Bracket Creep

Bracket creep is the increase in tax liability that follows from inflation rather than from any improvement in a taxpayer's circumstances. The Congressional Research Service defines it as what "happens when someone's tax liability increases because of a larger increase in that taxpayer's nominal income than their real income." Congress described the harm in a statutory heading rather than defining the term: section 1(f) of the Internal Revenue Code is titled, in part, "adjustments in tax tables so that inflation will not result in tax increases."

The mechanism is arithmetic. Tax is computed on dollar amounts, and inflation raises the number of dollars a person receives without raising what those dollars buy. If the thresholds in the tax law do not move by the same proportion, a taxpayer whose purchasing power is flat is measured against the law as though they had become better off, and is taxed accordingly. Nobody voted for the increase and nothing in the taxpayer's life changed.

The phrase is not in the tax code, in the regulations, or in any IRS publication. It is an economics term used by the Congressional Research Service, the Congressional Budget Office and Treasury. It is also used loosely for two related but distinct effects, and separating them is most of the value of understanding it.

The two things the phrase is used for. The first is inflation-driven creep: nominal income rises with prices, real income is flat, and an unmoved threshold taxes the illusion. The second is sometimes called real bracket creep: income genuinely grows faster than inflation, so a taxpayer moves up the schedule legitimately, and thresholds indexed only to prices do not keep pace with a rising standard of living. Only the first is a distortion; the second is the rate schedule doing what a progressive schedule is designed to do. Coverage that conflates them will describe an indexed system as still producing bracket creep and be half right for the wrong reason.

The historical case, and a date almost every source gets wrong. Before indexing, high inflation made this a first-order problem, and it was the direct reason Congress acted. The Economic Recovery Tax Act of 1981 introduced the indexing machinery. It did not take effect that year. The Act's own effective-date provision, section 104(e), states that "the amendments made by this section … shall apply to taxable years beginning after December 31, 1984." So the first indexed tax year was 1985, four years after enactment. The Congressional Research Service's own report says the brackets "have been indexed for inflation since 1981," which is accurate about the authorizing legislation and misleading if read as the date the adjustments began, and it is repeated as the latter throughout secondary coverage.

Where creep survives, and this is the substance rather than the history. Indexing reaches what a statute says it reaches, and a great many dollar figures in the Code carry no adjustment provision whatsoever. Read at source and searched for inflation-adjustment language, each of the following contains none:

• The capital-loss deduction against ordinary income, section 1211(b): $3,000, or $1,500 on a separate return, unchanged since 1986.
• The exclusion of gain on the sale of a principal residence, section 121(b): $250,000, or $500,000 on a joint return, set in 1997.
• The net investment income tax thresholds, section 1411(b): $250,000 joint, $200,000 in most other cases.
• The additional Medicare tax thresholds, section 3101(b)(2): the same pair.
• The base and adjusted base amounts that govern how much of a Social Security benefit is taxable, section 86(c): $25,000 and $32,000, $34,000 and $44,000.
• The exclusion for employer-provided group-term life insurance, section 79(a)(1): $50,000.

These do not creep for a year and then catch up. They creep permanently and compound, because the figure is fixed in nominal dollars and every year of inflation moves more taxpayers past it. That is a materially different thing from what an indexed provision does, and it is why the honest answer to "has bracket creep been solved?" is that it has been solved for the rate tables and not for the Code as a whole. The Social Security base amounts are the clearest case: they were set in the 1980s and 1990s, and the share of retirees whose benefits are partly taxable has risen ever since without a single vote. Each of those provisions states its own non-indexing on its own page; the machinery that indexes everything else is covered on the page for the tax inflation adjustment.

A lag remains even where indexing applies, and it is statutory rather than administrative. Section 1(f)(6)(B) defines the index for a year as "the average of the C-CPI-U as of the close of the 12-month period ending on August 31 of such calendar year," and section 1(f)(1) requires the Secretary to publish the next year's tables by December 15. So the adjustment applied to a tax year is computed from prices measured over a window that closed sixteen months before that year ends. In a period of stable inflation this is immaterial. In a year of rapid acceleration a taxpayer is compensated for last year's inflation while living through this year's, and the shortfall is only made up the following year. The reverse holds when inflation falls sharply: the adjustment overshoots.

A second-order leak, which is a matter of index choice rather than of the mechanism. The index used for these adjustments changed in 2018, and the current one rises more slowly than the one it replaced. The Congressional Research Service puts the gap concretely: from December 2015 to December 2025 the chained index rose 33 percent while the older measure rose 37 percent. Slower indexing means thresholds rise by smaller amounts, which means more taxpayers cross them, which is bracket creep by a different route. The published tax bracket page names the index and the rounding convention; what belongs here is the consequence, which is that indexing removes most of the effect rather than all of it.

One more mechanical detail that always runs one way. Section 1(f)(7)(A) rounds an increase to the next lowest multiple rather than to the nearest one, so the adjustment systematically slightly under-compensates, every year, in every provision that rule reaches. The published tax bracket page carries the convention itself; the point that belongs here is the direction. The amount is trivial in any single year and it never reverses.

The effect is easiest to see by holding purchasing power constant and letting only the dollar figures move.

A hypothetical with an unindexed threshold, because that is where the effect is now. Suppose Ottoline's other income plus half her Social Security benefits gives her provisional income of $24,000, just under the $25,000 base amount in section 86(c)(1), so none of her benefits are taxable. Over the next few years inflation runs at 3 percent a year, her benefits rise with the annual cost-of-living adjustment and her pension is indexed too, so by the fourth year her provisional income is roughly $27,000 in nominal terms while buying almost exactly what $24,000 bought before. The $25,000 threshold has not moved, because section 86 contains no inflation adjustment. She has crossed it. Part of her benefits is now taxable income, and she is worse off in real terms than she was, having become better off in no respect at all.

Extend that to its conclusion and you have the durable point about unindexed figures: the threshold is not merely lagging, it is falling in real terms every year and will keep falling until Congress changes the number. At 3 percent inflation, a fixed nominal threshold loses roughly a quarter of its real value in a decade.

The same story with an indexed threshold, for contrast. Take a rate-band boundary instead. If inflation runs at 3 percent and the boundary is adjusted by approximately that, a taxpayer whose income also rose 3 percent stays in the same position on the schedule and owes the same real tax. The creep here is confined to three residues: the measurement lag, because the adjustment reflects prices through the previous August 31; the rounding, which always goes down; and the index choice, which rises more slowly than the older measure. Small, one-off or slowly accumulating, and nothing like the unindexed case.

What a person can actually do about it. Very little about the mechanism, and something about the exposure. The levers that exist all involve knowing which specific threshold is in play and managing income against it: the timing of a retirement account withdrawal or a Roth conversion where provisional income is near a base amount, the realization of capital gains where net investment income tax thresholds are near, and awareness that a $3,000 capital-loss deduction absorbs less of a real loss every year. The general observation that inflation raises tax is not actionable. The specific observation that a particular unindexed threshold is approaching usually is.

Bracket creep is a defect rather than a product, so what follows is what the current arrangement handles well and where it leaves the problem in place.

What has genuinely been fixed

• The rate bands and the standard deduction are adjusted every year without Congress voting, so the largest version of the problem is gone.
• The adjustment is formula-driven and published in advance, so the following year's figures are knowable rather than discretionary.
• Because rates themselves are untouched by the adjustment and only the boundaries move, indexing cannot be used as cover for a rate change.
• The Congressional Research Service counts more than fifty indexed tax items, so the coverage is broad even though it is not complete.

What remains, and why it is getting worse rather than better

• Several significant provisions carry no adjustment at all, so their thresholds fall in real terms every year and compound indefinitely.
• The Social Security base amounts are the most consequential of those, and they reach ordinary retirees rather than high earners.
• The index used since 2018 rises more slowly than the measure it replaced, so even indexed thresholds drift behind a household's own experience of prices.
• The measurement window closes on August 31 of the previous year, so a year of accelerating inflation is compensated late.
• Rounding is always downward, which is a small permanent shortfall rather than a symmetrical rounding rule.
• Because none of this requires legislation, the increase is invisible, and an invisible tax increase is harder to argue about than a voted one.