We have revised our forecasts. While we typically adopt a highly conservative approach and update our projections only when strictly necessary, the current environment is characterized by more frequent and pronounced shocks. As a result, our models have adjusted the forecasts more frequently to reflect these dynamics.
Below, we present the updated forecasts derived from both our bottom-up and top-down models for CPI and PCE price inflation. These revisions incorporate the latest available data, including January CPI, PPIs, import prices, and credible announcements regarding tariffs.
Please refer to the conclusion of this note for our commentary on how to interpret and use the results of these models.
CPI space
Bottom-up forecast. We forecast the YoY of core CPI at 3.05% in December 2025, and the YoY of headline CPI at 2.9%. Reminder: this is the forecast we construct with a bottom-up approach covering 22 sectors and using ad-hoc sector models. The MoM of this forecast is what we circulate as a CPI preview each month. The YoY forecast of core CPI is marginally lower than post CPI following some minor rethinking and re-estimates.
Machine Learning model. The model is forecasting core CPI at 3.4% (YoY) in December 2025 (see figure), with the YoY of headline CPI expected at 3.1% (see figure). An Excel file containing the MoM and YoY forecasts for core CPI is here and for headline CPI is here. The model documentation is here.
“Main” model. The model forecast (Q4/Q4) for core CPI is 3.1% in 2025, 2.8% in 2026, and 2.6% in 2027 (see figure). As for headline, the Q4/Q4 model forecast is 2.5% in 2025, 2.2% in 2026, and 2.3% in 2027 (see figure). An Excel file with the forecast for core CPI is here and for headline is here. Reminder: this is a medium-term model based on Detmeister et al. (2014).
PCE space
Translation of our CPI bottom-up forecast. We forecast the YoY of core PCE at 2.75% and the YoY of headline PCE at 2.5% in December 2025.
“Main” model. The model forecast (Q4/Q4) for core PCE is 2.7%, 2.6% in 2026, and 2.44% in 2027 (figure here). For headline PCE, the model forecast is 2.2% in 2025, 2.0% in 2026, and 2.1% in 2027 (figure here). An Excel file with the forecast for core PCE is here and for headline is here. Reminder: this is a medium-term model based on Detmeister et al. (2014).
How to use our models
Our approach to modeling is straightforward: the models are tools we consistently run and closely monitor. However, we acknowledge that each model has inherent limitations and is optimized for specific objectives. Most of our models are designed to minimize one-year-ahead root mean square forecast errors, rather than focusing on shorter horizons such as one-month or three-month periods. Consequently, the models are best suited for medium-term forecasting. Our primary focus is core inflation.
For example, the “main” model’s October 2023 forecast for core PCE growth from Q4/Q4 2024 (see here) was remarkably accurate, aligning perfectly with actual outcomes. Furthermore, the Q4/Q4 forecast for 2025, generated in October 2023, deviates by only one-tenth from the current projection—despite being significantly above consensus at the time.
In summary, we recommend consulting the models consistently, using them primarily for medium-term forecasts. However, they can also be applied to near-term (MoM) projections when their information is supplemented with real-time data, such as gasoline prices for the prior month (“t-1”) available pre-CPI report each month.
At present, we find that the bottom-up and “main” model forecasts for CPI are well-positioned. However, we believe the machine learning model is likely overstating inflationary pressures. Regarding PCE forecasts, we consider the core PCE projections to be well-positioned, while headline PCE forecasts appear slightly underestimated—largely due to the models’ challenges in accurately projecting food prices.