US CPI: Headsup – Wrong Seasonal Factors in Macrobond

TL;DR: Be CAREFUL. Check your seasonal factors.

Ahead of tomorrow’s CPI release, we would like to flag a potentially important issue with the seasonal factors used in forecasting this report, as our checks over the past few days suggest that some forecasters may be relying on inappropriate assumptions and that some data providers are displaying figures that are wrong.

Consider Macrobond, our own data provider, and the CPI energy commodities series (Figure 1), for which an October 2025 seasonal factor is available despite the absence of published CPI index levels for that month. The reported factor looks markedly different from those for October in previous years (98.930 vs 100.294), while a similar issue appears in other components, including apparel, shown in the third column of the screenshot below, where Macrobond reports an October 2025 seasonal factor of 102.847, compared with 101.092 in October 2024. Forecasters who mechanically carry these October 2025 factors forward into tomorrow’s release could obtain materially misleading results.

The correct approach, instead, is to start with the BLS spreadsheet containing the 2026 seasonal factors for the elementary components and aggregate them to obtain the factors needed for our forecasts, an exercise we have already completed in the accompanying file. (here) For energy goods, for example, our calculation gives an October seasonal factor of 100.294, consistent with the value reported by Macrobond for October 2024, rather than October 2025. We suspect that the missing October 2025 index levels may have affected the calculation or presentation of these series, although we cannot determine the precise source of the discrepancy without knowing how Macrobond produced the reported values.

Interpolating the missing seasonally adjusted and unadjusted index levels and then taking their ratio (which seems to be what some people are doing to get a YoY tomorrow…) would also introduce an unsupported assumption into the forecast, because the resulting factor would depend on the interpolation procedure applied to two unobserved values. That ratio cannot be treated as an observed seasonal factor or assumed to provide an appropriate adjustment for tomorrow’s report without further validation.

The potential impact is material, as using the Macrobond factors instead of those we calculated would have lowered our forecast by approximately six basis points relative to the figure we published.

Our full forecast is available on our website, where we encourage you to review the latest updates, including downward revisions to our core CPI projections for the final months of the year. Ahead of tomorrow’s release, we strongly encourage analysts to check the seasonal factors embedded in their Excel files and codes, verify how those factors were obtained, and validate them against calculations based on the BLS source file, particularly wherever October 2025 values have been carried forward mechanically; as always, we are available to discuss our calculations and answer any questions.

Figure 1. Macrobond

Note: the Macrobond tickers used to get this table are: bls_cusr0000sace_sf, 100*uspric1340/uspric2379, bls_cusr0000saa_sf.

Contact

Want something more tailored?

We provide tailored consulting on ad-hoc projects.