The official figure says food is up about three per cent over the year. Almost nobody who does a weekly shop believes it.
They are not being irrational, and the figure is not being massaged. Beef is up more than twelve per cent and coffee close to twenty. Both statements are true at once, and the reason they can be is worth understanding, because it generalises far beyond groceries.
The specimen always in reach
Michel de Montaigne spent twenty years writing about a single subject on the grounds that it was the one he could actually examine: himself. His defence of the method was not modesty. It was that a general claim about how people are is assembled out of report and custom, whereas the one case in front of you can be inspected directly — and that each person carries the whole form of the human condition, so the particular is not as unrepresentative as it looks.
Applied here, that produces a specific instruction: stop arguing about the index and take your own reading.
An index is a weighted average over a representative basket. The weights are the shares of a typical household's spending, and the resulting number is a real, carefully constructed measurement of a thing that exists — the cost of that basket.
Your basket has different weights. If you drink coffee daily and buy beef weekly, those two items carry a share of your spending far above their share in the index, and your personal rate of increase is a multiple of the published one. If you are vegetarian and drink tea, it is below.
So the two numbers are not in conflict and neither is wrong. They are measurements of different objects, and the object the index measures is one that nobody actually buys.
Take the reading
The exercise takes about twenty minutes, once, and then five minutes a month.
Pull the last two or three receipts and pick your ten highest-spend items — not the ten you think about, the ten with the most money against them, which are rarely the same list. Note the current unit price of each. Find what you paid for them a year ago, from an old receipt, an order history, or a supermarket's own price archive.
Weight each by what you actually spend on it, and add it up. That is your rate. It is the only figure that describes your household, and it is generally the end of the argument about whether you are imagining things.
The second thing it produces is more useful than the first. The list will show that two or three items are doing most of the damage, and that is a fact you can act on in a way that the national figure gives you no purchase on at all. A general sense that everything is more expensive leads nowhere. Knowing that a fifth of the increase sits in one item is a decision.
Watch out for
His failure is generalising from a specimen of one, and he committed it regularly: his appetites, his aches and his comfortable income were not everybody's, and he sometimes mistook the habits of a Gascon gentleman for the human condition itself.
The version to avoid here is the one currently in wide circulation. My bill went up by a fifth, therefore the official figures are false. That does not follow, and it is exactly his error. Your basket is evidence about your household and about nothing else. The index is not a claim about your household and never was. Anyone using a personal receipt to argue that a national statistic is fabricated has taken a good instrument and pointed it at a question it cannot answer — and the people making that argument loudest are rarely offering you a better measurement in its place.
There is a second thing he would notice, and it is less comfortable. Some of what is on your list is there by habit rather than by choice — the thing bought weekly because it has always been bought weekly. A personal price index is also a record of which habits have quietly become expensive, and that is worth seeing even if you decide to keep them.
Answer this next
Take your last receipt, pick the three largest lines, and find out what those three cost a year ago.
That is the number your household actually runs on. Nobody publishes it, and nobody else can.
