One observation, one lineage
Each indicator has a stable site identifier and an exact producer code. Source, frequency, adjustment, geography, unit and price basis are kept with its observations. An acquisition file has a SHA-256 checksum and a UTC retrieval timestamp. The September 30, 2026 dataset is an acquired snapshot, not a claim of complete agency vintage history.
Public pages, chart controls, historical tables and downloads use the same approved JSON snapshot. Source publication and site acquisition are separate fields. Reviewed BEA files in this snapshot say September 30, 2026; the acquisition date does not replace that date. The H.15 XML contains observation dates but does not supply a publication timestamp for each reading.
Documented arithmetic
Absolute change = ending value − starting value. Percentage change = 100 × (ending value ÷ starting value − 1). Year-over-year growth compares the same reference period one year earlier; weekly claims compare with 52 weeks (364 days) earlier to preserve the weekday. A percentage-point difference subtracts two rates expressed in percent. Basis points equal 100 times that percentage-point difference.
Quarterly annualized growth = 100 × ((quarterly level ÷ previous quarterly level)^4 − 1). The real GDP default uses BEA’s published growth series where available. Index-to-100 = 100 × value ÷ base value. Comparison indexing uses a shared positive observation date for series at the same frequency.
Purchasing-power equivalent = amount × ending price index ÷ starting price index. Real wage growth = (1 + nominal growth) ÷ (1 + price growth) − 1, using decimal growth inputs over the same interval. Calculations reject invalid denominators. Source rounding can make calculated growth differ from a published headline.
Compatibility and missing values
Seasonally adjusted and unadjusted inputs are never combined within one ratio. CPI monthly adjusted values and yearly unadjusted values remain separately selectable variants. Rates, levels and growth values carry different unit labels. Missing source observations are stored as null, not zero. No monthly interpolation of GDP or automatic resampling is applied.
A stock, a flow and a rate need different conversion rules. This release provides no generic frequency converter, chained-dollar summation, suppressed-cell reconstruction or custom region aggregator. Different-frequency comparisons are limited to separate raw-level panels.
Review before publication
Source acquisition writes a candidate. The operator validates schema, duplicate periods, series identity, dates, units, scales and missing-value flags, then reviews new observations and revisions against source releases. Large changes trigger review rather than automatic replacement. Publication requires a successful build and site check; a failed update leaves the previous approved version visible.
Regional ranking and geographic calculations
State profiles use BEA SAGDP1/2, SAINC1 and SARPP plus BLS LAUS and state CES. Annual GDP growth uses consecutive annual real totals; payroll growth uses the same month a year earlier. GDP per capita converts millions of dollars to dollars and divides by the same-year BEA resident population. Regional price equivalence multiplies an amount by the destination/origin parity ratio for one shared year.
The rank universe is the 50 states, with DC separately unranked. Equal values at the stated ranking precision receive competition rank and skip the following position. Missing observations are excluded and counted; alphabetical order within ties does not break the tie. Industry profiles select non-overlapping broad nominal-output sectors without summing chained-dollar components.