A state incentive buys a location decision, not national capacity

A state incentive package is priced against a location decision. That is the transaction: the same project moves across a line on a map, from a county in one state to a county in another, and the winning state pays for the move. The case put to a legislature is almost always larger than that. It is told as national capacity, as an industrial base rebuilt, as a supply chain brought home, and the award is defended on the ground that the country now holds a plant it would otherwise not have had. My position is that the second claim is not what the money buys, and that the national series a state could read before signing do not yet show it.

What the national record shows

Construction spending on manufacturing plant is the series that moved. It reads 75 billion dollars in 2020, 82 in 2021, 125 in 2022, 202 in 2023, 244 in 2024 and 215 in 2025.Each figure is a seasonally adjusted annual rate rather than one month's spending, so 244 is roughly a year's spending. Between 2020 and the 2024 high the annual rate rose by 169 billion dollars. The 2025 reading is 29 billion lower than 2024, the only fall in the six.

Three series that would register a result sit below their own earlier readings. Manufacturing output, on a 2017 base of 100, runs 93 in 2020, 97.6 in 2021, 98 in 2022, 97.3 in 2023, 96.4 in 2024 and 97.3 in 2025, so three years on it has not regained the 98 it touched in 2022. Manufacturing employment runs 12.1 million workers in 2020, rises to 12.9 million in 2023 and falls to 12.6 million in 2025, three hundred thousand below its own peak. Capacity utilisation was 77.6 percent in 2022 and 75.3 percent in 2025, 2.3 percentage points lower across the years when plant spending was at its highest.

The obvious objection is time, and it is a good one. A slab poured in 2023 is not a production line in 2025. Equipment is installed, qualified and staffed over a period the construction series cannot show and annual points cannot resolve, so the spending of 2023, 2024 and 2025 sits largely ahead of its own output. Anyone arguing from these four series that the building produced nothing is arguing past the evidence, and I would not sign that sentence.

Read the objection the other way and it stops being a defence. If the national gain cannot yet be seen, it cannot yet be audited, and a state signing a long-dated award in 2026 is paying now against a quantity nobody publishes. No series in the national file states a level of capacity. The record holds an output index, a utilisation ratio and a headcount, and each of those is a proxy that fails in its own particular way. An award defended by national capacity is defended by a word that names no published number.

The opposite claim cannot be proven here either, and the limit is worth stating plainly. Whether a given plant was created or moved is a question about two states, and the state metrics this brief carries hold one observation and one prior period each, and the collections file holds a single fiscal year. There is no state history in them, so there is no honest way to follow a project from the place that lost it to the place that won it. What the national aggregates do is put a ceiling on what all of them together added to the country: whatever passed between states from 2020 to 2025, the output index finished 4.3 points above where it started and 0.7 below its 2022 reading.

What a state should change

Three changes follow, and none of them asks a state to stop competing. Know whose revenue is being spent. Where a package abates property tax, the collections file shows that the state is largely not the government losing the money: fourteen of the 51 jurisdictions levy no state property tax at all, and among the 37 that do, the median property-tax line is 1.19 percent of that state's total collections. In those places the revenue given up belongs to a county, a municipality or a school district, and the file that records state collections does not record its loss. A state can still decide the trade is worth making. It should not describe the cost as its own.

The second change is to write the award against a number both sides can read. State manufacturing employment is published monthly, in thousands, for 49 of 51 jurisdictions: Indiana's 512 thousand manufacturing jobs in July 2026 stand against 513.3 thousand in June 2026, and neither figure needs anybody's estimate to arrive. Alabama and the District of Columbia carry no value for that series at all. A state in that position should say so in the term sheet and name a measure it does hold, rather than write a performance clause against a quantity it cannot observe.

Then there is the input that does not travel with the project. Production occupations in Indiana employed 353,630 people in 2025 at a median annual wage of 47,360 dollars, with a location quotient of 2.01, a little above twice the national concentration of that kind of work. A plant arriving there is not creating that pool. It is hiring from it. The pool was assembled by other employers, by the schools and by the people who stayed, and no single negotiation transfers it to another state. It is also the part of the transaction that outlives the sponsor: equipment is sold and leases lapse, and the trained workforce is still in the county.

The wage complicates that recommendation and should be said rather than left out. Indiana's median wage across all occupations is 47,860 dollars. The median for production work is 47,360, five hundred dollars lower. In the state where production work is the most concentrated of the 51 jurisdictions, the occupation an incentive is defended by pays slightly less than the state's middle job. A state competing on workforce is competing to supply labour at that price, and the case for doing it has to rest on the count of jobs and on what else the pool attracts, not on the wage.

None of this says a state should refuse to bid. It says the bid should be priced against what the state can see. The record from 2020 to 2025 shows a building boom and an output index that has not moved with it, and several years will pass before it can show anything else. In the meantime a state is choosing between a claim about the country that it cannot verify and a set of numbers about itself that is published every month. I would sign against the second, and leave the first to be argued in Washington.