Tax Inflation Adjustment

A tax inflation adjustment is the yearly revision of a dollar figure in the tax law to reflect a change in prices. The Code's own name for it is the "cost-of-living adjustment," which is the caption of section 1(f)(3), and the IRS's newsroom calls the annual release "inflation adjustments." The descriptive name used here is what readers search for; the statutory term is the one to cite.

The structure is worth grasping before any of the detail, because it explains why this is a subject rather than a footnote to the tax bracket page. Section 1(f) imposes a duty on the Treasury Secretary to publish adjusted rate tables each year and defines the arithmetic for doing it. Dozens of other provisions across the Code then reach into section 1(f)(3) by cross-reference and use the same adjustment for their own dollar amounts: the standard deduction, the earned income credit, retirement plan contribution limits, the gift and estate exclusions, health savings account limits, the alternative minimum tax exemption, and many more. It is one machine with many customers rather than a rule about brackets.

The consequence is that "is this figure indexed?" is a question about the individual provision, not about the tax system. Two amounts sitting side by side on the same return can behave completely differently, and there is no way to tell from the outside which is which. That is the question this page exists to answer.

The duty, and what it may and may not change. Section 1(f)(1) requires that "not later than December 15 of 1993, and each subsequent calendar year, the Secretary shall prescribe tables which shall apply in lieu of the tables contained in subsections (a), (b), (c), (d), and (e) with respect to taxable years beginning in the succeeding calendar year." Section 1(f)(2) then constrains how: by increasing each bracket's minimum and maximum dollar amounts by the cost-of-living adjustment, "by not changing the rate applicable to any rate bracket as adjusted," and by adjusting the printed tax amounts to match. Rates are Congress's and boundaries are the formula's. That division is why an annual adjustment can be administrative rather than legislative.

The arithmetic, including the piece almost every summary drops. Section 1(f)(3)(A) defines the adjustment as the percentage by which the chained index for the preceding calendar year exceeds "the CPI for calendar year 2016, multiplied by the amount determined under subparagraph (B)." Subparagraph (B) is a splice ratio: the chained index for 2016 divided by the older index for 2016. Its purpose is to bridge two different price series so that a base year measured on the old index stays comparable with current readings on the new one. Without it the formula would compare unlike quantities. It is one clause, and it is what makes the whole thing internally consistent.

Which index, and the switch that dated every older description. Section 1(f)(6)(A) defines the C-CPI-U as "the Chained Consumer Price Index for All Urban Consumers (as published by the Bureau of Labor Statistics of the Department of Labor)," and fixes the values used as the latest published as of the date the Bureau publishes its initial August value for the preceding year. Section 1(f)(6)(B) sets the measurement period as the twelve months ending August 31. The switch to that index was made by section 11002(a) of the 2017 tax act, which "added par. (3) and struck out former par. (3)," whose text had measured the adjustment against "the CPI for the calendar year 1992." Section 11002(e) applies the change "to taxable years beginning after December 31, 2017." So: the older index off a 1992 base through 2017, the chained index off a 2016 base from 2018. The published tax bracket page carries the index name and the rounding convention in detail.

Base-period re-anchoring, and why it moves thresholds in the direction nobody expects. Because the formula divides by the base-year index, a later base year is a larger denominator and therefore a smaller cumulative adjustment. Provisions with different base years therefore sit at different heights even when they have been indexed for the same length of time, and section 1(f)(3)(C) exists to handle exactly that case, telling you how to apply the formula where another provision substitutes a year after 2016. Section 2010(c)(3)(B), for instance, indexes the estate and gift exclusion by "substituting 'calendar year 2025' for 'calendar year 2016.'" The intuition that a later base year means a bigger, more up-to-date number is backwards, and it is the single most reversible fact in this area.

Rounding varies by provision, so there is no site-wide rounding rule to learn. Section 1(f)(7)(A) rounds an increase "to the next lowest multiple of $50," and (7)(B) substitutes $25 for a married individual filing separately. Note two things about that. It rounds down, so indexing systematically slightly under-compensates. And it is not general: subparagraph (A) reaches only increases "determined under paragraph (2)(A), section 63(c)(4), section 68(b)(2) or section 151(d)(4)." Other provisions set their own, and they differ in both size and direction. Section 2010(c)(3)(B) rounds "to the nearest multiple of $10,000." Section 7872(g)(5) rounds to the nearest multiple of $100, and adds that an increase which is itself a multiple of $50 "shall be increased to the nearest multiple of $100," so that one rounds up. Writing that inflation adjustments round down to $50 is wrong about most of the Code.

A small illustration of how such a cross-reference can rot: section 1(f)(7)(A) still points at section 68(b)(2), and a References-in-Text note records that section 68(b)(2) "was omitted in the general amendment of section 68" by the 2025 tax act. The pointer survives the provision it points at.

The delivery vehicle, and how to read one. The adjustments arrive as an annual revenue procedure, published in the autumn for the following tax year. Three practical notes about using one:

• Prefer the Internal Revenue Bulletin version at irs.gov/pub/irs-irbs over the advance copy released at irs.gov/pub/irs-drop. The advance copy is pre-publication and is not the source of record; the two have disagreed.
• Section numbers within the procedure are not stable across years. The rate tables have appeared under three different section numbers in recent years, so search for the statutory cite rather than a section number.
• Not everything is in the autumn procedure. Health savings account and high-deductible plan limits come in their own procedure each spring, and the autumn one does not contain them at all.

And a category the procedures never contain: figures that change annually by statutory schedule rather than by indexing. The cap on the deduction for state and local taxes is the clearest current example. Section 164(b)(7)(A) sets it at $40,000 for 2025, $40,400 for 2026, then "101 percent of the dollar amount in effect … for the preceding calendar year" through 2029, reverting to $10,000 for years after that. It moves every single year and appears in no revenue procedure, so the usual habit of asking whether the annual procedure moved a figure gives the wrong answer for it.

The inventory of what is NOT indexed, which is this page's reason to exist. Each entry below was established by reading the section's own full text and searching it for cost-of-living or inflation-adjustment language. A zero result means the section contains no adjustment mechanism at all, not that the mechanism was found and is dormant.

There is a second, independent confirmation of the first six rows available in the statute book itself. Section 11002(e) of the 2017 tax act, which moved indexed provisions onto the chained index, lists by number every section it amended: sections 23, 25A, 25B, 32, 36B, 41, 42, 45R, 55, 59, 62, 63, 68, 125, 132, 135, 137, 146, 147, 151, 162, 179, 213, 219 to 221, 223, 280F, 408A, 430, 512, 513, 831, 877A, 911, 1274A, 2010, 2032A, 2503, 4161, 4261, 4980I, 5000A, 6039F, 6323, 6334, 6601, 6651, 6652, 6695, 6698, 6699, 6721, 6722, 7345, 7430, 7872, and 9831. Sections 1211, 121, 1411 and 86 are absent from that list, and the reason is that there was nothing in them to move. The presence of section 7872 in the list is a useful check on the method rather than a contradiction: that section does contain one indexed amount, a limit relating to continuing-care facilities, and the $10,000 and $100,000 thresholds sit outside it.

What follows from the inventory is a practical rule rather than a complaint. Each of those provisions states its own position on its own page, and the general observation that inflation raises tax is not something a household acts on. What is actionable is knowing which specific threshold nearby is fixed in nominal dollars, because that one is moving toward you every year. The effect itself, and what to do about it, is the subject of the page on bracket creep.

The annual cycle runs like this. The Bureau of Labor Statistics publishes its initial chained-index value for August. That completes the twelve-month average ending August 31 that section 1(f)(6)(B) requires. Treasury computes what the statute calls a cost-of-living adjustment under section 1(f)(3), a label the Code shares with the very different Social Security benefit increase, applies it to the rate-band boundaries and to every provision whose own text borrows the same adjustment, applies each provision's own rounding rule, and publishes the results in a revenue procedure. Section 1(f)(1) sets the deadline as December 15, and the figures apply to tax years beginning in the following calendar year. Nothing in the process changes a rate, and no vote is taken.

Reading the output. A revenue procedure is organized by provision, each section of it naming the Code section it adjusts. To find a figure reliably, search for the statutory cite rather than for a section number in the procedure, because the procedure's own numbering moves between years. Confirm you are reading the Bulletin version rather than the advance release. And check whether the figure you want is even in that document, since health savings account limits are published separately in the spring.

A worked illustration of the base-year point, because it is the one people get backwards. Two provisions are both indexed, both use the machinery in section 1(f)(3), and both have been adjusted every year for the same length of time. One measures its adjustment from a 2016 base and the other from a 2025 base, because its own text substitutes the later year. The formula divides by the base-year index, so the provision with the 2025 base is dividing by a larger number and receives a smaller cumulative uplift. Nine years of inflation are simply outside its measurement. The later base year does not mean the more current or the more generous figure; it means less accumulated adjustment. That is why the question "when is this provision's base year?" is a real question about size, and not a technical detail.

A second illustration, on the difference between an indexed and a scheduled figure. Take the cap on the state and local tax deduction. It is $40,400 for 2026, it was $40,000 in 2025, and section 164(b)(7)(A)(iii) puts it at "101 percent of the dollar amount in effect … for the preceding calendar year" for each year through 2029 before it reverts to $10,000. So a taxpayer checking whether their figure changed this year would find no revenue procedure entry for it, would reasonably conclude nothing moved, and would be wrong. The right question about any dollar amount in the tax law is not "did the annual adjustment change it?" but "what, if anything, in this provision's own text makes its number move?" There are three possible answers: an indexing cross-reference, a statutory step schedule, or nothing at all.

This is a mechanism rather than a product, so what follows is what indexing does well and where the design leaves problems in place.

What the machinery gets right

• It removes the largest source of unlegislated tax increases without requiring an annual vote, and it has done so since the mid-1980s.
• Rates cannot change through it. Section 1(f)(2)(B) permits only the boundaries to move, so the adjustment cannot be used as cover for a rate increase.
• The figures are published before the tax year begins, which is what makes deliberate timing of income and deductions possible at all.
• One formula serves dozens of provisions, so a change to the method reaches all of them consistently rather than provision by provision.
• The splice ratio in section 1(f)(3)(B) preserves comparability across a change of price index, which is a genuine piece of careful drafting.

The limits and the costs

• Coverage is opt-in by cross-reference, so a significant list of provisions gets nothing, and there is no way to tell from the outside which is which.
• Rounding under section 1(f)(7)(A) goes to the next lowest multiple, so the adjustment always slightly under-compensates and never over-compensates.
• Rounding rules differ by provision, so there is no single convention to rely on.
• The measurement window closes on August 31 of the preceding year, so an acceleration in inflation is compensated a year late.
• The index used since 2018 rises more slowly than the one it replaced, so indexed thresholds drift behind many households' own experience of prices.
• Figures on a statutory step schedule change annually while appearing in no revenue procedure, which defeats the natural way of checking.