Defence capacity is a tooling and workforce problem procurement reform does not reach

Defence industrial capacity is a tooling and workforce problem. Procurement reform is worth doing and it does not reach the constraint. That is my position, and I hold it knowing what this site cannot see. It carries no defence budget, no contract awards and no count of defence-sector workers. What it carries is the civilian industrial base that defence orders from, and over the last four years that base was given the clearest test anyone is likely to get.

Between 2022 and 2025 money arrived in American manufacturing plant at a rate far above anything else in the window. Construction spending on manufacturing plant, stated as a seasonally adjusted annual rate, ran 125 billion dollars in 2022, 202 in 2023, 244 in 2024 and 215 in 2025. The window opens at 75 billion in 2020 and 82 in 2021, so the 2024 figure is more than three times the first reading and falls six billion short of double the 2022 one.

The line that should have moved first

Core capital-goods orders are published as a monthly average in billions of dollars. Across those same four years they ran 77, 75, 74 and 77. The 2025 figure is the 2022 figure. Manufacturing employment ran 12.8 million in 2022, 12.9 in 2023, 12.8 in 2024 and 12.6 in 2025, leaving the count three hundred thousand below its own peak. Capacity utilisation ran 77.6 percent, then 76.6, 75.4 and 75.3, falling at every step and ending 2.3 percentage points below where the four years began. Manufacturing output, on a 2017 base of 100, ran 98, 97.3, 96.4 and 97.3.

The obvious objection is lag. A plant paid for in 2024 makes nothing in 2024, and an employment count or an output index will register it late if at all. That objection holds against output and it does not hold against orders. Equipment is ordered before a building is finished, not years after it opens, and the order is the earliest point at which a capacity programme becomes visible in any published series. Four years in which core capital-goods orders moved only between 74 and 77 is not a lagging series catching up. It is the leading indicator declining to lead.

Two series, two opposite kinds of number

The two central figures sit on different bases and nothing here compares their levels with each other. Factory construction is published as a seasonally adjusted annual rate, so 215 billion dollars is a yearly rate of spending, and reading it as a single month's outlay would put a year of factory construction at roughly twelve times its actual level.Core capital-goods orders are published as a monthly average, so 77 billion dollars is a month's figure. Those two warnings point in opposite directions because the two series are of opposite kinds. Every comparison below and above is a series against its own earlier readings.

The wage is the recruiting problem

The workforce half of the argument sits in a different file. In the 2025 occupational wage estimates, the median annual wage in Production Occupations is below the all-occupation median of its own state or district in 41 of 51 jurisdictions. Ten are at or above it. Concentration does not rescue the result. Indiana holds the highest location quotient in production work of any state at 2.01, with 353,630 people in those occupations, and the production median there is 47,360 dollars against a state median of 47,860, five hundred dollars below.

A sector paying under the local median in 41 of 51 jurisdictions is not losing a bidding war for labour at the margin. It is not in the bidding. No contract vehicle changes that number. Nothing in a multi-year award reaches it, and nothing in a faster award reaches it either, because the person being competed for is looking at a wage table and not at an appropriation.

What procurement reform actually buys

The case on the other side is not empty and I will not pretend it is. A firm multi-year order changes the discount rate a company applies to a machine it will own for twenty years, and uncertainty about next year's appropriation is a real reason not to buy one. Fixing that is worth doing. The argument is about what fixing it delivers. Certainty is an input to a tooling decision. It is not the tool, the queue in front of the tool, or the machinist.

Take those four years as the strongest available test of the alternative view; they run against it. Capital reached the sector at the highest rate in the window. It did not convert into orders for equipment, into workers, or into capacity in use, and two of those three moved the wrong way while it was arriving. Whatever was binding sat downstream of the money.

What would move me is a defence-specific orders series showing machine-tool purchases rising with award volume. This site holds no such series, and I would rather name that than write around it. On what it does hold, the money moved and the tools and the people did not. Until a better series exists, the place to check whether a capacity programme is working is the order book and the wage table, not the award notices.