Official labor statistics say most trades are near equilibrium. Employers on the floor say they can't hire. Both can't be right. The gap is a modeling choice: BLS projections assume supply rises to meet demand, so "openings" understate a structural shortage. The Arklight Demand Model drops that assumption and builds the two sides independently.
What is the Arklight Demand Model?
It is a bottom-up estimate of a skilled-workforce shortage: annual demand, built in three layers, measured against the number of hire-ready workers the pipeline actually produces. Where the official model assumes the market clears, ours asks a narrower question — how much work is there to do, and how many people can actually do it on day one?
How is demand built? The three layers.
| Layer | What it captures | Anchored to |
|---|---|---|
| 1 — Replacement floor | Retirements and attrition from the existing workforce | Workforce age structure (ACS), BLS separations |
| 2 — Organic growth | Baseline expansion of the occupation | BLS Employment Projections / OEWS |
| 3 — Project-driven | Incremental demand from real capital: AI/data-center capex, CHIPS fabs, grid build-out, reshoring, and the defense industrial base | Bottom-up from published capex, project trackers, and hiring commitments |
Layers 1 and 2 are the floor most models stop at. Layer 3 — modeled bottom-up from actual workload rather than assumed away — is where the Arklight gap opens up.
How is supply measured?
Supply is hire-ready entrants, not total credentials issued. We count completions from formal pipelines — NCES IPEDS (by CIP code), DOL RAPIDS apprenticeship completers, and military transitions — and apply a readiness discount, because informal and on-the-job entrants are not production-ready on day one. The result is the number an employer could actually put on a critical-path job this year.
Why do the numbers diverge from BLS?
Because BLS assumes equilibrium by construction and bakes in automation offsets the floor doesn't deliver. When you model demand from the workload and supply from hire-ready output, the gap is routinely several times the official "openings" figure. In the machinist briefing, that divergence is roughly 2.5×. That is the headline finding, not noise — and it is why we publish the method openly.
How is the economic impact calculated?
Two figures accompany each gap. Cost-of-vacancy uses BEA value-added per full-time-equivalent worker by industry (NAICS). The downstream multiplier uses IMPLAN Type-I plus published EPI and NAM multipliers, weighted higher on critical-path AI, fab, and defense builds. Modeled trajectories are labeled in-brief. Treat the range and direction as robust; treat any single point estimate as directional.
A worked example: electricians
Hire-ready supply of credentialed electricians runs about 10,000 per year (IPEDS + RAPIDS, readiness-discounted). Three-layer demand — replacement + organic growth + the AI/data-center/fab/grid build-out modeled bottom-up — lands near 97,000 per year. The gap is ~87,000 unfilled seats a year, at a modeled direct cost of $20–28B and roughly 250,000–385,000 downstream jobs blocked. See the full electrician briefing, or the machinist, fabricator, and nuclear briefings.
The bottom line
The Arklight Demand Model is deliberately conservative on supply and explicit on demand, so the gap it reports is a floor, not a ceiling. When you count the work that has to be done and the people who can actually do it, the shortage isn't a rounding error — it's the binding constraint on American reindustrialization.