Friday, July 31, 2009

Soybean trend yields from the model

In the post below are the estimated annual yield growth for corn, so here are the numbers for soybeans. All results are expressed in bushels/acre/year from 1986-2008 estimation period, with condition variables and planting progress also used as explanatory variables.
AL 0.35 bu/ac/yr
AR 0.65
GA 0.56
IA 0.54
IL 0.49
IN 0.63
KS 0.39
KY 0.33
LA 0.60
MI 0.44
MN 0.25
MS 0.63
MO 0.49
NE 0.71
NC 0.20
PA 0.35
ND 0.40
OH 0.44
OK ~0
SC ~0
SD 0.40
TN 0.38
WI 0.73

I've found some historical performance graphs from previous years I'll post. I'll also post some old performance relative to USDA, of course USDA predicts out of sample while I'm comparing it to my in-sampe results.

Wednesday, July 29, 2009

Corn trend yields from the model

Calculating simple trend yields is fraught with danger. The selection of the starting or ending year can have significant influence. There are still issues with using conditions to 'correct' for this but in the model I do estimate an underlying annual trend based on the 1986-2008 period. What follows are the estimated annual trend yield improvements for corn out of the model. I'll follow up with soybeans and cotton as I take the opportunity to use this blog to explain the model.
All results are expressed as bushes of corn/acre/year
AL 1.65 bu/ac/year
CO ~0
GA 2.06
IL 2.67
IA 2.74
IN 2.16
KS 1.43
KY 1.93
MI 1.96
MN 2.48
MS 2.75
MO 2.22
NE 1.95
NC 1.16
PA 0.97
OH 2.18
SC ~0
SD 2.84
TN 2.02
TX 1.95
WI 1.83

Tuesday, July 28, 2009

O.A. Cleveland sees 12.5 million bale cotton crop

weblink to article -> Analysts see 12.5 million bale cotton crop

Dr. Cleveland at Mississippi State puts the cotton crop at about 12.5 million bales, a good 0.75 million bales below my current estimate. Now my cotton model has not proven to be nearly as accurate over the last several years as the corn and soybean model, and when someone of Dr. Cleveland's stature wades in it is definitely worth a look.

O.A. Cleveland gives the following crop estimates

AR 1.1 million bales
MO 0.637 million bales
GA 1.6 million bales
NC 0.600 million bales
US 12.5 million bales

For week 17 my comparative yields in the corrected model are
(see my Week 17 post below to download the full spreadsheet)
AR 1.0 million bales
MO 0.599 million bales
GA 1.656 million bales
NC 0.631 million bales
US 13.35 million bales

So those are fairly close, the unknown is of course always Texas. Another heavy hitter, Dr. Anderson out of Texas A&M adds that he currently expects abandonment in Texas to be about 27%. Currently I'm just over 24%. A larger abandonment number would of course bring the two estimates together.

Monday, July 27, 2009

New format for sharing results

(click on 'comments' below each post)
When this model was first constructed, based on work by John Kruse, the intent was to create a parsimonious model that could be quickly updated and give weekly estimates of crop yields and production for corn, soybeans and cotton. In the first few years I sent out the file to a small group of people and we occasionally engaged in some discussion of the results. Over time the model has grown some and so has the distribution list, to the point that I'm sharing it with about 3 dozen people and send it out under a blind carbon copy. This of course kills discussion of the results. The file size has also increased to 3mb and I'm concerned about clogging email inboxes. So,to improve file delivery and to spur a bit of discussion, I've created this blog. You can download the same spreadsheet in a link in my weekly posting. I would hope you would leave any comments about your thoughts on the estimates.

YOU CAN CHOOSE TO RESPOND ANONYMOUSLY AND DO NOT HAVE TO SIGN UP FOR ANY ACCOUNT. So you can leave your comments, positive or negative, and if you want to identify yourself, you can still submit it anonymously and include your name in the post. I may also find out if anybody actually looks at the numbers.

Week 17

I show little change from last week in both corn and soybeans. I continue to show both the corrected and uncorrected results although I've been having some discussion about deleting the uncorrected results, at least from the summary presentation. To reiterate, the uncorrected model shows what the yield with be if we finish the year with current conditions. Of course conditions change and more often than not they decline, and so a crude correction can be applied to predict the final yield based both on current conditions and the 'average' decline. In some years this decline never comes, or is at least lest pronounced. These two estimates will of course converge as we approach late September. Please leave your comments

In late June the corrected model showed record corn yields (compared to 2000-2007) for IA, MT, NE, OH and WI. Conditions have slipped for some states and while I still show record yields for IA, MT and OH, only Ohio is more than a slight improvement on the record. The current corn yield estimate for Ohio is 169.6 bu/ac which is over a 6% improvement from the record yield of 159 bu/ac from 2006. This seems like quite a jump, and could be an issue given the variability in Ohio corn yields over the 2000-2007 period.

Recent talk I've heard would put me on the low side of soybean yields and I've yet to hear much about cotton yields to date. Any comments can be made anonymously, so let me know what your thoughts on this year's estimates. Download the spreadsheet you normally get by email at the link below.

Monday, July 20, 2009

Week 16

We will see continued convergence between the uncorrected and corrected models through mid/end of September. So conditions changed little in corn, but because it was another week closer to harvest, the needed correction declines, giving us some increase in yield and production. I don't think I'm alone in showing above trend yields in corn. I might be moving up above consensus, but I've seen reports higher than my corrected estimate.

For beans, I suspect I'm on the low side of consensus. The biggest change this week was in cotton where improving conditions increased both yield and harvesting rates which are running at 87.5% harvesting rate in the corrected model. The average of the lst 10 years for the US is 89.4% with a high of 96.9% in 2005 and a low of 81.6% last year. So at the moment, I'm below average harvest rates.

The key to this is abandonment in Texas which as fluctuated from 4% to 40% abandonment over the last 11 years. Right now I'm at just over 20% abandonment in the state.

Week 16 Spreadsheet (3mb)