We have updated all of our models to reflect the recent developments regarding tariffs. In doing so, we have adopted a working assumption that broadly follows estimates from the Tax Foundation. In particular, we assume that the new measures imply an effective tariff rate of approximately 12 percent, declining to around 6–7 percent once Section 122 expires.
Overall, the model results remain broadly consistent with previous runs. The new assumptions translate into a modest downward revision—of roughly one tenth of a percentage point—to the medium-term forecast.
A note of caution regarding the current model results. The environment in which we are operating remains highly volatile, and as a consequence the models are revising their forecasts more frequently and more materially than they would under more stable conditions.
A few considerations are in order for each model:
- The FRB model has not yet been updated, as the Board has—somewhat surprisingly—not released an updated dataset. We have been in contact with some former colleagues, who have indicated that a new dataset should be made available as soon as possible. Nevertheless, the (huge) delay is difficult to explain and reinforces the view that many changes are needed at the Board.
- Second, our machine learning (ML) model, based on past experience, tends to struggle in fully capturing movements in food/energy futures. However, it has proven particularly useful in forecasting core CPI, especially for the front month. In recent months, the ML model has in fact outperformed our bottom-up approach. The practical implication is that, at this stage, we would place somewhat less weight on the model’s headline signal and relatively more on its core signal.
- As for the other models, both the bottom-up framework, the CPI-to-PCE translation, and our “main” PC model remain, in our view, the most reliable tools in our suite. These models have continued to send a consistent message in recent months: inflation should remain broadly contained going forward, but there is still no indication that it has returned to target.
Given the highly volatile environment, we will be updating all models again shortly. We will keep you informed as new results become available.
All models forecasts
Table 1. All models’ forecasts for core, headline CPI and PCE.
Please, find below all models’ forecasts for core, headline CPI and PCE. The colors in the table indicates the left-hand-side variable. The numbers in the table can be found here.
Note: “ML” refers to “machine learning” model (see documentation here), “Bottom up” refers to our set of sectoral models, “main” model refers to our Phillips curve model (see here), and “FRB-US” stands for “Federal Reserve Board – US model” (see here). All MoM refer to the seasonally-adjusted figures.