August's 4.9% annualized CPI rests on one 0.40% month

Core CPI in August was close to a pure services reading. Solve for the weight that turns core services at 4.0% annualized and core goods at 1.3% into a 3.5% core print and you get about 81%; do it on the monthly figures, 0.33% against 0.10% producing 0.29%, and the implied weight is 83% ^[17].

There is a smaller puzzle inside core services. The aggregate rose 0.33% in August, and each of the three components Wolf Richter itemizes rose less than that: supercore 0.31%, rent of primary residence 0.23%, owners' equivalent rent 0.19% ^[20]. Core services is housing plus supercore by construction, so either the housing components he does not break out rose faster than 0.33%, or the gap is rounding below the second decimal.

Richter's own argument is that the index understates services inflation, because OER tracks what a large panel of homeowners think their home would rent for while the costs they actually pay, insurance and property taxes and repairs, sit outside CPI entirely ^[13][24]. OER "is a fundamentally flawed metric in the CPI and should be replaced by the actual costs that homeowners face," he wrote ^[15]. The same complaint covers the medical care complex, where health insurance prints at -8.5% year-over-year and medical drugs at -2.7% ^[14]. The Fed's preferred PCE index uses both components with smaller weights ^[13].

A 0.40% month compounded twelve times is 4.91%, which is where the 4.9% comes from ^[18]. It is one month, set against two annual rates that did not move from July ^[4][5].

The 1.4 points that separate the annualized headline from annualized core are energy, since food at home was unchanged month to month ^[19][11]. Year-over-year the split is wider: energy at 16.3% against core goods at 0.7%, a spread of 15.6 points ^[22]. Gasoline alone is over half the energy index ^[10].

Richter's piece does not discuss the rate path, and its only reference to the Fed is that PCE weights the two disputed components less ^[23]. On the evidence in this release, the case that disinflation has stalled rests on a single monthly print, and I would not reprice a cut on it. Two ways that view loses: the gasoline move from late August and September lands in the September index, as Richter expects, and pushes the headline up again ^[12]; or OER at 2.3% annualized is holding down a services number that a measure of actual homeowner costs would print hotter ^[7][24]. What would settle it is a September core services month at or above 0.33% that pulls the 3.0% annual rate up with it.