In September 2022, we published a note (available here) titled “The Fed Staff Is Probably Wrong Again. Here Is Why.” In hindsight, the models proved to be correct, while the Federal Reserve underestimated the persistence of inflation. In this note, we compare forecast errors from the Summary of Economic Projections (SEP) with those from our Main Model (MM), and explain why we once again believe that both the Fed staff and the FOMC are misjudging the inflation dynamics.
The implication is clear: if Chair Powell is indeed prepared to cut rates in the current environment—as he appears to be—then, in our view, inflation will remain well above target.
The SEP Forecast Is Systematically Biased to the Downside
Figure 1 compares realized core PCE price inflation (Q4/Q4, shown in black) with projections from the Summary of Economic Projections (SEP) and our Main Model (MM). SEP forecasts are represented by solid lines, while MM forecasts are shown as dashed lines. For each projection, we report the forecasts made during the June round of each year. For example, “MM21” refers to the forecast produced in June 2021 for core PCE inflation in Q4 2022 and Q4 2023.
Our goal is to assess the accuracy of SEP forecasts at 18- and 30-month horizons, relative to the MM. To ensure a fair comparison, we exclude 2021 from the chart so that both the Fed and our model would have already incorporated most of the inflation run-up—core PCE inflation was 3.9% year-over-year as of June 2021.
The evidence is unequivocal: the SEP systematically underestimates inflation at both the 18- and 30-month horizons. This is evident in Figure 1, where the SEP forecast lines consistently lie below the actual outcomes (the black line). In contrast, MM forecast errors are substantially smaller and, on average, close to zero – sometimes the model missed on the upside and sometimes on the downside.
Table 1 presents the corresponding statistics. Since the reemergence of inflation, the MM’s absolute mean forecast error has been approximately half that of the SEP. Specifically, at the 18-month horizon, the MM missed core PCE inflation by about 0.5 percentage points, compared to over 1 percentage point for the SEP. At the 30-month horizon, the MM’s forecast error is less than half that of the SEP (and it is even smaller than the MM error at 18-month).
Moreover, the MM’s average forecast error is close to zero—indicating an unbiased forecast—while the SEP exhibits a consistent negative bias. Repeating this analysis for the forecasts made in September and December yields similar results (see figures here and here, respectively. Summary statistics are reported in the lower panel of Table 1).
An Excel file containing the data of Figure 1 and Table 1 is here.
Figure 1. Realized core PCE vs SEP forecasts vs MM forecasts made in June
Note: the figure shows realized core PCE price inflation (Q4/Q4) – black solid lines. The figure shows also the SEP and our Main Model (“MM”) forecasts 18 and 30-month out. For instance, “SEP21” refers to the forecast made by the SEP in the June 2021 round for core PCE in 2022 and 2023.
Table 1. Forecast errors at 18 and 30-month horizon: SEP vs Main Model
Note: The table shows forecast errors statistics for the SEP and our Main Model (“MM”). “JUN (SEP) (DEC) forecast” refers to the forecast made in June (September) (December).
Why Is the SEP Forecast Consistently Too Low?
In short, the underlying reason lies with the Fed staff. For a detailed discussion, we refer the reader to our 2022 note. Below is the abbreviated version in bullet-point format:
- The SEP forecast is heavily influenced by the Fed staff forecast. See Peneva et al. (2025): a comparison between the chart on page 5 and the SEP confirms this alignment. When the Fed staff errs, the entire FOMC tends to follow.
- The Fed staff forecast is informed by models but ultimately judgment-based. This approach is reviewed in the recent paper by Peneva et al. (2025).
The mechanics of the Fed staff forecast process are as follows:
- The staff runs models to assess structural fundamentals, including trend (or “underlying”) inflation, denoted as pi*.
- The staff also uses models to evaluate risks—similar to our Main Model (MM) and sectoral frameworks.
- Ultimately, the forecast is judgmental. While model-informed, the staff can deviate from model outputs when justified.
Two core features define the Fed staff’s judgmental forecast:
- The pi* assumption. When the staff assumes a constant pi* (1.7% pre-COVID, now 2%), the forecast naturally converges back quickly to that level—unless the unemployment gap (U-gap) is large and persistent, which is not currently the case.
- Crucial point: In our experience (and consistent with recent SEP forecast errors), the staff does not revise the fundamentals for a given year until sufficient data from that year is available.
- Specifically, this means that when the staff issues a forecast in June of year t for year t+1, and assumes pi* = 2% with a U-gap near zero, the forecast will mechanically center around 2%.
- One might argue that, in light of recent shocks, the staff should have incorporated greater inflation persistence. But according to their process, doing so requires an explicit judgment—which the staff has consistently declined to make since 2021 (we are not sure why at this point, given the evidence).
- This, in our view—based on model results and past experience—is the key reason the SEP forecasts at 18- and 30-month horizons remain systematically biased.
Please feel free to reach out if you would like to discuss the Fed staff forecasting process in more detail.
Conclusion
Will Inflation Remain Above the Fed’s Target? Our Main Model (MM) projects core PCE price inflation to remain close to 3% over the forecast horizon. This contrasts with the SEP, which anticipates core inflation falling to 2.4% in 2026 and further to 2.1% in 2027 and to Powell’s speech in which he assumes that inflation will drop quickly. Given the Fed’s recent forecasting errors, if Chair Powell is either unwilling to acknowledge these misses—or unaware of them—and remains inclined to cut rates, then the implication is clear: barring a recession, inflation is likely to persist well above the 2% target in the years ahead.