The 2012 Census of Agriculture: A Capture-Recapture Analysis

摘要

The Census of Agriculture is conducted every 5 years, in years ending in 2 and 7. The Census list frame is incomplete, resulting in undercoverage. Not all operations on the list frame respond, and, based on the response, some misclassification occurs. In 2012, a capture-recapture analysis was conducted to adjust for undercoverage, nonresponse, and misclassification. This was the first time capture-recapture methods were used to produce official statistics for an establishment survey. The number of records on the Census Mailing List that were classified as farms was 1,382,099, and the published estimate of the number of farms was 2,109,303, a 34.5% adjustment. The adjustment was greatest for farms with low production levels and for specialty farms, both of which are difficult to identify and add to the list. The methods used are described. Challenges that arose in the implementation process are discussed. Areas for enhancement being targeted for the 2017 Census of Agriculture are highlighted. Supplementary materials accompanying this paper appear online.