Discover and read the best of Twitter Threads about #FDD
Most recents (24)
1. With the June Acreage report up on Thursday, good time to review what we know about revisions to USDA planted acreage estimates. My latest #FDD article.
2. Important first finding. Revisions to USDA March Prospective Plantings estimates are unbiased for major crops and total principal crops. Average revisions from March-January very, very close to zero. In other words, statistically unbiased. 
3. Unbiased does not mean unchanging. Not unusual for corn planted acreage revisions to be up or down as much as 2 million acres. That matters to the market. Key is that ups and downs tend to offset over time.
1. Now this is gonna catch your attention. Gary and colleagues have a very interesting new #FDD on projected corn and soybean returns in Illinois based on planting date and present prices and costs.
2. Recall from my #FDD last week that agronomic field trials show, on average, expected corn yields in Illinois by planting date look like this:
3. Here is the data on soybeans and planting date. Link to my article: farmdocdaily.illinois.edu/2022/04/what-d…
1. Do new crop corn and soybean futures contain a systematic pre-harvest "weather" premium? Something that @JosephJanzen covered in our IFES webinar last week. Before I review Joe's results on this question, some background.
@JosephJanzen 2. This argument has been around a long time. Back in the 90s, Bob Wisner of Iowa State and Dean Baldwin of Ohio State (better people you will never know) were on one side, arguing there was a pre-harvest weather premium could benefit crop farmers.
@JosephJanzen 3. Carl Zulauf and I were on the other side, basically arguing that under the efficient market hypothesis this did not make sense and that the statistical evidence was weak. Since Carl and I were at Ohio State and so was Dean Baldwin the different views were all in one place!
1. Back on crop acres with new #FDD article this week. Tried to figure out what happened to the 14.4 million acres that left US crop production after 2014. farmdocdaily.illinois.edu/2021/06/where-…
2. Here is my accounting of total US crop acreage. Notice that total was pretty stable from 1998-2014 around the average of 359.4 million acres. Then down down down through 2020. Total drop after 2014 of 14.4 million acres through 2020.
3. Started analysis with acres going into and out of CRP. Anything funny going on there. The answer is yes. Should be an inverse relationship between CRP and principal crop plus prevent plant acres. This held through 2014, then both declined. What gives?
1. This allows me the opportunity to implore my friends at NASS to avoid the mistake they made last year by revising Sep 1 stocks later on without any new data. They did this last January, which is what Karen mentions here.
2. To recap: last January, USDA revised 2018/19 production, feed and residual use, AND ENDING STOCKS. This was a violation of what I previously thought was an unbreakable principle for a statistical agency in constructing crop balance sheets.
3. What is the unbreakable principle? Never, ever, ever change the quarterly stocks estimates after the fact without having Census data (and who really cares about that 5 years down the road?) This is supposed to be the number measured with the least error. Leave. It. Alone.
1. And I thought I would have a normal day today! Going to focus on the implications of the SRE denial decision in this thread. Lots of unknowns about the E15 pump announcement. So sticking with what is known.
2. The implications of the SRE denial is far-reaching with regard to PAST and FUTURE implementation of the RFS mandates. Without the gap SREs, nationwide application of the 10th circuit decision means that very few of the SREs issued since 2011 or 2012 are valid. Poof.
3. I recall some in the oil refining business saying I was out to lunch that the 10th Circuit decision could blow up the SRE program. Done blowed up now! But honestly this was inevitable.
1. Implications of FSA Sep acreage totals for corn and soybeans planted acreage. I am simply updating tables found in this recent #FDD article: farmdocdaily.illinois.edu/2020/08/using-…
2. Given the reported FSA Sep 1 totals, have to make an assumption about the completeness of reporting and % of acres enrolled in FSA programs. The latter is very stable over time so pretty easy. The kicker is the completeness of reporting. We knew it was way behind in Aug.
3. I make two assumptions about completeness of FSA reporting. In this first table I assume the reporting % for this Sep is the historical Aug avg. Just sayin a month behind.
1. Take on the big issue with FSA acreage data in a new #FDD. What do the preliminary data released by FSA imply about possible revisions to NASS planted acreage estimates? June: 92 million for corn and 83.8 million for soybeans. Those are the targets. farmdocdaily.illinois.edu/2020/08/using-…
2. We look at historical ratios of preliminary to final FSA and final FSA to NASS acres. You can see those details in the article. The important part is the projections at the end of the article.
3. In the first set of projections we use historical averages for preliminary to final FSA acreage. In other words, we assume that the Aug FSA reported planted acreage of 81.1ma for corn represents 99.1% of final FSA acreage.
1. I am intentionally going to propose a controversial hypothesis in this thread. My hypothesis is that #pftour20 is more valuable now than in the past due to the change last year in USDA yield estimation methodology. My point is not related to wind damage. That is separate.
2. Recall that USDA last year dropped the objective yield survey from the Aug corn and soybean crop report. I have been worried about this decision ever since it was announced. The justification for dropping the objective yield survey for Aug was never really provided by USDA.
3. I presume that the USDA had grounds for believing that the objective yield survey did not add that much information early in the corn and soybean yield forecast season. I am certainly not opposed to cost saving if accuracy is not sacrificed.
1. Yea I went there in a new #FDD last week. Dove into the data on ethanol production profits during the COVID pandemic. Here is the link to the article farmdocdaily.illinois.edu/2020/08/ethano…
2. This is where everyone starts in thinking about financial hit ethanol industry has taken due to pandemic. Ethanol production off 48% at bottom, same as gasoline use.
3. This is what happened to ethanol plant prices. Even deeper trough than ethanol production. Down 62% at the bottom. How could the ethanol industry not have been absolutely decimated after seeing this?
1. New #fdd last week on forecast performance of simple crop weather models in forecasting US average corn yield. Interesting exercise farmdocdaily.illinois.edu/2020/07/the-ac…
2. Compared 2 models for May-Aug early season yield forecasts of corn: i) WAOB/USDA model, and ii) a similar model that TH and I developed. We reviewed the WAOB model is this article a week ago: farmdocdaily.illinois.edu/2020/07/unders…
3. Both models include trend, planting progress/late planting, June precip, July precip, and July temps. Some differences in data definitions and variables. Biggest is the WAOB use of a June precip shortfall measure. We use a standard quadratic.
1. WASDE yield projections in May, June, and July always a source of market conversations/controversy. Corn yield forecast was 178.5 bpa in May and June for the US. What will it be in the July WASDE that will be released on Friday, July 10?
2. Since 2013, WASDE yield projections for May-July have been based on a crop weather regression model. In our experience, the model and how it is used is not very well understood. So we wrote a #FDD last week to try to explain the model. farmdocdaily.illinois.edu/2020/07/unders…
3. First thing to get straight is that the WASDE/WAOB May-July corn yield forecasts are model-based, while NASS Aug-Nov yield forecasts are survey-based. Completely different methodologies. Only looking at WAOB model here.
1. This #FDD looked at sensitivity of soybean trend yield estimates for the US to different sample starting dates. Similar to what we did the week before for corn. farmdocdaily.illinois.edu/2020/06/how-se…
2. Here is the history of US average soybean yields since 1960. The question is when should we start our sample for estimating the the trend. Use the full 60 year sample to estimate the line shown. Sure looks like some kind of shift in the last decade.
3. We estimated linear trend regressions moving the start year ahead one year at a time. Then projected 2020 trend yield for each sample set. Here is what you get doing that.
1. New #FDD out this week on estimating 2020 corn trend yield for the US. It would seem simple to agree on the "right" trend but it is not. farmdocdaily.illinois.edu/2020/06/how-se…
2. History of US average corn yield since 1960. We ask what year should be the first in the sample used to estimate a trend line? Only discussed in bits and pieces as far as I could find. We take a systematic look.
3. We first estimated trend line for 1960-2019 and used that line to project a trend yield for 2020. Just project the line in the chart out one more year. This gives an "unconditional" trend estimate of 173.2 bpa.
1. Remember after the 10th Circuit SRE ruling when I said never under-estimate the ingenuity of the lawyers for crude oil refiners to figure out new ways to undermine the RFS? Well they have apparently really outdone themselves this time with "gap" SRE petitions.
2. My comments are prompted by an @OPISBiofuels article yesterday ***US Farm and Biofuel Groups Demand Answers From EPA on SRE 'Gap Petitions'
@OPISBiofuels 3. The 10th Circuit ruling held that only small refiners with a continuous exemption over the life of the SRE was valid. If applied nationwide this would effectively kill the current EPA SRE policy. See this #FDD for full details farmdocdaily.illinois.edu/2020/03/epa-an…
1. When thinking about early season crop condition ratings for corn and soybeans I think there is a topic that does not get enough attention. It is the question of statistical bias. We have covered this issue in several of the #FDD articles I tweeted our earlier today.
2. Here is an updated chart from a 2017 article. Last minus first US (18 state) G+E rating for corn and soybeans. Fascinating patterns here. Statistically unbiased through late 90s, then a clear bias develops towards high early ratings.
3. Since 1999, condition ratings for corn have almost never gone up after the initial rating. That is rather remarkable. On average, they have dropped by 7 percentage points. Except for 2012, decline is usually no more than about 10 points.
1. Crop condition ratings for corn and soybeans will start getting a lot of attention from here forward. Going to list 6 #FDD articles in this thread that may be useful references regarding some of the issues involved in using the ratings to forecast yields.
2. "When Should We Start Paying Attention to Crop Condition Ratings for Corn and Soybeans?" May 24, 2017 farmdocdaily.illinois.edu/2017/05/when-s…
3. "How Should We Use Within-Season Crop Condition Ratings for Corn and Soybeans?" June 1, 2017 farmdocdaily.illinois.edu/2017/06/how-to…
1. Good message here about early crop condition ratings. Just don't contain much information. This gives me the chance to review some previous #fdd articles on condition ratings.
2. First up is a pair of #fdd articles on high early season corn and soybean ratings from 2018. Corn farmdocdaily.illinois.edu/2018/06/what-t… and soybeans farmdocdaily.illinois.edu/2018/06/high-e…
3. These articles cover: i) degree of bias in early season corn and soybean condition ratings, and ii) degree of predictability in early season corn and soybean condition ratings.
1. Yesterday's Crop Progress report allows me to compute our measure of late planting for the US corn crop for 2020. We define this as % planted after May 20. This year = 16.6%, slightly below average of 17.6%.
2. Like all averages, this hides a lot of variation across states. Super fast planting in the western Corn Belt, not as fast in the eastern Corn Belt this year.
3. With the late planting measure so close to average in 2020, projected trend deviation for US average corn yield in 2020 is close to zero. Note this a one variable model. The main show for determining US corn yield is just getting started!
1. My next thread deals with the yield impact of late corn planting. We looked at three approaches to estimating impact on US avg corn yield in a #FDD article last week farmdocdaily.illinois.edu/2020/05/the-im…
2. The first method simply regressed trend deviation for US avg corn yield on the measure of late planting. Slope is -0.2, or decline of 2bpa for each 10% increase in late planting. Notice how well that fit the obs for 2019.
3. The explanatory power for this regression is low because we took trend out and only included one variable---late planting. Did not account for any weather impacts. Could result in a biased regression coefficient estimate due to "omitted variable bias."
1. It's now time to turn our attention to the flipside of US corn planting progress. That is, how much of the corn crop will be planted "late"? This is the key concept with regard to corn planting, not how much is "early."
2. Here is why I make that statement. This is drawn from a terrific #FDD article (as usual) by my crop science colleague here at UI, Emerson Nafziger. farmdocdaily.illinois.edu/2020/04/planti…
3. Notice for both corn and soybeans in Illinois two things: 1) the planting date/yield curve is very flat from early April through mid-May, and 2) the curve begins to curve down rapidly after mid-May. Early corn planting helps yield much less than late planting hurts yield.
1. It's planting progress Monday once again! Where will corn planting progress end up this week? Last week at 67% and proceeding at an extremely average pace so far in 2020.
2. I go back to the historical data to make an initial guess for week #20. 4.1 suitable field days and 3.5 % planting per suitable field days avg. over 1995-2019. Results in 14.5% planted for the week. Rounding that would put us at 82% planted.
3. A lot of rain over the weekend but reasonably open week in most of the Corn Belt last week. So, I am going to go with 4.5 suitable field days (really going out on a limb!) and 3.5% per day for a forecast of 16% rounded. This would put total progress at 83% in today's report
1. USDA Crop Progress out today. Time to compare to the forecast I made this morning. To recap, USDA reported corn planting progress at 51% as of Sunday. My simple method forecast was 43%. Where did I go wrong?
2. My simple method involves forecasting the two components of weekly corn planting progress: i) # of suitable field days, and ii) % progress per suitable field day. Basic idea is to focus on # suitable field days and assume rate per day is the historical average. I
3. I assumed 3 suitable field days and the historical average of 5.3% for rate of progress per day. This gave a gain in corn planting progress of 16% versus the actual of 24%. Note that US corn planting progress this year has been well above average last couple of weeks.
1. Next thread on corn planting progress. Actually forecasting planting progress for this week, which is week #18 on the USDS/NASS calendar. Start with these historical averages.
2. If last week was "average" we would have 3.9 suitable field days, plant 5.3% per suitable field day, and have total progress of 20.5% (3.9 X 5.3). Rounded this would put total US corn planting at 48% as of Sunday. This is our "base rate" forecast.
3. Our position is that the best forecast of rate of planting progress per day is the historical average. Maybe John Deere knows how much ground is being covered per day but I don't have any data to do better than the historical average of 5.3% per suitable field day.
