2027 Social Security COLA — Projected

The timeline: when this stops being a guess

First of three milestones reached — the July 2026 print landed this morning and is now in the arithmetic below.

• ✓ Aug 12, 2026Done. BLS released the July 2026 CPI at 8:30 a.m. ET: CPI-W 327.104. One of the three determining months is now known, and this page was rewritten around it the same day. The band narrowed from a full percentage point to 0.6 of a point.
• Sep 11, 2026Next. BLS releases the August 2026 CPI. Two of three known. At this point a careful projection is usually within about 0.1–0.2 points of the final figure.
• Oct 14, 2026BLS releases the September 2026 CPI. The formula now has every input it needs. SSA has historically announced the COLA the same morning, alongside the new taxable maximum and earnings test amounts. That taxable maximum has its own statutory formula and its own projection — see the projected 2027 wage base.
• Nov 2026CMS announces the 2027 Medicare Part B premium and IRMAA thresholds. Only now can anyone tell you what your net Social Security deposit will be. See our projected 2027 IRMAA brackets.
• Jan 2027The COLA appears in benefit payments, starting with the December 2026 SSI payment and the January 2027 Social Security payments.

One caution on that October date. The 2026 announcement was scheduled for October 15, 2025 and actually came on October 24, 2025, because the government shutdown delayed the September 2025 CPI release by nine days. The same shutdown cancelled the October 2025 CPI outright — a month that simply does not exist in the published series. A COLA that rests on exactly three monthly prints has no redundancy if one of them is disrupted.

The statutory formula, quoted

The baseline — official, and now one third of the numerator with it

These are published CPI-W readings, not estimates. As of this morning that includes July 2026, the first of the three months in the 2027 numerator.

A check you can run yourself. Divide the two published columns: 317.2647 ÷ 308.7290 = 1.027648, an increase of 2.7648 percent, which rounds under §415(i)(1)(D) to 2.8 percent — exactly the COLA SSA announced for 2026 on October 24, 2025. The same arithmetic applied to the same series reproduces the official answer to the digit. That is the whole reason to trust the method here, and to keep treating the August and September 2026 cells as what they are: inputs that do not yet exist.

Where the July 2026 figure comes from. 327.104 is the July 2026 value of BLS series CWUR0000SA0, published in the BLS Consumer Price Index news release of August 12, 2026 and read from the BLS public data API the same morning. Against July 2025's 316.349 that is a year-over-year rise of 3.400 percent, down from June's 3.523 percent. Month over month it is a rise of only 0.01 percent from June's 327.075, against a five-year average July step of about +0.17 percent — so July was a soft print by seasonal standards, and that is the reason our central case fell rather than rose.

What the trend looks like so far

Year-over-year CPI-W through the last published month. Green is published; amber is a month the formula needs and nobody has.

The shape matters more than any single number. CPI-W ran near 2.2 percent through February, accelerated hard into a 4.44 percent May, gave a chunk of it back in June at 3.52 percent, and has now eased again to 3.40 percent in July — the first of the three months that actually count. A run that volatile is precisely the case where extrapolating the remaining two months forward still deserves a band rather than a confident point estimate.

What August and September now have to do. July is fixed at 327.104, so the quarterly mean is (327.104 + Aug + Sep) ÷ 3 and only two numbers are still free. For the quarter to land on our central 3.4 percent, the quarterly mean has to be 328.05, which means August and September have to average 328.53. The bottom of our band, 3.1 percent, needs a quarterly mean of 327.10 and an August–September average of 327.10. The top, 3.7 percent, needs a quarterly mean of 329.00 and an August–September average of 329.95. Because the soft July print is now locked into the average, the two remaining months have to run hotter than July to reach any of those levels.

Central projection and band

We ran six carry-forward rules against the same published data rather than picking one and calling it the answer. Before this morning each rule had to fill all three of July, August and September 2026. Now July is a fact, not a rule output — every row below uses the official 327.104 for July and only carries August and September forward. Each row then averages the three and divides by 317.265.

The band, and why it narrowed by exactly a third. The six rules now span 3.29 to 3.66 percent, tighter than the 3.26–3.96 spread they showed yesterday, because they no longer disagree about July. We publish a band of 3.1% to 3.7% — still wider than the spread of the rules themselves, on purpose, for the same reason as before: the rules share one dataset and one blind spot, and none of them knows what actually happens in the months it is filling.

The width follows the page's own rule for cushioning unobserved months rather than a fresh judgement call. Yesterday's cushion was half a point on each side, allowed for three unobserved months — a full percentage point of width. One of those three is now observed, so two-thirds of the exposure remains and two-thirds of the cushion should go with it: 0.5 × 2⁄3 = 0.33 points a side. We publish that as ±0.3 around the rounded central case, because §415(i)(1)(D) rounds the COLA itself to a tenth of a point and quoting a band to hundredths would imply a precision that a two-unobserved-month projection does not have. So the band runs 3.1% to 3.7% — 0.6 points wide against yesterday's 1.0, with one third of the uncertainty resolved by one of three prints. The test that matters is whether the cushion still covers the disagreement between the rules, and it does with room on both sides: the six rules span 3.29% to 3.66%, comfortably inside 3.1%–3.7%. It is a proportional narrowing, not a claim that the answer got easier — the same one-index-point surprise in August or September still moves the COLA by about 0.105 points.

Corroboration, treated as a data point rather than a verdict. The Senior Citizens League, which publishes a monthly CPI-W-based estimate, put the 2027 COLA at 3.8 percent in its early-August 2026 update. That figure was computed before this morning's July print, at a time when all three determining months were still unknown, and it now sits just above the top of our narrowed band. TSCL updates after each CPI release, so expect its number to move too. We are recording the disagreement rather than smoothing it: we computed our figures independently from the BLS series and did not adjust them toward theirs, in either direction.

What a 0.1-point miss is worth

The average monthly benefit for retired workers was $2,084.40 in June 2026, per SSA's Monthly Statistical Snapshot. Every figure below applies the COLA to that average; your own number scales with your own benefit.

So 0.1 percentage point is worth $2.08 a month, or $25.01 over a year, on that average benefit. The full width of our narrowed band — 0.6 of a percentage point — is now $12.51 a month, about $150 a year, down from the $20.84 a month and roughly $250 a year that the old one-point band covered. Scale it: a $3,500 benefit moves $3.50 a month per 0.1 point, roughly $252 a year across the band.

What today's print was worth in dollars. The central case moved from 3.5% to 3.4%, one tenth of a point, so the projected raise on that average benefit fell from about $72.95 a month to $70.87 — roughly $2.08 a month, or $25 over a year. That is the whole cash value of the first of three prints, which is a fair measure of how little any single month can be leaned on.

How little it takes to move it again. Because the quarter is an arithmetical mean of three months, a surprise of one index point in a single month shifts the quarterly average by a third of a point — which is 0.105 percentage points of COLA, or about $2.19 a month for life on the average benefit. That arithmetic has not changed just because July is in. Two months are still open, so two such surprises are still available, and one month coming in a couple of tenths hot or cold is entirely ordinary. That is the mechanical reason the answer genuinely changes again on September 11 and October 14.

And the COLA is not the whole story of what arrives in January. It is a gross figure. What lands in the account is the COLA less whatever the 2027 Medicare Part B premium turns out to be, which CMS does not announce until November.

How the COLA collides with Medicare: the hold-harmless rule

Most people meet the COLA as a net number — the deposit went up by this much — because the Part B premium is withheld from the benefit. The rule governing that interaction is at 42 U.S.C. §1395r(f), and it works in a way that surprises people.

Who is left outside it. The protection is narrower than its reputation. It reaches only people whose Part B premium is deducted from a Social Security or Railroad Retirement benefit, and §1395r(f) expressly withholds it from several groups: those enrolling for the first time that year, people whose premiums are not withheld from a benefit at all, dual eligibles whose premiums are paid by Medicaid, those enrolled solely for immunosuppressive-drug coverage, and — the one that matters most to the readers of a 2027 planning page — anyone whose premium is adjusted under subsection (i), the IRMAA surcharge. Cross an IRMAA threshold and you lose hold-harmless entirely, in both directions.

Why 2027 probably does not test it. Hold harmless only bites when the COLA is small relative to the Part B increase. At our central 3.4 percent the COLA is roughly $71 a month on the average benefit — several times any plausible year-on-year Part B premium rise. It is the low-COLA years that trigger it, and when it does trigger, the cost does not disappear: Part B is funded to a target, so the increase the protected majority does not pay is loaded onto the minority who are excluded. IRMAA payers are in that minority. That is the asymmetry worth planning around, and it argues for watching the projected 2027 IRMAA thresholds alongside this page rather than instead of it.

The second-order effect nobody budgets for. A larger COLA raises your gross benefit, which raises the taxable portion of Social Security, which raises MAGI — and MAGI is what IRMAA is measured on, two years later. The base amounts that decide how much of your benefit is taxable ($25,000 and $32,000) have never been indexed since 1984, so every COLA pushes a little more of the benefit across them permanently. A 3.4 percent COLA is a raise and a slow tax increase at the same time. See how much you can earn before Social Security is taxed and the tax torpedo for where that lands.

Common questions

When is the 2027 Social Security COLA announced?

The Social Security Administration has historically announced the figure on the morning the September CPI report is released. The September 2026 CPI is scheduled for October 14, 2026, so expect the 2027 COLA that day. Treat that as a strong expectation rather than a booked date: the 2026 announcement was scheduled for October 15, 2025 and landed on October 24, 2025 because the government shutdown pushed the September CPI release back nine days.

What is the 2027 COLA projected to be right now?

Our central projection is 3.4 percent, with a band of 3.1 to 3.7 percent, updated August 12, 2026 with the July 2026 CPI-W print of 327.104. One of the three CPI-W months that determine the answer is now published and two are not, so the band is two-thirds as wide as the one we carried before the July print. The previous central case was 3.5 percent on a 3.0 to 4.0 percent band. The Senior Citizens League published 3.8 percent in early August 2026, before the July print landed; that figure now sits just above our band.

How is the Social Security COLA calculated?

Under 42 U.S.C. §415(i), the COLA is the percentage increase in the Consumer Price Index for the calendar quarter ending September 30 over the same quarter of the most recent year that produced an increase, rounded to the nearest one-tenth of one percent. §415(i)(1)(G) defines a quarter's index as the arithmetical mean of the three monthly readings. So the 2027 COLA is the July, August and September 2026 CPI-W average measured against the July, August and September 2025 average of 317.265.

Why is a 2027 COLA projection made in August so weak?

Because only one of the three determining months exists. July 2026 CPI-W published August 12, 2026 at 327.104; August 2026 publishes September 11, 2026; September 2026 publishes October 14, 2026. A projection made today still extrapolates a trend across two months of unknown data. A one-index-point surprise in a single one of those months moves the COLA by about 0.105 percentage points, and July itself landed about half an index point below what our carry-forward rules had assumed, which is why the central case fell from 3.5 to 3.4 percent.

Does a larger COLA raise my Medicare Part B premium?

Indirectly, yes. The hold-harmless rule at 42 U.S.C. §1395r(f) caps the dollar increase in your Part B premium at the dollar amount of your COLA, but only if your premium is deducted from your Social Security benefit. It does not apply to new enrollees, to anyone paying an IRMAA surcharge under subsection (i), to dual eligibles, or to people who pay Medicare directly. A COLA near 3.4 percent produces roughly $71 a month on the average retired-worker benefit, far more than any plausible Part B increase, so hold harmless would not bind for most people in 2027.