Why 0.3 Is Worse Than 3.4
Two numbers, same month
Every inflation report hands you two versions of the same fact. One compares prices to a year ago — that is the year-on-year rate. The other compares them to last month — the month-on-month rate.
This week's print had a year-on-year core rate of 2.4%. That is the number the headline writers used, and it sounds like the problem is solved. The month-on-month core rate was +0.3%, against +0.2% expected. That is the number the market traded.
The second one is worse and it is not close.
Why the ugly number is the honest one
Year-on-year is arithmetic with a long memory. It divides today's price by the price from 12 months ago. If something blew up last year and then went quiet, the year-on-year number keeps falling even though nothing is getting better — you are just dividing by a bigger base. Statisticians call it a base effect, which is a polite way of saying the number is partly about last year.
Month-on-month has no such memory. It tells you what happened in the last 4 weeks. It is noisier — one airline fare or one oil spike can move it — so it is a worse signal on its own.
So why trust it here? Because of direction. Year-on-year for core has been drifting down for a while. Month-on-month just came in hotter than expected. When the slow number says fine and the fast number says hot, the fast one is usually telling you about the future and the slow one is telling you about the past.
Do the annualisation, because that is what the Fed does
Here is the trick that makes 0.3% legible. If prices rose 0.3% every month for a year, compounding, you would get roughly 3.7% annual inflation. At 0.2%, you get about 2.4%.
One tenth of a percentage point per month separates on-target from off-target. That is the entire ballgame. It is also why the market moved several times more than the headline number's apparent size would suggest — a 0.1pp monthly surprise is a 1.2pp annual one, and a 1.2pp annual inflation surprise is worth a lot of interest-rate repricing.
The vocabulary they assume you have
Six terms do most of the work in any CPI story:
- Headline — everything. This week: +0.4% m/m, 3.4% y/y.
- Core — everything except food and energy. Strips out the two most volatile categories so you can see the underlying trend. This week: +0.3% m/m, 2.4% y/y.
- Supercore — core services excluding housing. The Fed watches this one most closely because services inflation is driven by wages, and wages are sticky. This week: +0.5% m/m, 3.0% y/y — accelerating.
- Base effect — the arithmetic hangover from what happened a year ago. It can make a bad number look good.
- Basis point (bp) — one hundredth of a percentage point. 0.25% is 25bp. You will see this everywhere; it exists so people can say "3 basis points" instead of "zero point zero three percent."
- Supercore, again — worth repeating because it is the one people skip, and it is the one that moved.
What actually broke
Energy is running +16.3% year-on-year. Core PCE — the Fed's preferred measure, a different index published later — sits at 3.3%, above the 2% target for roughly five and a half years.
Put together: the disinflation story was real, and it has stalled. The components still falling are the ones that already fell. The ones still rising are the ones tied to domestic wages and shelter, which is exactly what a central bank cannot fix by waiting.
Market response within the hour: the 2yr Treasury yield, which is the purest read on expected Fed policy, rose to 4.56%; the 10yr to 4.95%, a 3yr high; the 30yr to 5.37%.
If a single monthly decimal can do that to the price of government debt, it is worth understanding.