Weekly Forecasts 20/2026
Is U.S. inflation about to head to the 'stratosphere,' or ease?
Contents:
Developments in U.S. inflation.
Oil crisis scenario forecasts for U.S. inflation.
Oil crisis + supply chain shock scenario forecasts for U.S. inflation.
I discovered a data error from our U.S. inflation forecasts. For some reason, some of the CPI (Consumer Price Index) data was unadjusted for seasonality, while most was adjusted. About a year ago, we made comparisons with non-seasonally adjusted and seasonally adjusted CPI data. The data error most likely arises from there.
The above means that basically all of our U.S. inflation forecasts have been biased or erroneous in a statistical sense. Although the error is likely small because the same process drives both seasonally adjusted and not seasonally adjusted series, there has been one nonetheless. This discrepancy may also explain why our CPI forecasts have been missing their mark, while our core inflation forecasts have been more accurate.
My apologies.
In this week’s weekly forecasts, we naturally correct this error. Our model changes a bit as a result, as do our forecasts. Our baseline scenario forecasts, assuming no shock, see U.S. inflation cool rapidly but continue at an elevated level for the next 12 months.
Our oil crisis scenario forecasts see the U.S. inflation falling first and then continuing on its upward trajectory. Our scenario forecasts, assuming also a supply chain shock, foresee the U.S. inflation heading to the ‘stratosphere,’ with a projected annualized change in the CPI reaching 17%. The likelihood that the U.S. enters this last “apocalyptic” scenario naturally grows by the day the traffic through the SoH remains strictly limited, and the continuation of the “kinetic” phase of war would practically guarantee its manifestation.
Tuomas
Recent developments in U.S. inflation
Below you find a summarization of U.S. inflation by Grok Expert. It gathered the data from the U.S. Bureau of Labor Statistics, Trading Economics, and CNBC.
— —
The most recent U.S. inflation data is the Consumer Price Index (CPI) for April 2026, released by the Bureau of Labor Statistics on May 12, 2026.
Headline CPI (all items): Rose 0.6% month-over-month on a seasonally adjusted basis (down from +0.9% in March) and 3.8% year-over-year (up from 3.3% in March). This was the highest annual rate since May 2023 and slightly above consensus forecasts.
Core CPI (excluding food and energy): Increased 2.8% year-over-year (up from 2.6% in March) and 0.4% month-over-month.
Key drivers included a sharp rise in energy prices (+17.9% year-over-year, largely from gasoline), while food prices rose a more moderate 3.2%. Shelter and other services also contributed to the annual pickup, though some categories like new vehicles and medical care saw monthly declines.
For context on the Fed’s preferred gauge, the Personal Consumption Expenditures (PCE) price index (latest available: March 2026, released April 30) showed headline PCE at +3.5% year-over-year and core PCE around +3.2% (both in line with expectations at the time).
The most recent U.S. Producer Price Index (PPI) data—for final demand—is the April 2026 report, released by the Bureau of Labor Statistics on May 13, 2026.
Headline PPI (final demand): Rose 1.4% month-over-month on a seasonally adjusted basis (the largest gain since March 2022, up from a revised +0.7% in March and +0.6% in February). On an unadjusted basis, it increased 6.0% year-over-year (the largest 12-month rise since December 2022, up sharply from 4.3% in March).
Core PPI (final demand less foods, energy, and trade services): Advanced 0.6% month-over-month (largest since October 2025) and 4.4% year-over-year (largest since February 2023).
A broader core measure (excluding only food and energy) rose 1.0% month-over-month.
Key drivers were broad-based but led by energy and services:
Final demand goods rose 2.0% (more than three-quarters from energy prices +7.8%).
Final demand services increased 1.2% (nearly 60% of the overall headline gain), with trade services margins +2.7% and transportation/warehousing +5.0% (e.g., truck freight).
Intermediate demand also showed pressure: processed goods +2.7%, unprocessed goods +4.1% (with unprocessed energy materials +9.2%).
This PPI report shows a clear reacceleration in wholesale inflation—mirroring the April CPI pickup—driven heavily by energy costs but with broadening services and core pressures. It came in well above consensus expectations (0.5% m/m forecast) and has raised concerns about sticky inflation feeding into consumer prices.
In short, inflation reaccelerated in April 2026 after cooling earlier in the year, primarily due to energy costs, though core measures remain lower than headline. Upstream inflation surged in April 2026 after earlier moderation, with both headline and core measures hitting multi-year highs.
— —
So, like we forecasted, the U.S. inflation continued to pick up speed in April, but not by so much as our (partially erroneous) forecasts anticipated. Tuomas already commented on the figures and analyzed the success of our forecasts. We add here just that the big jump in the PPI indicates that inflation pressures are building rapidly again (we are about at the level of May 2021). In other words, this inflation shock is just starting, even if the war would end tomorrow and the Strait would be fully opened.
U.S. inflation forecast: Baseline
Like noted above, we will divide the forecast into two segments. In the first, we present our non-scenario or baseline forecasts. In them, we let the model do its work without adding, or “forcing,” any crisis dummies into it. In the second segment, we build scenario forecasts assuming first an oil crisis and then the combination of an oil crisis and supply chain shock.
Due to the data error, we needed to run our whole analysis again. The fact that the data series changed from not seasonally adjusted to seasonally adjusted in the middle of the series implies that it brought unexplained stochasticity into the model. This may have fooled the unit root tests and altered the autocorrelation function.
Autocorrelation functions (not shown) of the three U.S. inflation measures, CPI, core inflation, and supercore, still show very high persistence, with the autocorrelation of, e.g., the MoM CPI series dying down only after 80 lags (months). It is the characteristic of a unit root series that the memory of the series is “infinite” or very close to it. This “infiniteness” is manifested in the autocorrelation function as a very persistent autocorrelation (that does not seem to “die out”). This is what the autocorrelation function of MoM CPI essentially shows. Moreover, the autocorrelation of core inflation does not fall under the 5% statistical significance even after 100 lags.
However, unit root tests reject the I(1) non-stationarity of all three series with a 5% level of statistical significance (but not at a 1% level of statistical significance). The Johansen Trace Test also indicates that all three series would be stationary.1
We’ve been “dancing” around this topic, i.e., the stationarity vs. non-stationarity of the U.S. inflation series for some time. At some point, we tried the VAR-GARCH models to control for the heteroscedasticity of the series but returned to the VECM framework earlier this year, when we decided to approximate the statistical properties of the U.S. inflation series with a random-walk process. Back then, the results of testing were more supportive of this assumption, possibly due to the error in the data.2
Due to all the hassle, we return to the basics now. We assume that the U.S. inflation series is stationary (like the unit root and cointegration tests indicate). Hence, we apply the vector autoregression model, or VAR, to forecast the U.S. CPI, core inflation, and super-core inflation. We also did some other adjustments to the modeling, which we explain below.
Figure 1 presents our “baseline” forecasts for month-over-month U.S. inflation measures using a VAR model.3
