Is inflation understated?

A 'theoretical fixed basket' doesn't reflect what people buy

Question: Is inflation understated?
Position1 of 3›: No, inflation is overstated
Argument1 of 3›: A 'theoretical fixed basket' doesn't reflect what people buy

When prices rise, consumers buy cheaper goods rather than higher priced goods.

The argument

Rising prices lead to people substituting certain more expensive goods with cheaper ones. For example, when the price of steak rises, households will often buy more chicken and fish instead. This means that the basket that consumers actually purchase is generally lower than a theoretical fixed basket. Economists refer to this as a substitution effect. The Consumer Price Index in the late 1990s started incorporating the substitution effect using geometric weighting, but it still displays the tendency for inflation to lean towards being overstated.

Premises

[P1]When prices rise, consumers substitute cheaper goods for dearer ones—buying chicken and fish when steak gets expensive—so the basket people actually buy costs less than a theoretical fixed basket, an effect economists call the substitution effect. [P2] Although the Consumer Price Index began incorporating this substitution effect via geometric weighting in the late 1990s, it still tends to overstate inflation. [C] Because the fixed-basket method overlooks real substitution behaviour, inflation is overstated rather than understated.

Counter-arguments

The argument describes a bias that statistical agencies identified and corrected, then asserts it survives without saying why. Its own text concedes that substitution was incorporated through geometric weighting in the late 1990s, and further changes followed — rotating outlet and item samples, hedonic adjustment for quality change, and chained indices that update weights continuously rather than holding a basket fixed. The fixed-basket criticism therefore describes a superseded methodology, and the conclusion that the tendency "still" persists is stated rather than argued. Substitution also cuts both ways. If a household switches from steak to chicken because steak has become dearer, its spending is lower but so is what it consumes; treating the substituted basket as equivalent assumes the cheaper good delivers the same satisfaction, which is exactly what is in dispute. The switch is a response to a price rise, and a measure that nets it out records a smaller increase than the household actually experienced — on which reading substitution adjustment makes the index understate, the second sibling position's claim. The argument finally names one direction of measurement error while several run the other way: new goods enter indices with a lag, quality adjustment is contested in both directions, and outlet substitution and package downsizing are imperfectly captured. Reviews of the evidence have generally found residual bias small and its sign uncertain, which supports the third sibling position that the measure is broadly accurate.

Rejecting the premises

[Rejecting P1] Substitution is real, but the premise concedes the correction: geometric weighting was introduced in the late 1990s, and chained indices, rotating samples and quality adjustment followed, so the fixed-basket criticism describes a superseded methodology. [Rejecting P2] Netting out substitution assumes the cheaper good delivers equivalent satisfaction, which is what is disputed, since a household switching in response to a price rise is worse off — so the adjustment can make the index understate, while reviews of measurement error find residual bias small and of uncertain sign.