Scope and Purpose
NASS is responsible for preparing State and National crop estimates. Estimates on this season’s final acreage and production for various commodities, an early estimate of Winter Wheat and Rye planted for the following year’s crop; and estimates of the supply of grain in storage and the movement of grain to elevators thru disappearance are derived from the December Agricultural Survey (DAS). This is done by obtaining the basic agricultural data from farmers and ranchers all over the Nation. Estimates derived from the December Ag Survey supply basic information needed by farmers to make decisions for both short term and long-range planning. Reports received from individual farmers and ranchers remain confidential and are used only in combination with other reports to arrive at State and National estimates.
The use of crop acreage, production, and stocks information is extensive and varied. It helps producers find the best market opportunities for their commodities. Often, recommendations and forecasts presented in agricultural magazines, news releases, etc. are based on data from Agricultural Surveys found in NASS reports. Uses of data by farm organizations, financial institutions, insurance companies, agribusinesses, State and National farm policy makers, and foreign buyers of agricultural products may range from maintaining a basic data series to preparing marketing campaigns and determining needs and rates on farm loans and insurance. Government agencies at various levels are important users of statistics. Federal farm programs require information on acreage, production potential, stocks, prices, and income. Agricultural statistics are used to plan and administer Federal and State programs in areas such as consumer protection, conservation, foreign trade, education, and recreation.
Survey Timeline
The December Ag Survey is conducted annually in month of December. Data collection starts two business days prior to the reference date which is December 1st and continues for approximately seventeen days. Estimates from the states are due two business days after the data is summarized. The Crop Production Annual Summary is released the 12th of January. This summary contains annual U.S. data for acreage, yield, and production by crop.
Sampling
Sampling
The target population for the December Crops/Stocks Survey is all agricultural establishments with row crops (corn, soybeans, cotton, sorghum, rice, peanuts, etc.) planted on the land operated. NASS uses a dual frame approach, consisting of list frame and area frame components, to provide complete coverage of this target population. The December Crops/Stocks Survey is conducted in every state except Alaska and Hawaii.
The list frame includes all known agricultural establishments. A profile, known as control data, of each establishment is maintained on the list frame to allow NASS to define list frame sampling populations for specific surveys and to employ efficient sampling designs. Only list frame records with positive planted acres of the desired commodities are included in the list frame population. A lower boundary, such as 50 acres of total cropland or 1000 bushels of grain storage capacity, is used for some states to establish the population. The list frame row crops population includes approximately 825,000 farms and ranches and covers approximately ZZ% of row crops acreage in the U.S.
For all quarters of the Crops/Stocks Survey, there are three main subpopulations, the General, Row Crops, and Small Grains which are used in all States. The General population includes all
farm operators, and the purpose is to give everyone a chance of selection each quarter and to have a consistent NOL (area) domain. The other populations are created for groups of crops that need to have questions asked for the same time period. For example, crops in the small grains population will have questions for harvested acreage and production in September, and crops in the row crops population will have questions for harvested acreage and production in December.
Any individual operation can be selected for none, one, two or all three samples depending on the list frame control data. For each, there also is a replication scheme that is used reduce respondent burden and to provide indications of change by comparing reports from the same farm operators. Each sampled record is randomly assigned to replicate one, two or three. Replicate 1 is used in June and for quarters (including December) with ratios back to June. Replicates 2 and 3, both used in December, are the stocks panel.
The area frame contains all land in the state and, as such, is complete. The land is stratified according to intensity of agriculture using satellite imagery. The land in each stratum is divided into segments of roughly one square mile. Segments are optimally allocated and sampled to effectively measure crops and livestock. The sampled segments are enumerated in June. All farms and ranches found operating in these segments are checked to see if they are included in the list frame row crops population. The farms and ranches that are not included in the list frame row crops population are sampled for the December Crops/Stocks Survey so that the target population is completely represented. The area frame component of the December Crops/Stocks Survey covers approximately ZZ% of the row crops acreage in the U.S.
The December Crops/Stocks Survey list frame sample is selected using a multivariate probability proportional to size (MPPS) sampling scheme. Each list frame record is assigned a measure of size based on the control data for multiple specified commodities. The MPPS design makes it very easy to target samples sizes for the commodities of interest. The desired number of samples for each commodity can be controlled with a minimum overall sample size. The MPPS design makes it easier to change samples to meet the needs of the crops program changes. The MPPS design is a more efficient design because operations will have a more optimal probability of selection based upon their individual commodities and size.
The U.S. list frame sample size for the December Crops/Stocks Survey in recent years is approximately 75,000. The December Crops/Stocks Survey area frame sample is selected from area frame farms and ranches not included on the list frame population using a stratified sample design based on data collected in the June Area Frame Survey. The area frame sample size is approximately 7,000. Each list frame and area frame sampling unit is assigned a sampling weight which is used to create the survey estimates.
The strata from a previously used stratified design are not being used for sampling, but will still be used for nonresponse adjustments. Stratification for the crops/stocks sample is based on the control data items for total cropland, on-farm grain storage capacity, rice acreage (rice estimating States), potato acreage (potato estimating States), and some rare or specialty crops (for each State)
After the list frame samples are drawn, the sample weights are calibrated so the sum of the weighted, targeted commodities in the sample of interest for each quarter equals the sum of the list frame data for the targeted commodities. For example, the sum of the weighted list frame data for Winter Wheat equals the sum of the population list frame data, and is the same for each of the four quarters. The sum of the weighted Barley data for the sample equals the sum of the population list frame data for June, September and March because Barley data is not collected in December.
Data Collection and Editing
The December Ag Survey has three primary data collection methods. Most surveys are initially mailed to all respondents. The follow-up on non-response is decided at the discretion of the states. They can choose not to follow-up, have it phoned by the NASDA office enumerators or be sent to the field for a NASDA representative to follow up with a personal visit or call from their home phone.
Survey Edit
As survey data are collected and captured, they are edited for consistency and reasonableness using automated systems. Reported data are typically first edited as a “batch” of data when first captured. The edit logic ensures the coding of administrative data follows the methodological rules associated with the survey design. Relationships between data items on the current survey are verified and in certain situations those items may be compared to data from earlier surveys to make sure certain relationships are logical. The edit will determine the status of each record to be either “dirty” or “clean”. Dirty records must be updated and reedited or certified by an analyst to be clean. If updates are needed, they are reedited interactively. Only clean records are eligible for analysis tools and summary.
Analysis Tools
Edited acreage and stocks data are processed through an interactive analysis tool which displays data for all reports by item. The tool provides scatter plots, tables, charts, and special tabulations that allow the analyst to compare an individual record to other similar records within their state. Outliers and unusual data relationships become evident and Field Office staff will review them to determine if they are correct. The tool also allows comparison to a farm’s previously reported data to detect large changes in the operation. Suspect data found to be in error are corrected, while data found to be correct are kept.
Nonsampling Errors
Nonsampling errors are present in any survey process. These errors include reporting, recording, editing, and imputation errors. Steps are taken to minimize the impact of these errors, such as questionnaire testing, comprehensive interviewer training, validation and verification of processing systems, detailed computer edits, and the analysis tool.
Summary Methodology
Nonresponse Adjustment
Response to the December Agricultural Survey is voluntary. Some producers refuse to participate in the survey. Others cannot be located during the data collection period and some submit incomplete reports. These nonrespondents must be accounted for if accurate estimates of crop acreage and stocks are to be made. For this survey, nonrespondents are accounted for by imputing data where there are missing values.
The imputation program imputes for missing survey data using reported survey data and control data from “similar” reports with complete data. Imputation occurs at the reporting unit level. The algorithm defines “imputation groups” as Agricultural Statistics Districts (ASD) within the strata assigned at the time of sampling. Open ended strata do not form homogeneous groups and are not eligible for machine imputation. If multiple follow ups do not produce a response, office statisticians are required to manually impute.
For all other strata, the algorithm will first impute cropland for the nonrespondent. When available, previously reported cropland is used. Otherwise, the ratio of current survey cropland to the list frame control data value for cropland is calculated from the respondents in an imputation group. This ratio is applied to the nonrespondents’ frame cropland acres to derive the imputed value for the current survey. Missing crop acres are imputed similarly using the respondents’ ratio of crop acres to cropland within each imputation group. Similar calculations are made to calculate yields to impute for missing production.
An imputation group must have five or more respondents before it is used. Those with insufficient response are collapsed across ASD and, if there is still insufficient response, collapsed with adjacent strata.
Estimators
Two kinds of estimators are used in the December Agricultural Survey: direct expansions and ratio estimators. Direct expansions are used to estimate totals such as planted acres, harvested acres, production, stocks on hand, and capacity. For the list frame, direct expansions are calculated by summing the reported or imputed commodity values multiplied by the original sample weights. For the nonoverlap portion of the area frame (NOL), the direct expansion is calculated by summing over all segments the total farm data for each tract operation multiplied by the original sample weights adjusted for the proportion of the operation=s total farm land found in the segment. The multiple frame direct expansion is the sum of the direct expansions from the list frame and the area frame NOL component.
The ratio estimator takes the form of a ratio of two direct expansions which are calculated by summing over the total sample (list + NOL) the reported commodity values multiplied by the original sample weights adjusted for usability status. The ratio estimator is used for all within and across-survey ratios (e.g. Harvested to Planted Ratio, Yield, and Current to Previous). This estimator relies exclusively on reported data. In the case of the yield estimate, both the numerator (production) and denominator (harvested acres) must be usable and positive. If either of these components is missing, the sampling unit is excluded from the yield estimate and the weights of the usable records are adjusted accordingly. For the survey to survey ratios, both the current and previous survey data must be usable (usable positive or usable zero) to be included in the ratio.
The reweighting of the record level sample weight is made within the strata. The adjustment is calculated by summing the weights for all sample records within the strata and then dividing by the sum of the weights from the usable records. This ratio is applied to the weights of the usable records. This adjustment assumes that the yields of the nonrespondents are similar to the yields of the respondents.
Estimation
When all samples are accounted for, all responses fully edited, and the analysis material is reviewed, each Field Office executes the summary for their state. When all Field Office have run summaries, Headquarters executes the national summary. Since all states conduct identical surveys, the samples can be pooled and national survey results computed. The summary results provide multiple point estimates and their standard errors for each data series being estimated. It also provides information used to assess the performance of the current survey and evaluate the quality of the survey estimates, such as strata level expansions, response rates, and percent of the expansion from usable reports.
The December Agricultural Survey supports the end of season estimates of planted acres, harvested acres, production, and yield for late season crops, such as corn for grain, soybeans, cotton, and sorghum. It also supports the quarterly on-farm grain stocks estimates.
Field Offices are responsible for performing a detailed review of their survey results. Any irregularities revealed by the summary must be investigated and, if necessary, resolved. Using the historical relationship of the survey estimates to the official estimate, Field Offices must interpret the survey results and submit a recommended estimate to Headquarters for all data series for which they are in the NASS program. The data are viewed in tabular and graphical form and a consensus estimate is established. Field Offices see their survey results only and do not have access to other states’ results. For some data series, information from other sources is .
For the national estimates, NASS assembles a panel of statisticians to serve as the Agricultural Statistics Board which reviews the national results and establishes the national estimates. Since larger sample sizes yield more precise results, NASS employs the “top-down” approach by determining the national estimates first and reconciling the state estimates to the national number for acreage, production, and stocks. The “Board” also enjoys an advantage in being able to examine results across states as well as the state recommendations. utilizes a number of sources of information for establishing official estimates, but the primary data come from surveys conducted throughout the year. Each survey conducted generates a number of individual estimators; in many cases several for each item of interest (i.e. planted acres of corn). These estimators are evaluated over time to determine accuracy and bias using various tools to compare each estimator with the final estimate for the item of interest. This analysis is conducted in table form (difference tables) and graphic form (time-series charts). Every 5 years NASS conducts the Census of Agriculture, which is an exhaustive data collection effort for all known farm operations across the U.S. The information gathered from the Census of Agriculture is used to establish “bench mark” levels by which the survey estimators can be compared and bias determined. Survey based estimators can also be impacted by “outliers” – individual reports that have “excessive influence” on the results due to either improper classification or extremely unusual data for a given operation (i.e. operation is not representative of other operations). NASS thoroughly reviews the survey data to identify these situations and consider their impact on the survey results when establishing the official estimates. External information (administrative data) is also utilized in the process of setting estimates. In order to be considered, these data must be deemed to be reliable and come from unbiased sources. The most common administrative data is the certified acreage data from USDAs Farm Service Agency; however data from many different sources are utilized. NASS also employs a balance sheet approach whenever possible to ensure that estimates are as accurate as possible. This approach typically is limited to National-level estimates.
Contact Information
|
Unit |
Phone |
|
Estimation |
Crops Branch |
202.720.2127 |
HQ_SD_CB@nass.usda.gov |
Data Collection |
Survey Administration Branch |
202.690.4847 |
HQ_CSD_SAB@nass.usda.gov |
Questionnaires and Editing |
Data Collection Branch |
202.720.6201 |
HQ_CSD_DCB@nass.usda.gov |
Sampling |
Sampling Branch |
202.720.3895 |
HQ_CSD_SB@nass.usda.gov |
Analysis and Estimators |
Statistical Methods Branch |
202.720.4008 |
HQ_SD_SMB@nass.usda.gov |
Dissemination and Webmaster |
Marketing and Information Services |
202.720.1707 |
HQ_DAPP_MISO@nass.usda.gov |
Last modified: 03/11/11
