When USDA Gets it Wrong
Agency's estimates are the best in the world, but no system is perfect.
Arlan Suderman 
Published: Jan 14, 2011
Long ago my father told me not to count my chickens before they hatch. Yet, USDA is asked to do that each year when it develops monthly supply and demand estimates during the growing season.
That task necessitates an estimate of the size of that year's crops. The reports are intended to provide farmers, end users and those responsible for crop storage and transportation a the best information available for anticipating crop fundamentals so that they can effectively market and make storage preparations based on changing crop fundamentals.
USDA provides the most thorough and accurate domestic and global supply and demand estimates in the world. Yet, as it found last year, no system is perfect.
Quality data necessitates sound statistical methodology as well as an understanding of agronomics and forecasting ability. Yet, USDA's ability to forecast corn production in 2010 proved to be the most frustrating since 1995, which coincidentally was another warm humid growing season.
Sound methodology
USDA begins the process of production forecasting in March each year, surveying producers from a pool of 50,000 operations across the country regarding their planting intentions. The results of that survey are reported on March 31, or the last business day prior to that date. It follows that up with another producer survey during the first two weeks of June to find out how many acres of each crop farmers actually planted, which may include a few acres they still intend to plant. USDA then begins a monthly survey process to prepare production estimates starting with the August crop report and ending with the final crop estimates in January.
The process has two facets. The first is a producer survey that asks farmers to reveal their assessment of yield prospects for the crops they grow. The second is an objective yield survey where statisticians walk previously identified random fields to statistically evaluate yield potential, assuming normal temperature and precipitation for the remainder of the growing season. Data from both surveys are collected during the last week of the previous month and the first two or three days of the survey month in which the estimates will be reported.
Surveyors return each month to the same field locations through the remainder of the growing season. As crops are harvested, statisticians must rely heavier on the producer surveys to develop their production estimates.
What went wrong in 2010?
In essence, what went wrong was USDA's devotion to its statistical methodology, assuming normal weather for the remainder of the growing season. 2010 was a warm humid summer for most of the Corn Belt, with elevated overnight temperatures due to the high humidity readings. Data over the past 50 years confirms that high overnight readings divert energy away from kernel development toward plant maintenance, while also speeding the crop toward maturity, shortening the time period that it has to make grain.
These summers produce crops that look excellent in August with high ear counts due to nearly greenhouse growing conditions, but result in disappointing yields.
USDA believes it is risky to forecast weather for the remainder of the growing season, and it is correct. As such, it commits to assuming normal temperature and precipitation, even though the growing season may be "locked" in a pattern that would be expected to impact yields. USDA found that kernel and ear weights didn't measure up to expectations in its fall field surveys because the corn plants had not packed as much energy into the kernels as would be expected in a normal growing season.
The number of warm humid growing seasons experienced in recent years is not large, providing a rather small statistical pool to draw from. Yet, a look at the data finds that USDA typically reduced its average corn yield by another one to two bushels per acre in the final January production estimates. Soybeans on the other hand are influenced more by day length and rainfall, providing few clear trends for yields attributed to warm humid summers.
Permalink: Click here
Tagged: usda, soybeans, Corn Belt, corn production, corn yield
|