How Do You Measure a Whole Economy's Prices? Building a Number From Many

Prices do not move together, so summarising them means weighting some more heavily than others, and the weights are where the difficulty lives. Weighting by how much people actually buy seems obvious and immediately raises a question, since people buy different amounts at different times and partly because prices changed. Using the quantities from the earlier period tends to overstate the rise, because it ignores the substitution people make away from items that became expensive. Using the quantities from the later period tends to understate it for the mirror-image reason. Neither is wrong exactly, and the gap between them is real and measurable. Various compromises exist, including taking a geometric average of the two, and the choice between them shifts the published figure by amounts that matter.

Building the index requires choices that are judgements rather than calculations:

• •Which items to include and how to keep the basket representative as habits change
• •How often to update the weights, since stale weights drift away from what people buy
• •How to handle goods that improve, since a better product at the same price is a price fall in real terms
• •How to treat housing, which can be counted as rent, as purchase cost or as an imputed rent
• •Which population the index represents, since the poor and the old face different baskets
• •How to handle items that disappear from the market entirely

Goods that improve over time pose a genuine difficulty. A computer costing the same as one from a decade ago is enormously more capable, so treating the price as unchanged overstates inflation, and statistical agencies therefore adjust for quality using techniques that estimate how much of a price difference is attributable to measurable characteristics. Those adjustments are large in some categories and are contested, since they systematically reduce measured inflation and the estimation involves assumptions. A commission reporting in the United States in 1996 concluded that the consumer index there overstated inflation by around a percentage point annually, largely through substitution and quality effects, which sounds small and compounds into an enormous sum given how much government spending is indexed. Methods changed in response, and the debate about whether the correction was right continues.

The method extends well beyond prices and the same difficulties follow it. Economic output at constant prices requires deflating current values by an index, so every measurement problem in the price index propagates into the growth figures, and international comparisons of output require a further index converting currencies by what they actually buy. Wage indices, house price indices and share indices all face the weighting and composition problems in their own forms, with share indices additionally having to decide what happens when a company leaves or enters. Composite indicators ranking countries by development, competitiveness or freedom combine unlike quantities using weights that are chosen rather than derived, which makes them considerably more arbitrary than a price index and considerably more widely quoted without qualification.

The published figure is not merely descriptive, since large amounts of money move automatically according to it. Pensions and benefits are uprated by it in many countries, so a lower measured inflation figure reduces payments permanently. Wage agreements reference it. Index-linked government bonds pay according to it, which means the issuing government has an interest in the measurement. Tax thresholds are adjusted by it. Central banks target it directly, so monetary policy responds to the number rather than to prices as such. That combination makes the methodology politically consequential, and changes to it have been introduced and reversed amid substantial argument, including a switch in the United Kingdom between two different formulas that produced a persistent gap and affected pensions, rail fares and student loan interest for years.

Summarising many prices requires weights, and using early quantities overstates the rise while using later ones understates it, with the gap being real and measurable. Quality improvements must be adjusted for, which is large, contested and systematically reduces the measured figure. Pensions, wages, bonds and tax thresholds move automatically with the result, which is why the method is fought over.