USDA markers

I added the USDA markers to my graphics. I'm in the unusual position of doing decently well in cotton and being farther off from USDA in corn and soybeans. This is not normally the case. USDA lowered corn yields, but I'm still quite far off, but they did break the trend of a big crop getting bigger. In soybeans they certainly moved it up and it was in the upper end of the range of trade estimates while I'm in the lower 1/2. For cotton, my abandonment percentage is different for Texas (and probably Arkansas) but the yield has been spot on this year, or at least it has followed the USDA estimates. I may get conditions next week, but I know some harvest decisions are being made and this is the latest we have had most of this info, so my numbers are unlikely to change. I'll continue to post if I have conditions, otherwise I'll look at the few votes on what crops to add and I'll get started on an additional crop or two for next season. I'll be looking forward to the January numbers as well.

Week32

We are over 50% harvest so I'm no longer getting conditions which is why my prediction is flat. I've updated this because I wanted to show the range in the trade at the moment, it is all over the board and has grown each week (the spread). Compare this to corn and you will see the spread has been a bit more consistent.

Week 32 Spreadsheet (3mb)

Week 31 Crop production and yield estimates

Week 31 Spreadsheet (3mb)

Any interest in wheat or other crops?

I'm considering constructing a similar model to run this spring for wheat, but I'd like to gauge interest before I spend any time doing it. Check out the poll on the right and leave your opinion. It is possible to do crops which are not listed by using conditions of a crop that are reported. I do that for some very minor states now (for example using soybean conditions to determine corn yields).

Week 30

For a discussion of why I no longer am publishing a corrected soybean condition, see the posts that follow this one. Also check the poll on the right. I'm about done with updates for the year. Once 50% of the crop is harvested conditions are no longer reported. This is the latest I've gotten soybean conditions since the data started in 1986 (in my week breakouts) and corn conditions should also be ending soon. You can clearly see in the estimates that harvest issues and concerns about test weights (?) are having an effect on conditions and therefore yield and production.

The yield numbers for cotton have fallen substantially since the first part of September. You will see in earlier posts where I think my errors are coming from on the production side (TX and AR). The yields, however, have tracked USDA reasonably well and so if you fit the model changes to USDA points we might see some sizable reductions in the USDA cotton production numbers in August. I should note that New Mexico has not reported conditions yet for the week so I'm using last weeks conditions carried over. Given the small size of New Mexico and the likelyhood of large changes in thier conditions from one week, this is probably not having a noticable effect on the outcome. You will be hardpressed to see the change when they finally release thier numbers.

I'll continue to post as long as I have conditions and then again near the time of the USDA November report, maybe the December report for cotton and then after the January report where numbers are usually 'final' or at least change little from that point for production and yield. Again I'll solicit any comments you have concerning this blog.

Week 30 spreadsheet (3mb)

Spreadsheet link fixed

It appears that the link the the spreadsheet was having some issues, it seems to be working now. Here is the link.

Week 29 Spreadsheet (3mb)

Why no corrected number for soybeans this week?

The closer I get to harvest, the more observations I'm missing for that week when I calculate my 'correction'. This is because the last date the condition information is published depends on harvest progress. So, if I publish a SOYBEAN correction this week it would be with only a handful of observations. In the US report, I'd only have conditions for this week in 1990, 1992 and 1993. This is clearly related to the delayed harvest we are undergoing.

I probably should have gone ahead and published the correction, but it would require a bit of an explanation. Let me go ahead and do that and see if you readers out there think it should be investigated further. So with potentially fewer observations (not calculated how many) the adjustment this week would be significant. I'd be down to 41.2 bushels an acre. Down about a full bushel from last week. The adjustment in Iowa and Illinois would be about 2 bu/ac. This the last week I'd get any adjustment at all, we have not had bean conditions past week 29 in my calendar

Now it could simply be some fluke of the small number of observations available this late in the year or it could be that conditions tend to significantly over estimate yields when harvest is delayed, or the delay itself results in lower yields. I certainly don't have enough information to draw any conclusion.

It also looks like I trimmed off LAST weeks soybean corrected numbers, so here they are again for completeness. They do NOT include this weeks adjustment with just three observations.

Week 29

Corn: My estimates for corn production and yield are holding steady. As soon as harvest reaches 50% USDA will stop reporting conditions. My adjustment to the corrected model also becomes a bit more suspect this late in the year because the date of the last reported conditions vary. No big deal but worth reporting. My post last week outlined where I'm different. Much of the difference is in 3 cornbelt states (primarily Illinois) where my planting date variable is limiting terminal yields. I've heard increasing stories about low test weights, so maybe I'll end up closer to the final USDA number than I am now. See the earlier post for a discussion of this issue. While conditions may not fully pick up on this late season weather effect on test weights, it may be that the planting date does. So perhaps I shouldn't be so quick to throw it away.

Soybeans: I've consistently been a bit higher than USDA all season, but by less than 1/2 a bushel per acre. What is worth noting that the estimated yield slid again this week and has been on a downward trajectory since mid-September.

Cotton: I'm actually pleasantly surprised how the model has tracked along with USDA estimates on cotton yields. The reports of worsening crops and harvest difficulties appear to be showing up in conditions and if the model is truly tracking well, we might see a sizable downward revision in yield numbers come November. Now my crop production number has consistently been above the USDA number. I'd say 1/2 of this is due to abandonment differences in Texas. The model just does not seem to do well in predicting harvested acres in Texas. The other big difference is in Arkansas where current conditions are well outside of the range seen since 1986. So when determining parameters for the lowest category, the model isn't distinguishing yields much. This will change when I re-estimate after the first of the year. So I'm well above USDA.

Week 29 spreadsheet (3mb)

So where did it 'all go wrong' ?

So I'm pretty far away from the USDA estimates and this last October report moved them even further away from me. The model has performed quite well in previous years and so this of course make me curious as to what was going on and where the differences were. I should point out here that I DO NOT adjust the model during the season, but I do re-estimate the parameters each year.

Well it is clear that one big source of difference is the Illinois corn yield. I currently have 163.1 bu/ac in my corrected model while the USDA is running 179 bu/ac. Is it conditions? Well conditions don't tell much of the story. In 2004 IL yield was 180, in 2008 it was 179 so while trend yields are important here (I'm running 2.7 bu/ac growth annually) lets set that aside for the moment and look at conditions for this week in history.

2004: VP-1 P-2 F-13 G-57 E-27
2008: VP-1 P-4 F-23 G-52 E-20
2009: VP-1 P-6 F-28 G-52 E-13

So conditions are worse than those previous years. So while conditions could still be it, lets look at what I'm getting from the other variable, planting progress. Because this is just a hobby, I search for the most significant planting progress percent among a number of weeks for each state. I try to make the date later as I go north in the states. Well I used the week May 6-12 percent planted for Illinois this year. Now the high in this week over the last five years is 95%, the low is last year at 60%...This year? 10%. This is giving me one heck of a yield hit. If I simply go in and put the 5 year average for planting progress in for Illinois (78.8% planted) my new Illinois yield is 177.7 bushels an acre. Within 1.5 bushels of the USDA estimate. While that doesn't solve everything, it does put me closer to USDA, here is my total if I make that simple change. Now this week, the numbers from the USDA STILL put me quite far off, but I'm a more reasonable distance from the estimate.


Using the average planting progress of the last 5 years JUST for Illinois

So, clearly the USDA estimate suggests I have far to0 much of an impact on terminal yield based on planting progress. This could be because I have the wrong 'week' as the indicator, but planting was delayed across all the weeks to some extent. Maybe it is because farmers can get into the fields and get it all planted more quickly now than before, or maybe the seed technology is different. Or it could be that in previous years, delayed planting resulted in freeze issues that are not materializing this year, or, was the weather just cooperative this year and we didn't have the 'hot and dry' spell that would have crushed a late planted crop?

The same issue is present for IN and MI, but I'm using a bit later week for OH and planting progress was about 'normal' this week and thus you will note that I'm actually above USDA. While there are still several weeks left, I'm not sure I hold much hope to close that much of a gap. What I can say is that planting progress is playing a bigger role this year than I anticipated it would.

Week28

I remain pretty far away form the USDA for corn. For soybeans I'm very close.

Week 28 Spreadsheet (3mb)

Comparison with USDA