Inflation debates often turn on which measure gets the spotlight. Headline CPI captures prices across the consumer basket, while core measures exclude selected categories such as food and energy. That distinction can help analysts isolate persistent trends. It can also leave households wondering why the number being discussed feels so different from the prices they pay. The history of U.S. inflation statistics is a history of changing definitions, baskets and methods—and of persistent arguments over what those choices mean.
The view advanced here is that many changes have tended to restrain reported inflation, sometimes for defensible statistical reasons and sometimes in ways that served political or fiscal aims. That is a pointed interpretation, not a settled account of every revision. The record, however, shows that the choice of index and the treatment of its components can have consequences for public debate, cost-of-living adjustments and the Federal Reserve’s stated inflation target.
From wartime cost of living to the CPI
The dispute reaches back to World War II. Critics of the wartime Cost-of-Living Index argued that it understated price increases as goods became scarce, wages were frozen and black-market prices spread. The source estimates that the index missed inflation by four percentage points from 1941 to 1944. In 1945, the Bureau of Labor Statistics renamed the series the Consumer Price Index and described it as a fixed-basket measure.
A broad revision followed in 1953, rebuilding the basket using a 1950 survey. It was the first comprehensive update since the 1934–36 survey. Those updates reflect a basic challenge for any price index: consumer spending changes, new products arrive, and the prices people face can diverge from a basket that takes time to revise.
The 1961 Stigler Committee recommended reframing the CPI as a cost-of-living index. The related methodological debate covered substitution between goods, changes in product quality and the treatment of new products, including hedonic adjustments. These methods can make an index better reflect changing consumption and product features. They can also produce a lower measured rate than a fixed basket would, particularly when consumers switch to less expensive alternatives.
Why “core” became a policy focus
The BLS had published an “all items less food and energy” series since 1957, but core inflation gained prominence during the stagflation of the 1970s. After food prices jumped in 1972–73, Federal Reserve Chair Arthur Burns directed staff to exclude food, describing its price movements as driven by weather rather than monetary policy. Energy was later stripped out following the oil embargo. In the author’s account, exclusions expanded at times to categories such as used cars, mobile homes and children’s toys.
Core measures can help policymakers distinguish volatile price swings from broader inflation trends. The author’s objection is that repeated exclusions risk making the measure feel detached from household budgets. His 2005 label for that problem was “Inflation Ex-Inflation.” The same tension appears in later alternatives: the Cleveland Fed’s median and trimmed-mean CPI, introduced in 1994; the Dallas Fed’s trimmed-mean PCE in 2005; Atlanta’s sticky-price CPI in 2010; and the “supercore” services measure excluding housing discussed by Jerome Powell in 2022.
These gauges answer different questions. A trimmed mean limits the influence of unusually large price changes, while a measure excluding housing focuses on a narrower slice of services. They can be useful analytical tools. But their readings shouldn’t be mistaken for a complete account of the cost increases consumers experience.
Shelter measurement and the housing cycle
The adoption of owners’ equivalent rent (OER) in the CPI-U in January 1983 was a major change. Before then, the index treated homeownership through house prices and mortgage-interest costs. With mortgage rates reaching 18% in 1981, those costs contributed to a sharp inflation reading and large cost-of-living adjustments. The revised approach measured the value of owner-occupied housing as the rent a homeowner would pay to occupy an equivalent home. The Bureau of Economic Analysis had used rental equivalence in the PCE measure already.
The rationale is that a house is an asset, while the shelter it provides is a consumer service. Yet rental equivalence can move differently from home-sale prices. The author argues that the 2002–07 housing boom was muted in CPI shelter readings as many households moved into homeownership, including through no-documentation and adjustable-rate mortgages. He also points to a lag in shelter inflation during 2021–23. That gap matters: a measured shelter series can trail shifts in housing costs, complicating the interpretation of current inflation.
Bias estimates and fiscal stakes
In 1995, Federal Reserve Chair Alan Greenspan told Congress that the CPI might overstate inflation by 0.5 to 1.5 percentage points. The Boskin Commission—Michael Boskin, Ellen Dulberger, Robert Gordon, Zvi Griliches and Dale Jorgenson—reported in 1996 an estimated upward bias of 1.1 percentage points a year, with a range of 0.8 to 1.6. The estimate attributed roughly 0.4 points to upper- and lower-level substitution, 0.1 to outlet substitution, and 0.6 to quality changes and new goods.
The article reads the commission’s work as politically useful because a lower CPI path could reduce cost-of-living adjustments and the effects of bracket creep in the tax code. It cites an estimated $1 trillion in deficit reduction over roughly a decade as the fiscal stakes. The BLS implemented methodological changes administratively. The author argues that these revisions may also have shaped perceptions of inflation during the 2003–07 housing and credit boom, when official CPI inflation remained subdued.
His concern is that quality adjustments and substitution can obscure the budget impact of rising prices. A car that costs more may also have improved features; a consumer who switches from steak to chicken may hold spending down, but that doesn’t mean steak became affordable. These methods have a statistical rationale, yet the resulting index can tell a different story from the one implied by a consumer’s previous purchases.
Which index guides policy?
The Federal Reserve shifted its preferred inflation gauge from CPI to PCE in February 2000. PCE covers a broader range of spending, uses a chained formula and draws on business surveys that include third-party health-care spending. In 2004, the Fed turned to core PCE for its inflation focus. Since 2012, the central bank’s 2% target has been defined in terms of PCE. The source estimates that this choice produces an inflation rate about half a percentage point lower each year than CPI.
That difference changes the context in which inflation performance is judged. It doesn’t make one index inherently right and the other wrong: they cover different spending and use different methods. But comparisons need to specify which measure is being used, especially when a target is discussed alongside household-facing CPI data.
In 2017, the Tax Cuts and Jobs Act adopted chained CPI-U for tax-bracket indexing, following an earlier proposal in President Barack Obama’s fiscal 2014 budget. The source says the proposal was withdrawn under pressure from the left. Chained CPI is used for tax brackets, while Social Security benefits continue to use CPI-W. Because the chained measure accounts for substitution, its effect on bracket indexing can be to raise tax revenue over time as income growth pushes taxpayers into higher brackets.
Recent changes and data gaps
More recent revisions reflect the strain of keeping measures current. CPI weights shifted to annual updates in 2023 after pandemic disruptions changed spending patterns. The source also describes adjustments to health-insurance measurement and a new-tenant rent index intended to address shelter lags. In 2025, the BLS replaced survey collection with vendor transaction data for leased vehicles in April and wireless services in July.
The author makes a more serious charge about 2025: staffing cuts allegedly led to suspended CPI collection in several metropolitan areas, with missing observations imputed. He reports that Commissioner McEntarfer was fired on August 1 after a weak jobs report, and that a subsequent nomination of Antoni did not proceed. The account says the October–November shutdown resulted in no October CPI release because the information could not be collected retroactively; the November report omitted monthly changes for affected items. Those claims, as presented, mark a distinction between revising a measurement method and losing the ability to collect the underlying data.
The long-running argument over inflation statistics is therefore not simply about whether one measure runs hotter than another. It concerns what each index includes, how quickly its components respond and whether missing or delayed observations leave the signal incomplete. For analysts, the practical takeaway is to identify the gauge, inspect its components and account for known lags before drawing conclusions from a single inflation print.
Previously:
The History of Inflation ex-inflation (September 23, 2005)
For Lower Inflation, Stop Raising Rates (January 18, 2023)
CPI Increase is Based on Bad Shelter Data (January 11, 2024)
CPI Shelter Measures: 6-12 Month Lag (February 15, 2024)
__________
1. The source notes that federal cost-of-living adjustments were not automatic in 1961. Social Security indexation arrived in 1972 and tax-bracket indexation in 1985; at the time, union escalator clauses were the main CPI-linked payments. It argues that many claims made by the Boskin Commission in 1996 had roots in the Stigler Committee’s 1961 work.
2. The 1978 revision introduced CPI-U, covering all urban consumers—about 87% of the population—alongside the wage-earner index, and renamed the Wholesale Price Index as the Producer Price Index. The author says he doesn’t view this change as manipulation, while remaining open to argument.
3. Analyst Josh Rosner wrote that “A Home Without Equity is Just a Rental with Debt.” He described such arrangements as “Renting with the right to default.”
4. The source links to Greenspan’s testimony and quotes: “It’s not a lie if you believe it.”
5. The commission’s estimated annual bias was divided into approximately 0.4 percentage points from upper- and lower-level substitution, 0.1 from outlet substitution, and 0.6 from quality change and new goods.
6. The source quotes Bailout Nation on the role of Boskin-related changes in perceptions of inflation during 2003–07: “During most of the 2003-2007 run-up in prices, the consumer price index (CPI), the official measure of inflation, hardly budged. Credit the Boskin Commission for contributing to this illusion of low inflation. The Boskin reforms changed not only the basket of goods measured for inflation but how we measure them. It allowed “hedonic adjustments” for the improvement of quality. Forget what the window sticker says; Your new car isn’t really more expensive—it’s better.
Then there was substitution—for example, when the price of steak rises, consumers can replace it with chicken at the previous low prices. In the topsy-turvy world of Boskin, substituted prices remained the same. In the real world, inflation just drove steak out of your price range.”
7. President Obama’s fiscal 2014 budget proposed C-CPI-U for Social Security adjustments and tax indexing, then withdrew the proposal under pressure from the left. The 2017 Tax Cuts and Jobs Act adopted chained CPI for tax-bracket indexing; Social Security benefits still use CPI-W. The source says the change makes the index closer to the Boskin model and raises tax revenue through bracket creep.
8. The source says the October–November 2025 shutdown led to no October CPI release, because the data could not be collected later, and that the November report omitted monthly changes for affected items.
The post The Long History of Underreporting Inflation appeared first on The Big Picture.