Twice a month the calendar produces an event that can move every asset class at the same instant. One of them is the Federal Reserve meeting. The other is the US Consumer Price Index, released mid-month at 8:30 a.m. New York time by the Bureau of Labor Statistics.
The reason CPI carries that weight is not the number itself. It is what the number implies about the path of interest rates, and through that, about the discount rate applied to every future cash flow priced in dollars.
What the CPI actually measures
The CPI tracks the average change in prices paid by urban consumers for a fixed basket of goods and services: food, rent, gasoline, clothing, medical care, transport, recreation. If the basket cost $100 a year ago and costs $103 now, CPI inflation is 3%.
The BLS collects tens of thousands of price quotes every month from retailers, service providers and rental units across urban areas, then aggregates them using weights that reflect what households actually spend.
Two versions of the change are published every month, and they answer different questions:
- Month-on-month — the change since last month. Noisy, but it is where a turn in the trend appears first.
- Year-on-year — the change over twelve months. This is the number the media quotes, and it is smoother, because it carries eleven months of history with it.
The gap between the two is a common source of confusion. Annual inflation can fall while monthly inflation rises, simply because a large increase from twelve months ago has dropped out of the window. Base effects are arithmetic, not economics, and traders who forget them misread the print.
Why shelter dominates the index
The basket splits into eight major groups, and one of them overwhelms the rest. Housing carries roughly a third of the total weight, transport around a sixth, food and beverages a little less. Medical care, education and communication, recreation, apparel and everything else share what is left.
Inside housing sits shelter, and shelter is where the index gets strange. It combines actual rents with an estimate called owners' equivalent rent — what homeowners would theoretically pay to rent their own home. That estimate is surveyed rather than transacted, and it updates slowly by construction.
The consequence matters for anyone trading the release. Shelter lags the real rental market by roughly six to twelve months. When market rents turn, CPI keeps printing the old trend for months afterwards. Analysts routinely calculate what the index would show with a live rent measure substituted in, and the difference can be more than a full percentage point. The official number is not wrong; it is measuring something that responds late.
Headline against core
Headline CPI includes everything. Core CPI strips out food and energy.
The reason for stripping them is not that food and energy do not matter — they matter more to household budgets than almost anything else. It is that their prices are driven by weather, geopolitics and supply shocks that reverse. Gasoline can jump 20% on a Gulf headline and give it all back the following month. Including that noise in a measure meant to guide a policy that takes eighteen months to work produces bad policy.
Core is the trend measure, and it is the one the Fed and the market weight most heavily. A headline reading pushed up by crude, with core steady, is treated as a shock passing through. Core rising steadily is treated as inflation embedding itself in wage and price setting, which is the thing central banks genuinely fear.
Watch both, and know which one is driving the move. Markets frequently react more to a core miss of a tenth than to a headline miss of three tenths.
Supercore: services without shelter
During the 2022–2023 inflation episode a third cut entered the mainstream, and Fed officials cited it directly. Supercore is core services excluding shelter — restaurants, transport services, medical services, personal care.
The logic is to isolate the part of inflation that responds to domestic labor costs. Goods prices reflect global supply chains. Shelter reflects a lagging survey. Non-housing services reflect wages, because that is what they mostly are. If wage growth is running hot, supercore is where it shows up first and most cleanly.
For a trader, supercore is the measure that tells you whether disinflation is real or whether it is being carried entirely by falling goods prices that cannot fall forever.
CPI is not what the Fed targets
Here is the detail that catches people out. The Fed's 2% target is defined on a different index: the Personal Consumption Expenditures price index, published by the Bureau of Economic Analysis, not the BLS.
The two differ in three ways that matter:
- Scope. PCE includes spending made on consumers' behalf — most importantly medical costs paid by insurers and government programs. CPI counts only what households pay directly, which is why healthcare has a much smaller weight in CPI.
- Substitution. PCE updates its weights continuously and allows for consumers switching between goods when relative prices change. CPI uses a more fixed basket, which tends to overstate the cost of maintaining a standard of living.
- Weights. Shelter carries far less weight in PCE, which is why the two indices can tell noticeably different stories when rents are moving.
PCE typically runs a few tenths below CPI. It is also released around two weeks later, which is why CPI still dominates the immediate market reaction: it is the first hard read on the month, and the market cannot wait.
How the market actually reacts
The mechanical chain runs: CPI surprise → repricing of the expected rate path → repricing of everything priced off that path.
A hotter-than-expected print typically means a later or shallower easing path, which supports the dollar, pushes bond yields up and prices down, and pressures equities — growth names hardest, because their value sits furthest in the future. Gold usually falls, since it pays nothing and competes directly with real yields. A cooler print does the reverse.
The word doing the work in that paragraph is surprise. The level of inflation is known and priced; only the deviation from consensus is new information. A 3.0% print is bullish if the market expected 3.3% and bearish if it expected 2.7%. The absolute number tells you nothing about the reaction without the forecast beside it. Our economic calendar shows the consensus and the previous reading for exactly this reason.
The same logic explains why a "bad" inflation number sometimes produces a rally. If the market had positioned for something worse, relief is a directional force in its own right.
Trading it, or deciding not to
In the first seconds after the release, major currency pairs can move fifty to a hundred pips, gold twenty to forty dollars, index futures hundreds of points. Spreads widen by a multiple, slippage becomes severe, and a stop loss is a request rather than a guarantee.
For most retail traders, the correct position going into CPI is a small one or none. Closing exposure fifteen to thirty minutes before the release and waiting for the dust to settle is not timidity; it is refusing to take a binary bet with an execution cost you cannot measure in advance.
Traders who do work the release generally trade the second move, not the first. The initial spike is algorithmic and often reverses within two or three minutes, once the composition of the report — core versus headline, which categories drove it, what the revisions did — is digested by people rather than parsers. The reversal is the informed move. Chasing the spike puts you on the wrong side of it.
Whatever the approach, the discipline is the same one that applies to the monthly payrolls report: know what is priced, size for the volatility, and accept that the first print is not the whole report.
Three ways to read it wrong
Reading headline and stopping. The Fed does not act on a gasoline spike. Core, and increasingly supercore, is where the policy-relevant signal lives.
Ignoring the consensus. The print alone is meaningless. Without the expectation, you cannot know whether the number is hawkish or dovish, and you will be surprised by the direction of the move.
Treating one month as a trend. Single prints are noisy, seasonal adjustment is imperfect, and revisions happen. Central banks look at three- and six-month annualized rates, and so should anyone trying to anticipate them. A single CPI report almost never changes policy. A run of three in the same direction does.
The number under the number
CPI matters because it constrains what the Federal Reserve can do, and what the Fed does sets the price of dollar money for everyone. Knowing how to read it means knowing which components moved, whether the move is likely to persist, and how far it sits from what the market had already assumed.
That last part is the difference between understanding inflation and trading it. The economy is described by the level. The market is moved by the gap between the level and the forecast — and that gap is usually small, occasionally large, and always the whole story on the day.
