USDA's Economic Research Service and National Agricultural Statistics Service have surveyed farmers to obtain information about farm financial conditions and production practices since 1975. Earlier surveys were

• Farm Costs and Returns Survey (FCRS), which collected enterprise-level financial data on cost of production for selected crop and livestock commodities, and
  
  
• Cropping Practices Survey (CPS), which collected enterprise-level chemical use, nutrient, pest, and soil management practices on field crops.

ARMS is an annual, State-by-State survey of farms and agricultural commodities conducted to obtain information about:

•   The status of farmers' finances
•   Production practices for specific commodities
•   Use of natural, physical, and financial resources
•   Household economic well-being.

See ARMS content.

ARMS phases

The ARMS survey is conducted in three phases over the course of the survey year, which runs from June through April.

In the first phase, conducted during the summer of the reference year, farmers selected for inclusion in the survey sample are screened to verify their operating status and to determine whether they are producing commodities targeted that year for data collection.

In the second phase, conducted in the fall and winter of the reference year, farmers from phase I are interviewed to collect information on their production practices and chemical use. Phase II mirrors the former Cropping Practices Survey conducted prior to 1996.

In the third phase, conducted in the spring of the year following the reference year, a nationally representative sample of farmers is interviewed to obtain information on their costs and returns during the reference year. Farmers that reported production practices for specific commodities in phase II are also contacted to obtain information on their costs and returns, including data needed to estimate the costs of production associated with their production practices. Phase III replaces the former Farm Costs and Returns Survey conducted prior to 1996.

Major crop and livestock enterprises (corn, soybeans, wheat, cotton, dairy, pork, and beef) were covered about every 5-8 years, while additional commodities (other feed grains, sugar crops, rice, peanuts, tobacco, and poultry) are covered on a less frequent interval. (See ARMS coverage.)

sample design

The ARMS survey is technically described as a multiframe, stratified, probability-weighted sampling design. What do these three characteristics of the sample design mean?

Multiframe sampling—Sampling frames provide the universe from which observations are drawn in a statistical sample design. Two sampling frames are used by NASS to select farms for the survey. The primary sample is derived from the NASS List Frame. NASS maintains a list of farm operations that exhibit certain characteristics. The lists are constructed and maintained from many different sources, including the Census of Agriculture, other NASS surveys, administrative data from other government agencies including the IRS. Because some information is already known about these farms, the list can be stratified (sorted) according to farm types and size classes. The list frame is an efficient sampling source because it contains most of the large farms (those with the highest production and sales), as well as those expected to have the target commodities.

The second sampling frame for ARMS is the NASS Area Frame. This is used only to capture farms not on the List Frame, and consists of randomly selected agricultural land segments that represent all land in farms. Each year, NASS conducts a spring survey selected from the area frame to estimate crop acreage and land use. This survey identifies all land uses within the segment, and it can be used to stratify target crops for follow-on surveys. A sample for ARMS is then selected from the spring survey results. Only those farms not on the list frame (nonoverlap) are retained for sampling.

Of 18,608 sample contacts in 2000, 16,238 came from the list frame and 2,370 from the area frame. However, only 10,309 respondents provided complete, usable data for the final ARMS dataset.

Stratified sampling—Strata are divisions within the sample frames that have particular characteristics. Farms in different strata are sampled with a different probability of selection. For example, large farms in the list frame have a higher probability of selection than small farms because many small farms are not included in the list. Within a stratum, the weight (expansion factor) is based on the probability of its selection. In the area frame, land use or crop type can be used to stratify target crops for follow-on surveys.

Probability-weighted sampling—Because of the complexity of the design, each observation has a different weight, or expansion factor, to reflect its probability of selection, and therefore what part of the universe sampled it should represent. Appropriate sample weights (expansion factors) are provided to prepare population estimates from the survey results. Population estimates are constructed by weighting each sample with the appropriate expansion factor. A sophisticated statistical estimation procedure, known as the delete-a-group jackknife procedure, has been determined to be an efficient and "nearly unbiased" method for constructing variance estimators. Fifteen replicate weights are constructed to apply this procedure.

More information about the sampling procedures can be found in the NASS survey administration manuals or by contacting NASS directly. The ERS staff paper "Variance Estimation with USDA's Farm Costs and Returns Surveys and Agricultural Resource Management Study Surveys" and the NASS research report "Using the Delete-A group Jackknife Variance Estimator in NASS Surveys" discuss the analytic methods for constructing variance estimators from the ARMS data.

data collection

Trained enumerators conduct personal interviews with farm operators to collect data about their farm operations for the ARMS survey. An interviewer's manual outlines detailed enumeration procedures for each phase of the survey. These documents provide specific directions on how the interview is to be conducted and insight into how to interpret each question. Examples of appropriate responses are included.

data quality control

NASS provides enumerator training just prior to the survey through a series of State schools. NASS and ERS provide training materials to the State statisticians who conduct the schools.

After questionnaires are completed by the enumerators, each questionnaire is reviewed by supervisory enumerators for completeness, inconsistent responses, or errors, and then transferred to a NASS State office. Supervisory statisticians also review each questionnaire before it is keyed into an electronic format. A computerized edit routine is then used to identify other potential errors or inconsistencies, checking that responses fall within expected ranges. When responses are anomalous, State supervisory statisticians correct or verify the responses.

Additional quality control procedures are used as the data are summarized. Summary statistics on all questions are provided to each State office for final review and clearance. The information includes response distributions, data ranges, variance measurements, and identification of possible outliers.

A survey administration manual provides specific details about survey administration and data processing procedures. (See Detailed Documentation).

for more information, contact: Mitch Morehart
web administration: webadmin@ers.usda.gov
page updated: November 21, 2003