What is the difference between PCE and CPI?

CPI, from the Bureau of Labor Statistics, measures what urban households pay out of pocket for a fixed basket. PCE, from the Bureau of Economic Analysis, covers everything consumed by or on behalf of households, including health care paid by employers and government, gives shelter much less weight and adjusts for substitution between goods.

Why does the Fed use PCE instead of CPI?

Because PCE covers a broader share of spending, updates its weights as people switch between products and can be revised to stay consistent. The Fed made it its main gauge in 2000, and its 2% target, adopted in January 2012, is defined in PCE terms.

Is PCE inflation always lower than CPI?

Usually, by a few tenths of a point, but not always. In August 2026 core PCE rose 3.0% from a year earlier while core CPI rose 2.4%, mainly because slowing shelter inflation weighs about a third of CPI but only about 15% of PCE.

Guide · Markets

PCE vs CPI: Why the Fed Watches a Different Inflation Number

The Fed targets 2% inflation on the PCE price index, not the better-known CPI, because PCE covers more spending and adjusts for substitution. Here is how the two differ in scope, weights and formula, why core PCE ran at 3.0% in August 2026 while core CPI was 2.4%, and what to look for when September's PCE arrives on October 29.

Oct 27, 2026

In August 2026 the United States had two official inflation rates for underlying prices, and they did not agree. On September 11 the Bureau of Labor Statistics reported that the consumer price index rose 3.4% in the year to August, and 2.4% excluding food and energy. On September 30 the Bureau of Economic Analysis published the personal consumption expenditures price index for the same month: 3.4% overall as well, but 3.0% excluding food and energy. The headlines matched; the core measures were 0.6 percentage point apart. And the higher of the two, core PCE, is the one the Federal Reserve uses to judge whether it is meeting its 2% goal.

The two indexes start from different questions. CPI asks what urban households pay out of their own pockets for a fixed basket of goods and services, with weights taken from surveys of what those households buy. PCE asks what is consumed by and on behalf of all households, so it also counts spending that someone else pays for: the employer's share of health insurance, Medicare and Medicaid payments to doctors and hospitals, and the services of non-profit institutions. CPI comes out first, around the middle of the following month; PCE follows about two weeks later inside the monthly personal income and outlays report.

Those definitions produce three practical differences. The first is weights. Shelter, which is mostly rent and the imputed rent of owner-occupied homes, makes up about a third of CPI but only about 15% of PCE, while health care carries more than twice as much weight in PCE as in CPI because PCE includes what insurers and government programmes pay. The second is the formula. CPI updates its basket weights once a year, so when people switch from an expensive item to a cheaper one, the index catches up only slowly; PCE is a chained index that reflects that substitution as it happens, which is one reason it has usually run a few tenths of a point below CPI. The third is revisions. CPI, before seasonal adjustment, is never revised, while PCE is revised as better source data arrive; the August 2026 report itself carried BEA's annual update, which revised figures back to January 2021.

The Fed switched its main inflation gauge from CPI to PCE in 2000, and when it adopted a formal 2% target in January 2012 it defined that target in PCE terms. The reasons it gives are the ones above: PCE covers more of what the economy actually consumes, it adapts to changes in spending patterns, and its history can be revised to stay consistent. That is why a Fed chair asked about inflation will usually quote core PCE, even on a day when the news is dominated by a CPI release.

The usual ordering, CPI above PCE, flipped in 2026, and the weights explain most of it. With shelter inflation in CPI down to 3.0% a year in August, the category that dominates CPI was pulling its core rate lower. In PCE, where shelter matters less, faster-rising services outside housing, health care among them, carry more weight, so the same price changes added up to a higher core figure. September showed the gap had not closed: on October 14 the BLS reported headline CPI up 3.7% and core CPI up 2.6% from a year earlier. Anyone comparing a CPI number with the Fed's 2% target is, in 2026, measuring with a ruler that reads low.

The release itself tends to move markets less than CPI does, and for a mechanical reason. By the time PCE is published, economists have already translated the CPI and producer price reports into a PCE estimate, using the components the two share and the producer-price series, such as airfares, portfolio management fees and medical services, that feed into PCE directly. Surprises are therefore usually small, and they often come from revisions to earlier months rather than from the new one. September's report is due at 8:30 a.m. Eastern on Thursday, October 29, the morning after the Fed's decision and alongside the first estimate of third-quarter GDP. It cannot change the October vote, but it is the inflation number the committee will have in front of it when it meets again in December, and the one its statement means when it says inflation remains elevated.