Tax Bracket

A tax bracket is a band of taxable income taxed at a single rate. The federal income tax currently has seven of them, at rates of 10, 12, 22, 24, 32, 35, and 37 percent, and income moves through the bands in order rather than being assigned to one of them. The formal name for the instrument is different from the popular one: the tax code calls the underlying provisions rate brackets, the annual guidance that sets the dollar figures calls them tax rate tables, and almost everyone else says "tax bracket." All three describe the same thing. The dollar boundaries change every year, and the current ones are published by the IRS at IRS.gov rather than restated here, because a figure copied into a reference page is a figure that goes stale in January.

Where the numbers come from. Congress wrote the tables into section 1(j)(2) of the Internal Revenue Code, stated in 2018 dollars. In the married filing jointly table, for instance, the statute puts the top of the 12% band at $77,400, the top of the 22% band at $165,000, and the start of the 37% band above $600,000. Nobody has paid tax on those exact figures since 2018, because section 1(f)(1) then requires that "not later than December 15" of each year the Secretary shall prescribe replacement tables for the following year. Those replacements arrive as an annual revenue procedure. The practical consequence is that a year's brackets are settled and published before the year begins, which is what makes tax planning in December possible at all.

There are five tables. Section 1(j)(2) contains one for married couples filing jointly and surviving spouses, one for heads of households, one for unmarried individuals, one for married individuals filing separately, and one for estates and trusts. The last is the outlier and it surprises people: subparagraph (E) contains only four rates, 10, 24, 35, and 37 percent, and in the statute's 2018 dollars it reaches 37% above $12,500 against $600,000 for a married couple. That ratio holds after indexing, so a trust pays the top rate roughly fifty times sooner than a couple does. It is the main reason a trustee with discretion tends to distribute income to beneficiaries rather than let the trust retain it.

Why the boundaries move, and why more slowly than they used to. Section 1(f)(3) keys the annual adjustment to the Chained Consumer Price Index for All Urban Consumers, usually written C-CPI-U, which replaced the older CPI measure for this purpose in 2018. Section 1(f)(4) and section 1(f)(6)(B) define the index for a year as the average over the twelve months ending August 31, so the reading is complete in early autumn and the tables can be issued in October. Chained CPI rises more slowly than the traditional measure because it accounts for consumers substituting between goods as prices change. Slower indexing means bracket boundaries climb more slowly than incomes do in an inflationary period, which is the mechanism behind bracket creep. Each year's increase is rounded down to the nearest $50, or to the nearest $25 for a married person filing separately and for a single filer. Heads of household round to $50 like a married couple, because the statute's $25 rule reaches an unmarried individual only where that person is neither a surviving spouse nor a head of household.

The asymmetry introduced for 2026, which is the part almost nobody mentions. The One Big Beautiful Bill Act of 2025 did two separate things to this subsection. Its section 70101(a) struck the end date that would have returned the pre-2018 rates after 2025, so the seven-rate structure is now permanent rather than scheduled to expire. Its section 70101(b) then changed how the tables are indexed. Section 1(j)(3)(B)(i) sets the base year the adjustment measures from, and the Act inserted a qualifier into it: "solely for purposes of determining the dollar amounts at which any rate bracket higher than 12 percent ends and at which any rate bracket higher than 22 percent begins." Before that amendment, the later base year applied to every boundary in the table. Confining it to the upper boundaries leaves the top of the 10% band and the top of the 12% band measured from a base one year earlier, and an earlier base produces a larger cumulative adjustment.

Three things follow, and the third is the one that gets overstated. The bottom two boundaries are permanently higher than they would have been under the old rule, by roughly a single year's inflation. The uplift is a fixed percentage, because the two adjustment factors differ only by the ratio between two frozen index readings, so in proportional terms it is a one-step change carried forward rather than a gap that compounds. In dollars, though, that fixed percentage is applied to a figure that keeps growing, so the dollar advantage does widen year by year like everything else the index touches. And the split reaches only the four individual tables. The estates and trusts table has no 12% bracket at all, and its lowest boundary is the point at which a 24% bracket begins, so every boundary in it still measures from the later base year and nothing about it changed. The amendments apply to tax years beginning after December 31, 2025, so the 2026 tables are the first to show any of this.

Capital gains have their own brackets. The 0, 15, and 20 percent rates on long-term capital gains and qualified dividends run on separate thresholds under section 1(h), as modified by section 1(j)(5), and those thresholds do not line up with the ordinary bands. A person can sit in the 22% ordinary bracket and the 15% capital-gains bracket on the same return. Neither table describes the 3.8% net investment income tax or the 0.9% additional Medicare tax, which are separate charges with their own statutory, non-indexed thresholds.

The annual cycle runs in one direction. Congress sets a table in the statute. The Bureau of Labor Statistics publishes the chained price index each month. The Treasury averages the twelve readings ending August 31, compares that average against the statutory base year, applies the resulting percentage to every boundary in every table, rounds each increase down, and publishes the finished tables in a revenue procedure before the year starts. The rates themselves are never adjusted by this process; section 1(f)(2)(B) says so directly, and only Congress can change a rate.

A hypothetical illustration of the base-year split, using invented index values so the arithmetic can be checked by hand. Suppose the chained index averaged 100.0 in the earlier base year, 102.0 in the later base year, and 130.0 in the calendar year before the one the tables apply to, which is the reading section 1(f)(3)(A)(i) actually uses. The top of the married filing jointly 12% band starts from the statutory $77,400 and is measured from the earlier base, so it is multiplied by 130.0 divided by 100.0, giving $100,620; rounding the increase down to the nearest $50 makes it $100,600. The top of the 22% band starts from the statutory $165,000 but is measured from the later base, so it is multiplied by 130.0 divided by 102.0, giving $210,294.12, which rounds to $210,250. Had the upper boundary used the earlier base as well, it would have been $165,000 multiplied by 1.30, or $214,500. The split therefore leaves the upper boundary about $4,250 lower, and the ratio between the two adjustment factors is exactly the index step between the two base years, 1.02, in every future year as well.

Pros of a bracketed, indexed structure

• Annual indexing means inflation alone does not raise anyone's real tax rate, and it happens without Congress having to vote on it each year.
• Because the tables are published before the year begins, liability is computable in advance, which is what makes deliberate timing of income and deductions possible.
• One published table per filing status applies to everyone in it, so the computation is transparent and checkable rather than discretionary.

Cons and limits

• Chained CPI understates the inflation many households actually experience, so some bracket creep survives the indexing.
• The tables price ordinary income only. Long-term capital gains, qualified dividends, the net investment income tax, and the additional Medicare tax all sit outside them, and the last two are not indexed at all.
• Taxable income is the input, and the deductions and exclusions that produce it change far more often than the bracket structure does.
• State income tax brackets are entirely separate, and several states index theirs partially or not at all.
• Estates and trusts reach the top rate at a tiny fraction of an individual's threshold, which makes retained fiduciary income unusually expensive.