How Good is USDA’s Crystal Ball?

Since the USDA released its August production estimates two weeks ago, the trade has largely turned its attention to predicting this year’s yields. The Pro Farmer Midwest Crop Tour was in Nebraska last week and reported a potential corn yield of 172.5 bushels per acre and an average soybean pod count of 1,210.8 pods.

The USDA August estimates forecast an average yield of 186 bushels per acre in Nebraska for corn, and 58 bushels per acre for soybeans, both numbers exceeding the estimates of the Pro Farmer Tour. Last year the Pro Farmer Tour estimated corn yield at 179.2 bushels and the soybean pod count at 1,299.1 pods. Actual average for corn and soybeans last year was 192 bushels and 59 bushels respectively.

What is the accuracy of USDA August yield estimates over time? Figure 1 graphs USDA August forecasts for Nebraska average corn yields (orange line), along with actual yields (blue line) for 1980-2018. The grey line plots the differences between actual yields and August yield estimates. A point on the grey line above zero indicates actual yield exceeded the August estimate. A point below zero indicates actual yield fell below the August estimate. In percentage terms, the differences between actual yields and the August estimates averaged just -0.51 percent of actual yields over the period. In other words, on average, USDA August estimates tended to overestimate yields by 0.51 percent of actual yield, or 0.23 bushels per acre. Interestingly, the biggest difference between actual yield and the August estimate occurred in 1993 when the USDA overestimated yield by 20 bushels per acre. If you recall, excessive rain and flooding plagued parts of Nebraska in 1993 much like this year.

History has shown USDA August estimates of average corn yield in Nebraska are remarkably accurate. The August estimates missed the mark by more than 10 percent in only 2 years over the past 38 years, 1983 and 1993. The August estimates do tend to overestimate yields slightly, and this tendency coupled with this year’s delayed planting could mean the USDA projection of 186 bushels per acre might be a bit high. Most experts believe the USDA will lower its yield projections the closer harvest looms.

Source: USDA NASS and NEFB calculations

Figure 2 graphs the same information for soybean yields between 1980-2018. In percentage terms, the difference between actual yields and the August estimates as a percent of actual yields averaged 1.33 percent, meaning USDA estimates tend to underestimate yields by 0.90 bushels per acre on average. The biggest absolute differences between actual yields and the August estimates occurred in 1983 and 1984 when actual yields were 7.0 and 7.5 bushels per acre less than the estimates respectively. On the other hand, USDA overestimated yields by 6.5 bushels per acre in both 2005 and 2013. USDA August estimates missed the mark by more than 10 percent in 10 of the 38 years, or 26 percent of the time. And, the August estimates tend to underestimate yields. History would suggest then that the August soybean yield estimate of 58 bushels per acre is more likely to be off. But then again, last year the USDA nailed yield perfectly.

Given this year’s track record thus far, the surprises are probably not over. Unless the USDA grossly underestimated yield in its August projections, Nebraska’s total corn and soybean production this year will be less than last year. Last week’s Pro Farmer tour only reinforces this notion. This means crop receipts will be less this year too.>

Source: USDA NASS and NEFB calculations