Results from the 2002 Classification Error Study

The 1997 Coverage Evaluation program was designed to measure four components of error in the census farm counts: 1) undercount due to farms not-on-the-mail list (NML); 2) overcount due to farms duplicated or enumerated more than once; 3) undercount due to farms incorrectly classified as nonfarms; and, 4) overcount due to nonfarms incorrectly classified as farms. The NML component is by far the largest contributor to coverage error. It utilizes the NASS area frame to assess the number of farms not on the list. The other three components account for misclassification error. In 1997, the Classification Error Study (CES) was conducted to evaluate the feasibility of using the NASS area frame to measure misclassification error and replace the Classification Error Survey (re-interview approach) in place at the time. The results of the 1997 CES, conducted on the 11 Western states, compared favorably to the re-interview approach estimates. Recommendations were to replace the classification error survey (re-interview approach) with the Classification Error Study (area frame approach) for the 2002 Census of Agriculture. This report outlines the results of the 2002 Classification Error Study (area frame based approach). The 2002 CES was implemented operationally for the 48 conterminous states. Its main objective was to determine the relative size and likelihood of classification errors to warrant future CES studies.