Is the CPI Accurate?

Somewhere in America right now, a government employee is weighing a bag of chips in his hand. This is not a cry for help. A colleague is reading a cereal box more carefully than anyone has ever read one. He isn’t on a diet either. He’s checking whether the box got smaller, which is a real part of his job. Across town, another colleague is pricing a funeral. The job has quite a range, and it’s how the country measures inflation.

Every month the Bureau of Labor Statistics sends people like them into supermarkets, dry cleaners, car dealerships and doctors’ offices to write down about 80,000 prices. Then it boils them into one number, and the internet explains that the number is a lie.

Is the CPI accurate? Mostly, for the question it’s built to answer: how prices change, on average, for the average city household. Most of the famous complaints about it are myths. But the average household isn’t you, the index has a few real soft spots, and there’s a second question it was never built to answer: what’s happening to the money itself.

Already know how the index is built? Skip to the myths.

How is the CPI calculated?

The index was born in a shipyard. During World War I, prices in the towns building ships for the war rose so fast that wages needed cost-of-living adjustments, and nobody could say what the cost of living was. So the bureau studied family spending in 92 industrial centers from 1917 to 1919, published indexes for 32 cities in 1919, and launched a national index in 1921. It priced six groups: food, clothing, rent, fuels, house furnishings and “miscellaneous,” which has been carrying the heavy boxes ever since.

A century later, a close relative of that index decides how much every Social Security check grows each year. That is why people care so much whether it’s right, and why your grandmother has opinions about seasonal adjustment.

Call her the Collector. She isn’t a real person. She’s a composite of the people the Bureau of Labor Statistics sends out to record prices, and she’s here to make the machine visible. She wears sensible shoes and carries a list of specific items in specific stores. Not “bread,” but a particular loaf, in a particular size, on a particular shelf, under a particular flickering light in a particular city. Next month she’ll come back and price the same loaf. If it’s gone, she finds the closest substitute and notes exactly how it differs, like a detective whose only suspect is rye.

That’s half the index: what things cost. The other half is how much each thing matters to your wallet. A 10% rise in the price of salt and a 10% rise in rent are not the same event, unless you eat salt by the shovelful. So the bureau weights every item by how much households actually spend on it, using surveys in which thousands of families report their spending and keep diaries of every purchase, down to the gum at the checkout.

Into the weeds: the spending surveys

The weights come from the Consumer Expenditure Survey: more than 20,000 households report their spending each quarter, and about 12,000 more keep two-week diaries of every purchase.

Multiply the price changes by the weights, add them up, and you have the CPI. The “all items” index stood at 334.980 in August 2026, against a base of 100 in 1982 to 1984, according to the bureau’s release. In plain English: the average basket costs about three and a third times what it did when E.T. was in theaters. E.T. wanted to go home. At these prices, he’d have to rent.

So much for the machine. Now the rumors.

Does the CPI swap steak for hamburger?

The story goes like this. When steak gets expensive, the statisticians assume you switch to hamburger. When hamburger gets expensive, you switch to beans. When beans get expensive, you switch to a glass of tap water and a positive attitude, and the bureau announces that dinner costs no more than it did. It would be an outrage, if it happened.

It doesn’t. Since January 1999 the bureau has used a formula that allows for switching within a narrow category, such as one kind of ground beef for another in the same city. If one brand of ground beef gets more expensive, people buy the other brand. That’s the whole theory.

Steak is a different category. In the bureau’s own words, “hamburger and steak are in different CPI item categories, so no substitution between them is built into” the main index.

The steak stays steak. It’s priced as steak, every month, by the Collector, who would be very surprised to hear she’d been writing down hamburger and would like a word with whoever started the rumor.

(There is a separate index, the chained CPI, introduced in 2002, that does allow broader switching between categories. It runs a little lower and isn’t the headline number.)

What is hedonic adjustment, and does it hide inflation?

“Hedonic” is a word that sounds like something happening at a party you weren’t invited to. It means adjusting a price for a change in quality. If a television costs the same as last year but has a much better screen, the bureau counts part of that as a price drop, because you’re getting more for the money.

Critics call this a way to make inflation vanish. The bureau’s answer is unusually specific. For shelter and clothing, where quality adjustments are used, they usually raise the index. For other durable goods, the items adjusted this way make up only about one percent of the CPI. Their combined effect has been to raise the all-items index by about five thousandths of a percentage point a year.

Five thousandths. If that’s a conspiracy, it’s the least ambitious one in history. A plotter who kept it up for an entire forty-year career would move the index a fifth of a point and retire in disgrace, with a farewell sheet cake the bureau would adjust for quality.

Why does the CPI use rent instead of house prices?

Now the complaint with real teeth. The biggest single item in the American CPI is something nobody pays. It’s called owners’ equivalent rent: the rent a homeowner would pay to rent their own house. At the end of 2025 it made up 26.2% of the entire index, according to the bureau, more than food, more than gasoline, more than health insurance. Of every $100 the index tracks, about $26 is a rent check nobody writes, to a landlord who is also you, for a house you already own.

Try this at home. Walk out your front door, turn around, knock, and ask yourself what you’d charge you. Be firm but fair. You’ve been a quiet tenant, apart from the drum kit. Now raise your rent and watch yourself take it personally.

The logic is better than it sounds. A house is two things at once: a place to live, which is spending, and an asset that may rise in value, which is investing. The CPI wants only the living-in part, and the cleanest way to price that is to ask what the house would rent for. So the bureau surveys rents and uses them to estimate what owners are, in effect, “paying” themselves. It’s a defensible method, used by many countries, and when it was adopted in 1983 it actually raised the shelter index.

But two side effects need to be said plainly. House prices themselves aren’t in the CPI at all. And rents move slowly, because leases last a year, so the index catches a housing boom late and lets go of it late, like a smoke alarm on a six-month delay.

If you bought a house recently, the CPI isn’t measuring what happened to you. It’s measuring what would have happened to you if you had rented it from yourself. Picture slipping a rent check under your own bedroom door every month, then complaining to yourself about the leaky faucet.

Was the October 2025 CPI ever published?

No. This is where the real weaknesses live, in the unglamorous business of sending people out to write prices down.

Then came the shutdown. From October 1 to November 12, 2025, the federal government was closed, and nobody collected prices. Not one loaf. The bureau did not issue an October 2025 CPI at all, and when later months needed an October figure to build on, it carried September’s prices forward. In effect, October was canceled and everyone was asked to use September again. The rents were worse: October’s were carried forward from April.

Which is how an official price index came to record a month in which American rents did not change. This is real. The bureau published a page of questions and answers about it. Somewhere a landlord read that and had to sit down.

The bureau had also thinned its sample earlier in 2025, closing collection in three cities and cutting about 15% elsewhere, and it said so openly. Nobody cooked the books. The data got thinner, and the label said so. But “the CPI is a precise number” and “the October rent figure came from April” can’t both be true, and only one of them is.

Into the weeds: the 2025 collection cuts

The bureau suspended collection entirely in Lincoln, Nebraska and Provo, Utah, in April, and in Buffalo, New York, in June, and cut roughly 15% of the sample in the other 72 areas. Its own testing found the effect on the national figure tiny, less than a hundredth of a percentage point on the 12-month rate. It also said, honestly, that it had not measured the effect on city-level and individual item indexes.

Does the CPI understate inflation?

Brace for the twist that annoys everybody. The most serious official review of the CPI concluded that it overstated inflation.

In 1996 a commission of economists appointed by the US Senate, known as the Boskin Commission, estimated that the index was running about 1.1 percentage points a year too high. The main reasons: it didn’t account well for people switching stores and products, for improving quality, or for new products. Several of the changes critics now complain about, including the 1999 substitution formula, came out of that period.

Britain kept a working example of an index that ran high: its old Retail Prices Index, whose averaging formula nudges it upward. It was stripped of its official status in 2013 and is being brought into line with the newer index. So “the CPI is rigged low” isn’t the consensus of the people who study it for a living. Every index makes choices. Choices push both ways.

Which suits us fine. The case for sound money never needed the CPI to be rigged. It needs only what the CPI leaves out, and that list is coming up.

Into the weeds: the British formula gap

About 0.9 percentage points of the RPI’s typical 1.15-point annual gap over the newer CPI comes from its averaging formula alone, according to the UK’s statistical record. The RPI lost its status as an official national statistic in 2013. It is still used in some contracts and government bonds, and it will be brought into line with the newer index from 2030.

Why does the CPI not match my own bills?

Because you aren’t the average urban household. Nobody is. The average household is a statistical creature. It rents and owns at the same time, drives a little and flies a little. It owns a small slice of a boat. It spends a precise fraction of its income on pet food whether or not it has ever met a dog.

Take restaurants. Since the 1982 to 1984 base period, the all-items CPI has risen to about 335. The index for food away from home has risen to 397.868, according to the bureau’s August 2026 release. Restaurant prices have roughly quadrupled while the average basket has a bit more than tripled. Put a number on it: a $20 restaurant tab from the early 1980s now runs about $80. If restaurants had only kept up with the average basket, it would be about $67. The extra $13 is roughly the tip, so your family has been tipping twice since the early ’80s: once on the table and once on the menu.

If you eat out a lot, your inflation has run above the headline for forty years. If you never eat out, it has run below it in that category. Both of you are correctly described by the same index, which is describing neither of you.

Housing works the same way. A renter whose lease just reset, a homeowner with a paid-off mortgage and a first-time buyer in 2026 live through three different inflations. The CPI averages them into one, because that’s its job. It’s a weather report for the whole country, and you live on one street. “Mostly sunny nationwide” is no comfort when the hail is denting your car.

How does your country measure inflation?

Every country argues about its own index, and the arguments are mostly about the same thing: what to do with the house. Here’s how some of the countries this site serves handle it.

Sources: BLS; UK index history; European Central Bank, July 8, 2021; Statistics Canada; ABS and ABS, July 23, 2025.

Now put one family in each country, with the same house and the same mortgage, and let interest rates jump. The Canadian index feels it as mortgages renew. The British RPI feels some of it. The American index feels it slowly, through rents. The euro-area index officially doesn’t feel it at all.

Same house. Same family. Four thermometers, four different fevers.

What does the CPI not measure?

The CPI measures the prices of things people buy. It doesn’t measure what’s happening to the currency itself, and it was never meant to.

Houses, stocks and other assets are left out by design. So when new money flows into asset prices rather than store prices, the index doesn’t see it. A family can watch the CPI report modest inflation while the house it hopes to buy, and the retirement it hopes to fund, drift out of reach. The arithmetic is fine. The question is just smaller than the one you’re asking.

The Collector can tell you, very precisely, how many dollars a loaf costs. She has no way to tell you what a dollar is worth, because her clipboard, her pencil and her paycheck are all denominated in the thing she would be measuring. It’s like asking a ruler how long it is.

That’s why this site always shows two measures side by side: the CPI, for what has happened to the shopping, and gold, for what has happened to the dollar against something nobody can print. The fair objection is that gold swings far more than the CPI from month to month, which makes it a poor thermometer for a single year. Over decades, though, it’s the one ruler on the shelf that no government issues and no agency samples. When the two measures diverge, the gap is the story. Our dollar tool runs both from any year you choose.

So, is the CPI accurate?

The Collector, if she existed, would say yes. She priced the same loaf in the same store, and the loaf cost what she wrote down. Her steak was steak. Her quality adjustments, in the categories that use them, raised the index more often than they lowered it.

Buy her a coffee and she’d admit the rest. Her list has no line for your house, your restaurant habit or your city. One October, through no fault of hers, it didn’t go out at all. And every price on it is written in dollars, which means the one thing she can never price is the dollar. For that you need something nobody can print.

The coffee, by the way, is in the index. Food away from home. It has gone up faster than average. She wrote that down too, right under the tip.

Sources

• How prices and weights are collected (about 80,000 prices a month, the Consumer Expenditure Survey): Bureau of Labor Statistics, CPI questions and answers.
• Substitution, hedonic adjustment and rental equivalence: BLS, Common misconceptions about the CPI.
• Owners' equivalent rent and its December 2025 weight: BLS, Owners' equivalent rent and rent.
• August 2026 index levels: BLS, Consumer Price Index news release, September 11, 2026.
• The October 2025 gap: BLS, 2025 federal government shutdown impact on the CPI.
• The 2025 collection reductions: BLS notice, July 29, 2025.
• The chained CPI: BLS, Chained CPI questions and answers.
• The 1996 commission: Boskin Commission, with the report’s later review by the National Bureau of Economic Research.
• The other countries: see the sources under the table above.

Figures read October 3, 2026. Index levels are from the August 2026 release; the restaurant comparison is our arithmetic on the bureau’s published index levels. Added October 25, 2026: the $20 tab ($20 × 397.868 ÷ 100 ≈ $80, against $20 × 334.980 ÷ 100 ≈ $67) and $26 of every $100 (owners’ equivalent rent’s 26.2% weight).

Questions people actually ask

Is the CPI accurate?

Broadly, yes, for what it measures: the average price change for a fixed basket bought by urban US households. The Bureau of Labor Statistics records about 80,000 prices a month. Its weak spots are thinner data in some periods, slow treatment of housing, and the fact that no real household spends like the average one.

How is the CPI calculated?

The Bureau of Labor Statistics records the prices of the same specific items in the same stores each month, about 80,000 prices in all, then weights each item by how much households spend on it, using a survey of more than 20,000 households a quarter plus spending diaries. The weighted price changes are combined into one index.

Does the CPI assume people switch from steak to hamburger?

No. Since 1999 the CPI has allowed for substitution only within narrow categories, such as one kind of ground beef for another in the same city. The Bureau of Labor Statistics says hamburger and steak are in different item categories, so no substitution between them is built into the main index.

What is hedonic adjustment in the CPI?

Hedonic adjustment is a way of separating a price change from a quality change, such as a better screen on a television that costs the same. The Bureau of Labor Statistics says that for shelter and clothing it usually raises the index, and that for other goods it affects about one percent of the CPI.

What is owners' equivalent rent?

Owners' equivalent rent is the rent homeowners would pay to rent their own homes. The CPI uses it instead of house prices to measure the cost of living in a home you own. At the end of 2025 it was 26.2% of the CPI, the largest single item, according to the Bureau of Labor Statistics.

Why was there no October 2025 CPI?

Prices were not collected during the federal government shutdown from October 1 to November 12, 2025, so the Bureau of Labor Statistics did not publish an October 2025 CPI. It carried September prices forward into October, and carried October rents forward from April 2025.

Does the CPI overstate or understate inflation?

Both claims have been made. A 1996 commission of economists appointed by the US Senate, the Boskin Commission, estimated the CPI overstated inflation by about 1.1 percentage points a year, and several methods were changed after it. Critics today argue the opposite. Most specific claims of understatement, such as steak-to-hamburger substitution, do not match how the index works.

What is the difference between CPI, CPIH and RPI in the UK?

The CPI is the measure the Bank of England targets at 2%. CPIH, the lead measure since 2017, adds owner-occupiers' housing costs using equivalent rent. The older RPI includes mortgage interest and uses a formula that runs higher; it lost its status as an official national statistic in 2013 and will be aligned with CPIH from 2030.

Does euro-area inflation include house prices?

No. The Harmonised Index of Consumer Prices, used for Germany, France, Austria, the Netherlands, Ireland and the rest of the euro area, leaves out owner-occupied housing costs. The European Central Bank said in 2021 that including them would better represent household inflation, and described that as a multi-year project.

Why does my inflation feel higher than the CPI?

Because the CPI measures an average household, and your spending is not average. Restaurant prices have roughly quadrupled since 1982 to 1984 while the overall index has a bit more than tripled, so frequent diners have faced higher inflation. House prices are also left out; the CPI counts owners' equivalent rent instead.

What is the chained CPI?

The chained CPI is a separate index, introduced in 2002, that allows for consumers switching between categories of goods when relative prices change. It is published alongside the main CPI and generally runs a little lower. It is not the headline inflation number.

When did the CPI start?

The Bureau of Labor Statistics began publishing price indexes for 32 cities in 1919, after rapid World War I price increases, especially in shipbuilding centers, made cost-of-living adjustments to wages necessary. Regular publication of a national index, the U.S. city average, began in 1921.