Numerous commentators have noted anomalies in the latest CPI release. Smith/Bloomberg quotes
“Lost in Translation,” according to TD Securities. “Delayed and Patchy,” per William Blair, and a “Swiss Cheese CPI report” from EY-Parthenon.
In contrast, the downside surprise was hailed by credulous NEC Director Kevin Hassett as “astonishingly good”.
Here are some visualizations to illustrate the anomalies in the report.
Figure 1: Quarter on quarter CPI inflation annualized (bold black), Bloomberg consensus (blue triangle), Cleveland Fed nowcast (tan). Source: BLS, Bloomberg, Cleveland Fed, and author’s calculations.
Figure 2: Quarter on quarter Core CPI inflation annualized (bold black), Bloomberg consensus (blue triangle), Cleveland Fed nowcast (tan). Source: BLS, Bloomberg, Cleveland Fed, and author’s calculations.
How unexpected are these realizations? Consider core CPI? I estimate a ARIMA(1,1,0) of (log) core CPI over the 2022-2025 period. In Figure 3, I plot the dynamic forecast along with ± 2 standard error.
Figure 3: Core CPI (bold black), ARIMA(1,1,0) (blue), ± 2 standard error prediction interval (gray), 1982-84=100, all s.a. Source: BLS and author’s calculations.
The November realization is just at the outside 95% band. In that sense, the core CPI number is quite “astonishing”.
The treatment of shelter costs has been identified as a key issue. Essentially, no rent and OER inflation was assumed for October, dragging down the CPI level in November. In Figure 4, q/q CPI and CPI ex-shelter inflation is compared.
Figure 4: Quarter on quarter CPI inflation annualized (bold black), CPI ex-shelter inflation (green). Source: BLS and author’s calculations.
Other issues include the truncated sampling period for November — i.e., the latter half of the month, when holiday sale prices would be pervasive, thereby biasing surveyed prices downward.
In other words, it makes sense to be wary of the November numbers.
