Column: Jarring U.S. corn stock revisions are making USDA’s numbers unpredictable

FORT COLLINS, Colo. (Reuters) - Anyone looking for fireworks following the U.S. Department of Agriculture’s Tuesday forecasts was not at all disappointed, but for most market participants, the recent trend in some of the numbers might be downright unsettling.

Farmer Roger Hadley sitting in his John Deere combine harvests corn from his fields in Woodburn, Indiana, U.S., October 16, 2020. REUTERS/Bing Guan

The U.S. corn figures were the most surprising in the data landslide published by USDA on Tuesday, particularly for quarterly stocks and 2020 crop yield. Both of those numbers fell considerably below the trade range of estimates, which were historically wide for yield.

That locked most-actively traded Chicago corn futures up the daily 25 cents per bushel limit at $5.17-1/4 where it finished the session, and traders reported the contract was trading synthetically through options up another 8 cents. That would make for the highest price in the most-active contract since July 2013.

Quarterly U.S. corn stocks seem to be getting harder to predict, and that appears largely driven by revisions to the previous quarter’s stocks, which is part of USDA’s standard procedure. But the revisions are getting larger, setting up even more uncertainty for future reports.

USDA placed Dec. 1, 2020, U.S. corn stocks at 11.322 billion bushels, well below the trade average of 11.951 billion and the lowest estimate of 11.59 billion. That was by far the trade’s worst miss of Dec. 1 stocks in either direction since at least 2005.

But the extremely large 76 million-bushel reduction in Sept. 1, 2020, corn stocks to 1.919 billion bushels was what the market could not have prepared for. That is the third-largest previous-quarter corn stock revision since at least 2010, behind those of June 1, 2020, and Sept. 1, 2019.

Prior to the Sept. 1, 2019, revision that occurred in January 2020, the average absolute adjustment to the previous quarter’s corn stocks since 2010 was about 6.4 million bushels, with one outlier of 50 million in 2017. But since then, the average adjustment has been 90 million bushels with a high of 205 million.

The National Agricultural Statistics Service (NASS), the USDA agency responsible for the survey-based quarterly stocks, said on Twitter on Tuesday that the Sept. 1, 2020, corn adjustment was the result of updated reports from facilities. Changes to off-farm stocks accounted for 100% of the decline.

NASS has said the most common reason for the previous quarter revisions is late-reported data, but it usually results in very small moves. It is less common for the re-analysis of past data to determine that a revision is needed, but those moves can be more impactful, and this was the reasoning behind the June 1, 2020, adjustment.

Late-reported data might have been more of a problem in 2020 than in past years due to the pandemic, but the large Sept. 1, 2019, stock adjustment observed a year ago was in the pre-virus era, so that explanation might not suffice.

USDA critics have pointed out the circular logic seemingly employed in the quarterly stock revisions. The quarterly stocks are surveyed numbers that are supposed to help imply statistics that cannot be implicitly measured such as feed and residual use, but NASS is turning around and using the implied feed and residual to then adjust stocks.

NASS was asked during the Data User’s Meeting in October why larger adjustments are suddenly being made to the previous quarter’s corn stocks, and officials did not know the answer. But this has become something worth investigating, or else market chaos could ensue with each report.


There has been clamor by many industry members since 2019 that USDA overestimated that year’s corn crop, and it was expected that the issue would eventually be revealed in quarterly stock reports.

Corn stock adjustments for the previous four quarters beginning with Dec. 1, 2019, sum up to a reduction of 343 million bushels. But USDA never reflected a significantly smaller 2019 U.S. corn crop even after the review period.

The official USDA balance sheet suggests suspiciously high feed and residual corn use for 2019-20, leading many to believe the leftover corn was simply stuffed into this category for lack of a better option.

Estimation of the 2020 U.S. corn crop was apparently tricky as well, because none of the survey-based estimates issued last fall came very close to capturing the yield NASS reported on Tuesday. That put corn production some 1.8 billion bushels lower than what was estimated in June. A change in harvested acres did not play a big role in the latest prediction.

NASS placed 2020 U.S. corn yield at 172 bushels per acre, down 2.2% from the November estimate. That is easily the biggest percentage change between November and January since the 1993 crop, and the yield was well below the average trade guess of 175.3 bpa, making it analysts’ worst January miss of U.S. corn yield in at least 17 years.

Some of the statewide adjustments were extremely large, such as the 10 bpa drop in No. 4 corn state Minnesota between November and January. When asked about the unusually large moves on Twitter, NASS chalked it up to being “one of those years where actual yields at harvest were just not as good as they appeared most of the growing season.”

Both 2020 U.S. corn and soybean production have the potential to change again in September when NASS does its full revision of the 2020-21 marketing year after collecting the Sept. 1 stocks data.

The final chance for adjustment to all the numbers, including the quarterly stocks, will come in the 2022 U.S. Census of Agriculture. That data may not become available until early 2024.

The opinions expressed here are those of the author, a market analyst for Reuters.

Editing by Matthew Lewis