On October 14, 2026, the Bureau of Labor Statistics reported that the Consumer Price Index rose 3.7 percent in the 12 months through September, while the index excluding food and energy, the core rate, rose 2.6 percent after 2.4 percent through August. Shelter, the largest single block in both measures, rose 3.2 percent over the same 12 months. For anyone reading the two rates side by side, that one component pushes them in opposite directions: it lifts core from 2.0 percent to 2.6 and holds the all-items rate at 3.7 instead of 4.0, so the claim that housing is pushing inflation up is true of one number and false of the other. The figures come from the BLS Consumer Price Index news release for September 2026 and its first three tables. This post follows the shelter line through them: what the index counts as shelter, why September's monthly shelter increase of 0.4 percent is not a rent number, and how much of each 12-month rate shelter accounts for once the rounding comes off. T...
On September 11, 2026, the World Agricultural Outlook Board approved WASDE-675. Its U.S. corn table in the September 2026 World Agricultural Supply and Demand Estimates puts the 2025/26 marketing year, which ended on August 31, at 18,602 million bushels of total supply and 16,680 million bushels of total use, leaving ending stocks of 1,922 million bushels. The marketing year is over. The number looks settled.
It is not a measurement. Nine days from now, on September 30, the National Agricultural Statistics Service publishes its quarterly Grain Stocks report, and one of the figures in it is corn stocks in all positions on September 1, 2026. For the 2025/26 marketing year, that is the same quantity as ending stocks. NASS will measure it. WASDE has so far inferred it.
The gap between those two things has to go somewhere, and on a corn balance sheet there is only one place it can comfortably go. This piece sizes that place, walks the reconciliation of the previous marketing year release by release, and sets out what the September 30 print would have to show for each branch to open.
A Line Whose Published Name Contains the Word Residual
USDA's own framing is unusually frank. The Economic Research Service Feed Grains Database documentation describes the accounting directly: "Supply data are disaggregated into subcategories reflecting stocks, production, and imports. Disappearance data are disaggregated into subcategories reflecting: food, alcohol, and industrial use (FAI), seed use, feed and residual use, total domestic use (the sum of FAI, seed use, and feed and residual use), and exports." It then states the closing identity in one sentence: "The ending stocks variable reflects the difference between total disappearance and total supply."
The same page fixes the calendar: "Corn begins September 1 and comprises four quarters (September-November, December-February, March-May, and June-August)." That is why the September 1 stocks survey and the marketing-year ending stocks figure are the same object.
Now read the use side of the September table. Food, seed and industrial use is 6,905 million bushels, of which ethanol and by-products account for 5,550. Exports are 3,425. Feed and residual is 6,350. Three of those four lines have independent collection behind them. The fourth is named for what it is.
NASS is equally plain in the Grain Stocks report itself, where quarterly use is never called use. The June 30, 2026 Grain Stocks report says: "The March - May 2026 indicated disappearance is 3.74 billion bushels, compared with 3.50 billion bushels during the same period last year." Indicated. The figure is read off the change in stocks, not collected from the operations doing the feeding.
This is not a flaw hidden in a footnote. It is the published architecture, and the balance sheet could not close without it. The consequence is arithmetic rather than scandal: when a measured stocks figure arrives for a completed marketing year, and production, exports and industrial use for that year are already collected, the adjustment has almost nowhere else to land.
Two Septembers, Two Gaps
The cleanest way to size the September 30 event is to compare, for the same marketing year, what the September WASDE carried against what the Grain Stocks report then published. Two such pairs are available in the recent record.
| Sept 1 measured | September WASDE estimate of that year's ending stocks | Grain Stocks, as first published | Gap |
|---|---|---|---|
| 2024 (2023/24) | 1,812 (WASDE-652, Sept 12, 2024) | 780 on-farm + 980 off-farm = 1,760 | -52 |
| 2025 (2024/25) | 1,325 (WASDE-664, Sept 12, 2025) | 643 on-farm + 888 off-farm = 1,531 | +206 |
The September 30, 2024 Grain Stocks report states: "Old crop corn stocks in all positions on September 1, 2024 totaled 1.76 billion bushels, up 29 percent from September 1, 2023." The September 30, 2025 report states: "Old crop corn stocks in all positions on September 1, 2025 totaled 1.53 billion bushels, down 13 percent from September 1, 2024."
Two observations are not a distribution, and this article does not treat them as one. What they do establish is direction. In 2024 the measurement came in below the WASDE estimate, and feed and residual for 2023/24 was subsequently revised up, from 5,775 to 5,832. In 2025 the measurement came in above, and feed and residual for 2024/25 was revised down, from 5,675 to 5,492 in the next published WASDE. The sign of the gap and the sign of the feed and residual revision were opposite in both cases, which is what the identity requires.
What a Full Year of Revisions Did to Three Corn Balance Sheets
A wider view is available if the comparison is stretched to twelve months: the September WASDE that first estimated a completed marketing year against the September WASDE one year later, by which point the stocks survey, its own revisions and the annual crop production estimate have all been absorbed.
| Line, million bushels | 2022/23 | 2023/24 | 2024/25 |
|---|---|---|---|
| Ending stocks, first September estimate | 1,452 | 1,812 | 1,325 |
| Ending stocks, one year later | 1,360 | 1,763 | 1,551 |
| Change in ending stocks | -92 (-6.3%) | -49 (-2.7%) | +226 (+17.1%) |
| Change in feed and residual | +61 (+1.1%) | +57 (+1.0%) | -237 (-4.2%) |
| Change in production | -79 | -1 | +25 |
| Change in exports | -3 | -35 | +43 |
| Change in food, seed & industrial | -47 | +24 | -5 |
| Change in total use | +11 | +46 | -199 |
Sources are WASDE-640 of September 12, 2023, WASDE-652 of September 12, 2024, the WASDE-664 of September 12, 2025 and WASDE-675 of September 11, 2026.
One pattern holds in all three years: the feed and residual line moved further than total use did. Sixty-one against eleven, fifty-seven against forty-six, two hundred thirty-seven against one hundred ninety-nine. Every time, other use lines partly offset it rather than reinforcing it.
A second pattern is about scale asymmetry. Feed and residual moved 1.1 percent, 1.0 percent and 4.2 percent. The ending stocks line moved 6.3 percent, 2.7 percent and 17.1 percent. The ratio of the stocks percentage move to the feed and residual percentage move ran 5.7, 2.7 and 4.1. There is no fixed multiple here and this article does not claim one; what the three years support is only that a small proportional move in a large use line lands as a larger proportional move in a much smaller stocks line.
The table also contains something that cuts against a tidier story, and it should be said plainly rather than buried. In 2022/23 the production line moved 79 million bushels while feed and residual moved 61. For that year, production was the larger single mover, not the residual. Anyone arguing that feed and residual is always the dominant revision channel is contradicted by the first column of this table. The narrower claim survives: among the use lines, feed and residual moved most in each of the three years.
The 2024/25 Chain, Release by Release
The most recent year shows the mechanism in full, including a disruption worth remembering.
The November 14, 2025 WASDE carries a note that matters for any calendar-based plan: "Due to lapse in government funding from October 1 through November 12, some U.S. data sources that are typically used were not available for the November 2025 WASDE." There was no October 2025 WASDE. The reconciliation still happened, six weeks late.
The January revision is the other half of the lesson. NASS states the policy in each issue: "On-farm and off-farm stocks are subject to revision the quarter following initial publication and again in the following December 1 Grain Stocks report published in January each year." The January 12, 2026 report put September 1, 2025 corn stocks at 1,551,286 thousand bushels against the 1,531 million first published. The measurement that resolves the estimate is itself provisional, and the same relationship between a first print and a later benchmark shows up in other series this blog has taken apart, from payroll benchmark revisions landing on a single month to a capacity utilization denominator re-estimated every year.
NASS also publishes the precision of the surveyed component, and is careful about which component it refers to: "The on-farm stocks estimates are subject to sampling variability because all operations holding on-farm stocks are not included in the sample. This variability, as measured by the relative standard error at the United States level, is approximately 1.7 percent for corn, 3.1 percent for soybeans, and 3.6 percent for all wheat." That figure applies to on-farm stocks, not to the all-positions total. On the 643 million bushel on-farm figure for September 1, 2025, 1.7 percent is roughly 11 million bushels, before any of the revision channels above.
Where a September 30 Surprise Goes
Define D as the September 1, 2026 corn stocks figure NASS publishes on September 30, minus the 1,922 million bushels now carried in WASDE. The branches below are exhaustive and mutually exclusive over every possible value of D.
The boundary values belong where the diagram puts them. A gap of exactly -100 is Box B, not Box A. A gap of exactly +100 is Box C, not Box D. Of the two directly comparable cases in the record above, the 2024 gap of -52 falls in Box B and the 2025 gap of +206 falls in Box D. Boxes A and C hold no observation from those two years. That is a statement about a sample of two, not evidence that those outcomes are unlikely.
The pass-through to the new crop year is mechanical. WASDE-675 carries 2026/27 beginning stocks at 1,922 and projects 2026/27 ending stocks at 1,567 on total supply of 17,747 and total use of 16,180. Holding the 2026/27 use forecast unchanged, a repeat of the 2025 gap would carry 2026/27 ending stocks to roughly 1,773, about 13 percent higher, while a repeat of the 2024 gap would take them to roughly 1,515. The use forecast will not in fact be held unchanged, which is exactly why the October report is the one that matters, not the September 30 print alone.
What Would Invalidate This
Several things would break the argument, and one of them is already visible in the evidence above.
Production, not the residual, may dominate. The 2022/23 column shows production revised by 79 million bushels against 61 for feed and residual. For a marketing year whose crop was finalized in the preceding January, that channel is largely closed by now, though not completely. If the October report revises the 2025 crop, the residual carries less of the load than this piece implies.
Exports are not fixed either. They moved 35 million bushels for 2023/24 and 43 for 2024/25 across the twelve-month windows. Those are export figures for completed years, revised after the fact from trade data. A large export revision alongside the stocks print would split the adjustment.
The residual may be real feed. Nothing here establishes that the residual is error. Livestock numbers, feeding rates and on-farm shrink all sit inside that line legitimately. A large revision may be a correct restatement of how much corn was actually fed, not a plug absorbing measurement noise. The honest position is that the line cannot distinguish the two, which is the point.
The reconciliation may not arrive on schedule. The 2025 precedent is direct: no October WASDE was published, and the September 30 stocks figure sat unreconciled until November 14. Any plan that assumes a fixed gap of a few weeks between measurement and reconciliation has a live counterexample within the last twelve months.
The measurement may move. The off-farm component of the September 1, 2025 figure changed by 20 million bushels between September and January. A reader who treats the September 30 print as final is repeating, one level down, the same mistake as treating the September WASDE as final. The problem described in a monthly construction series built from annual forecasts recurs here in a different agency's accounts.
Concrete Framework
A monitoring checklist for anyone reading the corn balance sheet as a demand signal.
| Check | What to look at | What it tells you |
|---|---|---|
| 1. Separate inference from measurement | Ending stocks in any WASDE for a year whose September 1 stocks are unpublished | That figure is supply minus use, not a survey result |
| 2. Compute D on release day | Published Sept 1 stocks minus the WASDE figure it replaces | Sign gives the direction of the feed and residual revision |
| 3. Check the other use lines before attributing | Exports and food, seed and industrial in the next WASDE | Whether the residual absorbed the gap or shared it |
| 4. Watch the crop line too | Production for the completed year | 2022/23 showed production can move more than the residual |
| 5. Re-check in January | Revised Sept 1 stocks in the January Grain Stocks report | Off-farm moved 2.3 percent between prints last year |
| 6. Do not assume the calendar holds | Whether the next WASDE is actually published | The October 2025 report was not |
What to Watch Next Week
The NASS release listing schedules Grain Stocks for Wednesday, September 30, 2026 at 12:00 p.m., alongside the Small Grains Summary. The corn figure to read is stocks in all positions on September 1, 2026, with its on-farm and off-farm components printed separately. Subtract 1,922 from it, and the sign of the result determines which side of the diagram above opens.
The reconciliation itself is not in that release. WASDE-675 states the remaining 2026 calendar: the next reports are scheduled for October 9, November 10 and December 10. The October 9 report is where the revised 2025/26 feed and residual, the restated 2025/26 ending stocks and the new 2026/27 beginning stocks appear together. Until then the balance sheet carries a measured number on one side and an inferred one on the other, and the size of the discrepancy is the story.
The narrow lesson generalises past corn. Where a statistical system publishes a line whose name contains the word residual, that line is the shock absorber of the whole account, and the number everyone quotes is usually the one sitting on the other side of it.
Comments
Post a Comment