What transfers between places is the method, not the programme

Where this is written from

This is written from a narrow position, and the narrowness is the point of it. I am an analyst working from published statistics and public documents, and from nothing else. Every figure below is an annual observation in a federal series or a value in a state-level file, retrieved on a stated date and held on this site. What the position gives is comparability. The same measure, taken on the same basis, exists for all fifty states and the District of Columbia, and where an agency did not publish a value the file says so instead of supplying one.

It gives nothing at all about the inside of a decision. No file here records an incentive award, a site-selection criterion, a negotiation, a project cost, or whether a plant was ever finished. I cannot observe why a firm chose one county over another. I cannot observe the counterfactual, which is the thing anyone judging a development policy actually wants. A reader who needs that will have to get it elsewhere. What the published record carries is structure, and structure is the subject here.

The spending has no address

Take the money first. Construction spending on manufacturing plant, in billions of dollars at a seasonally adjusted annual rate, runs 75 in 2020, 82 in 2021, 125 in 2022, 202 in 2023, 244 in 2024 and 215 in 2025. The 2024 peak is 169 billion above the first reading. The most recent point is 29 billion below that peak. Set against all nonresidential construction, on the same basis and from the same agency, which runs 856 in 2020 and 1,309 in 2025, factory plant went from 8.8 percent of that total to 16.4 percent, having touched 18.8 percent in 2024.

Those are national totals and nothing more. The series carries no state, county or project breakdown, and no other file held here supplies one. So the most visible fact about American regional development policy in this window, a step change in the building of factories, cannot be put on a map from the published record at all. Whether it worked in a given place is not a question the spending series can be asked. That leaves only what the places themselves are measured on.

Building is not yet producing

Manufacturing output, indexed to 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. The 2025 reading is 2.7 points below the base and 0.7 below 2022. Manufacturing employment, in millions of workers, peaked at 12.9 in 2023 and stands at 12.6 in 2025, 0.3 million lower than the peak and 0.5 million above 2020. Plant spending at its peak was 169 billion dollars a year above its 2020 level. Neither of the other two series moved with it. That is not evidence the spending failed. A plant financed in 2023 and built in 2024 has no obligation to appear in a 2025 annual average, and an annual average is what the output and the employment series each give. The narrower statement is the defensible one. The national record does not yet contain the result, and a judgement taken from it now is being taken early.

The base is occupational, and it is already measured

Occupational data is where this record distinguishes places by degree rather than by kind. The Occupational Employment and Wage Statistics for 2025 give each of the fifty-one jurisdictions the same twenty-two occupational major groups, and for each group a location quotient. The file defines that plainly in its own header: the occupation's share of that state's jobs divided by its share of national jobs, so above 1 means more concentrated than the nation. For production occupations the range runs from 2.01 in Indiana to 0.13 in the District of Columbia. Indiana holds 353,630 people in production work, 110.466 for every thousand jobs in the state. The District of Columbia holds 4,990, or 7.103 per thousand. The file is a single 2025 cross-section. It carries no history, so it cannot date either concentration or say what produced one, and I will not assert either. What it does fix is the spread: measured the same way in the same year, the two jurisdictions sit at opposite ends of the national scale for the same kind of work. That spread is on the public record before any instrument is written against it, which is the only sense in which it comes first.

Cost sits beside concentration, and the two are not independent. The same file gives an all-occupation annual median wage for each jurisdiction. The highest is the District of Columbia at 91,540 dollars; the highest state is Massachusetts at 63,590; the lowest is Mississippi at 40,120. Between Massachusetts and Mississippi the gap is 23,470 dollars, a ratio of about 1.58 to one, and between the District of Columbia and Mississippi it is 51,420. Across the fifty-one jurisdictions the correlation between the production location quotient and that median wage is about minus 0.59, and minus 0.55 with the District of Columbia set aside: the places most concentrated in production work tend to be the cheaper places to hire in. Mississippi carries a production quotient of 1.53 and Massachusetts 0.7. That is a median across every occupation in a state rather than a manufacturing wage, so it describes the labour market a project enters and not the payroll it will run. It is still the cost level a project meets on arrival, and it is not a level a development authority sets.

An instrument has to fit the tax it credits

Fiscal structure decides which instruments are available at all. The Census Bureau's survey of state government tax collections for fiscal 2025 records, for each of the fifty-one jurisdictions, which of five taxes it levies. Six states record no individual income tax: Alaska, Florida, Nevada, South Dakota, Texas and Wyoming. Four record no corporation net income tax: Nevada, Texas, Washington and Wyoming. Three of those, Nevada, Texas and Wyoming, record neither. Four record no general sales and gross receipts tax: Alaska, Delaware, Montana and New Hampshire. Fourteen record no state property tax, and that last count needs the qualification the source itself supplies, since it covers state government collections only and county, municipal and school district taxes lie outside it, which is where local property taxation and most abatement sit. A credit written against a tax a state does not levy is not a weaker incentive there. It is not an incentive at all, and that is visible long before any question of implementation arises.

What transfers is the method

So the programme is not the portable part. An abatement, a credit, a training subsidy or a site grant is written against a particular base and a particular tax code, and it arrives in the next jurisdiction as a different instrument or as none. The method is portable. A location quotient is a method, and so is the threshold this workforce file applies to its own speciality lists, which admits an occupation only when it employs at least a thousand people in the state, so that a high concentration in a tiny occupation is not mistaken for a capability. So is the treatment of gaps. The state metrics file records exactly two absences, Alabama and the District of Columbia in the state-level manufacturing employment series, and shows them rather than filling them; those two happen to be the jurisdiction with the second-highest concentration of production work and the jurisdiction with the lowest, which is a fair warning about what a complete-looking table can cost a reader. A place able to measure its own base can decide what to accept. A place unable to will take what is offered and learn the answer afterwards. From where I sit I cannot tell you which places did which, because nothing held here records a decision or an outcome. I can say that the conditions which make an instrument work are mostly not policy variables, are visible in advance, and are already published.