The inflation figure that leads the news is the consumer price index. The inflation figure the Federal Reserve actually targets is a different one — the personal consumption expenditures price index — and it is consistently lower. The gap is not a rounding difference. It is structural, it is well understood, and it regularly runs to several tenths of a percentage point.
Anyone comparing a CPI print against the Fed’s 2 percent objective is comparing the wrong number to the target.
Four reasons they differ
Weights. CPI weights are derived from a survey asking households what they spent. PCE weights come from business surveys recording what was sold. The second captures spending made on households’ behalf — most consequentially, employer-provided and government-funded healthcare. Healthcare therefore carries roughly three times the weight in PCE that it does in CPI, and housing carries considerably less.
Substitution. This is the most important structural difference. CPI uses a basket that is updated periodically but assumes within-period that consumers buy roughly the same things. PCE uses a formula that allows the weights to shift continuously as relative prices change — if beef becomes expensive and consumers buy more chicken, PCE reflects the shift as it happens.
Because consumers do substitute away from things that become expensive, allowing for it produces a lower measured inflation rate. This accounts for a persistent portion of the gap and it runs in the same direction essentially always.
Scope. CPI measures out-of-pocket spending by urban consumers. PCE measures all consumption in the economy, including rural households and spending by non-profits serving households.
Revision policy. CPI is essentially never revised once published. PCE is revised routinely as more complete source data arrives, and the revisions can be meaningful. This matters for anyone reconstructing what policymakers were looking at when they made a decision — the figure they saw is often not the figure in the database today.
Why the Fed chose the one it did
The switch was made explicitly and the reasoning was published. Three arguments carried it.
- It covers more of what people actually consume, including the substantial healthcare spending that never passes through a household’s hands.
- The substitution formula is closer to a true cost-of-living measure. The question monetary policy cares about is what it costs to maintain a standard of living, not what a fixed basket costs.
- The weights update continuously, so it adapts to changing consumption patterns without waiting for a periodic revision.
The counter-argument is straightforward and worth stating. CPI is what most people experience. Wage negotiations, pension indexation and inflation-linked bonds are tied to it. A central bank targeting a measure that runs systematically below the one governing household contracts is targeting something the public does not feel — and the credibility of an inflation target depends on the public believing it describes their circumstances.
Why “core” at all
Both indices have core versions excluding food and energy. The exclusion strikes many people as evasive — food and energy are precisely what households notice — and the objection deserves an answer.
The case for core is not that food and energy do not matter. It is that they are volatile and driven substantially by supply factors that monetary policy cannot address. A central bank cannot raise interest rates to increase the oil supply. Reacting to an energy price spike by tightening would suppress demand across the whole economy to offset a shock that will likely reverse on its own.
Core is used as a predictor of where headline inflation is heading once temporary shocks pass. Historically it has done that job better than headline predicting itself.
The limitation is real, though. If energy prices rise persistently rather than temporarily, they pass through into core over time via transport and production costs. Core is not immune to energy; it is lagged. Sustained headline inflation eventually becomes core inflation.
The measure officials actually watch
Beyond core PCE, policymakers frequently reference core services excluding housing. The logic of that construction is worth understanding.
Goods prices are heavily influenced by global supply chains and exchange rates. Housing is measured through a methodology that lags observable market rents by a year or more. Strip both out and what remains — services other than shelter — is the category most closely tied to domestic wage costs, and therefore the part of inflation most responsive to domestic monetary policy.
When an official says inflation is proving persistent despite a falling headline rate, this is usually the series they mean.
The practical rules
- Compare like with like. The 2 percent target refers to headline PCE. A CPI print is not directly comparable to it.
- Expect PCE to run below CPI. When it does not, something unusual is happening in the components where they differ — usually healthcare or housing.
- CPI arrives first. It is released earlier and several PCE components are constructed from CPI and PPI source data, which is why desks estimate PCE from those two releases before it is published.
- Neither is your personal inflation rate. Both describe an average basket. A household spending a third of its income on rent experiences something quite different from the published figure.
The wider point is that there is no single inflation number, only a family of measures built on different assumptions for different purposes — a theme we developed in how to read inflation data without being misled.
