COLA Ledger

An independent reference — not affiliated with the Social Security Administration. Every figure traced to the government publication that produced it.

Corrections and disclosures

This site publishes under a brand rather than a personal byline. This log is what stands in place of one: a dated, public record of every figure changed, every conflict between sources found, and every limit that could otherwise have been left unsaid.

Entries in the log

9

All dated 10 August 2026 — the record opens with the disclosures found while building the datasets, rather than starting empty.

What counts as an entry here

A published figure that changed. A place where two official sources disagree. A limit on what this site can compute. A scope decision that made the published set smaller than planned. Typographical fixes and wording changes are not logged; changes to numbers always are.

The log

10 August 2026 — BLS and SSA disagree on one CPI-W month

Source conflict

Across the 628 overlapping months checked, the Bureau of Labor Statistics series and SSA’s republished table agree exactly on all but one, where the two published values differ by two tenths of an index point.

Resolution: BLS is the producer of record for the series and SSA republishes it, so the BLS value is committed. The month feeds no adjustment computation under any of the three regimes, so no published figure on this site changes either way. The ingest gate is strict where it matters: a disagreement in a month that does feed a computation fails the build outright rather than picking a side.

10 August 2026 — 6 historical adjustments cannot be reproduced from the current index

Disclosure

Payable years 1976, 1978, 1979, 1980, 1984, 1985. SSA computed these from the CPI-W as published on a 1967 base; the 1988 rebasing means the values now served for those months are one-decimal conversions, which discards enough precision to flip the final tenth.

Resolution: SSA’s published figure is what those pages state. Each page carries the disclosure rather than claiming the recomputation matches, and the build gate allow-lists precisely those years and bounds each at a tenth of a percentage point, so the list cannot grow silently. The table of differences.

10 August 2026 — The hold-harmless premium on a $600 benefit is $201.00, not $201.80

Correction

The capped premium is the prior standard premium plus the beneficiary’s dollar adjustment — and that adjustment is the difference between two floored monthly benefit amounts, because benefits floor to the whole dollar before anything compares them. On a $600 benefit the floored adjustment is +$16, giving $201.00. Using the unfloored increase gives $201.80.

Resolution: the computed figure is published and is unit-tested. Where a computation and a piece of prose disagree on this site, the computation wins and the prose changes.

10 August 2026 — The hold-harmless boundary is an integer boundary, not $639.29

Correction

The continuous break-even of $639.29 is correct as derived arithmetic, but it is not where the cap actually starts and stops binding — for the same flooring reason as above, the real boundary sits between two whole dollars a little higher.

Resolution: both figures are published, the ladder rungs where the cap binds are computed from each year’s own data rather than hardcoded, and the break-even is never presented as the boundary. The hold-harmless page.

10 August 2026 — The quarterly rounding rule is era-dependent, and a single rule gets six years wrong

Disclosure

Determinations through 2006 round the quarterly CPI-W mean to the nearest 0.1; from 2007 they round to 0.001 under 20 CFR 404.275. The switch is keyed to the determination year, not to the precision of the months themselves.

How we know: SSA’s own footnote states the 1999 adjustment was originally determined as a figure that only the tenth-rounding rule reproduces. A flat three-decimal rule would make the statutory override that followed look like no override at all. The rule, with both answers for the same quarter.

10 August 2026 — The benefit ladder steps in $100 because the source data does

Scope change

The ladder was originally planned at half that step. SSA’s published distribution is banded at $100 and no finer table exists, so the ladder was widened to match rather than the percentile context being interpolated to fit a page count.

Consequence: the published set is smaller than planned, and every rung carries a percentile band that comes straight from the source.

10 August 2026 — Medicare income-related tables are published for 14 years only

Disclosure

Two documented reasons. The 2007 and 2008 tiers are phase-in figures that cannot be validated against the statutory cost-share multipliers. Several later Federal Register notices state the tier percentages in prose but publish no bracket table at all — verified against both the raw Federal Register text and the govinfo HTML, which are byte-identical.

Resolution: no table is hand-keyed from a secondary source. The standard premium and deductible for those years are committed and complete, and the net-of-Medicare arithmetic this site is built on needs only the standard premium. The missing years, with the reason in each case.

10 August 2026 — October 2025 CPI-W does not exist and never will

Disclosure

BLS did not publish an October 2025 Consumer Price Index. Due to a lapse in appropriations, BLS could not collect October 2025 reference period survey data and was unable to retroactively collect these data.

Resolution: the month is stored as an explicit null rather than a zero or a missing key, the function that averages a quarter throws if any member is null, and every twelve-month comparison crossing it renders a dash rather than a number. The full account.

10 August 2026 — One figure is corroborated rather than primary-sourced

Disclosure

The Medicare Part D national base beneficiary premium is announced in a CMS document this site cannot retrieve — cms.gov blocks automated access, and PDF mirrors of CMS fact sheets have been observed to parse into plausible but wrong tables.

Resolution: the value is recovered from the published income-related amounts themselves. The statutory formula reproduces all five exactly at $38.99 and at no other base, and SSA’s programme manual publishes the same ladder. The identity is asserted as a build gate, so a wrong base fails the build. The arithmetic.

Standing conflicts we do not expect to resolve

Known disagreements between published sources, and how each is handled
DisagreementHow this site handles it
SSA’s two pages state the base-quarter rule differentlyThe precise formulation is used. The two readings coincide except after a zero-adjustment year, where the loose one is wrong by half a percentage point. Worked in full.
SSA’s table year is one behind the public name for each adjustmentBoth keys are stored on every record. Pages use the payable year, because that is what people search for, and a build gate fails on any mismatch. The full problem.
SSA’s SSI table labels the July 1983 increase as a percentageIt was a flat statutory dollar increase, not a cost-of-living adjustment — no adjustment was payable in 1983 at all. Encoded as a documented override. The 1983 page.
Part B-ID income-related amounts sit about ten cents below the Part B onesA real artifact of computing against a different actuarial rate, preserved rather than “corrected”. The Part B-ID page.

Report an error

Email support@inventum.com.au. A figure with a source attached will be checked against the primary publication. If it is right, the figure is corrected, the affected pages are rebuilt, and an entry is added here with the date and the reasoning. If it is not, we will say why.

Related

Sources and freshness

Adjustment for 2026: 2.8%. Log opened 10 August 2026.

Page figures last verified against the sources above on 10 August 2026. BLS data retrieved 10 August 2026. BLS.gov cannot vouch for data or analyses derived from these data after retrieval. Corrections: the correction log · support@inventum.com.au