How Social Security COLA Is Calculated: The CPI-W Formula

TL;DR. Every October, the Social Security Administration announces the following year’s cost-of-living adjustment (COLA) by comparing the average of one specific inflation index — the Consumer Price Index for Urban Wage Earners and Clerical Workers (CPI-W) — in July, August and September of the current year to the same quarter in the last year that had a COLA. The percentage increase, rounded to the nearest tenth, is the raise. If CPI-W didn’t rise, benefits stay flat — they never go down.

The 2026 COLA of 2.8% was calculated exactly this way. In this explainer, we’ll walk through the statutory formula, the Q3 rule, the “computation quarter” concept, the safety net that stops benefits from ever falling, and how to project the next COLA before SSA publishes it.

The formula, in one line

Section 215(i) of the Social Security Act defines the COLA as:

COLAnext year = (Average CPI-W in Q3 of current year ÷ Average CPI-W in Q3 of the last COLA year) − 1

Two things do the heavy lifting in that formula:

  • The index is CPI-W, not CPI-U or PCE. When Congress amended the Act in 1972 to make COLAs automatic starting in 1975, CPI-W was the only CPI series that existed. BLS has since developed the broader CPI-U (and the Fed prefers the PCE price index), but SSA still uses CPI-W by statutory precedent.
  • “The last COLA year” is called the computation quarter. Usually that’s just last year’s Q3, but if CPI-W fell (a deflation year like 2009 or 2015), SSA skips the COLA and the benchmark stays at the last quarter that actually saw an increase. That’s why the 2010, 2011 and 2016 COLAs were 0%.

A worked example: how the 2026 COLA was calculated

SSA published the mechanics in the November 3, 2025 Federal Register notice. The exact CPI-W index values SSA used:

Quarter July August September Arithmetic mean
Q3 2024 (last computation quarter) 308.501 308.640 309.046 308.729
Q3 2025 (current) 316.349 317.306 318.139 317.265
Source: SSA, “Cost-of-Living Increase and Other Determinations for 2026,” 90 FR 49047, Nov 3, 2025. CPI-W series CWUR0000SA0.

Plug it in: 317.265 ÷ 308.729 − 1 = 2.7647%. Round to the nearest 0.1 percentage point and you get 2.8%. That’s what monthly Social Security checks went up by starting with the December 2025 benefit (paid in January 2026).

How the 2026 Social Security COLA was calculated from CPI-WBar chart showing CPI-W index values for July, August and September of 2024 and 2025. The Q3 2024 average is 308.729; the Q3 2025 average is 317.265; the increase is 2.8 percent, which is the 2026 Social Security COLA.Worked example: how the 2026 COLA was computed305310315320308.501Jul ’24308.640Aug ’24309.046Sep ’24316.349Jul ’25317.306Aug ’25318.139Sep ’25avg 308.729avg 317.265Q3 2024 (last computation quarter)Q3 2025 (current)2.8% COLA = (317.265 / 308.729 − 1) × 100, rounded to 0.1 → 2.8%

Why CPI-W (and not CPI-U or PCE)?

CPI-W and CPI-U are built from the same underlying price data but cover different households:

  • CPI-U tracks all urban consumers — about 93% of the US population — including retirees, professionals, and the self-employed.
  • CPI-W is a subset of about 30% of the population: households where at least one earner is in a wage or clerical job. That makes it a working-family basket. Retirees are, ironically, not in the CPI-W sample even though the index sets their benefits.

Both indexes use the same category structure. The Bureau of Labor Statistics publishes relative-importance weights for each. Shelter dominates:

CPI-U category (Dec 2024) Weight
Shelter 35.5%
Transportation 16.6%
Food and beverages 14.5%
Medical care 8.3%
Energy 6.2%
Education and communication 5.7%
Recreation 5.3%
Other goods and services 2.9%
Apparel 2.5%
Source: BLS, “Relative importance of components in the Consumer Price Indexes,” December 2024. CPI-W weights are similar; the two indexes diverge mainly in medical care and shelter shares because retiree spending differs from wage-earner spending.

Because CPI-W under-weights medical care relative to what retirees actually spend, advocacy groups have long argued that the SSA should switch to a “CPI-E” (elderly) index. Congress has considered but never adopted the change. For now, the law is the law: CPI-W it is.

Two things that can override the CPI-W number

1. The no-decrease rule. If CPI-W falls year over year, there is no COLA — benefits stay flat, they do not shrink. And critically, the benchmark also stays put: the next COLA is measured against the last quarter where CPI-W actually rose. That’s why 2016 recorded a 0.0% COLA even though CPI-W had ticked up slightly from 2014 — 2015’s Q3 was still below the 2014 Q3 benchmark left behind by the 2015 zero-COLA year.

2. The OASDI fund-ratio brake. Section 215(i) contains a rarely used provision: if the combined OASI and DI trust-fund reserves at the start of the year are less than 20% of that year’s expected outgo, the COLA is capped at the growth in the national average wage index instead. This has never triggered — the 2025 fund ratio is 169.0%, per SSA’s own 2026 determination — but it is the statutory reason trust-fund exhaustion projections matter for benefit growth, not just benefit continuity.

Twenty years of COLAs

Benefit year COLA Announced
2026 2.8% Nov 2025
2025 2.5% Oct 2024
2024 3.2% Oct 2023
2023 8.7% Oct 2022
2022 5.9% Oct 2021
2021 1.3% Oct 2020
2020 1.6% Oct 2019
2019 2.8% Oct 2018
2018 2.0% Oct 2017
2017 0.3% Oct 2016
2016 0.0% Oct 2015
2015 1.7% Oct 2014
2014 1.5% Oct 2013
2013 1.7% Oct 2012
2012 3.6% Oct 2011
2011 0.0% Oct 2010
2010 0.0% Oct 2009
2009 5.8% Oct 2008
2008 2.3% Oct 2007
2007 3.3% Oct 2006
Source: Social Security Administration annual “Cost-of-Living Increase and Other Determinations” notices, published in the Federal Register (docs 2005-21444 through 2025-19763). Three years (2010, 2011, 2016) recorded 0.0% because CPI-W did not exceed the prior computation quarter.

Annual Social Security COLA, 2007 through 2026Bar chart of the Social Security cost-of-living adjustment applied each calendar year from 2007 to 2026. Three years (2010, 2011, 2016) show 0% because CPI-W did not rise above the prior benchmark. The 2023 COLA of 8.7% is the highest in the window.0%2%4%6%8%Annual Social Security COLA, 2007–2026 (percent)3.320072.320085.8200902010020113.620121.720131.520141.72015020160.320172.020182.820191.620201.320215.920228.720233.220242.520252.82026Year benefit increase applies (announced October of prior year; paid starting January)

Two things jump out. First, big COLAs cluster around inflation shocks: 5.8% for 2009 after the 2008 oil spike, 5.9% for 2022 and 8.7% for 2023 after the post-pandemic surge. Second, the 20-year median is closer to 2.5%, so 2023’s 8.7% is an outlier, not a new normal.

The COLA calendar every year

Once you know the mechanics, the annual timing is predictable:

  • Mid-October: BLS releases the September CPI report. This is the final data point SSA needs. Section 215(i)(2)(D) requires SSA to publish the new COLA within 45 days of the end of Q3 — in practice, within hours of the September CPI release. Watch the BLS CPI release schedule.
  • Late October / early November: SSA publishes the full “Cost-of-Living Increase and Other Determinations” notice in the Federal Register with the new benefit amounts, wage base, and Medicare Part B interaction.
  • December (statutorily): The increase technically applies to benefits for December of the current year.
  • January payment: That December benefit is paid in January, so the first check with the new COLA lands in January of the following year.

How to project the next COLA before SSA announces it

You don’t have to wait. The CPI-W index is public, so you can compute the running COLA the moment the July, August, and September readings drop:

  1. Pull CPI-W monthly values from FRED (series CWUR0000SA0) or BLS Table 1.
  2. Average the three Q3 months of the current year.
  3. Divide by the average of the three Q3 months of the last COLA year (usually last year, unless the last year was 0.0%).
  4. Subtract 1, multiply by 100, round to the nearest 0.1.

After the August CPI release, you have two of the three data points — market economists routinely publish an “implied COLA” that has been within about 0.1 percentage points of the final figure. See our tracking of the 2027 COLA projection after the August 2026 CPI report for a live example.

Common mistakes to avoid

  • Don’t use CPI-U or headline CPI. They’re close, but not identical. In some years the difference between CPI-W and CPI-U is 0.1–0.2 percentage points — enough to round the COLA up or down a step.
  • Don’t use seasonally adjusted data. SSA uses not seasonally adjusted CPI-W. Seasonally adjusted values are for economists modeling monthly changes.
  • Don’t compare Q3 to the immediately prior Q3 in a zero-COLA year. If the last COLA was 0%, the benchmark is still the earlier Q3, not last year’s.
  • Remember the round. The math is rounded to 0.1 percentage points at the very end. A 2.74% raw result becomes 2.7%; a 2.75% result becomes 2.8%.
  • Medicare Part B is separate. Rising Part B premiums are deducted before your net Social Security check, so a nominal 2.8% raise can feel smaller. That’s a Medicare policy question, not a COLA formula question — see our breakdown of the different inflation measures for why headline inflation and retiree inflation drift apart.

Related concepts and what to learn next

Sources

Disclosure: This article is for informational purposes only and is not investment advice.

Leave a Comment