Conflicting signals, not at target. On the one hand, the distribution shows some progress: less dispersed, lower median. On the other hand, the CI model estimates a very strong common component, one of the highest readings of the year. We have learned to discount the news when the models send conflicting signals. Therefore, we do not give particular weight to today’s report. Translated: we are not at target, and we are likely to remain around 2.5% (in PCE space) going forward (reminder: we estimate pi* at 2.4% right now in PCE space). This report does not clarify anything else.
As for the medium-term model, the forecast is little changed, as our Q3 nowcast turned out to be accurate. The model remains above the Fed target going forward. In our view, the FOMC is likely to revise (back) up its 2024 core PCE forecast to 2.8%.
A PPT containing all relevant CPI/PCE charts can be downloaded here.
Evidence from the distributions
Not consistent with target. This month, the distribution has made progress and it is less dispersed than last month (Figure 1); having said that, in this highly volatile environment we have learned to take little signal from a single month. Zooming out, in the last 3 months (Figure 2) there are some signs of progress, but the distribution remains different than pre-Covid. Finally, the median (Figure 3) remains volatile and ticked down again this month. Overall, we are still careful taking signal because of possible residual seasonality and because, as mentioned, the distributions are still not consistent with target.
We expect core PCE prices to grow at an average of 18bps in the remaining months of the year, and the YoY at 2.8% in December 2024.
Figure 1. Distribution of PCE excluding food and energy items changes (%, a.r.)
Note: The Figure shows the fitted Kernel (Epanechnikov) distribution of MoM percent changes at annual rate of PCE prices excluding food and energy items. The colors indicate the percentiles: 0-10pct, 10-25pct, etc. The dashed line shows the median of the distribution.
Figure 2. Kernel of PCE excluding food and energy items changes (%, a.r.)
Note: The Figure shows the fitted Kernel (Epanechnikov) distribution of MoM percent changes at annual rate of PCE prices excluding food and energy items.
Figure 3. Median PCE price increase
Note: The Figure shows the median (MoM %, a.r.) of the distribution of PCE prices changes excluding food and energy items (left panel) and the YoY (right panel).
Evidence from our Common-Idiosyncratic (CI) model
Common component around target. Figure 4 shows the decomposition of the MoM of core PCE in the “common” component (blue bars) and the “idiosyncratic” component (yellow bars). The model estimates that this month the common component increased by 28bps, while the idiosyncratic shock is a small negative (-3bp). The 3m/3m ar of the common component (Figure 5) is now running around target but we expect some tick up in the coming months. Overall, the signal of the common component (Figure 5) seems roughly in line with the one of the distributions and the latest estimate of pi* (at 2.4%).
Figure 4. Contributions to MoM changes of PCE excluding food and energy items (CI-C model)
Note: The Figure shows the decomposition of the MoM percent changes of PCE prices excluding food and energy items. The contributions are estimated using our CI-C model, a 2-stage OLS-LASSO regression model. The “Covid” effect is identified with price variations outside the 10th-90th percentiles of each item pre-Covid price change distribution.
Figure 5. Estimated “Common” component: YoY, 3m/3m a.r. and 6m/6m a.r.
Note: the Figure shows the 3m/3m at annual rate (green line), the 6m/6m at annual rate (red line), and the YoY (blue line) of the “common component” estimated using our CI-C model.
Implications for the medium-term forecast of core PCE price inflation
The medium-term forecast is little changed. Today’s data imply no change to the medium-term forecast. The current (Q4/Q4) model forecast is: 2.8% in 2024, 2.4% in 2025, 2.4% in 2026, and 2.4% in 2027.
Note: The figure shows the latest run of our “main” Phillips curve model. The confidence intervals (C.I.) are estimated using quasi-out-of-sample methods (estimate the model over a sub-sample, forecast, and calculate the root mean squared forecast errors).