What changed in the reviewed data?

Two acquisitions, four kinds of change

This comparison matches original snapshot 2026-09-30 with reviewed snapshot 2026-09-30-d6e32ee5. It finds 527 changed numeric observations or published-growth values across matching source identities, and 15 changes to qualifiers with the value unchanged. Values are compared at stored source precision.

13,130 matching-series periods extend earlier historical coverage, while 7 extend beyond the previous coverage endpoint. Earlier history added today is not newly occurring economic activity. Eight newly added concepts cover retail, production, utilization, claims and construction. Changes to ECI and the unemployment variant are source corrections; they are excluded from ordinary value revisions.

National values that changed for the same period

The complete numeric-change table includes default and adjustment variants plus GDP’s published growth companion. The old column comes from the preserved earlier file; the reviewed column comes from the new one. A numerical difference is a candidate source revision or source precision change, not automatically a statistically significant economic change.

Corrections are labeled separately

The original ECI page described civilian compensation while using private-industry code CIS2010000000000I. The reviewed source is civilian code CIS1010000000000I. Those populations differ, so changing from one to the other is not an ordinary revision of the same measure. The legacy view identifies the problem and retains the original data.

The original unadjusted unemployment variant used invalid code LNU14000000 and contained no observations. The corrected variant is LNU04000000, which reports 4.3% for August 2026 in the acquired source. Empty advertised variants now fail validation; the legacy viewer does not offer its empty variant.

The GDP article displayed millions under a billions label. Its corrected table divides both endpoints by 1,000 and has a separate corrected CSV . The cumulative percentages did not change. The original CSV with the known label error remains frozen for audit purposes.

What this evidence can establish

The report matches stable site identifiers, exact producer codes and reference periods. Changes to unit, frequency, geography, price basis or identity are flagged separately. New periods are distinguished from same-period differences; unavailable values remain null. We do not reconstruct an earlier value from a current revised history.

Both acquisitions occurred on September 30. Earlier producer file dates do not make this a backdated record of what was known intraday in April or August. This is a reproducible comparison of acquired source versions, rather than a complete agency vintage archive. Comparisons within each snapshot use its own consistent observation history.

Source files, acquisition timestamps and SHA-256 checksums are retained. Original national snapshot · Reviewed national snapshot · Revision policy · Corrections contact.