CPI-W: the index that sets the COLA

Every Social Security cost-of-living adjustment since 1975 comes out of this one series. It is not the inflation number in the news — that is CPI-U — and the difference in who each one measures is the whole argument about whether the adjustment is adequate.

Series

CWUR0000SA0all items, U.S. city average

Base period

1982-84=100

Monthly readings

630from January 1974

Permanent gaps

1October 2025

Which series, exactly — and how we know

CPI-W measures the spending of urban wage earners and clerical workers: a working-age, wage-earning population. The version used is not seasonally adjusted, and no official page states that in words, which is worth being honest about. The proof is two strands and both are checkable.

1. Numeric identity. SSA published the three third-quarter 2025 readings behind the 2026 adjustment as 316.349, 317.306, 318.139. Those are exactly the not-seasonally-adjusted values. The seasonally adjusted series carries different numbers.
2. BLS escalation guidance. BLS states plainly that using seasonally adjusted data in escalation agreements is inappropriate, because seasonal factors are recalculated annually and adjusted data is revised for up to five years. A revisable index cannot underpin a statutory benefit computation.

The recent series

Only three months of the year actually matter

Eleven of the twelve monthly releases move nothing. The adjustment is decided by July, August and September, averaged — and in practice by September, because the other two are already known when it lands. The September 2026 reading is released 08:30 ET Wednesday 14 October 2026.

The date is soft and the event is not. In 2025 the September release was scheduled for mid-October, an appropriations lapse began on 1 October, and BLS recalled staff to publish it nine days late — SSA announced the adjustment the same morning it landed. September data is always collected before the fiscal year turns, which is exactly why September 2025 survived and October 2025 did not.

The data-quality question, stated rather than hidden

BLS suspended CPI collection entirely in three metropolitan areas during 2025 and reduced the sample in the others. Imputation in the commodities and services survey has run far above its historical norm for more than a year.

There is a useful asymmetry in what BLS has said about the effect. It affirmatively measured the impact on the national all-items series and found it negligible, and it explicitly states that it did not measure the impact on subnational or item-level indexes. Scoping this site to national all-items CPI-W keeps it inside the one zone BLS has vouched for — which is a binding reason never to add metro-level or component indexes here. The measurement changes in detail.

Related

• How the COLA worksfour steps
• The missing monthOctober 2025
• Data qualitywhat changed in 2025
• CPI-W vs CPI-Ethe proposals
• Against inflationdid it keep up
• 2027 projection0 of 3 months