Reminder: Riccardo Trezzi will be in NY and DC soon. Get in touch to schedule a meeting!
As Expected.
The August reading came in as expected and left our Q3 nowcast unrevised. Once again, our Excel-based translation proved to be perfect (within 1bps) for both, core and headline PCE.
From the models perspective: The distributions of price changes are progressively less friendly, the common component of our CI model is above target, as well as the medium-term forecast. Put it simply: not at target and no signs of improvement.
As for the Fed: the September SEP was very clear. For now, we expect Powell to focus on the labor market, despite the risks flagged by the models.
A PPT containing all relevant CPI/PCE charts can be downloaded here.
Evidence from the distributions
Not consistent with target.
This month’s distribution appears a bit less favorable compared to last month (Figure 1), which itself was already notably less favorable than the preceding month. A broader perspective (Figure 2) reveals that the current distribution remains significantly different from the pre-COVID period; indeed, recent months have shown a deterioration, characterized by increased dispersion and a more pronounced right tail. Furthermore, the median (Figure 3) continues to exhibit considerable volatility and has moved up again this month.
In summary, we are not at target and there are no signs of improvement.
Figure 1. Distribution of PCE excluding food and energy items changes (%, a.r.)
Note: The figure displays the fitted Kernel (Epanechnikov) distribution of the MoM percent changes at an annualized rate for PCE prices excluding food and energy items. The colors represent different percentiles (0-10th, 10-25th, etc.), while the dashed line indicates the median of the distribution.
Figure 2. Kernel of PCE excluding food and energy items changes (%, a.r.)
Note: The figure presents the fitted Kernel (Epanechnikov) distribution of the MoM percent changes at an annualized rate for PCE prices excluding food and energy items.
Figure 3. Median PCE price increase
Note: The figure displays the median MoM % (annualized rate) of the distribution of PCE price changes excluding food and energy items (left panel) and the YoY rate (right panel).
Evidence from our Common-Idiosyncratic (CI) model
Common component above target.
Figure 4 illustrates the decomposition of the MoM core PCE into the “common” component (blue bars) and the “idiosyncratic” component (yellow bars). This month, the model estimates that the common component increased by 21bps, while the idiosyncratic shock was small and positive (2bps). The 3m/3m annualized rate of the common component (Figure 5) is currently running above target at 2.8%.
Overall, the signal from the common component (Figure 5) are roughly in line with the one from the distributions and our latest estimate of pi* (at 2.6%).
An Excel file containing the results of the CI model shown in Figure 4 is available here.
Figure 4. Contributions to MoM changes of PCE excluding food and energy items (CI-C model)
Note: The figure presents the decomposition of the MoM percent changes in PCE prices excluding food and energy. Contributions are estimated using our CI model.
Figure 5. Estimated “Common” component: YoY, 3m/3m a.r. and 6m/6m a.r.
Note: The figure displays the 3m/3m annualized rate (green line), the 6m/6m annualized rate (red line), and the YoY rate (blue line) of the “common component,” as estimated using our CI model.
Implications for the medium-term forecast of core PCE price inflation
The medium-term revised down. The data released today (and the new assumption for average tariff rate) had only a minimal impact on the medium-term projection. The forecast is essentially unrevised.
The model’s latest Q4/Q4 forecasts are as follows: 3.1% in 2025, 3.1% in 2026, 2.7% in 2027, and 2.6% in 2028. This forecast remains above the latest SEP at the end of the horizon.
Note: The figure presents the latest run of our “main” Phillips curve model. The confidence intervals (C.I.) are estimated using quasi-out-of-sample methods—specifically, by estimating the model over a sub-sample, generating forecasts, and calculating the root mean squared forecast errors.