Inflation calculator
One of the years shown is not a full-year average. The 2026 figure averages only the months published so far and will move as the rest are released. The 2025 figure averages eleven months: the October 2025 index was never published, so that year has no twelve-month average and never will.
Based on the CPI-U for all urban consumers, US city average, all items. It measures a national basket, so it will not match the price of any particular thing — housing, tuition and electronics have all moved very differently from the average.
Inflation is measured by comparing a price index between two years. Multiply the amount by the later index and divide by the earlier one. Using the US CPI-U, $100 in 1990 has the buying power of about $250 in 2026; a total rise of roughly 150%, or an average of about 2.6% a year.
How to use this calculator
The average annual rate is a compound rate, not the total divided by the years. Over the 36 years from 1990, prices rose 153.4% in total, which is 2.616% a year compounded. Dividing would give 4.26%, which is not a rate the money ever grew at: compounded across the same 36 years it turns $100 into $449 rather than the $253 that actually happened. What makes the mistake stick is that 4.26% looks like a real year — 1991 came in at 4.21% and 2023 at 4.12% — so the artefact of the division is easy to read as a fact about prices. The compound rate is the one that lets you compare periods of different lengths: 1970 to 2026 is 753.6% in total but only 3.903% a year, and 2020 to 2026 is 28.0% in total and 4.195% a year, so the short recent run was the faster of the two despite being much the smaller number.
Two of the years in this series are not full years
The 2026 figure averages the months published so far and moves as the rest arrive. The 2025 figure averages eleven months and always will, because there is no October 2025 index to average. The Bureau of Labor Statistics states that it "did not collect CPI data from October 1, 2025, through November 12, 2025" during the lapse in appropriations, and that the missing data affected the October and November 2025 indexes. Price data for a month that has passed cannot be collected afterwards. Any calculation touching either year is therefore an estimate rather than a settled comparison, and the tool says so on screen when you select one.
A national average is not your basket
CPI-U tracks what urban consumers as a group buy, weighted by how much of it they buy. Nobody spends in those proportions. A household paying rent in a tight market, or one with a child in higher education, or one whose spending is dominated by medical care, has experienced a different rate from the index for years at a time. The index is the right instrument for restating a sum of money in comparable terms and the wrong one for arguing about the price of any particular thing.
What people use it for
- Restating an old salary or a childhood price in today’s money
- Checking whether a run of pay rises kept pace with the index
- Making sense of a cost quoted in a book written decades ago
- Getting the average annual rate between two specific years
- Uprating a long-running contract or budget line so it holds its value
- Separating a large total change from a modest annual one
Questions
CPI-U: the Consumer Price Index for All Urban Consumers, US city average, all items, not seasonally adjusted. It is the series the BLS publishes monthly and the one most commonly quoted.
$253.39 in 2026 terms, using the annual averages. That is a 153.4% total rise, or 2.616% a year compounded.
To 1913, the first year of the series. $100 then is $3,345.25 now, an average of 3.155% a year, but comparisons across a century are indicative at best because the basket has changed beyond recognition.
Because it compounds. A 153.4% rise over 36 years is 2.616% a year, not the 4.26% that dividing gives. Dividing overstates any rising series, and the longer the span the worse it gets: over the six years from 2020 it returns 4.66% against a true 4.195%, and over the 36 from 1990 it returns 4.26% against 2.616%.
All of them. Each year here is the mean of that year’s monthly index values, so a comparison between two years is a comparison of two twelve-month averages and not of two specific dates.
No. The figure for this year averages the months released so far and will shift as the rest are published. The tool flags it when you select it.
Because the October 2025 index was never published. BLS did not collect CPI data between 1 October and 12 November 2025, and prices for a month that has passed cannot be collected later. The 2025 figure is the mean of the other eleven months and will stay that way.
Not by much, but it is not nothing, and the direction is unknowable. An eleven-month mean weights the months it does contain slightly differently from a twelve-month one. Treat any comparison involving 2025 as carrying a small unquantified error.
Because CPI averages a whole basket. Housing, healthcare and university fees have risen much faster than the index; clothing and electronics have risen far more slowly or fallen.
The 1970s, on this series, by a wide margin. 2020 to 2026 averages 4.195% a year. The 56 years from 1970 average 3.903% a year overall, and the early part of that span ran far hotter than either.
Yes. Put the old salary in the earlier year and compare the figure it returns against what you are paid now. Anything below it is a real-terms cut regardless of the headline increase.
Yes, and this page uses one of them. The BLS also publishes the CPI for Urban Wage Earners and Clerical Workers (CPI-W) and, since August 2002, the Chained CPI for All Urban Consumers (C-CPI-U). They rest on the same price collection but cover different populations or aggregate differently, so they do not report identical rates. A figure quoted elsewhere may not be from this series.
Only if the contract names an annual average of CPI-U, which is what this page computes. Escalation clauses more often name a specific month or a seasonally adjusted series, and either of those will give a different number from the one here.
Because the series begins there. Nothing before it exists to compare against in this index, and the older reconstructions that do exist are built on very different methods.