Predicting the future with precision is tough unless you are a psychic, a trait that most grain analysts, including the U.S. Department of Agriculture, do not possess.

Last Thursday and Friday the USDA released preliminary projections for 2016/17 U.S. crops at the annual Agriculture Outlook Forum, where the corn and soybean forecasts are always a sought-after matter.

With the spring crop season still far in the future, some components of the balance sheet have more skill than others. The current market and economic situation allow for some meaningful insight to planted area, which is probably the most informed element at this early stage.

Early insight on yield is desirable, but yields are highly vulnerable to summer weather. As such, USDA’s current predictions are nothing more than a simple trend, and this method is not too bad on the whole. USDA’s average deviation from February to final is 5 percent on corn and soybeans, and 6 percent on wheat (tmsnrt.rs/1ThfMnP).

While 5 percent does not seem a lot, this is the difference between a 162- and 170-bushel-per-acre corn yield. If corn and soybean yield were to change significantly, this would likely not be reflected until USDA’s August report, in which state-level projections based on ground surveys are first released.

Otherwise, large changes to yield are generally not observed at the national level prior to August unless there is a really obvious weather issue, similar to 2012 when yield had already been cut by 11 percent by July due to a historical drought.

The general rule of thumb is that items towards the top of USDA’s preliminary 2016/17 balance sheets are more predictable than those towards the bottom. With this in mind, plug the ending stocks into your balance sheets at your own risk.

Acreage: Steady tendencies

Due to large inventories and low global prices, USDA expects total planted acreage for wheat, corn, and soybeans to decrease slightly for the second year in a row. However, they expect that corn will be the big winner of the three, gaining 2 million acres over last year, based on lower fertilizer and fuel prices.

Predictions for the 2016/17 U.S. wheat crop were already somewhat revealed in mid-January’s Winter Wheat Seedings, in which winter wheat area was expected to decline nearly 2.9 million acres on the year. Last week, USDA placed all wheat area down 3.7 million acres, meaning that spring wheat is not expected to pick up any of that slack.

For corn and soybean acreage, the two major reports the market looks to during the year are Prospective Plantings, released at the end of March, and June Acreage, at the end of June. Both reports incorporate data from farmer surveys on intentions in March and reality in June.

Given historical estimates back to 1998, Prospective Plantings is unlikely to move corn or soybean planted area more than 1.4 percent and 1.8 percent in either direction, respectively, from the initial figures.

There is not a clear report-to-report bias over the last 18 years on the acreage predictions, though USDA’s February estimate often follows closely to the final figure from the previous year. But corn acreage from February to June was overestimated consistently over the past three years, and USDA still predicts an area increase over last year (tmsnrt.rs/1Thg6Dk) (tmsnrt.rs/1ThfZaR).

Over the last two years, soybean acreage estimates made considerable gains in the March and June reports, though this ultimately backfired last year. Current soil moisture is generally well above the last two years and if this continues, soybeans could receive a boost, as reported March/June/final soybean acreage tends to track higher in years with higher early soil moisture.

If the soils become too wet by spring, corn acreage is likely to lose ground to soybeans as the planting window for corn closes earlier. This will have to be something to watch over the next couple of months as planting approaches, but at this point, drought-impacted planting seems a less likely scenario.

Carryout: Darts at the dartboard?

Of all the balance sheet components projected in February, the lowest skill would probably reside with ending stocks. This is not surprising though, since it relies on everything that happened during and after the actual harvest, and is therefore the furthest prediction into the future.

Carryout is prone to swing wildly from initial to final estimates, sometimes more than doubling (or halving) over the whole time frame. In terms of the actual percentage change from February to final since 1998, corn has been the most volatile and wheat the least, but neither of them have a particular bias toward the high or low side.

USDA has had a remarkable tendency to overestimate soybean carryout in February, as this was true in 13 of the last 18 years. However, the predictions seem to get better following years with larger carryout, meaning 2016’s estimation of 440 million bushels after 2015’s 450 is likely not overestimated, barring a significant supply disruption this summer.

But historical February-to-final carryout becomes more interesting when considering the actual volumes. On average since 1998, USDA has missed corn ending stocks by 439 million bushels. For perspective, this volume is representative of the entire corn crop produced by the state of Missouri.

USDA’s largest miss in this regard came in 2004, when corn carryout was initially estimated 1,293 million bushels below final, or a corn volume similar to that produced by Minnesota last year. The differential on soybean carryout has been similar in magnitude to that of corn, but the largest miss was 275 million bushels in 1999. Soybean carryout has exceeded this volume itself only five times since 1998.

While early projections on wheat ending stocks are the most reliable of the three crops, the disparity between initial and final can still be quite large, to the tune of an entire winter wheat crop in Oklahoma – 5 percent of overall U.S. wheat production.

However, USDA had initially under-predicted wheat carryout the last two seasons, during which global supply ballooned and prices subsequently fell, a pattern that we are very much still in heading into 2016/17. Even though the U.S. wheat crop is expected to fall on the year, USDA predicts ending stocks at a 29-year high, influenced by slow sales and a falling stocks-to-use ratio. Judging by the past two years, this estimate is sound and if the crop is particularly good, it will push even higher.