Identifying Sources of Survey Errors: The 2007 Classfication Error Survey for the United States Census of Agriculture

摘要

For the 2002 Census of Agriculture (COA), the number of operations misclassified (either as farms or non-farms) in the COA was estimated. Operations in NASSs June Agricultural Survey (JAS) and the COA were matched and their answers compared. Misclassification estimates were based on the assumption that the JAS was truth. The misclassification rate was small but it was clear that the JAS assumption was not always justified. The 2007 Classification Error Survey focused on understanding why operations reported differently in the JAS and COA, rather than estimating misclassification. Operations in the 2007 COA and JAS were matched but neither report was assumed as truth. Instead, operations were reinterviewed and respondents asked to resolve discrepancies. More errors were found in the JAS than in the COA, and were related to respondents, enumerators and NASS procedures. A multipart solution will be required to address them.