«Crop Forecasts Office of the Chief Economist, World Agricultural Outlook Board Miscellaneous Publication No. 1554 Understanding Crop Statistics. ...»
Sample corn, soybean, cotton, and spring wheat fields are selected from those identified in the area-frame sample portion of the June Agricultural Survey. Winter wheat sample fields are selected from the fall area-frame survey. Observations within each selected field are made in two randomly located plots.
Plots include two or three adjacent rows of predetermined length.
Harvested yield can be thought of as biological or gross yield minus harvest loss. Counts, measurements, and other observations from each sample plot are input to statistical models based on historical data to predict final number of fruit and final weight per fruit. A forecast of gross yield is calculated by multiplying these two components together and dividing by land area. Figure 1 shows the forecast variables used to predict the gross yield components for each crop.
Plant characteristics used as prediction variables change as the crop maturity progresses. At an early stage, plant counts may be the only data available for forecasting the number of mature fruit. As the crop matures, actual fruit counts can be used, and weights and measurements of the immature fruit are used to predict final weight per fruit.
The same plots are revisited each month until the crop is mature. At that time, the plots are harvested and final counts and weights are obtained. After the entire field has been harvested, the sample field is revisited and two more plots are laid out. The grain left on the ground in these plots is picked up and weighed to provide a measure of harvest loss.
When harvest is complete, the farmers who operate the sample fields are re-contacted to obtain final harvested acres and yield for the sample field.
The estimate of number of soybean pods per acre from the Objective Yield Survey is usually very consistent from month to month and accurate when the bloom period has ended. Record pod counts have been occurring in recent years as a result of a shift to narrower rows. Pod count forecasts usually stabilize with the September survey.
Average pod weights prior to crop maturity are based on historical averages. In normal years, much of the soybean crop has matured by the October survey, so current-year pod weights are used.
Corn objective yield survey forecasts are based on estimates of number of ears and average ear weight.
The ear count forecasts are accurate early in the season. When the crop is late developing, the August projection of ears is based on a model using plant population. Historical average ear weights are used until ears are present to measure. Kernel row length models are then used to project ear weight until crop maturity.
Cotton objective yield forecasts also contain two components, number of bolls, and forecasted weight per boll. The boll count model is in effect a growth model that uses current survey counts compared to historic counts of squares, blooms, small and large bolls to forecast open bolls. Each month during the survey visit to each plot, open bolls are picked and sent to a laboratory where they are dried, weighed, and the lint seed ratio determined. When 20 percent of the forecasted number of bolls for the sample plot have been harvested, the current measured boll weights are used for that sample. Otherwise, 5-year historical average boll weights are used.
Identical procedures are used to forecast yields for both winter and spring wheat. Number of head forecasts are based on different models depending upon crop maturity. These models have a considerable forecast error until emerged heads are present to count. Historical average head weights are used until fertile spikelets and actual filled grains are available to be counted.
Potential accuracy of each month’s forecast for these crops is dependent on the crop maturity at the time of the forecast and future weather. When maturity lags normal patterns, number of pods, ears, etc., is based on number of plants and fruiting positions rather than actual number of fruit. Thus, when maturity lags, the forecasts become more variable because the expected number of fruit can differ from the final.
However, the primary source of forecast error occurs when final end of season fruit weights differ from the historic average because fruit weight cannot be fully determined until crop maturity.
NASS will revise estimates of harvested acres if necessary during the forecast season. Again, the goal is to make the production forecasts as accurate as possible. The production forecasts are based on projecting the acres that will be harvested and the final yield per harvested acre. If acres are lost during the forecast season because of weather or disease problems, those yields drop to zero, the acres are classified as planted but abandoned, and acres for harvest reduced. For this reason, it is possible for the production forecast to be reduced without a corresponding drop in forecast yield per acre. It is also possible for the yield per acre to increase during adverse periods if acres for harvest are abandoned and classified as not for harvest. Data on which to base changes in harvested acres come from the yield forecast surveys when sample fields are taken out of production or the operator reports acres no longer being considered for harvest.
Grain in Storage: Quarterly grain stocks estimates are based on surveys conducted during the first two weeks of December, March, June, and September. Separate surveys are conducted to obtain the on-farm and off-farm estimates. The on-farm stocks survey is integrated with the quarterly agricultural surveys.
These producers are asked to provide the total quantities of grain on their operations as of the survey date. This includes all whole grains and oilseeds stored whether for feed, seed, or sale as well as any stored under a Government program.
The off-farm stocks survey is an enumeration of all known commercial grain storage facilities. This includes approximately 12,000 facilities with about 8.50 billion bushels of storage capacity. An effort is made to obtain a report from all facilities. Reports of stock holdings are normally received from operations covering about 90 percent of the capacity. Estimates are made for missing facilities to make the survey complete.
Considerable data available from other organizations--both private and public--are used to evaluate the accuracy of production estimates and to determine the final estimates. These sources of information become available during the marketing year, but often after the preliminary production estimates are determined. Some examples of administrative data follow.
Acres Planted for Selected Crops. Prior to 1996, farmers who participated in government programs and wished to be eligible for deficiency payments or disaster benefits must have recorded the acres they planted, by crop, with the USDA Farm Service Agency. These data became available at the end of the growing season and provide a measure of the minimum amounts planted. These data affected revisions in acreage estimates if they were above the levels shown by the survey data.
Utilization Data. Information about imports, exports, soybean crush, cotton ginnings, and industrial use become available during the year. These data are used in a balance sheet that starts with carryover stocks from the previous year and the current production estimate to give a measure of total supply. The subtraction of the utilization data at the end of the marketing year from the total supply established at the beginning of the crop year should correspond closely with the ending stocks. If there is a large unexplained difference or residual, the previous year’s acreage, yield, and production survey and stocks data are reviewed to determine where revisions within the range of the survey sampling errors can be made to minimize the residual in the balance sheet.
Preparing Supply and Demand Estimates
The process used by USDA to make supply and demand balance sheet estimates differs from that used for NASS crop production forecasts. NASS forecasts, as seen above, are based on surveys of farmers and on examination of crops in the field. Estimating crop production for foreign countries often involves greater subjectivity because objective farm surveys or other ground truth may not be available. And, forecasting how available supplies will be allocated between domestic use, trade, and stocks in more than 100 countries also entails considerable economic analysis and judgment.
Several features of the process used to derive supply and demand estimates help compensate for the paucity of foreign forecast data. All available information sources are compared. USDA's own resources include weather analysis, country reports from our agricultural attaches abroad, and evidence from satellite imagery. In addition, private and public information sources are considered. This broad information base is reviewed by analysts from several agencies who bring diverse expertise and points of view to bear. To arrive at consensus forecasts, alternative assessments of domestic and foreign supply and use are vetted at meetings of Interagency Commodity Estimates Committees convened and chaired by the WAOB. Throughout the growing season and afterwards, estimates are compared with new information on production and utilization, and historical revisions are made as necessary.
The Balance Sheet Concept
USDA supply and demand estimates reflect a full balance sheet for each commodity and country.
Separate estimates are made for beginning stocks, imports, and prospective production to determine the total supply of a crop that will be available for the new marketing year. The demand side of the balance sheet reflects domestic use, exports, and ending stocks. Domestic use may be further subdivided to the extent that such data are available from other sources. For example, USDA obtains data on U.S. wheat ground for flour, oilseed crush, and cotton mill use from the Bureau of the Census. It should also be noted that the demand side of the balance sheet may include a category for “residual” or “unaccounted” disappearance to account for disappearance or usage that cannot be verified or cross-checked against another objective information source.
The balance sheet disciplines individual estimates: total supply must equal domestic use plus exports and ending stocks. Prices tie both sides of the balance sheet together by rationing available supplies between competing uses. Prices also shape planting decisions for the out year, providing a link between current and future years. The process of forecasting price and balance sheet items is a complex one involving the interaction of expert judgment, commodity models, and in-depth research by Department analysts on key domestic and international issues.
This process plays out each month in Interagency Commodity Estimates Committees, where analysts from the World Agricultural Outlook Board, the Economic Research Service, the Foreign Agricultural Service, the Agricultural Marketing Service, and the Farm Service Agency meet to evaluate current forecasts, new data from NASS and other sources, new information on foreign markets from the FAS staff at foreign posts and other sources, and important U.S. policy developments.
Commodity models, together with new research on commodity market issues, provide the framework for assessing this information. For example, committee members use a variety of models developed in the Economic Research Service and the Farm Service Agency to analyze the new crop information available each month. Expert judgment then provides a reality check and evaluation of model results, and has the final say in determining the Departmental estimates.
Critical supply and demand relationships change over time as policies and structure change. A quality Departmental forecasting process requires a strong research program to ensure that understanding of markets keeps up with changing conditions. For example, the 1996 Farm Act has increased the responsiveness of planting decision to changes in market prices. It has also influenced how the market determines prices. USDA analysts are responsible for keeping the Department’s information base and models abreast of changing market relationships. Research conducted by ERS is a critical part of the process of maintaining the quality and objectivity of Departmental supply, demand, and price estimates.
Table 1 shows the balance sheet for U.S. soybeans published in the August 12, 1998, WASDE report.
USDA made several offsetting adjustments in projected soybean supply and demand for 1998/99. On the supply side, both domestic soybean production and beginning-year stocks were lowered 5 million bushels from projections made in July. Production was lowered to reflect new information from NASS’s first survey-based forecast of the 1998 crop. Also, the import projection was increased 1 million bushels.
Since these three changes reduced projected supply by 9 million bushels, offsetting changes had to be made on the demand side of the balance sheet. Domestic crush was increased 15 million bushels, despite low crush margins, as demand for soybeans by foreign processors diminished. Seed usage was increased 8 million bushels, following a data series revision for 1997/98. Exports were reduced by 25 million, as foreign competition and declining demand lowered prospects abroad. The residual (unexplained) use, was reduced 2 million, in line with historical indications. Finally, ending stocks were lowered 5 million bushels to balance total use with total supply. The small change in ending stocks resulted in no change in the 1998/99 price forecast range.
Table 1. Projected 1998/99 U.
S. Soybean and Product Supply and Use 1/
Source: August 12, 1998 World Agricultural Supply and Demand Estimates report, WASDE-341.
Revisions made in Table 1 are reflected in Table 2, Projected 1998/99 World Soybean Supply and Use.
However, data now appear in metric tons, the international unit of measure. Table 2 includes adjustments made to foreign supply and demand.