The Social Security Administration is expected to announce the 2027 cost-of-living adjustment on October 14, 2026, the same day the September CPI-W is released. Current estimates cluster between 3.4% and 3.6% — the Senior Citizens League at 3.6%, AARP at 3.5%, independent analyst Mary Johnson at 3.4% — against a 2026 COLA of 2.8%.
What gets far less attention is how that number is manufactured. The COLA is not a survey of retiree costs or a forecast. It is arithmetic: the average CPI-W reading for July, August, and September 2026, compared with the same three months of 2025. Two of those three months are already published. The CPI-W rose 3.4% over the 12 months ending in July 2026, and August data arrives September 11.
Where the Numbers Actually Come From
The Bureau of Labor Statistics does not observe every price in the index. When a price is missing, it imputes one — using price changes for similar items in the same area, then the same item in a wider region, and finally, if nothing comparable exists, carrying last month's price forward unchanged.
That has always happened. What changed is the scale. BLS staff historically estimated roughly 10% of the cells in the CPI calculation. Apollo Academy's tracking of BLS data put the estimated share above 30% as collection shrank.
The shrinkage is documented in BLS's own notices. Amid a federal hiring freeze, BLS suspended CPI collection entirely in Lincoln, NE, and Provo, UT, in April 2025, and in Buffalo, NY, that June. Roughly 15% of the sample across the remaining 72 areas was suspended as well. Then the 2025 appropriations lapse wiped out the October housing survey completely — every rent and owners' equivalent rent observation for the month.
How Much Does an Estimate Move the Answer?
BLS researchers Mark Bowman and Craig Brown quantified it. In a May 12, 2026 Monthly Labor Review article, they rebuilt the October 2025 shelter index four different ways and compared each to the carry-forward method BLS actually used.
The spread was not trivial. Backcast and interpolation approaches produced 0.24%–0.27% differences in 12-month changes; forecast and categorical-average approaches produced 0.45%–0.55% differences. Shelter is the single largest component of the index.
Two caveats matter for retirees. First, the authors found all approaches reconverged by April 2026, when the housing panel was collected again — the distortion was real but self-correcting. Second, October 2025 falls outside the COLA's third-quarter windows entirely. The lesson is not that your COLA is wrong. It is that the gap between competing reasonable methods can reach half a percentage point in a single component, which is the same order of magnitude as the difference between a 3.4% and a 3.6% COLA.
What to Do With This
- Do not budget to the decimal. The spread across current forecasts is 0.2 points — roughly $4 a month on a $2,000 benefit. Plan on a range.
- Wait for the net number. The 2027 Medicare Part B premium is announced separately and is deducted before your check arrives. The gross COLA is not the raise you receive.
- Remember COLAs compound permanently. Each adjustment resets the base for every future year, which is why a small measurement gap matters more at 85 than at 67.
- Own an inflation hedge the CPI does not define. TIPS, I Bonds, and hard assets respond to price levels directly rather than to a survey's sampling budget.
Sources: BLS CPI imputation methodology and 2025 collection-reduction notices; BLS Monthly Labor Review, Bowman and Brown, May 12, 2026; BLS Consumer Price Index Summary, July 2026; The Motley Fool, September 2, 2026; Apollo Academy.