Very stable this week, added trade estimates

Just an FYI, the folks sent a picture to Rome of an ear out of their field in western Iowa, the slow way by mail. 18 rows around and full length. If things haven't changed, they will have a decent crop at $8+ a bu.

I've added the trade numbers for comparison. My harvested area for corn is roughly consistent with the middle of the trade range. Yield and harvesting rates should be inversely related in their estimates. I haven't adjusted my model to use percent deviations as discussed and I doubt I'll have the time to do it in advanced of September 12th. Soybean numbers are in the middle of the trade range and cotton production is roughly consistent with USDA production numbers. Cotton yields are significantly too high but interact inversely with area (similar to corn this year) and while I look at these when estimating the equations I do not adjust them during the year and tend to pay more attention to the production numbers in total.


Corn area harvested   85.205 million acres
USDA estimated corn area harvested 87.400

Corn Yield:  139.5 bu/ac

Corn Production:  11,888 million bushels

latest USDA Production: 10,773 million bushels 

Soybean Yield:  37.2 bu/ac

Soybean Production: 2,760 million bushels

latest USDA Production = 2,692

Cotton Yield:  838 lbs/ac

Cotton Production 17.974 million bales 

latest USDA Production = 17.650 million bales

Slightly better conditions in corn and soybeans, lower for cotton

The corrected and uncorrected model are now converging (for a full explanation of the difference see previous years posts). I've added a graph on cotton area harvested as well to show the (over) sensitivity of the model to abandonment. Again, the chosen response of the model (read bushel deviations instead of percent deviations)  to abiotic stress (read drought) has cause the model to be significantly off this year for corn. If I have time in the off season this year, I'll do a percent deviations model, which has some issues that make it slightly more technically challenging, but I might also try to get the wheat model back up and running considering the interest in all grain markets at the moment. 

Soybean conditions improved so I've got just a bit more production. I was very close last week in predicting the USDA but I'm not up a bit and probably in the mid to bottom end of the pre-report trade estimates.

For cotton  I lost about another 140,000 bales this week. I was very close to the estimate by USDA and so I'm down just a bit further from there. As you can see, I''m probably too high on yields and perhaps too low on harvesting rates, which gets me back to the USDA estimate. I know yields are too high, but harvested area isn't as certain. This would say there is some additional downside potential for cotton area. I'll have to check in to see what the legends in Texas and Mississippi have to say. 

Corn Yield:  140.2 bu/ac

Corn Production:  11,990 million bushels

latest USDA Production: 10,773 million bushels 

Soybean Yield:  37.2 bu/ac

Soybean Production: 2,764 million bushels

latest USDA Production = 2,692

Cotton Yield:  832 lbs/ac

Cotton Production 17.452 million bales 

latest USDA Production = 17.650 million bales

Flat production estimates.

I was way off on corn, but as discussed last week, you could see this coming (absolute vs. percent deviations of the model structure), however, I was very close on both soybean (within 0.3% of reported production) and cotton (within 0.5% of production ). Things haven't changed much since that time. 

I've put up a state by state comparison with USDA so you can look more closely at how the states compare to USDA estimates. I've also put up a graph showing the harvesting rate coming out of the model compared to USDA. If I used their harvesting rate I would be even further off on production than I am now (the model is specifically designed to incorporate harvesting losses). I was a bit surprised at the number of acres they held into production but with $8 corn and $30 harvesting costs, it doesn't take many bushels to make harvest worth while. 

Corn Yield:  139.2 bu/ac

Corn Production:  11,957 million bushels
USDA Production: 10,773 million bushels 

Soybean Yield:  36.5 bu/ac
Soybean Production: 2,719 million bushels
USDA Production = 2,692

Cotton Yield:  821 lbs/ac
Cotton Production 17.593 million bales
USDA Production = 17.650 million bales

I've been asked why I don't fix the maize model if I know it is going to be wrong (re-specify it with % deviations). This is a very good point. My only reason is I don't like to change the model during the growing season most years as in other situations I've found I would have made inappropriate adjustments and the model was doing just fine (not the case this year for maize!). Secondly this is all for fun at the moment so to think about the best way to re-estimate the equations in short order might take away from my day job. 

Notice the numbers for Oklahoma cotton. The large difference is a lack of proper synchronization between the harvesting equation and the yield equation. So the harvesting falls off rapidly but this increases the yield at a pace which isn't realistic and thus you end up with production which is far too high.

Here, the US harvesting rate is holding at just over 90%. We saw much lower harvesting rates in 1988 and 1993 but the argument goes that the number and location of livestock was different and we won't see the same type of maize chopping we did during those years.

Crop production and What are the RIN markets telling us?

Edited to add: Yes I think my maize yields are high, and I suspect it has something to do with the way I structured the model years ago (see discussion below). I always try to make few/no changes to the model during the season, but switching to % deviations at the end of the year will be investigated, depending on how this one turns out. The August report from the USDA has an error predicting final yields with a standard deviation of over 6 bu/ac, and doesn't really close on the final number until the October report. This year the September number may be better given the early harvest going on. 

I'm going to try to cover a few topics today with one post as I've had little time in the evening this week to get posts together. So I'd like to talk about some simple observations of the RIN markets that I hope you will scrutinize and criticize, tell you why I think I'm wrong about the corn yield and then give you my predictions anyway. 

What are the RIN markets telling us or maybe I don't speak the language?

First let me start out by saying I don't clearly understand the 2013 RIN prices and what these represent. I'm trying to figure out if these are forward quotes or some other specification. I'll make some comments on them but these are entirely contingent on them being forwarded RINs of some type. If anybody has clarification please let me know.

Conventional RINs

2011 = $0.01

2012 = $0.05

2013 = $0.08

In looking at conventional RINs, I think they suggest several things. With respect to the mandate either 1) the  low price suggests the mandate isn't all that binding and removing it might not have the large effect some might thing (See Irwin and Good here ) and/or the RIN prices include some probability that a waiver will come and devalue existing RINs. I would suspect it is a combination of the two. 

The 2013 value jumps another few cents, reflecting the drought and risk premium, and perhaps even a bit of a tightening blend wall. When I look at maize and oil prices in the futures market, maize prices are down (which would be negative for RIN prices) and oil prices are flat (higher oil prices should lower RIN prices). So this suggests it is either risk or the blend wall, at least to me. 

A back of the envelope calculation might suggest taking the RIN price, multiplying it times the number of gallons per bushel of corn and calculating how much the mandate is supporting the price of corn, overly simplistic? Probably for the reason it might reflect expectations of  a waiver already, but it is my starting point 

Advanced RINs

2011 = $0.49 

2012 = $0.49

2013 = $0.51

Advanced RINs are mostly flat all the way out, reflecting, perhaps, adequate sugar markets. I would argue the RINs give a good approximation to relative prices of ethanol in US and Brazilian markets and tell us all else equal (which it never is!) how much maize prices would have to rise with mandates in place until imports would start displacing maize ethanol in the overall mandate. 

First we might expect that 

Brazil ethanol price + $0.30 for transport - US maize ethanol price - conventional RIN = Advanced RIN

using round numbers for the day these RINs (a day or two old)

$2.85 + $0.30 - $2.65 - $0.05 = $0.45    So in the ball park for expectations.

Again, how much could corn rise all else constant before sugarcane ethanol and maize ethanol come into competition? (again the 2.8 is gallons of ethanol per bushel)

(advanced RIN - Conventional RIN)*2.8  = (0.49-0.05)*2.8 or about $1.23. Yes we are importing now and faster than last year, but we also have a 490 million gallon advanced mandate gap to plug in 2012, 3 times the size of last years. 

Biodiesel RINs

2011 = $1.08

2012 = $1.10

2013 = $ 1.23

For biodiesel, this tells me it is a long way from helping to fill the advanced gap by producing beyond its mandate. Keep in mind if you want to try the simple calculation to the price of soybean oil that the biodiesel mandate is on physical gallons but it produces 1.5 RINs per physical gallon. This should also tell you something about the expected production effects of a re-instatement of the biodiesel blenders credit!

Feel free to beat me up about this string of pure conjecture! 

On to crop yields and production.

Corn Yield:  139.2 bu/ac
Corn Production:  11,930 million bushels

Soybean Yield:  36.3 bu/ac
Soybean Production: 2,706 million bushels

Cotton Yield:  814 lbs/ac
Cotton Production 17.697 million bales

Here is where I paper up my excuses for being wrong on Friday! The graph just below is a illustration (not real model results) of what I mean when I said the model uses absolute vs percent deviations and how that is likely to effect the result. In the model the difference between a poor acre and an very poor acre is a fixed number of bushels, over time this becomes a smaller and smaller percentage of overall yields, thus it implies that the maize crop is better (substantially better) at dealing with abiotic stresses such as drought. This year has certainly called that into question. An alternative specification might use percentages in each class so the same conditions in 1988 would give you a similar percent deviation from the trend line and thus a MUCH lower maize yield than I currently have. I had hoped to run some of this analysis this week but as you can see, I've run out of time.  

I'm clearly above the trade estimates for maize, but I'm below the trade estimates for soybeans, or at least at the bottom end of the range. For cotton I dropped nearly 1 billion bales this week in production but I'm probably still well above the trade estimates for production and yields.