How each number is actually produced.

Ownership sectors

Industry employment below the all-industries total is published only by ownership sector — federal, state, local and private — with no combined row. Selecting All ownerships sums the four, then recomputes average weekly wage and the location quotient from the summed totals, rather than averaging the per-sector figures. Averaging averages of differently sized populations produces a number that is not the wage of anything.

Location quotient

Local employment share in an industry or occupation, divided by the national share for the same industry or occupation. Above 1.2 reads as genuine concentration; near 1.0 means the region looks like the country; below 0.8 means the activity is under-represented. It is a share ratio, not a growth measure — a region can be highly concentrated in a shrinking industry, which is usually the finding that matters.

Shift-share

Five-year regional employment change decomposed into three parts: national growth (what the region would have gained if it grew at the national rate), industry mix (the effect of being weighted toward faster- or slower-growing industries), and a competitive residual (what is left, attributable to local conditions). The residual is the part worth acting on; the first two are context.

Estimated annual openings

Annual openings — the sum of growth and replacement demand — are officially published nationally only. To produce a local figure, the national openings rate for an occupation (openings ÷ base employment) is applied to that occupation's local employment. The basis is stated wherever the number appears.

This assumes local separation and replacement behave like the national average for that occupation. It is the standard practice, and it is an assumption, not a measurement.

Regional occupational outlook

Rather than extrapolating a noisy small-area survey series, the state's official projected growth rate for an occupation is applied to the area's current employment base. This is the standard regionalization practice used by state workforce agencies, and it keeps the projection anchored to an official figure instead of to a trend line fitted through survey noise.

Occupational similarity

Cosine similarity over occupational importance ratings — skills, knowledge, abilities and work activities — that are z-scored element by element across all occupations first. This standardization is not cosmetic. Raw importance ratings sit in a narrow band, so an unstandardized cosine puts nearly every pair of occupations above 0.95 and discriminates nothing. Standardizing makes each profile encode what is distinctive about an occupation rather than what is true of all work.

Five lenses are available — skills, knowledge, abilities, work activities, and a composite — plus element-level gap detail for the closest matches, which is what makes the result usable for curriculum design and recruiting rather than just interesting.

Modeled forecasts

Damped log-linear OLS with an 80% interval derived from the residual spread. Used only where no official projection exists: county and metro industry employment, regional labor force, and program completions.

The damping matters. Undamped extrapolation of a fast-growing ten-year window produces absurd ten-year values — a metro that grew 9% a year for a decade does not grow 137% over the next one. Growth is damped geometrically. These figures are labeled as modeled everywhere they appear and are never mixed into official projection tables.

Staffing-pattern estimates

National industry × occupation staffing shares, applied to local industry employment. This fills in occupations that are suppressed at small geographies. It assumes the local industry hires the same occupational mix as the national industry, which is a reasonable first approximation and an approximation nonetheless.

Pipeline balance — the two corrections

Comparing completions to openings is the single most requested figure in program review, and the single easiest one to get badly wrong. Two corrections make it usable.

First, program-to-occupation mapping is many-to-many. One occupation is fed by several program families. Charging its full openings to each of them would inflate total demand roughly four-fold. Each occupation's openings are therefore split evenly across the program families that lead to it. The per-occupation column still shows that occupation's whole regional openings; only the aggregate is allocated.

Second, demand defaults to occupations whose typical entry route is a postsecondary award or degree. Without that filter, every construction laborer opening is charged against construction-trades completions — which is simply not how anyone enters that job. The Supply Pipeline module exposes the unfiltered basis as an option, because for some questions you do want it.

Validation

Twenty-six automated checks compare figures the platform serves against independent published totals. Where a figure has no external counterpart — nobody publishes "the sum of the ownership sectors", for example — it is checked against the control totals that are published alongside it.

Some checks legitimately do not assert exactness, and say why. Small occupational cells are suppressed, so staffing shares can fall short of 100% — the assertion is that they never exceed it. Some occupations are published only at a broader level. And some rates are derived from multi-week averages, so a single-week identity is close rather than exact.

---

Questions about a specific method, or want the detail behind a figure you saw in a demo? Ask us directly