The inflation figure is treated as a measurement, like temperature. It is a construction, and knowing how it is constructed explains why it frequently disagrees with personal experience.
The basket
Statistical agencies define a basket of goods and services intended to represent typical household consumption.
Prices for the items in that basket are collected, weighted by how much households spend on each, and combined into an index.
Which means the figure reflects average spending patterns, and any household whose spending differs from the average experiences different inflation.
A household spending a large share on rent and energy, in a period when those rose faster than everything else, experienced considerably more inflation than the headline figure showed. The figure was not wrong. It was measuring something else.
How the weights are set
Weights come from household expenditure surveys, updated periodically.
Which introduces a lag, since spending patterns change faster than surveys are conducted.
The pandemic period exposed this sharply — spending shifted abruptly away from travel and hospitality toward goods, while the weights reflected the previous pattern.
Agencies adjusted methodology in response, and the episode illustrated how much the weighting choice affects the result.
Quality adjustment
The most contested methodological element.
If a product improves while its price stays the same, the argument is that its effective price has fallen.
Statistical agencies adjust for this using various techniques, most notably for technology products where specifications change constantly.
Supporters argue that without adjustment, the index would overstate inflation by ignoring genuine improvements in what people receive.
Critics argue the adjustments are large, based on judgement, and systematically reduce the measured figure.
Both positions have substance. The adjustments are necessary and they are estimates.
Substitution
When a price rises, people buy less of that item and more of something else.
An index holding quantities fixed overstates the cost of maintaining living standards, since it ignores that people adapt.
Indices using formulas that allow for substitution produce lower figures than those that do not, which is why different index formulations of the same underlying data differ systematically.
Several countries publish more than one measure for exactly this reason, and which one is used for uprating pensions and benefits has been a live political question.
Housing
The hardest category and the largest for most households.
A house is partly consumption and partly an investment asset, and inflation indices are supposed to measure consumption.
Approaches include measuring rents paid, imputing a rent to owner-occupiers, or measuring the costs of ownership directly.
Each produces materially different results, and countries have made different choices.
Which means international inflation comparisons are less clean than they appear, since the largest component is measured differently.
Core inflation
Excluding food and energy produces a measure that is less volatile.
The rationale is that these components swing on supply factors that monetary policy cannot address, and that the underlying trend is more informative for policy.
The obvious objection is that people eat and heat their homes, and a measure excluding those is not measuring their cost of living.
Both are correct, because the two measures serve different purposes — one for policy, one for describing household experience — and problems arise when one is used for the other.
Why the gap with perception is persistent
Research consistently finds that perceived inflation exceeds measured inflation.
The explanations offered include that frequently purchased items are noticed more than infrequent ones, that price increases are remembered better than stable prices, and that the basket genuinely does not match any individual.
All three appear to contribute, and the gap is a durable feature rather than a recent development.
What to do with this
Treat the headline figure as what it is — a weighted average across an assumed basket, constructed with contestable choices.
It is genuinely informative about the general direction and it is not a measurement of your own costs, and no available statistic is.
Collection methods
How the underlying prices actually get gathered, which has changed considerably.
Traditional collection involved staff visiting shops and recording prices, which is expensive and produces a limited sample.
Scanner data from retailers now provides transaction-level prices and quantities at enormous scale, which allows weights to reflect actual purchasing rather than survey recall.
Web scraping captures online prices continuously, which matters increasingly as more spending moves online.
Each source introduces its own biases — scanner data covers only participating retailers, online prices may not reflect what is paid after delivery charges — and agencies combine them with adjustments.
Producer and wholesale measures
Separate indices track prices earlier in the chain, at the factory gate or at import.
These lead consumer prices to some degree, since costs pass through with a lag, which makes them useful as an early indicator.
The pass-through is incomplete and variable, since margins absorb some of the movement, so the relationship is directional rather than arithmetic.