Inflation is reported as a single number, which is the source of most misunderstanding about it. There is no single inflation rate. There are several indices, built on different baskets, using different methods, updated on different schedules, and they routinely tell different stories about the same month. Knowing which one you are looking at — and what it structurally cannot capture — is most of the skill in reading the data.
The Consumer Price Index
CPI is the headline measure in most countries. It tracks the price of a fixed basket of goods and services intended to represent what a typical urban household buys, with each item weighted by its share of spending.
Two features of that construction matter enormously.
First, the basket is fixed between revisions. If beef becomes expensive and households switch to chicken, CPI continues to price beef at its old weight for a period. This is the substitution bias, and it tends to overstate the true cost-of-living increase people experience.
Second, weights determine everything. Housing is the largest single component in most CPI baskets. A modest change in the shelter figure moves headline CPI more than a dramatic swing in a small category. When commentators say inflation was “driven by” some component, they usually mean it had a large weight, not that its price moved most.
Why shelter lags reality
Housing deserves separate treatment because it is where CPI most visibly diverges from lived experience.
Statistical agencies do not measure house prices in CPI — a house is an asset, and CPI measures consumption. Instead they measure the cost of shelter services: rents actually paid by tenants, plus an imputed figure for owner-occupiers, usually called owners’ equivalent rent, estimating what the owner would pay to rent the same property.
The sample includes all existing leases, not just newly signed ones. Since most tenancies run twelve months or longer, the index reflects rents agreed across the previous year. When market rents turn, the CPI shelter component follows with a lag typically measured in several quarters. Analysts watch new-lease rent indices to anticipate where official shelter inflation will be well before it appears.
Core inflation, and why it is not a trick
Core inflation is the headline index with food and energy removed. This reliably provokes the objection that food and energy are precisely what people buy — which is true, and beside the point.
Core is not an attempt to describe household experience. It is an attempt to extract signal. Food and energy prices are set substantially by weather, harvests, geopolitics and supply shocks — forces that are volatile, frequently reverse, and are entirely unresponsive to interest rates. A central bank raising rates cannot alter the price of oil.
Stripping them out gives a cleaner read on the underlying, demand-driven trend that policy can influence. Headline inflation tells you what happened to household budgets. Core inflation tells you what is likely to persist. Both are useful; they answer different questions.
Analysts increasingly supplement core with narrower cuts — services excluding housing, trimmed-mean and median measures that discard outliers at both ends — all attempting the same thing: separating persistent inflation from noise.
The Producer Price Index
PPI measures prices received by domestic producers for their output, rather than prices paid by consumers. It sits earlier in the supply chain, which is why it is often treated as a leading indicator.
Treat that reading with care. The pass-through from producer to consumer prices is real but incomplete and slow. Firms absorb cost increases in margin when competition prevents them raising prices, and expand margin when input costs fall without cutting prices. PPI also covers a different universe — it includes goods sold to other businesses and excludes imports, which form a substantial share of consumer spending.
A PPI spike signals cost pressure building. It does not reliably predict the size or timing of any consumer price response.
PCE and why central banks may prefer it
In the United States, the Federal Reserve’s stated target is the Personal Consumption Expenditures price index rather than CPI. The two differ in three structural ways.
- Scope. PCE captures spending made on households’ behalf — notably employer- and government-funded healthcare — which CPI largely excludes. This gives healthcare a much larger PCE weight.
- Weights. PCE updates its weights continuously rather than periodically, so it adapts to substitution as it happens.
- Formula. The two use different aggregation methods, which produces a persistent gap.
PCE typically runs somewhat below CPI as a result. Comparing a PCE target to a CPI print is a common and consequential error.
Base effects: the trap in year-over-year figures
Annual inflation compares today’s index to the same month a year ago. That means it is determined by two numbers, and the older one is fixed history.
If prices spiked twelve months ago, the annual rate will fall this month even if prices are currently rising briskly — the comparison base is simply high. This is a base effect. It is arithmetic, not disinflation.
The defence is to look at month-over-month changes, usually annualised, and at three- and six-month annualised rates. These reveal the current run-rate. When monthly momentum diverges sharply from the annual figure, the monthly series is describing the present and the annual figure is describing last year.
Seasonal adjustment and revisions
Most reported series are seasonally adjusted to strip predictable calendar patterns — holiday retail, summer fuel demand, new-year price resets. Comparing a seasonally adjusted figure to an unadjusted one produces nonsense.
Seasonal factors are themselves estimated from history and get revised. A month that looked alarming on release can look ordinary after revision. Any single print carries meaningful uncertainty; the trend across several months carries far more information than the latest number.
A practical reading order
- Identify the index and whether it is headline or core, adjusted or unadjusted.
- Read the month-over-month change before the annual rate.
- Check three- and six-month annualised rates for the current run-rate.
- Ask whether the annual move is a base effect.
- Decompose by contribution — weight times price change — not by which price rose most.
- Treat shelter as a lagging signal and read new-lease data for the leading one.
- Wait for confirmation across months before concluding a trend has turned.
None of this requires specialist tools. The statistical agencies publish the component detail alongside the headline, and the discipline of reading it is what separates a useful interpretation from a misleading one.
Related reading
For how central banks respond to these figures, see how central bank rate decisions reach the real economy.
This article is general information and journalism, not investment advice. See our Editorial Policy.