anyone believed an instrument would achieve. Nothing below attributes a movement in any series to any policy, and where a sentence reads as though it does, it does not.

An instrument is a thing a government can switch on

The instruments of economic statecraft are easy to list, and the list has never been the difficulty. A tariff raises the price of an import at the border. An export control denies a named good to a named buyer. A production subsidy pays part of the cost of building something at home. A development loan puts public money behind a project abroad on terms no commercial lender would offer. An investment screen blocks an acquisition. A stockpile buys a material and holds it. Each has an administering body, a legal authority, a switch, and a date on which the switch was thrown.

A doctrine is a different object, and the difference is the whole subject of this piece. A doctrine is a ranking. It states which end yields when two instruments pull against each other, and it states the ranking in advance, while the conflict is still hypothetical and before anyone has acquired an interest in the answer. A tariff on steel raises the input cost of the plant a subsidy is paying to build. A control that denies a component abroad removes a customer from a domestic supplier the same programme is trying to scale. Those are not administrative failures. They are the ordinary collisions of instruments acting on the same quantities, and only a ranking resolves them. A list, however long and however well funded, cannot.

The outlay is measured long before the result

Take the outlay first. Construction spending on manufacturing plant is published by the Census Bureau at a seasonally adjusted annual rate, so each annual point is a rate of spending per year rather than a month's worth of it. It ran at 75 billion dollars a year in 2020, 82 in 2021, 125 in 2022, 202 in 2023, 244 in 2024 and 215 in 2025. The 2024 figure is more than three times the 2020 one, and the final step is downward by 29 billion a year. Research and development inside GDP, on the same annual-rate basis from the Bureau of Economic Analysis, ran at 754 billion dollars a year in 2020 and 1,081 in 2025, higher by 327. That second series counts research wherever it is carried out, in firms, universities and government alike; it is not a budget line and no part of it is claimed here for any programme.

Now the other end of the same transaction. Manufacturing output is an index set to 100 in 2017. It read 93 in 2020, 97.6 in 2021, 98 in 2022, 97.3 in 2023, 96.4 in 2024 and 97.3 in 2025. The entire six-year range is 5 index points. The 2025 reading sits 2.7 points below the 2017 base and 0.7 below the window's own high of 98, and it is identical to the 2023 reading. Set that beside an outlay running in 2025 at nearly three times its 2020 rate. The honest reading is not that the money failed: a plant takes years to build and longer to qualify, an output index registers nothing until product leaves the door, and the two series are on different clocks by construction. The honest reading is that the outlay is legible now and the result is not, and a doctrine judged today will be judged on the legible half.

One line did move

There is an exception and it deserves stating as forcefully as the rest. U.S. semiconductor production, on an index also set to 100 in 2017, read 121 in 2021, 131 in 2022, 135 in 2023, 144 in 2024 and 165 in 2025. That is 44 points across four steps, 21 of them in the last step alone, leaving the index 65 points above its base year. No other production series cited here moves like it. The reason it is visible, though, is not that the instrument aimed at it was better designed. It is visible because semiconductors happen to carry their own index inside the industrial production system, and most of what statecraft is aimed at does not. Resilience has no index. Alliance depth has no index. A supply chain that would hold under blockade has no index.

Two prices the doctrine does not set

Every instrument that spends has to be financed, and the terms moved sharply across these six years. The ten-year Treasury constant maturity rate averaged 0.89 percent in 2020, then 1.45, 2.95, 3.96, 4.21 and 4.29 percent through to 2025. It rose at every observation in the window, 3.4 percentage points in all. The federal surplus or deficit over the same years read minus 3,096 billion dollars in 2020, then minus 2,774, minus 1,374, minus 1,687, minus 1,815 and minus 1,775. The narrowest of the six is 2022, and each of the three years since is wider than it, by 401 billion in 2025. A commitment weighed when the long rate was under one percent is carried and refinanced in a world where it is above four. No instrument chose that, and it changes what every instrument is worth.

The exchange rate is the second such price. The broad dollar index, set to 100 in January 2006, averaged 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. The highest of the six readings is 2024. From the 2021 trough the index stands 9.6 points higher, and 4.9 points above where the window opened. An index higher than it was means a domestically made good is dearer to a foreign buyer and a foreign good cheaper to a domestic one, which is arithmetic about prices and not a claim about what followed from it. Net exports of goods and services, at an annual rate, ran at minus 619 billion dollars a year in 2020 and minus 926 in 2025, wider by 307, with the widest of the six being 2022 at minus 938. Those two series are set side by side and neither is differenced against the other. Exports are about 11 percent of output on the latest single reading. An instrument acting on that share acts on a minority of the economy, while the exchange rate that reprices every unit of that trade is set by no instrument on the list.

What a doctrine would have to be able to state

Three things, and not one of them is a number. A ranking, first: when the tariff and the subsidy pull against each other, which yields, decided in advance and written down. Then a date, naming the year by which the capability the instrument was bought to create is meant to exist, fixed when the money is committed rather than reconstructed afterwards from whichever series happened to move. Then a measure, chosen before the result is known and published whether it flatters the programme or not. A doctrine that can state those three can be shown to be wrong, which is the point of having one. A doctrine that cannot state them can only be defended, and it will be defended with the outlay series, because the outlay series is the one that has already moved. Whether such a ranking exists is not visible from where I sit, and that is the honest limit of this vantage. It is also the argument for publishing one: a ranking on paper is a ranking an analyst outside the building can check, and no instrument would then have to be judged on the only line that happens to be legible.