In August, the USDA estimated the U.S. corn crop to be 175.4 bushels per acre. With harvest just weeks away, it raises the question: How much variability should we reasonably anticipate in the USDA’s yield estimates throughout the fall? Would a range of +/- 2 bushels be sufficient? +/- 5? +/-10?
The answer, it turns out, might be wider than most of us initially expected.
Error in Corn Yield Estimates
Figure 1 shows the difference between various USDA corn yield estimates and the final yield. As one might expect, throughout the fall months, the USDA’s estimates get more accurate – or have fewer errors. However, for any given month, the range of historic errors might be surprising.
For example, the average change between August and the final yield is 4.5 bushels. On eight occasions – or 38% of the time – final yields have been more than 5 bushels different than estimated in August. At the extreme, the August estimate has been more than 10 bushels wrong on three occasions, or 14% of the time.
The errors and range of possible outcomes shrink significantly by October and November. Even then, though, considerable uncertainty remains. For instance, in 2020, the November WASDE was 4.4 bushels higher than the final estimate.
Error in Soybean Yield Estimates
While the trend of improvement throughout the fall is the same with soybeans, there are three differences to keep in mind. First, the error of historic August yield estimates isn’t much improved from the May WASDE or a simple trend yield. This is to say we generally don’t know much about the size of the soybean crop until later in the fall.
Second, there is, historically, more error in soybean yield estimates. For instance, even in September, the average absolute error – expressed as a percentage – is higher for soybeans (3.6%) than what is observed with corn (2.4%). For nearly every estimate considered, the relative error is larger for soybeans.
Finally, it isn’t until October that the range of possible outcomes considerably narrows for soybeans. Even then, on two occasions – roughly 10% of observations – the final yields have been more than 2 bushels per acre different than the October estimate.
Wrapping It Up
At the beginning of this article, we challenged readers to ponder how wide of a range would be reasonable enough to capture how final U.S. corn yields might adjust from the recent August estimates. Clearly, “reasonable” is an ambiguous term, and there is no right answer, but it’s common to under-appreciate how much the production outlook can change throughout the fall.
This isn’t to say that will be the case in 2022, but instead is a reminder that the range of possible outcomes remains large. Given the tight stock situation, large changes in production estimates – higher or lower – could significantly shift the market outlook.