The Fed settled this argument in writing years ago and people still get it wrong. Since 2012 the FOMC's longer-run goals statement has defined the 2 percent objective in terms of the price index for personal consumption expenditures, PCE, published by the Bureau of Economic Analysis, and not the CPI that dominates the financial press. And yet CPI morning is the one where implied volatility spikes, where index options price some of the biggest moves on the macro calendar, and where half the traders I know flatten their books rather than hold through the release. So the Fed watches one number while the market trades another, and once you see why, both releases get much easier to use.
One nuance before the mechanics. The formal target is headline PCE, but the Fed leans on core PCE, which strips out food and energy, as its working signal, because food and energy are mostly noise over the horizon monetary policy operates on. When a Fed official talks about progress on inflation, core PCE is almost always the series behind the sentence.
Same economy, two different questions
CPI comes from the Bureau of Labor Statistics and answers a household question: what do urban consumers pay out of pocket for a basket of goods and services. PCE comes from the national accounts and answers a broader one: what is being consumed on behalf of households, whoever pays for it. Most of the divergence between the two traces back to that single design choice.
Housing is the clearest case. Shelter is roughly a third of the CPI basket, mostly rent and owners' equivalent rent. In PCE it carries roughly half that weight. So a shock to rents moves CPI about twice as hard as it moves PCE, and any stretch where shelter is the main inflation story will open a wide gap between the two indexes.
Healthcare is the mirror image. CPI counts only what you personally pay, so its medical weight is small. PCE counts what employers, insurers, Medicare, and Medicaid spend on your behalf, so healthcare carries a much larger weight there. It is also measured from a different angle. CPI medical prices are consumer-facing, while PCE leans on producer-side data that reflects what insurers and the government actually pay providers. The two can genuinely disagree on the direction of medical inflation in the same month, and that disagreement flows straight into the wedge between the indexes.
The substitution effect, and why PCE runs cooler
The formulas differ too. CPI is close to a fixed-basket index. Its weights get updated on a schedule, but between updates it keeps pricing the old basket, so when beef gets expensive and everyone quietly switches to chicken, CPI keeps billing you for beef. PCE uses a chained formula whose weights move with actual spending, so it captures the switch as it happens.
The practical result is that PCE inflation has historically run a few tenths of a percentage point below CPI inflation over long stretches. That means the Fed's 2 percent, translated into CPI terms, sits somewhere in the low twos. It matters when you read TIPS breakevens, because TIPS and inflation swaps settle on CPI, so a breakeven modestly above 2 percent can be fully consistent with the Fed hitting its goal. I have watched people call breakevens unanchored without ever making that adjustment.
Two smaller differences are worth keeping in your head. PCE covers the whole population plus nonprofits serving households, while CPI covers urban consumers only. And PCE gets revised, repeatedly, sometimes years later, while the non-seasonally-adjusted CPI is essentially never touched after release. The number the Fed reacts to in real time can quietly become a different number a year down the line.
Why markets trade CPI anyway
If the Fed targets PCE, why does CPI own the volatility? Three reasons, and none of them are irrational.
- Timing. CPI for a given month lands roughly two weeks before the PCE report covering the same month. It is the first serious look at the inflation data, and markets pay up for first looks.
- Plumbing. Inflation swaps, TIPS, and most cost-of-living adjustments key off CPI. There are mechanical flows tied to the print itself, independent of anything it implies about policy.
- Forecastability. By the time PCE is released, analysts have already mapped the CPI and PPI components into a core PCE estimate, usually accurate to within a few hundredths. Most of the information in PCE day has leaked out before it arrives.
The forecastability point is the one that shapes how I spend my attention. PCE surprises are rare because PCE is largely a function of two earlier releases. The interesting exceptions come from components that live in PCE but barely respond to CPI, like healthcare measured from the producer side, or portfolio management fees, which track the stock market closely enough that a big equity rally shows up as services inflation shortly after. When those wedge components move, the CPI-implied nowcast misses and PCE day briefly matters again.
How I actually use the two releases
My rule of thumb is simple: trade CPI for the volatility, read core PCE for the policy path. On CPI morning the sequence I care about is core before headline, and inside core, shelter versus everything else. The Fed knows CPI shelter lags market rents by something like a year, so a hot print driven by shelter gets discounted, while heat in core services outside of housing does not. A hot headline sitting on soft core-ex-shelter and a soft headline hiding hot core-ex-shelter are opposite signals wearing the same surprise.
Then PPI lands, usually within a day or two of CPI, and the desks publish their core PCE translations. That estimate is the number that shows up later in Fed speeches and projections, while the CPI print that moved the market fades into the background. If the CPI surprise came from categories that do not pass through to PCE, the market reaction and the policy implication have quietly diverged, and that gap is where the more interesting trades tend to live.
The failure mode I see most often is fading a hot CPI on the logic that the Fed targets PCE anyway, without checking where the heat came from. If it came from shelter, fine, the discount is real. If it came from categories that map almost one to one into core PCE, you are fading a number the Fed will see again two weeks later under its preferred name. This is roughly the logic behind how we tag inflation releases in Blockcircle's macro scorecards, with CPI treated as the volatility event and PCE as the policy signal, and the wedge components watched in between.
None of this makes the prints predictable, and I would not pretend my own CPI-day record is anything special. But knowing which release carries the volatility and which one carries the policy signal at least stops you from being surprised that the market cares loudly about a number the Fed only glances at.