How was the Census Conducted

< 1 2 3 4 >

Why are there two columns of 1997 numbers in the 2002 publication?

What statistical methodology is used to complete the Census? What is "Coverage Adjustment"?

How do these adjustments affect data at the County Level?

What is the accuracy of the results?

How do we account for farms that do not respond to the Census?

Why did the U.S. number of farms with 2,000 acres or more decrease in 1997 after adjusting for undercoverage?

Are the unadjusted numbers available like they were in the past?

Does NASS change its "Official Estimate" based on the Census results? Does NASS adjust the Census counts to match their official estimates?

Why weren't Alaska and Hawaii adjusted for coverage like all other states?

I've been tracking Census data for years. How can I interpret the coverage adjusted numbers?

We've used Agriculture and Economic Census data together to profile America. Has anything changed for the Economic Censuses and how do they differ?

My local government uses your Census data to make decisions in my local jurisdiction. How do changes to the way you conduct the Census affect me?

Where do you place crop acres, livestock and other data when a farm operates in more than one county? In addition, for Contractee Operations, are the data in the county of the contractor of contractee?

WHY ARE THERE TWO COLUMNS OF 1997 NUMBERS IN THE 2002 PUBLICATION?

Past censuses showed coverage measures only for a limited set of data items and these were not in the main tables. In order to provide the most complete picture of agriculture, all data in the 2002 Census of Agriculture reflect coverage adjusted totals. In order to have a bridge for historical comparisons, the 1997 numbers have also been adjusted and shown in these tables.

WHAT STATISTICAL METHODOLOGY IS USED TO COMPLETE THE CENSUS? WHAT IS COVERAGE ADJUSTMENT?

NASS collects census data from a list of all known potential agriculture operators. This list was assembled from previous census records, State and Federal agencies, trade associations and similar organizations that could be identified as associated with agriculture. However, the list is never complete. Producers go in and out of business every day and many small operations are simply never identified. In order to measure the incompleteness of the Census Mail List, i.e., determine how many producers are not on the list, NASS conducts an Area Frame Survey each year. This survey is based on sample tracts of land randomly selected throughout the conterminous U.S. Every agricultural operator discovered within the selected tracts is interviewed and all land in each tract is accounted for. This process ensures that every agricultural operation in the U.S. is eligible for selection. The results of this survey are then compared with the Census Mail List to determine the level of completeness of the list and characteristics of operations and operators that were missed. Census data can then be weighted based on the characteristics of operations who did not appear on the Census Mail List. All data included in census totals are reported by census respondents. Some records have a weight greater than 1 to account for similar operations missing from the list.

Classification and measurement errors were handled through the use of commodity targets in the coverage adjustment process. Attempts were made in the analysis and processing of data to identify and make corrections for duplication and misclassification. Operation weights by type and size were adjusted by the level of commodity incompleteness known from reliable sources. This could result in slight adjustments of numbers either up or down.

Return to Top

HOW DO THESE ADJUSTMENTS AFFECT DATA AT THE COUNTY LEVEL?

The coverage adjustments are generated and therefore most accurate at the state level. The state level data adjustments for undercoverage are then allocated among census records having reported data with appropriate characteristics in the counties.

WHAT IS THE ACCURACY OF THE RESULTS?

The data to be released in 2002 census publications are the result of a consistent statistical process designed to provide the truest picture of US agriculture. As such, those data present a more complete and representative picture of the agriculture and farming operations in each State and county than has been possible in past census publications. The more rare the commodity or characteristic included in the summary tables, the more chance there will be for anomalies resulting from higher weights on some records. Adjustments for undercoverage account for approximately 2.4 percent of land in farms and 2.9 percent of product market value.

Return to Top

HOW DO WE ACCOUNT FOR FARMS THAT DO NOT RESPOND TO THE CENSUS?

NASS maintains a list of agricultural producers which is used to generate the Census Mail List. Everyone on this list was mailed a 2002 Census of Agriculture report form. Although multiple attempts were made to obtain a response from each operator, some producers did not return their form. Accounting for farms that did not file a return is done through a nonresponse adjustment. A weight is applied to the good responses from similar operations to compensate for this nonresponse. The data in the 2002 Census of Agriculture are reported data adjusted for nonresponse. This adjustment has also been made throughout the history of the census.

The response rate has steadily increased over the course of the past several censuses. The overall response rate for the 2002 Census of Agriculture was 88 percent.

WHY DID THE U.S. NUMBER OF FARMS WITH 2,000 ACRES OR MORE DECREASE IN 1997 AFTER ADJUSTING FOR UNDERCOVERAGE?

This is an unusual result of the change in methodology. Nonresponse weighting in 1997 produced one set of weights for census records which were rounded to integers. In 2002, the 1997 data were summarized using the new methodology to be able to compare data with 2002 census data. The new weights rounded differently and resulted in a slightly smaller total (-0.25%) for farms with 2,000 or more acres.

Return to Top

ARE THE UNADJUSTED NUMBERS AVAILABLE LIKE THEY WERE IN THE PAST?

Data adjusted only for nonresponse for 2002 are not readily available and NASS does not plan to publish them. Census totals based solely on list records are subject to variation in list coverage between censuses and therefore may not be truly comparable. In other words, even though the same census process was used to count known operations in 1997 and 2002, the number of known operations included on the census list was different in 1997 than in 2002. Better indications of overall changes are reflected in the comparable complete data totals for 1997 and 2002. The "old" way loses comparability due to changes in list coverage. We would recommend against using these numbers. If data users have a specific request for unadjusted numbers this request should be submitted via normal data lab procedures. Each request will be reviewed and must be approved by the Associate Administrator.

DOES NASS CHANGE ITS "OFFICIAL ESTIMATES" BASED ON THE CENSUS RESULTS? DOES NASS ADJUST THE CENSUS COUNTS TO MATCH THEIR OFFICIAL ESTIMATES?

NASS continues to review census results to evaluate and revise official estimates. Revisions to major estimates are published in a series of historical bulletins. During 2004, twenty one historical bulletins will be published. Similarly, NASS has also always looked to other sources of reliable data (both internal and external) as a check during data analysis and processing of the census. This includes a review of NASS official estimates and estimates from other sources. For example, livestock slaughter provides an indication of earlier livestock inventory. . Census data, survey data, market data, and administrative data all serve to benchmark the best numbers for agriculture.

Return to Top

WHY WEREN'T ALASKA AND HAWAII ADJUSTED FOR COVERAGE LIKE ALL OTHER STATES?

No data from area frame samples were available in these States so there was no statistical basis to make a coverage adjustment.

I'VE BEEN TRACKING CENSUS DATA FOR YEARS. HOW CAN I INTERPRET THE COVERAGE ADJUSTED NUMBERS?

The 1997 census provides a bridge between the list only census approach (prior method not adjusted for coverage) and total census results (new method adjusted for coverage). Number of farms changed the most because of small farms missing from the list. Relatively small changes to land, production, and economic data resulted from coverage adjustments. Changes are reflected from 1997 to 2002 on a coverage adjusted basis. (See Appendix C for relative change resulting from nonresponse and coverage adjustments.)

Return to Top

WE'VE USED AGRICULTURE AND ECONOMIC CENSUS DATA TOGETHER TO PROFILE AMERICA. HAS ANYTHING CHANGED FOR THE ECONOMIC CENSUSES AND HOW DO THEY DIFFER?

The availability of more complete lists of businesses for other economic censuses gives those sectors an advantage over agriculture. The new 1997 and 2002 Census of Agriculture totals are now more comparable with other businesses when evaluating characteristics in different economic sectors.

MY LOCAL GOVERNMENT USES YOUR CENSUS DATA TO MAKE DECISIONS IN MY LOCAL JURISDICTION. HOW DO CHANGES TO THE WAY YOU CONDUCT THE CENSUS AFFECT ME?

Census results (as well as all NASS statistics) are publicly available to everyone. NASS considers ways in which data may be used when deciding what statistics to publish and how to generate them. In the past, coverage adjusted numbers were not available for all items. The new census numbers may not be directly comparable to data that users may have relied on in the past, but the new 2002 census provides the best data for decision makers. Impact resulting from the change to more complete representation for agriculture may be evaluated using the 1997 bridge year.

Return to Top

WHERE DO YOU PLACE CROP ACRES, LIVESTOCK AND OTHER DATA WHEN A FARM OPERATES IN MORE THAN ONE COUNTY? IN ADDITION, FOR CONTRACTEE OPERATIONS, ARE THE DATA IN THE COUNTY OF THE CONTRACTOR OR CONTRACTEE?

The county in which the largest value of agricultural products were raised or produced is the principal county in which the reported farm or ranch data will be published.. . Generally, when more than one county or State is reported, all data are allocated to the principal county. However, reports received with land in more than one county or State were split into two or more separate reports if the data published collectively are extreme principal county or state level data. For contractee operations, all data are published in the county which the commodity was produced.

Return to Top

< 1 2 3 4 >

Last modified: 12/09/08