On June 12, the USDA will release its latest World Agriculture Supply & Demand Estimate. I have previously written an article questioning the preliminary estimate of the USDA, in May, of a corn yield at 166.0 bushels/acre. The highest estimated point yield I found was 164.7 bushels/acre.
After thinking about it more, I realized that I was basing the results on the final yield for each crop year and not the estimated yield for the May report or June report. I still received statistically significant results with an R^2 around 80%. However, by using the estimated yield for each month, such as the month of June from 1993-2011, I was able to get a much more robust model with an R^2 of 97%!
In the June 2011 report, the USDA said that the yield estimate was based on:
Projected corn yield based on the simple linear trend of the national average yield for 1990-2010 adjusted for 2011 planting progress.
I went back and looked at each report since 1990 and used a linear trend and planting progress to best fit a regression line. One difference is that I could not find June reports from 1990-1992 so the regression based on yield estimates starts from 1993.
First, here are the results of the model that uses the June yield estimates with a time trend and planting progress. For planting progress, I used week 20 of each year.
One can see that plant progress, and time trend are statistically significant, with a t-statistic over 2 and p-value less than .05. Adjusted R^2 is over 97%. This model provides a "point" estimate of 163.0 bushes/acre with a 95% confidence interval of 157.5 - 168.4.
In 2011, this model, with the regression run from 1993-2010 gave a point estimate of 158.6 bushels/acre vs. the actual of 158.7 bushels/acre. Now this estimate is not fool-proof, there are risks, some statistical tests indicated the estimating values might have changed in the last 10 years. However, I generally prefer a model with more data as opposed to less; but for those inclined to think differently, the model from 1999-2011 gave a 165.7 bushel/acre point forecast.
We will move on to modeling a yield projection based on the final yield of each year's crop, instead of the yield estimated for that month. Here is the regression from 1990-2011.
One sees that plant progress is now not statistically significant and that the adjusted R^2 is much lower at just under 73%. This model gave a point forecast of 164.7 bushels/acre. The May 2012 report used a regression model that skipped the data for 2011, which I noted before as unusual, and just used 1990-2010. I also ran a regression over this period and the model gave a 166.7 bushels/acre point forecast.
In summary, most models give a lower yield forecast than the current USDA projection of 166.0 bushels/acre. I lean towards the forecast that uses the estimated yield for June each year, instead of the final yield because the R^2 is much higher. If the USDA did reduce the estimated yield by only 3 bushels/acre this would cut corn production by 267 million bushels.
Stock traders who do not have a futures account can use the CORN ETF to trade this market. Also, traders have to be aware that last month, the CMEGroup has changed its trading hours for the corn contract and now the corn market will be electronically trading when the report is released at 8:30 am EST. Be aware that placing orders right in front of this report can be very dangerous as liquidity disappears.
That said, my guess is that most of the news about a "bumper" corn crop has been priced into the market, and corn bears should recognize this. One last word of warning, corn trades the weather forecasts for most of June. The report will provide a reaction in the market, one way or the other, but after a few days, a "weather" market will provide short-term direction.
Additional disclosure: I have a position in corn futures.