Economists strip food and energy prices out of their main inflation reading because those two categories swing violently for reasons that have nothing to do with where prices are actually trending. What’s left over is core inflation, a steadier signal of where inflation is heading once the noise clears — not, as it’s often accused of being, a trick to make a bad number look better.

It is, admittedly, the single most resented thing economists do. Food and fuel are what households notice buying most visibly, so leaving them out of the headline conversation can look like sleight of hand. The reasoning behind it is more defensible than that, but the resentment isn’t entirely unfounded either.

What core inflation is actually measuring

The European Central Bank defines core inflation as the Harmonised Index of Consumer Prices excluding energy and food. The purpose isn’t to pretend those costs don’t exist — it’s to separate a temporary shock from a durable shift. The ECB excludes them specifically because they’re volatile components subject to frequent transitory shocks: a drought, an OPEC decision, or a pipeline outage can move the headline rate by a full percentage point and then reverse within months. A central bank that raised interest rates in response to that kind of spike would be tightening policy against a shock that had already passed by the time the tightening took effect.

Why not just wait for the shock to pass?

Because monetary policy works slowly. Rate decisions taken today affect prices over the medium term, so a central bank has to act on where inflation is going rather than where it has already been. Core inflation is the better predictor of where the headline rate will eventually settle, which is why it carries real weight in rate decisions even though no central bank actually targets it directly.

There are two ways to build it, and they disagree

The method matters more than most coverage admits. The exclusion method drops the same categories every month; this is the ECB’s approach, and it’s simple to explain and simple to check. Its weakness is that it assumes food and energy are always the volatile things, which isn’t true in every month or every economy. The trimming method instead drops whatever actually moved most that month: the Bank of Canada’s CPI-trim excludes components whose monthly price changes fall in the tails of the distribution, removing the bottom and top 20 percent by weight, while its CPI-median takes the single price change sitting at the 50th percentile of the distribution. Canada leans on trimming for a specific reason — changes in indirect taxes, such as a shift in sales tax, can move the total index on their own, and a fixed exclusion list would never catch that. The practical consequence is that two countries both reporting “core inflation” are often not measuring the same thing, and the numbers aren’t directly comparable.

Does leaving out food and energy make the number dishonest?

It would be, if core inflation were presented as the cost of living. It isn’t, and central banks are clear about this: what they actually target is headline inflation, the rate that includes everything. Core is a diagnostic tool used on the way to that target, not a replacement for it. That’s why a household can face a punishing rise in grocery and fuel bills in the same month that officials describe inflation as cooling — both statements can be accurate at once, because they’re describing different quantities, and only one of them is about your weekly shop.

Core inflation has its own blind spot

The ECB has warned that the tool isn’t working as cleanly as it used to. Recent major economic dislocations left a sizeable and possibly slow-reverting temporary element embedded inside underlying inflation measures, which makes them less informative about the medium term than they were designed to be. In other words, some of the noise that core inflation was built to filter out has partly leaked into core itself. Economists still use it, but with more supplementary analysis and less confidence than a decade ago.