The index was updated to current federal data across an occupation-taxonomy boundary. The methodology is unchanged: the same nine O*NET variables, three dimensions, and composite formula, validated by exact reproduction of every prior score before the rebuild.
| Component | Old (April 2026) | New (September 2026) |
|---|---|---|
| O*NET database | 23.1 (November 2018), O*NET-SOC 2010 taxonomy | 31.0 (August 2026), O*NET-SOC 2019 taxonomy |
| Employment weights | OEWS May 2024 | OEWS May 2025 (published May 15, 2026) |
| Occupations scored | 966 | 911 |
| Occupations with unique 6-digit BLS employment | 772 (122.6M workers) | 696 (119.5M workers) |
| Academic programs scored | 1,786 | 1,806 |
| Variables, dimensions, formula, tier thresholds | Unchanged | |
The occupation counts changed because the taxonomy changed, not because coverage was cut: the 2019 taxonomy merged, split, and renumbered occupations (1,110 codes became 1,016), and O*NET 31.0 carries all nine index variables for 911 of them. Programs grew from 1,786 to 1,806 because the CIP-SOC crosswalk matches more programs under the 2019-aligned codes.
Under the index's published update policy, a routine data refresh recomputes scores on newer vintages of the same variables. This update is more than a refresh in one respect: moving from O*NET 23.1 to 31.0 crosses the O*NET-SOC 2010 to 2019 taxonomy boundary, where occupations were merged (the two 2010 software developer codes became one Software Developers code), split, and renumbered, so April and September scores describe differently defined occupations. The update is therefore a re-baseline with a documented bridge, and the old version remains archived in full. The methodology version is unchanged at v2 because the construction itself did not change.
The bridge maps every old occupation to its September counterpart through O*NET's official 2010-to-2019 crosswalk: 1,164 links covering 44 occupations that split and 97 that merged. Across the 971 links where old and new both carry a score, the correlation is 0.943 and the mean absolute score change is 0.031, which is the expected scale of movement for eight years of newer survey data plus a taxonomy change. The largest movers are documented below, and the full bridge file is provided on request through the form at the end of this page.
| Occupation (2019 title) | April | September | Change |
|---|---|---|---|
| Writers and Authors | 0.606 | 0.806 | +0.200 |
| Medical Transcriptionists | 0.691 | 0.498 | −0.193 |
| Forest Fire Inspectors and Prevention Specialists | 0.555 | 0.363 | −0.192 |
| Cutters and Trimmers, Hand | 0.305 | 0.487 | +0.182 |
| Music Directors and Composers | 0.672 | 0.500 | −0.172 |
| Billing and Posting Clerks | 0.567 | 0.730 | +0.163 |
Movement of this kind reflects updated O*NET survey responses. Writers and Authors is the clearest example: incumbent ratings of computer interaction and information processing rose sharply in the post-2023 survey waves, and the September score moves the occupation from the AI-exposed band to the highly exposed band.
Before any new score was computed, the April 2026 pipeline was reproduced from archived O*NET 23.1 source data: all 966 occupation scores matched the published values exactly, on every dimension and the composite. The September build then applied the identical construction to O*NET 31.0. The build is fully scripted, and every input file is recorded with a checksum and source URL, so any published figure can be traced to the exact upstream release it was computed from.
The April 2026 pages and data are frozen at stable addresses so previously cited figures remain resolvable: the program index, the occupation index, the methodology, and the occupation-level dataset (CSV). The current version is always at the unversioned addresses, beginning with the Academic Program index.
On August 27, 2026, BLS published AI Exposure Categories alongside the 2025-35 employment projections: four ordinal exposure levels for 831 occupations. The BLS categories are not an input to this index, which preserves an independent measurement; a concordance analysis comparing the two is planned as a companion release.
The complete April-to-September bridge (1,164 crosswalk rows with both scores) and prior-vintage source extracts are provided on request. Requests are answered from the Opportunity Data inbox, typically within two business days.