September 10, 2001
HOW ACCURATE ARE THE SEPTEMBER
CORN AND SOYBEAN YIELD FORECASTS?
The USDA's September Crop Production
report will be released on September 12. A number of private forecasts
have been released and numerous yield estimates based on weekly
crop condition reports are cranked out every week. All of these
forecasts and estimates make for interesting reading, but it is
the USDA forecast that the market will react to. That is as it
should be. The USDA's methodology is the most objective and comprehensive,
by far, than that of any other source of crop yield forecasts.
Even so, the "final" yield estimate in January after
harvest often deviates from the September forecasts because it
reflects more information, including the impact of late season
weather, actual harvest results, and estimates of field loss.
Since 1979 (22 observations), the
September corn yield forecast has been within 0.1 bushel of the
August forecast 5 times, below the August forecast by more than
0.1 bushels 10 times, and above the August forecast by more than
0.1 bushels 7 times. The August to September decline in the yield
forecast exceeded 3 bushels on only 2 occasions 1983 (14.8
bushels) and 1995 (4.5 bushels). The average decline for the other
8 years was about 1.6 bushels. The August to September increase
in the corn yield forecast ranged from 0.6 to 4 bushels per acre,
averaging 2.3 bushels. This year, the September yield forecast
is generally expected to be smaller than the August forecast.
History would suggest that the decline will be relatively small.
Based on the August forecast of acres harvested for grain, the
yield forecast would have to decline by more than 3.8 bushels
per acre to put the production forecast under 9 billion bushels.
The corn yield estimate in January
after harvest has differed form the September forecast every year
for the past 22 years. The January estimate was above the September
forecast 14 times and below the September forecast 8 times. The
differences tended to be slightly larger than the August to September
changes. The yield reductions ranged from 0.4 bushels to 12.4
bushels (1993) and averaged about 4 bushels. Excluding the extreme
years of 1993 and 1995, the average decline was 2.2 bushels. The
yield increases ranged from 0.3 to 10 bushels and averaged 4.2
bushels. For all 22 years, the January yield estimate was within
1 bushel of the September forecast 5 times; within 2 bushels,
6 times; and within 3 bushels, 9 times. It is also interesting
to note that in the 10 years that the September yield forecast
was below the August forecast by more than 0.1 bushel, the January
estimate was below the September forecast 6 times and above 4
For soybeans, the September yield
forecast has been very close tot he August forecast in most years.
The difference was one bushel or more in only 6 of the 22 years.
Three of those 6 occurred in 1998, 1999, and 2000. The largest
difference was a 4.8 bushel decline in 1983. The largest increase
was 1.7 bushels in 1985. The September forecast was below the
August forecast in 12 of the 22 years. Opinions are mixed for
the expected change in the yield forecast in September this year.
Based on the August estimate of harvested acreage, the September
yield forecast will have to be a bushel or more below the August
forecast to result in a production forecast below 2.8 billion
The January yield estimate was below
the September forecast in 11 of the past 22 years, 6 of the past
8 years, and each of the past 4 years. The drop in the yield estimate
in those 11 years ranged from 0.3 bushels to 2.1 bushels and averaged
1.2 bushels. The decline were relatively large (1.4 to 1.7 bushels)
in each of the past 3 years.
In the 11 years of an increase in
the yield estimate in January, the increase ranged from 0.4 to
3.7 bushels per acre and averaged 1.6 bushels per acre. Increases
in yield estimates from September to January have tended to be
larger than declines in yield estimates.
For the most part, changes in yield
forecasts from August to September tend to be relatively small.
The changes also tend to be smaller than reflected by many private
forecasts. In particular, reductions in yield estimates tend not
to be as large as advertised by the private sector. This year
is not likely to be an exception. Changes from September to January
tend to be somewhat larger than changes from August to September,
but a large change is rare. Given the maturity of this year's
crops, particularly corn, the September yield forecast should
be a good predictor of the final yield estimate.
Issued by Darrel
University of Illinois