The second survey is the Objective Yield Survey (OYS) and is based on an area frame sample of cultivated land in 10 principal states for corn and 11 states for soybeans. Based on the June acreage survey, a random sample of fields is drawn in each state and enumerators visit these fields to take measurements needed to forecast yields in predetermined segments of the fields.

In 2011, these measurements were taken in 1,920 cornfields and 1,835 soybean fields. For corn, the number of plants and number of ears per segment are counted, and the size of the ears is measured. For soybeans, enumerators measure row width and count the number of plants, number of main nodes, lateral branches, dried flowers and pods and pods with beans in each segment. The data are used to forecast grain weight (yield) per acre. Each segment is visited in September, October and November (if not yet harvested) to take new measurements and counts to form new yield forecasts. Just before the operator harvests the field, each segment is hand harvested and weighed. The data from the two surveys are combined to forecast average yield and production.

Good says that the USDA makes new yield and production forecasts in September, October and November with final estimates released in January. “Because yield forecasts are limited by crop maturity and are influenced by subsequent weather, the August yield forecasts tend to have the largest deviation from the final estimates released in January,” he says.

For the period 1970 through 2010, the August yield forecast for corn ranged from an overestimate of 18.3% to an underestimate of 10.1%. The middle 50% of the forecast errors ranged from an overestimate of 1.3% to an underestimate of 5.1%.

For soybeans, the August yield forecast ranged from an overestimate of 16.2% to an underestimate of 10.2%. The middle 50% of the forecast errors ranged from an overestimate of 3.0% to an underestimate of 4.5%.