How CPI Is Measured (And Why It's Disputed)

One Number, Enormous Consequences

Every month, a single percentage — the Consumer Price Index change — dominates financial news, moves interest-rate decisions, and quietly adjusts millions of paychecks, pensions, and tax brackets. If you want to see what a given rate of inflation does to a specific amount of money, the Inflation Calculator handles that instantly. This article is about something different: how that headline number actually gets built, and why a formal government commission once concluded it was wrong by more than a full percentage point a year.

That's not a fringe claim. It came from a bipartisan panel of economists commissioned by the US Senate, and it reshaped how the Bureau of Labor Statistics calculates inflation to this day.

How the CPI Basket Is Actually Built

CPI isn't a single price — it's a weighted average across a "basket" of goods and services meant to represent what a typical household buys: food, housing, transport, healthcare, clothing, recreation, and more.

The basket and its weights come from large household expenditure surveys — the UK's Living Costs and Food Survey, the US Consumer Expenditure Survey — where thousands of families record their spending in detail. Whatever categories take up the largest share of average spending get the largest weight in the index; a 1% rise in housing costs moves the headline number more than a 1% rise in, say, spending on postage stamps, because housing takes up a much bigger share of the average budget.

This is reviewed and refreshed periodically — annually in the UK, every two years in the US — because spending habits genuinely change. Streaming services didn't exist in a 1995 basket; landline telephone charges mattered far more then than now.

The Boskin Commission: When Economists Formally Audited CPI

In 1995, the US Senate appointed a panel of economists — the Advisory Commission to Study the Consumer Price Index, chaired by Stanford economist Michael Boskin and universally known since as the Boskin Commission — to formally investigate whether CPI was measuring what it claimed to measure: the true cost of maintaining a constant standard of living.

Its December 1996 report, Toward a More Accurate Measure of the Cost of Living, concluded that CPI overstated true inflation by approximately 1.1 percentage points per year (with a plausible range of 0.8 to 1.6 points). That might sound like a small technical correction, but the Congressional Budget Office estimated the bias — left uncorrected — would have added roughly $1 trillion to the national debt by 2008, since Social Security payments, tax brackets, and other indexed programs all move with published CPI.

The Commission broke its estimate down into three main sources, which are each worth understanding on their own.

Substitution Bias, Explained With Beef and Chicken

The largest single piece of the Boskin estimate — about 0.4 percentage points — came from substitution bias.

Traditional CPI assumes households keep buying the exact same basket of goods in the exact same quantities, year after year, regardless of how relative prices shift. Real people don't behave that way. If beef prices spike sharply while chicken stays flat, a meaningful share of shoppers buy more chicken and less beef — softening the actual hit to their household budget. A fixed-basket index doesn't capture that adaptive behaviour, so it systematically overstates how much a real consumer's cost of living has risen.

A second, related effect — outlet substitution bias (worth about another 0.1 point in the Commission's estimate) — captures a similar blind spot: shoppers increasingly buying the same goods from cheaper retailers (discount chains, then later online marketplaces) rather than the stores originally surveyed, a shift the traditional methodology was slow to reflect.

The Chained CPI: A Fix, With Its Own Trade-offs

The direct policy response to substitution bias was the chained CPI-U (C-CPI-U), which the Bureau of Labor Statistics has published since 2002. Instead of fixing basket weights from a past survey period, it continuously updates weights to reflect how spending actually shifts between categories as relative prices change — directly addressing the substitution problem the Boskin Commission identified.

The trade-off is timeliness: because it needs data on current spending patterns, the chained CPI isn't fully finalised until roughly a year after the period it covers, so it's initially published as a preliminary estimate and revised later. Since 2000, the chained CPI has run about 0.25–0.3 percentage points lower per year than the traditional CPI-U on average — a gap that looks small year to year but compounds meaningfully: over roughly two decades, that's the difference between a cumulative price rise reported as 30% versus about 26%.

This is also why proposals to switch official government indexing (Social Security, tax brackets) to the chained CPI are politically contentious — a permanently "lower" measured inflation rate means smaller automatic increases to indexed benefits over time, which is exactly the mechanism the Boskin Commission's critics on the other side worried about.

Hedonic Adjustment: Pricing In a Better Product

The largest piece of the Boskin Commission's estimate — about 0.6 percentage points — came from how CPI handles quality change and new products.

Consider a smartphone: today's $1,000 phone has a dramatically better camera, processor, and battery than a $1,000 phone from a decade ago. If the price stayed literally the same but the product improved substantially, has "inflation" really been zero for phones? Statistical agencies attempt to separate genuine price inflation from quality improvement using a technique called hedonic regression — modelling how much of a price difference is explained by measurable feature changes, and attributing the rest to inflation.

This is one of the more contested areas of CPI methodology. Critics on one side argue the adjustments are too aggressive and can make genuine price increases disappear into "quality improvement," understating real inflation for consumers who didn't want or need the extra features. Critics on the other side (including much of the reasoning behind the original Boskin estimate) argue the opposite — that CPI historically under-adjusted for quality and new products, overstating inflation. Both critiques can't be fully right at once, which is itself a good illustration of how genuinely difficult this measurement problem is.

Shrinkflation: The Price Rise That Isn't Called One

A related, more recently prominent phenomenon is shrinkflation — when a product's shelf price stays the same but the quantity inside shrinks (a cereal box that quietly goes from 800g to 750g at an unchanged price). It's economically identical to a price increase per unit, just harder to notice.

Statistical agencies are generally aware of the practice and are supposed to track price per standard unit (price per 100g, for example) rather than price per package specifically to catch this. In practice, the correction depends on package-size changes being flagged and re-measured by the survey process, which introduces a lag — consumers on social media have repeatedly flagged shrinkflation on specific products well before it showed up clearly in official statistics, making it one of the more visible real-world examples of CPI measurement lagging lived experience.

Who's Right? The Case for and Against CPI

It's worth holding two things as true at once. First: CPI, imperfect as it is, remains the most rigorously constructed, internationally comparable measure of inflation available, built by professional statisticians using large representative surveys, reviewed and refined for over a century of methodological development. It is not an arbitrary or politically manipulated number in any of the major economies covered by this site.

Second: virtually every serious economist who has studied it agrees it is not a perfect cost-of-living index, and reasonable experts disagree about which direction the remaining biases point and by how much. A retrospective by economist Robert Gordon roughly a decade after the original Boskin report estimated the remaining bias had fallen to around 0.8–1.0 percentage points a year following the 2002 reforms — smaller than the original 1.1-point estimate, but not zero.

The practical takeaway isn't that CPI should be ignored — it's that "official inflation was X%" is a well-constructed estimate with known, studied limitations, not a physical measurement with no margin of error. Understanding those limitations is part of understanding what the number actually means.

Frequently Asked Questions

How is the CPI basket of goods actually decided?

Statistical agencies run large household expenditure surveys, asking thousands of families to record what they buy. The results determine which several hundred representative items go into the basket and how much weight each category gets, based on the average share of spending each takes. The basket and weights are reviewed periodically as spending habits shift.

What is substitution bias, in plain terms?

Standard CPI assumes people keep buying the same fixed basket even as relative prices change. In reality, if beef prices spike, many shoppers buy more chicken instead — but a fixed-basket index doesn't capture that switch, so it can overstate how much a consumer's actual cost of living rose. The Boskin Commission attributed roughly 0.4 percentage points of CPI's estimated 1.1-point annual overstatement to this effect alone.

What was the Boskin Commission, and what did it conclude?

A bipartisan panel of economists appointed by the US Senate in 1995 to formally audit CPI methodology. Its 1996 report concluded CPI overstated true cost-of-living inflation by about 1.1 percentage points a year, split roughly across substitution bias, outlet substitution, and under-adjustment for quality improvements and new products. The findings led directly to methodology changes, including the chained CPI introduced in 2002.

What is the chained CPI, and how is it different?

The chained CPI-U, introduced by the US BLS in 2002, updates its basket weights continuously to reflect actual substitution between categories, rather than using fixed weights from a past survey period. It consistently runs lower than traditional CPI — since 2000, the gap has averaged about 0.25–0.3 percentage points a year, which compounds to a meaningfully different total over a decade or more.

How does CPI handle products that improve in quality?

Through hedonic quality adjustment: statisticians try to separate a price increase into the portion caused by genuine inflation versus the portion caused by the product being better. It's most visible with electronics and cars. Critics argue the methodology is subjective and can cut in either direction — understating or overstating real inflation depending on how it's applied.

What is "shrinkflation," and does CPI capture it?

Shrinkflation is when a product's price stays the same but its quantity shrinks — effectively a hidden price rise. Agencies are supposed to adjust for it by tracking price per standard unit rather than per package, but consumers often notice shrinkflation well before it's reflected in official data, since the correction depends on the pack-size change being flagged and re-measured.

Summary

CPI is simultaneously the best tool economists have for measuring inflation and a number with well-documented, professionally studied imperfections. The Boskin Commission's 1996 audit put a number on those imperfections — about 1.1 percentage points of annual overstatement — and reshaped US inflation methodology for decades afterward, while genuinely difficult problems like quality adjustment and shrinkflation remain active areas of debate rather than solved problems.

None of this makes the headline inflation rate meaningless. It means treating it as a carefully constructed estimate, not an exact physical measurement — useful context whether you're reading the news or working out your own numbers.

If you want to see the practical impact of a given rate on your own money — salary, savings, or a historical price — that's what the Inflation Calculator is for.

More Finance Tools

• Compound Interest Calculator — see how compounding growth compares to inflation's compounding erosion
• What If I Invested Calculator — what a past investment would be worth today, inflation and all
• Real Hourly Wage Calculator — your true hourly rate once commute, costs, and time are factored in

---

This article summarises published economic research and methodology documentation for educational purposes. It is not financial advice.