Reviewed 12 August 2026 · Sourced from the BLS Handbook of Methods, the BLS CPI news release, the SSA and the Internal Revenue Code
The Consumer Price Index is the Bureau of Labor Statistics' measuring instrument for inflation: what a fixed basket of goods and services costs a specific reference population of urban consumers, published as an index level against a 1982–84 average of 100 and reported as a percent change.
It exists because too many dollar amounts needed to move with prices automatically. Benefit payments, tax brackets, Treasury bonds, union contracts and commercial leases all needed one number, published on a fixed schedule, that neither side of the transaction controls. That is what the CPI is for — and it is why the arguments about its methodology are not academic. They are arguments about money.
- The CPI is an instrument, not the thing it measures. It has published settings: one reference population, one basket, one formula, one base period — 243 item strata across 32 index areas, built from about 80,000 prices a month collected from roughly 23,000 establishments in 75 urban areas.
- CPI-U is the headline. CPI-W sets the check. CPI-U covers “over 90 percent” of the population and CPI-W “approximately 30 percent” — and Social Security and SSI cost-of-living adjustments are computed from CPI-W on third-quarter averages, not from the headline.
- Federal income tax parameters run on a third index. They are indexed to the chained C-CPI-U measured against the CPI for calendar year 2016 — 26 U.S.C. § 1(f)(3) and (f)(6).
- Owners' equivalent rent is 26.204% of the CPI-U, and it is not a survey of homeowner guesses. BLS states plainly that “owner-occupied units are not priced in the CPI Housing Survey.” The monthly movement comes from the tenant rental sample, with utilities stripped out.
- Geometric means arrived with January 1999 data, covering about 61 percent of the CPI-U at the time, and BLS published its own estimate that the switch would “reduce the annual rate of increase in the CPI-U by about 0.2 percentage point per year.”
- BLS declines to call the CPI a cost-of-living index. In its own words it “differs in important ways from a complete cost-of-living measure” and is “sometimes termed a conditional cost-of-living index.”
- July 2026: index level 333.918 (1982–84=100), all items up 3.4% over twelve months, all items less food and energy up 2.5%, one-month change seasonally adjusted +0.1% — released 12 August 2026.
What the CPI actually is
Three numbers land in the same week. The headline says prices are up 3.4% from a year ago. The benefit letter says this year's increase is 2.8%. The lease renewal says rent is going up 9%. Somebody is lying, or somebody made a mistake — that is the natural reaction, and it is wrong. All three are correct, and they disagree because they are not the same measurement.
The Consumer Price Index is not inflation. Inflation is the phenomenon: a sustained rise in the general price level. The CPI is one instrument for measuring it, published monthly by the U.S. Bureau of Labor Statistics. Every instrument has settings, and the CPI's are public. Once you know them, those three numbers stop conflicting.
The CPI tracks what a fixed basket of goods and services costs a specific reference population of urban consumers, expressed as an index level against a 1982–84 average of 100, and reported to the public as a percent change.
Every phrase there is doing work. Fixed basket: the composition is held still on purpose, so the change you see is a change in prices, not in what people bought. Specific reference population: somebody decided whose spending counts, and the answer is not “everybody.” Index level: the raw output is a number like 333.918, meaningless alone and useful only against another one.
The machinery is larger than most people assume. BLS sorts spending into more than 200 categories in eight major groups and calculates 7,776 basic indexes — 243 item strata across 32 index areas, from prices collected in 75 urban areas at approximately 23,000 retail establishments. In BLS's words, “we record the prices of about 80,000 items each month, representing a scientifically selected sample of the prices paid by consumers.” Far more than any private effort collects, and a rounding error against the transactions in a U.S. month. The tension between those two facts is where every honest criticism of the CPI starts.
CPI-U, CPI-W, and the index that sets your check
There is no such thing as “the” CPI. BLS publishes three indexes for urban consumers, and different dollars are legally wired to different ones. For anybody receiving a federal payment, this is the most consequential section on the page.
| Index | Reference population | What it governs |
|---|---|---|
| CPI-U All Urban Consumers | “Over 90 percent of the U.S. population” | The headline in every news story. TIPS indexation. Most private escalation clauses. |
| CPI-W Urban Wage Earners and Clerical Workers | A subset of the CPI-U population — “approximately 30 percent” | Social Security and SSI cost-of-living adjustments. |
| C-CPI-U Chained | Same population as CPI-U, different formula | Federal income tax parameters — brackets, the standard deduction and more. |
The benefit letter. SSA is explicit: “COLAs are based on increases in the Consumer Price Index for Urban Wage Earners and Clerical Workers (CPI-W).” And the comparison is not the twelve-month change in the latest month — it is quarterly averages: “A COLA effective for December of the current year is equal to the percentage increase (if any) in the CPI-W from the average for the third quarter of the current year to the average for the third quarter of the last year in which a COLA became effective.” The 2.8% for December 2025 benefits came out of that calculation, and the same increase reaches SSI: “Federal SSI payment levels will also increase by 2.8 percent effective for payments made for January 2026.” So when the COLA on your letter does not match the news, a different index, population and averaging window produced a different answer.
The tax brackets. Those move on the chained index: the statute sets the adjustment as the percentage by which “the C-CPI-U for the preceding calendar year” exceeds “the CPI for calendar year 2016.” BLS began publishing it with the July 2002 release, and it arrives unfinished — initial values are preliminary, take three interim revisions, and are final 10 to 12 months later. What makes it different from CPI-U is covered below.
The CPI is an urban index, and BLS states the exclusions flatly: “Not included in the CPI are the spending patterns of people living in rural nonmetropolitan areas, farm households, on military installations, religious communities, and in institutions, such as prisons and mental hospitals.”
Read that against who the index governs. If you live in a rural county, on a farm, on a base, or inside an institution, your spending is not in the basket at all — and the index still sets your benefit payment and your tax bracket. A published design boundary, not a scandal. But nobody mentions it in the headline.
How the basket gets built and weighted
Two entirely separate operations produce the CPI, and confusing them causes a lot of bad writing about it.
One survey decides what is in the basket and how much each thing counts — the Consumer Expenditure Survey. BLS: “This expenditure information from weekly diaries and quarterly interviews determines the relative importance, or weight, of the item categories in the CPI index structure.” Households keep diaries and sit for interviews about what they actually spent, and those answers become the weights.
A second operation collects prices. The sampled items are priced every month across the 75 urban areas. This side of the machine never asks what anyone bought, only what things cost.
The weights decide almost everything, because a category at 30% of spending moves the index thirty times as hard as one at 1% — which is why shelter dominates, as the next section shows. The full weight table lives on the inflation page. What matters here is that the mechanism has a lag in it.
The basket you are measured against is a couple of years old
BLS gives its own example: “There is a time lag between the expenditure survey and its use in the CPI. For example, CPI data in 2023 was based on data collected from the Consumer Expenditure Surveys (CE) for 2021.” Since January 2023 the weights are refreshed annually from a single year of expenditure data; before that, every two years from two years of data. Either way, the spending pattern the index holds fixed sits roughly two years behind the month being measured.
“Grocery prices jumped” and “the CPI jumped” are different statements. The effect on the index is the price change times the weight. A 10% move in a category carrying a 1% weight adds about a tenth of a point; the same 10% move in shelter adds roughly three and a half points. That is why the index can look calm during a period that felt anything but — arithmetic published in advance, not tampering.
How to read a CPI release
The raw output of the CPI is a level, and the level is close to useless on its own. BLS notes that “most CPI index series have a 1982-84=100 reference base” — set to 100 at the average of those three years. For July 2026 the CPI-U all items level was 333.918. Nobody quotes that, and nobody should. It exists so you can compute the change between any two months you like.
Percent change = ( Index later − Index earlier ) ÷ Index earlier × 100The same formula gives a twelve-month rate, a month-over-month rate, or a change across forty years. Only the two levels you pick change.
A lease says rent adjusts each August by the twelve-month change in the CPI-U, all items, U.S. city average, not seasonally adjusted. Current rent is $1,240 a month. The unadjusted index was 323.048 in July 2025 and 333.918 in July 2026.
Difference in the index: 333.918 − 323.048 = 10.870. Divide by the earlier level: 10.870 ÷ 323.048 = 0.03365, or 3.365%. Apply it: $1,240 × 0.03365 = $41.72.
New rent: $1,281.72 a month — $500.64 across twelve payments.
Now use the published headline instead. BLS reports the twelve-month change as 3.4%, rounded to one decimal, which gives $42.16 a month — $0.44 more, or $5.28 a year. Small here, not small on a large lease, and that is why escalation clauses are written against index levels rather than the rounded rate. The arithmetic in this box is ours, run on levels BLS published.
Headline and core, and why both exist
The July 2026 release reports all items up 3.4% and all items less food and energy up 2.5%. That second series is what everyone calls core. BLS does not use the word.
The rationale is real: food and energy prices swing on weather, harvests and geopolitics, hard enough to hide the trend in everything else, so stripping them out is an attempt to see the signal. The objection is equally real and not a gotcha: people buy food and gas. A household spending a large share of income on groceries and a commute lives in the headline, not in core. Neither series is honest alone. The inflation page covers the trimmed-mean and median alternatives that exist because core has its own blind spot.
Seasonally adjusted, and not
The release carries both, and the pairing is deliberate: the twelve-month change is published before seasonal adjustment, the one-month change (+0.1% for July 2026) after it. BLS's guidance splits the same way — users “interested in analyzing general price trends in the economy should use seasonally adjusted indexes,” but those using “the CPI in escalation agreements… should typically not.” Adjustment is not permanent either: “each January, the seasonal status of every index series is reevaluated,” with new factors in mid-February. So an adjusted monthly figure you read in March can look different a year later — one more reason contracts are written on the unadjusted series.
Owners' equivalent rent, the biggest piece and the most argued
Shelter is 35.625% of the CPI-U on the December 2025 weights, and owners' equivalent rent by itself is 26.204%. A quarter of the entire index is a dollar amount nobody is ever billed.
Why house prices are not in the index
BLS excludes home purchase prices for a conceptual reason: “the CPI program views owned housing units as capital (or investment) goods distinct from the shelter service they provide, and therefore not as consumption goods.” A house is an asset; the CPI measures consumption. So the index prices the shelter service the house delivers each month. For a renter that has an invoice. For an owner there is none, so BLS estimates one — owners' equivalent rent, by an approach called rental equivalence.
What almost every explanation of OER gets wrong
The standard description is that BLS asks homeowners what their house would rent for and puts that answer in the index. That is half right, and the wrong half is the important one.
The Consumer Expenditure Survey does ask exactly that: “If someone were to rent your home today, how much do you think it would rent for monthly, unfurnished and without utilities?” Those answers set the weight — the share of the index OER gets. They do not set the monthly price change. BLS is unambiguous: “Owner-occupied units are not priced in the CPI Housing Survey.”
The month-to-month movement comes from actual leases: “Using the sample of rental units, the CPI program calculates a measure of price change for each CPI index area for the rent and OER indexes.” BLS first derives “pure rents” by removing any utilities included in the rent, since owner-occupants pay their own, then reweights the units so the mix of structure types resembles the owner-occupied stock.
A utility-stripped, structure-adjusted measure of observed change in tenant rents, carrying a weight derived from what homeowners estimated their own homes would rent for. It is not a monthly poll of homeowner opinion, and it contains no house prices, mortgage payments, property taxes, insurance or maintenance.
The collection schedule, which explains the lag
BLS “collects rent data from each sampled unit every 6 months,” split into six rotating panels: “The rents for panel 1 are collected in January and July; panel 2, in February and August,” and so on. About one-sixth of the rental sample is replaced each year. Combine six-month repricing with the fact that a signed lease is not renegotiated monthly, and the consequence is mechanical: rent and OER reflect the average across all leases in force, not what a new lease costs this week. When the rental market turns, these series turn months later.
Why it is the most criticized component
Three objections get run together and deserve separating. The lag follows from the design above. The conceptual choice — that rent movement stands in for owner-occupied shelter, with mortgage rates, taxes and upkeep excluded — is a real disagreement about whether an owner is consuming shelter or holding an asset, and BLS answered it in the quote above rather than dodging it. The leverage is why both matter: at a quarter of the index, a decision inside OER moves the headline further than any other single choice in the CPI.
Substitution, hedonics and geometric means
Almost every claim that the CPI understates inflation traces back to one of three published methodological choices. All three have a real rationale and a real objection, and presenting one side without the other is how you get either a conspiracy theory or a press release.
Geometric means inside a category
Effective with January 1999 data, BLS began using “a weighted geometric mean of price ratios” for most item strata — the elementary building blocks of the index. A geometric mean lets the measured change within a category reflect consumers buying relatively less of whatever got relatively more expensive inside it. A limited number of strata keep a modified Laspeyres formula, a weighted arithmetic mean of prices: housing at school, some utilities and government fees, and some medical services.
BLS published the expected effect itself, in advance: the new formula would cover indexes making up “about 61 percent of the Consumer Price Index (CPI-U)” and would “reduce the annual rate of increase in the CPI-U by about 0.2 percentage point per year.” That figure gets quoted as though a critic exposed it. BLS announced it.
Substitution between categories
Within a category the CPI-U now allows substitution. Between categories it does not: BLS says the CPI-U “uses a formula that assumes consumers do not substitute across item categories,” while the C-CPI-U “reflects the effect of substitution that consumers make across item categories in response to changes in relative prices.” If beef gets expensive and households buy more chicken, CPI-U does not register the shift and C-CPI-U does. That is the whole reason the tax code moved to the chained version, and why chained indexation produces smaller bracket adjustments over time.
Hedonic quality adjustment
This one generates the most heat. BLS defines it as “a method of adjusting prices whenever the characteristics of the products included in the CPI change due to innovation or the introduction of completely new products,” done by decomposing an item into its characteristics, estimating the value of each, and using those estimates to adjust the price when quality changes. It is applied where quality moves fast — in BLS's words, apparel, “consumer appliances and electronics.”
The case for it. A $500 laptop in 2026 is not the $500 laptop of 2005. An index recording “no price change” would treat two different products as one. BLS frames the problem as bias that appears “if new version price changes are systematically different from the price changes of the unchanged goods” — doing nothing is also a methodological choice, and a worse one.
The case against it. The adjustment is an estimate from a statistical model, not an observed price. The quality it credits may be capability you did not ask for and cannot decline, and you cannot buy the 2005 laptop at the 2005 price, so the improvement is not optional the way the math implies. Every adjustment that credits quality reduces the measured price increase, which is why critics read a persistent downward pull.
Both sides describe the same mechanism correctly; the disagreement is about magnitude, and there we hit a wall. BLS does not publish a figure for the total effect of hedonic adjustment on the headline CPI, so it sits in the ledger below as unverified. For the theory underneath index construction, that is Economics for Traders.
What the CPI is not
Three things the CPI gets used for that it does not actually claim to do. BLS says so about two of them in print.
Not a cost-of-living index
This is BLS's own position, not a critic's. “The CPI frequently is called a cost-of-living index, but it differs in important ways from a complete cost-of-living measure,” and it is “sometimes termed a conditional cost-of-living index.” The Handbook agrees: “the CPI only approximates a cost-of-living index” because “the cost of living is affected by many things not captured in market transactions.” A longer commute, a worse school, water quality, crime, the closure of the only clinic in town — none of that is a market transaction and none of it is in the index. The irony stands: the adjustment on a benefit letter is called a cost-of-living adjustment and is computed from an index BLS declines to call one.
Not your inflation rate
Also BLS's own language: “The CPI does not necessarily measure your own experience with price change… BLS bases the market baskets and pricing procedures on the experience of the relevant average household, not of any specific family or individual.” Shelter is 35.6% of the index; if rent is 55% of your budget, a hard move in rent hits you half again as hard as it hits the index, and if you own outright it barely touches you while still moving the headline. Neither case is an error. The index was never measuring your basket.
Not PCE, and PCE is what the Fed targets
The second official U.S. inflation measure is the PCE price index, from the Bureau of Economic Analysis — different agency, scope, weights and formula. It matters because the Federal Open Market Committee's longer-run 2 percent objective is stated on the annual change in the PCE price index, not the CPI. Holding a CPI print against 2% measures with the wrong ruler, and it is done constantly. The inflation page covers the four reasons BEA gives for the two diverging; this site's economic pillar carries both side by side.
On a benefit letter as a COLA, from CPI-W. In a lease or service contract as an escalation clause, almost always the unadjusted CPI-U. Inside a TIPS bond, indexed to non-seasonally-adjusted CPI-U on a three-month lag written into Treasury's offering terms. In union contracts and pension formulas. In your tax brackets, through C-CPI-U.
And further out: the CPI print is one input the FOMC weighs when it sets the federal funds rate, which flows into every interest rate you are quoted — and the APY on your savings is nominal until you subtract a price index back out of it.
Confirmed, unverified, convention
House practice here is to label what is sourced, what is not, and what is merely customary.
| Status | Claim, and what establishes it |
|---|---|
| Confirmed | July 2026 CPI-U: level 333.918 (1982–84=100), all items +3.4%, core +2.5%, one-month seasonally adjusted +0.1%, shelter +3.2% — BLS release USDL-26-1378, 12 August 2026. |
| Confirmed | Population shares (“over 90 percent,” “approximately 30 percent”); the excluded rural, farm, military and institutional populations; and the scale — 7,776 basic indexes, ~80,000 prices monthly, ~23,000 establishments, 75 urban areas — BLS FAQ and Handbook of Methods. |
| Confirmed | Owner-occupied units are not priced in the Housing Survey; OER price change comes from the rental sample with utilities removed; rents collected every six months in six panels — BLS OER factsheet. |
| Confirmed | Geometric mean formula effective with January 1999 data, ~61 percent of the CPI-U, BLS's own estimate of about 0.2 percentage point per year — BLS. |
| Confirmed | Social Security and SSI COLAs on CPI-W third-quarter averages, 2.8% for December 2025 benefits — SSA. Tax parameters indexed to C-CPI-U against the CPI for 2016 — 26 U.S.C. § 1(f)(3), (f)(6). |
| Unverified | The total effect of hedonic quality adjustment on the headline CPI. Asserted confidently in both directions. We found no BLS figure for the aggregate contribution and are not estimating one. |
| Unverified | “Chained CPI runs 0.2 to 0.3 points below CPI-U per year.” BLS confirms C-CPI-U usually rises more slowly and that exceptions are infrequent, but we found no BLS statement of a standing average gap. Also unverified: the more precise 93 and 29 percent population shares that circulate, against BLS's published “over 90” and “approximately 30.” |
| Convention | Calling “all items less food and energy” core — shorthand BLS does not use. Quoting the twelve-month change to one decimal, when escalation contracts work from unrounded levels. Treating the CPI release as “the” inflation number, when it is one of several and not the one the Fed's target is written on. |
| Our arithmetic | Every calculation derived from index levels — the 3.365% twelve-month change, the escalator dollars, and reading 333.918 as about 3.34 times the 1982–84 average — is ours, run on levels BLS published. |
What trips people up
Comparing a CPI print to the Fed's 2% target. The wrong ruler — the FOMC's objective is written on the PCE price index, and the two routinely differ for the same month.
Expecting the COLA on a benefit letter to match the headline. It is CPI-W, not CPI-U, and it is a third-quarter average against a third-quarter average — not the twelve-month change in the latest month. See the SSI page for how the resulting figure lands on a payment.
Writing an escalation clause on the seasonally adjusted series. BLS recommends against it, and adjusted values get republished each February with newly estimated factors.
Reading the index level as a percentage. 333.918 does not mean 333.9% inflation. It means the basket costs roughly 3.34 times what it cost on average across 1982–84 — our arithmetic on the published level. The level is a ruler mark, not a rate.
Chaining twelve seasonally adjusted monthly changes and expecting the published twelve-month rate. Two different calculations on two different series. They will be close, they do not have to match, and neither is wrong when they don't.
Assuming your own basket matches. It does not, and BLS says it does not. For the version of this that is about your dollars rather than the national statistic, that is Stage 5 territory.
BLS collected no CPI data from 1 October through 12 November 2025 during the lapse in appropriations, and the missing October 2025 data also affected April 2026 rent and owners' equivalent rent indexes. Any twelve-month comparison spanning that window rests partly on prices nobody collected. The inflation page carries the detail, the BLS statement, and why a falling rate still means rising prices.
Frequently asked questions
What is the Consumer Price Index (CPI)?
The Consumer Price Index is a monthly price index published by the U.S. Bureau of Labor Statistics that measures the change in what a fixed basket of goods and services costs a specific reference population of urban consumers. BLS records the prices of about 80,000 items each month from roughly 23,000 establishments across 75 urban areas, producing 7,776 basic indexes from 243 item strata in 32 index areas. The output is an index level set against a 1982-84 average of 100 — 333.918 for July 2026 — which is then converted into the percent change reported in the news.
What is the difference between CPI-U and CPI-W?
They cover different populations. CPI-U is the Consumer Price Index for All Urban Consumers, which BLS says represents over 90 percent of the U.S. population, and it is the number in the headlines. CPI-W covers urban wage earners and clerical workers only, is a subset of the CPI-U population, and represents approximately 30 percent of the total population. The distinction has real dollar consequences: the Social Security Administration computes cost-of-living adjustments from CPI-W, not CPI-U, so a benefit letter and a news headline will normally show different numbers for the same period.
How is owners' equivalent rent calculated?
Not the way most explanations say. The Consumer Expenditure Survey does ask homeowners what their home would rent for monthly, unfurnished and without utilities, but those answers set the weight OER carries in the index, not its monthly movement. BLS states that owner-occupied units are not priced in the CPI Housing Survey. The month-to-month change is calculated from the sample of tenant-occupied rental units, with the value of any included utilities removed to produce pure rents, then reweighted so the mix of structure types resembles owner-occupied housing. No house prices, mortgage payments or property taxes enter it.
Why does the CPI exclude food and energy in core inflation?
Because food and energy prices swing on weather, harvests and geopolitics, and those swings are large enough to obscure the trend in everything else. BLS publishes the series as all items less food and energy; core is market shorthand the agency does not use. The obvious objection is also correct: people buy food and gas, and a household spending a large share of income on groceries and a commute experiences the headline number rather than core. Neither series is complete on its own. Headline is closer to what you paid; core is closer to what tends to persist.
Is the CPI a cost-of-living index?
No, and BLS says so directly. In its own words the CPI is frequently called a cost-of-living index but differs in important ways from a complete cost-of-living measure, and BLS describes it as sometimes termed a conditional cost-of-living index. The Handbook of Methods adds that the CPI only approximates a cost-of-living index because the cost of living is affected by many things not captured in market transactions. A longer commute, a worse school, deteriorating water quality and rising crime all change your cost of living and none of them appear in the index.
Does the Federal Reserve target CPI inflation?
No. The Federal Open Market Committee's longer-run objective of 2 percent is stated on the annual change in the personal consumption expenditures price index, published by the Bureau of Economic Analysis, not on the CPI. The two measures use different scopes, different weights and different formulas, so they routinely disagree for the same month. Comparing a CPI print to the 2 percent figure gives a systematically wrong answer about how far policy has to go. This site's live economic pillar carries both series so you can read them side by side.
Why is my personal inflation rate different from the CPI?
Because your spending is not the average household's spending, and BLS is explicit that the CPI does not necessarily measure your own experience with price change — it bases the baskets and pricing procedures on the relevant average household, not any specific family or individual. Shelter is about 35.6 percent of the CPI-U. If rent is 55 percent of your budget, a sharp rent increase hits you far harder than it hits the index; if you own your home outright, it barely touches you while still moving the headline. That gap is the design working, not failing.
Related terms
Where to go next
- Read the companion page on inflation itself — the year-over-year formula, the full weight table, CPI against PCE, and what a nominal return looks like after prices.
- See current CPI, core, PCE, rates and labor readings on the live economic pillar instead of the dated figures on this page.
- Go deeper on how price indexes are constructed and what central banks do with them in Economics for Traders.
- See what a rising index does to taxes, housing and income streams in Stage 5 · Build Wealth, or how COLAs land on a fixed benefit in the Disability Wealth Guide.
- Browse every definition in Learn the Lingo.
- U.S. Bureau of Labor Statistics, Consumer Price Index — July 2026 (USDL-26-1378), released 12 August 2026 — establishes the index level of 333.918 (1982–84=100), all items +3.4% over twelve months, all items less food and energy +2.5%, the seasonally adjusted one-month change of +0.1%, and shelter +3.2%. The July 2025 level of 323.048 used in the worked escalator example is from the same series via FRED, CPI-U All Items Not Seasonally Adjusted; the percent change and dollar figures derived from those two levels are our own arithmetic.
- U.S. Bureau of Labor Statistics, Consumer Price Index Frequently Asked Questions — the CPI-U “over 90 percent” and CPI-W “approximately 30 percent” population figures, the 1982–84=100 reference base, “about 80,000 items each month,” the “conditional cost-of-living index” language, the statement that the CPI does not necessarily measure your own experience with price change, and the Consumer Expenditure Survey's role in setting weights with its two-year lag.
- U.S. Bureau of Labor Statistics, Handbook of Methods: CPI Concepts and CPI Design — the excluded populations (rural nonmetropolitan, farm households, military installations, religious communities, institutions such as prisons), the “only approximates a cost-of-living index” statement, the 7,776 basic indexes from 243 item strata across 32 index areas, the 75 urban areas and the approximately 23,000 retail establishments.
- U.S. Bureau of Labor Statistics, Handbook of Methods: CPI Calculation — the weighted geometric mean of price ratios used for most item strata, the modified Laspeyres formula and the specific categories that retain it (housing at school; electricity, water, gas and vehicle registration; prescription drugs and physician, hospital, dental and nursing home services).
- U.S. Bureau of Labor Statistics, How the CPI measures price change of Owners' equivalent rent of primary residence and Rent of primary residence — that owner-occupied units are not priced in the CPI Housing Survey, that OER price change is derived from the rental sample, the “pure rents” utility adjustment, the six-month collection interval and the six-panel rotation, and the treatment of owned housing as a capital good rather than a consumption good.
- U.S. Bureau of Labor Statistics, Quality Adjustment in the CPI: Questions and Answers — the definition of hedonic quality adjustment, the decomposition-into-characteristics method, the item categories where it is applied (apparel, consumer appliances and electronics), and the bias rationale.
- U.S. Bureau of Labor Statistics, Frequently Asked Questions about the Chained CPI-U and New CPI estimator expected to lower inflation rate by 0.2 percent — the C-CPI-U substitution formula, the July 2002 start, the preliminary-and-revised publication schedule, the January 2023 exception; and the January 1999 effective date of the geometric mean estimator, its roughly 61 percent coverage and the 0.2 percentage point per year estimate.
- U.S. Bureau of Labor Statistics, Seasonal Adjustment in the CPI — that users analyzing general price trends should use seasonally adjusted indexes, that escalation agreements should typically not, and that the seasonal status of every series is reevaluated each January with new factors published in mid-February.
- U.S. Social Security Administration, Latest Cost-of-Living Adjustment — that COLAs are based on CPI-W, the third-quarter averaging rule, the 2.8% adjustment for December 2025 benefits, and that federal SSI payment levels rise by the same 2.8% for January 2026. Legal Information Institute, 26 U.S.C. § 1 — subsection (f)(3) indexing to “the C-CPI-U for the preceding calendar year” against “the CPI for calendar year 2016,” and (f)(6) defining C-CPI-U.
The figures on this page are checked against the source that publishes them, and dated. Published rates move after the release named above — the linked source always carries the current number. This page explains a term; it does not recommend a product.