Consumer Price Index, Year over Year

What the number is, and what it is not

The consumer price index is a fixed basket of goods and services priced every month across US cities. The headline figure quoted everywhere is the twelve-month change in that basket. It answers a narrow question: how much more does the same basket cost than it did a year ago. It does not answer how much prices rose last month, and it does not answer what any individual household experienced, because no household buys the basket.

The distinction between the twelve-month rate and the one-month rate matters more than it sounds. A year-over-year reading is an average of twelve monthly changes, so a single large month drops out of the calculation twelve months later. That mechanical roll-off is why inflation can appear to fall sharply without any prices actually falling: the comparison base moved. In mid-2022 the monthly changes were running near one percent; a year later the same monthly changes produced a much lower annual figure, and the difference was arithmetic rather than economic.

Shelter is the other thing to know before reading any CPI print. It is roughly a third of the index, and it is measured by asking tenants what they pay and by estimating what an owner would pay to rent their own home. Those answers move slowly and lag market rents by a year or more, because leases turn over slowly. So when market rents surged in 2021, the CPI shelter component did not peak until 2023, and it kept the headline elevated long after the surge had passed.

Why the Federal Reserve watches a different index

The Fed's stated target is two percent on personal consumption expenditures, not on CPI. The two differ in construction and in weight: PCE covers a broader set of goods, gives less weight to shelter, and more to health care, where prices are administratively set rather than market-clearing. Over long spans PCE inflation runs a few tenths below CPI, which is why a CPI reading of three percent is compatible with a PCE reading closer to two and a half.

This page publishes CPI because it is the number households and contracts actually reference. Social Security adjustments, many wage agreements and most inflation-linked bonds outside the TIPS market are tied to it. If you want to know what the Fed is looking at, CPI is a proxy; if you want to know what a household budget is facing, CPI is the closer measure.

What the long record shows

The series starts in 1948. Its median twelve-month change is a little under three percent, which is a useful anchor: half of the post-war record sits below that, half above. The top tenth begins around seven percent, a level reached in the 1970s, briefly in 1980, and again in 2022. The maximum is 14.6 percent in March 1980, at the end of the second oil shock and the start of the Volcker tightening that broke it.

The minimum is minus three percent in August 1949, during the post-war demobilisation, when price controls were lifted and military demand collapsed at the same time. Deflation of that kind has not appeared since. The closest modern approach was 2009, when the twelve-month rate fell to minus two percent for a few months without anything like the 1949 contraction in output.

Two episodes are worth comparing because they show how differently inflation can end. In 1980 it ended with a policy rate near twenty percent and two recessions. In 2022 it ended without a recession, as supply chains normalised and the policy rate rose to five and a half percent. The second outcome was not obvious at the time and is still not fully explained.

Reading the percentile with care

The percentile on this page compares the current reading with every month since 1948. That is a relative measure, and relative measures can mislead when the distribution has changed. A reading in the 67th percentile sounds high, and against the post-war record it is: two thirds of months were lower. But the post-war record includes a decade of double-digit inflation that no longer describes the policy regime. Against the 1990 to 2019 sample alone, the same reading would sit near the 95th percentile.

We publish the full-history percentile because it is the only one that does not require us to choose a sample. The distribution table below the chart lets you see where the cut points actually fall, which is more informative than a single number.

How we calculate it

We take CPIAUCSL, the seasonally adjusted all-items index for all urban consumers, and divide each month's level by the level twelve months earlier. The result is rounded to two decimals. Nothing is smoothed and nothing is annualised: the number you see is the plain twelve-month change that the Bureau of Labor Statistics publishes as its headline.

The index level itself is also shown in the summary line, because it makes the base effect visible. A twelve-month change of three percent on an index of 336 is a different dollar amount than the same percentage on an index of 200, and readers comparing decades sometimes forget that.

The source is a Bureau of Labor Statistics release in the public domain, distributed through FRED. The percentage change is computed here rather than copied from a licensed data vendor.

Questions

Related

• 10-year breakeven inflation
• Real 10-year yield
• M2 money supply growth
• Unemployment rate
• Plans and full-data access