Weekly Forecasts:
U.S. inflation cooled further, but inflation expectations skyrocketed.
We conduct an impulse-response analysis on drivers of U.S. inflation.
Our forecasts indicate U.S. inflation will accelerate.
The U.S. inflation cooled in April, but there were (continued) disturbing movements in the underlying dynamics. Month-over-month rates increased, and inflation expectations of households simply exploded.
We cover these underlying currents this week by concentrating on the month-over-month changes in U.S. inflation measures instead of the year-over-year changes. This brings more concise results. Our impulse-response analysis suggests that household inflation expectations are a significant long-term driver of U.S. core inflation series, in particular. The impact is also visible in our forecasts, which indicate that we would see an acceleration of U.S. inflation, especially during the fall and winter. Such long-term forecasts naturally carry within themselves a notable dose of uncertainty. However, it is interesting to see what comes and how our forecasts hit their mark.
Tuomas
U.S. inflation cools further
The U.S. Bureau of Labor Statistics published its April inflation figures on Tuesday. Here’s the breakdown.
The consumer price index (all items, all urban consumers, U.S. city average, seasonally adjusted), or CPI, rose by 2.3% annualized, 0.2% month-over-month (MoM). The core index, deducting food and energy prices from the CPI, rose by 2.8% annualized, 0.2% MoM. The “super-core,” i.e., services less energy costs, rose by 3.6% annualized, 0.3% MoM. The pace of monthly increases in all three main measures of U.S. inflation accelerated from March. In March, CPI declined by 0.1% (MoM), while core and super-core rose by 0.1% (MoM). Our forecasts for month-over-month changes, presented below, imply that this acceleration would continue.
We have been rather outspoken that the inflation “crisis” is not over yet. Most of the decline in the U.S. CPI is due to energy costs. Gasoline and energy costs have fallen by 11.8% and 11.5%, respectively, during the year. The oil price is dropping, which hints at an approaching global recession, or at least a downturn. The price currently hovers near the 10-year trend of $63.8. In mid-April, we also noted that the price increases by tariffs can be smaller than many think, because “in the short term, companies usually try to bury price increases into their contribution margins.” These create powerful downward pressure for prices.
In mid-February, we also asked, “Will DOGE crush inflation?” Our answer was:
Inflation outlook crucially depends on:
The depth of the cuts to domestic federal spending.
The tariffs.
Geopolitics.
If the first one are deep enough, they will bring about an U.S. recession, which will kill inflation. However, tariffs can re-invigorate it, in combination with possible escalation in the Middle East, with the worst-case scenario being stagflation.
The main worry currently is the skyrocketing inflation expectations of U.S. households. We have never seen such a decoupling from the CPI. The question is, what are the U.S. households seeing? We will analyze the implications of this below.

Impulse-response analysis
Like we noted last month, the analysis of the series of the 12-month percent change of the CPI is complicated by its unclear time-series properties (see further explanation below). This is why we turn to analyzing the month-over-month changes of the U.S. inflation series. They seem to provide both an unambiguous conclusion on the time-series properties of the series and also more robust forecasts.
We start by analyzing the relationship between the one-year-ahead price expectations of the University of Michigan (UMICH), month-over-month changes in the CPI, core inflation, and ‘supercore’ inflation. We tested the variables, and each one was found to be stationary, implying that they do not have a time-structure like, for example, I(1) non-stationary or random-walk processes have (see the April Black Swan Outlook).
We used the Vector Autoregression model (VAR) to estimate the relationship between the variables.1 Then we ran the impulse-response functions with (bootstrapped) confidence intervals,2 using orthogonalized impulse-response functions, which allow us to analyze the effects between variables that affect each other contemporaneously.3 We started by shocking the UMICH inflation expectations series. This is what it yielded.
Keep reading with a 7-day free trial
Subscribe to GnS Economics Newsletter to keep reading this post and get 7 days of free access to the full post archives.