Reading an Inflation Report Without Being Misled by It

An inflation release is a single dataset that reliably produces contradictory headlines. One outlet reports prices cooling; another reports them accelerating. Often both are accurate, because they are quoting different measures of the same thing.

The measures are not interchangeable, and each answers a different question.

Headline and core

Headline inflation covers everything a household buys. Core strips out food and energy.

Excluding the things people most notice sounds perverse, and it is regularly criticised on that basis. The rationale is narrow: food and energy prices are volatile and driven largely by supply — weather, conflict, production decisions — rather than by domestic demand. A spike caused by a pipeline outage says little about whether the economy is running hot, and a central bank raising rates in response would be tightening into a supply shock it cannot fix.

So headline is the better measure of what households experience. Core is the better measure of underlying pressure. Neither is the honest one; they answer different questions.

Monthly and annual

Year-on-year inflation compares this month to the same month last year. It is stable and easy to interpret, and it is slow — it will keep reporting elevated inflation for a year after prices stop rising, simply because the comparison point is old.

Month-on-month captures what is happening now, and is noisy. A single month tells you very little. Three consecutive months tell you a good deal.

Base effects

Because annual figures compare to a point twelve months back, they move when that old month drops out of the window — even if current prices did nothing. If a month a year ago saw an unusual spike, the annual rate will fall mechanically when it rolls off.

This is a base effect, and it produces some of the most misleading coverage in economics. An annual rate falling for this reason is not disinflation; it is arithmetic.

The shelter problem

Housing is the largest single component of US consumer price indexes, and it is measured with a substantial lag. Statistical agencies survey rents on a rotating schedule and include owner-occupied housing through an estimate of what owners would pay to rent their own homes.

The consequence is that shelter inflation reflects rental agreements signed over the previous year or more. When market rents turn, the index follows months later. Analysts often compute measures excluding shelter for precisely this reason — not to flatter the number, but because the shelter component is describing the past.

CPI and PCE

The US publishes two main measures. The consumer price index is more widely reported. The personal consumption expenditures index is what the Federal Reserve targets.

When the Fed says it targets 2%, it means PCE. Comparing a CPI print to that target is comparing two different things.

Weights, and why your inflation differs from the index

A price index tracks a basket weighted to represent average spending. Almost nobody spends like the average.

A household that rents in an expensive city, does not own a car and has no children faces a different effective inflation rate from one that owns outright, drives daily and pays tuition. When shelter and fuel move in opposite directions, the two experience opposite realities while the published figure reports a single number.

This is not a flaw in the index — an aggregate has to aggregate. But it explains the persistent gap between reported inflation and what people report feeling, and that gap is not usually evidence that the statistics are wrong.

Sticky and flexible prices

A useful cut of the data separates prices that change frequently from those that do not.

Flexible prices — fuel, food, some goods — respond quickly to conditions and are volatile. Sticky prices — rents, insurance, services with contracted rates — adjust slowly, often only annually.

The distinction matters for forecasting. Because sticky prices reset infrequently, firms setting them build in expectations about the future. A rise in sticky-price inflation therefore carries more information about where inflation is heading than an equivalent rise in flexible prices, which may reverse next month.

It also means inflation has momentum. Once it works into contracts, wage agreements and annual price reviews, it persists beyond whatever caused it — which is the main argument central banks make for responding early rather than waiting for confirmation.

Trimmed measures

Beyond headline and core, statistical agencies and regional central banks publish trimmed measures that discard the largest price moves in both directions each month, rather than always excluding the same categories.

The logic is that whichever category happens to be volatile this month is the one distorting the average, and that is not always food or energy. A trimmed mean adapts, where core does not.

These measures tend to be steadier than either headline or core, and they are useful precisely because they are boring: when a trimmed measure moves, it is harder to attribute to one unusual category, which makes it a more credible signal of broad pressure.

How to read a release

A reasonable sequence: look at month-on-month core first for current momentum; check whether the annual change was driven by a base effect; look at what is actually contributing, since a broad-based rise means something different from one category moving; and check the revisions to prior months, which are frequent and sometimes larger than the new number.

Then hold the conclusion loosely. One month is a data point with meaningful measurement error, not a trend.