Inflation: CPI and PCE
Lesson 9 · about 11 min
Inflation data decides what the central bank does next, so in any regime where the bank is focused on prices, the inflation release is the most important scheduled event of the month. There are two US measures that matter and they are not the same thing.
CPI
The Consumer Price Index is published by the Bureau of Labor Statistics (BLS), usually between the 10th and the 15th of the month for the prior month, at 8:30 a.m. Eastern. It measures the price of a fixed basket of goods and services bought by urban consumers.
Four numbers come out at once and the market reads them in this order:
| Series | What it is | Why it matters |
|---|---|---|
| Core CPI m/m | Month-on-month change excluding food and energy | The market's primary focus; forecast to a tenth of a percent, so a 0.1 miss is a big surprise |
| Headline CPI m/m | Month-on-month change including everything | Matters for consumers and for headlines; energy makes it noisy |
| Core CPI y/y | Twelve-month change excluding food and energy | The level the central bank compares to its target |
| Headline CPI y/y | Twelve-month change, all items | The number the public sees |
Within core, traders have learned to look at the components that the Fed has flagged. Shelter (rent and owners' equivalent rent) is about a third of CPI and moves slowly. "Supercore," services excluding shelter and energy, was the Fed's stated focus in 2023 because it tracks wage pressure. Used cars, airfares and medical services have each been the swing factor in individual months.
The unrounded number matters. Core CPI of 0.3% could be 0.25% rounded up or 0.34% rounded down, and those are different signals. BLS publishes to three decimals, and the desk reads the third decimal.
PCE
The Personal Consumption Expenditures price index comes from the Bureau of Economic Analysis (BEA), usually in the last week of the month, at 8:30 a.m. Eastern. It is the Fed's official target measure: the 2% goal is defined on headline PCE, and the Fed's practical focus is core PCE.
PCE differs from CPI in three ways that matter:
- Weights. PCE weights come from what people actually spend, updated continuously, so when beef gets expensive and people buy chicken, PCE captures the switch and CPI does not. PCE shelter weight is lower than CPI shelter weight.
- Scope. PCE includes spending made on behalf of consumers, notably employer-paid health insurance and Medicare, which CPI excludes. Health care is a larger share of PCE.
- Level. Because of the above, core PCE tends to run a few tenths below core CPI over time.
The consequence for trading is that PCE is usually less of a surprise than CPI. By the time PCE comes out, CPI and the Producer Price Index (PPI) for the same month have both been released, and economists map the components across. The consensus for core PCE is often accurate to 0.1, and the market move is typically a fraction of CPI day's. The exception is when the CPI-to-PCE mapping breaks, which happens when the components that differ between them (health care, financial services, airfares) move sharply.
Key idea: CPI is the surprise, PCE is the target. CPI day moves markets because it is the first read; PCE day moves less because CPI and PPI have already told you most of the answer.
The rest of the inflation set
| Release | Source | When (typical) | Time (ET) | Market weight |
|---|---|---|---|---|
| CPI | BLS | 10th-15th | 8:30 am | Very high |
| PPI | BLS | Day after or a few days around CPI | 8:30 am | Moderate; used to refine PCE forecast |
| PCE (with personal income and spending) | BEA | Last week of month | 8:30 am | Moderate, higher if CPI-PCE mapping is uncertain |
| Import prices | BLS | Mid-month | 8:30 am | Low |
| Michigan inflation expectations | University of Michigan | Prelim mid-month Friday, final end of month | 10:00 am | Low to moderate; spikes get quoted by the Fed |
| Employment Cost Index | BLS | Quarterly, end of Jan/Apr/Jul/Oct | 8:30 am | Moderate; the Fed's preferred wage measure |
Outside the US: euro area flash HICP at the start of the month (Eurostat, 11:00 CET), UK CPI mid-month (ONS, 7:00 London), and Japanese national CPI near month-end (with Tokyo CPI a few weeks earlier as a leading read).
How the market reacts
On a CPI surprise the front end of the curve moves first, then the dollar, then index futures. A hot core CPI (above consensus) in an inflation-fighting regime typically means 2-year yields up, dollar up, ES and NQ down, gold down. A cool print reverses all four. The size is discussed in the last lesson of this module; the caveat is that regime matters enormously, and there were stretches in 2019 and 2021 when CPI barely moved anything.
The second-order read is the mix. A hot headline driven by energy with a soft core is treated as mostly benign. A soft headline with a hot core and hot supercore is treated as hawkish, even if the newspaper says inflation fell.
Base effects
Year-on-year inflation is the sum of twelve monthly changes. When a large monthly print from a year ago drops out of the window, y/y falls even if the current month is unremarkable. This is a base effect. Economists know about it in advance, so it is in the consensus, but it explains why y/y can fall while m/m rises. When you are following the level (Module 1), watch the three-month and six-month annualised rates, which show the current trend without the base-effect distortion.
Try it: For the last CPI release, find the consensus and actual for all four headline numbers plus the unrounded core figure. Note which components the press coverage blamed. Then write one sentence on whether the mix was hawkish or dovish, independent of the headline surprise.
Recap
- CPI (BLS, mid-month, 8:30 ET) is the market's main inflation surprise; core m/m to the third decimal is the key number.
- PCE (BEA, month-end) is the Fed's target measure but is largely predictable from CPI and PPI.
- CPI and PCE differ in weights, scope and level; core PCE tends to run below core CPI.
- Read the mix: shelter, supercore, and one-off components decide whether a print is hawkish beyond the headline.
- Use three- and six-month annualised rates to see the trend through base effects.
See it drawn
Original diagrams for the ideas on this page. Illustrative, not real market data.