Friday, July 13, 2012

Comparison to 1988?

Having pulled earlier than current weeks, the point I was trying to make is that I almost can’t get the kind of deviations seen in 1988 because of the way I’ve structured the model. The conditions are equated to ABSOLUTE yield drops. So when an acre goes from Fair, to Poor, this means that acre drops by a yield volume, not a percent of trend yield. Will that be important this year? I don’t know. This only becomes important in the most extreme years. So to get a 39% drop in yields this year would take far worse conditions than in it did in 1988 given the way the model is structured. 

This comparison to 1988 with the large reduction in trend yields got me thinking last night so I did some calculations last evening with the model, which for whatever reason predicted the USDA maize yield exactly for this month’s report (and so it provides a good starting point for a comparison to 1988).

First let me state the problem with a simple comparison to 1988 was that conditions at this time of the year were better than they are currently for this point in July. No! they declined substantially in the July 10th report to the following....... 

For this week in July in 1988, conditions were as follows

Very Poor-18%

They currently stand at
Very Poor-12%

After the big drop the crop didn't change much after that
and this was the condition of the crop in August-September and I used this in the following calculation. 
Very Poor-16%

Definitely things ended up worse than where things stand now.

So I did some simple experiments taking out trends in yields, putting in 1988 conditions (for states that were reporting then as maize area has expanded, if no data were available I included current conditions) and also putting in 1988 area (as the geographic mix isn’t irrelevant given maize’s movement and the 42% larger corn area in 2012 than in 1988!  )

The ACTUAL yield in 1988 was 84.6 bu/ac.

When I take the model and put in available conditions from 1988, take out the trend back to 1988, and put in the area from 1988 I get 85.7 bu/ac so the model seems to replicate history fairly well, off by only 1 bushel.

If we come forward, putting back in the trend, put in the current planted area, so the only  thing that is different is I’m putting in conditions in 1988 in the model for those states available I get the following

Early September conditions from 1988 = 137.2 bu/ac  (Before the big drop in conditions in 1988 I would get about 157 bu/ac). 

I don’t get anything close to a 39% decline from trend discussed in the article even with conditions as ‘’bad’’ as 1988, a significant part of this would be the assumption embedded in the model which assumes absolute deviations based on conditions instead of % deviations, itself an assumption that the crop is better able to handle abiotic stressors than it was 1988. The outcome for crop yield distributions (the coefficient of variation) is a heavily debated one*

Given historical USDA behavior and accumulation of crop damage, we could possibly see yields fall further, but I would be shocked if we hit 1988 style deviations. Of course as I do this primarily as a hobby I could be way off!

(the specification of the model implies constant absolute deviations based on condition, not percentage deviations based on conditions, which shows my own personal bias)
(the model calculations above are the average of either weeks 9-10 for current conditions or weeks 19-20, mid September, for final conditions. )
(I did NOT adjust planting progress to 1988 numbers which does impact final yields and probably gives a small unwarranted boost to yields for this year). 

No comments:

Post a Comment