Online Program

Effects of Changes to the 2017 Census of Agriculture Questionnaire on Capture-Recapture Methodology (307972)

Keywords: capture-recapture, imputation, list frame, area frame, logistic regression, simulations

The U.S. Department of Agriculture’s (USDA) National Agricultural Statistics Service (NASS) conducts a Census of Agriculture (COA) every five years, in years ending in 2 and 7. For the 2017 COA, capture-recapture methods were used to adjust the COA for undercoverage, nonresponse, and misclassification of farms/non-farms. NASS's June Area Survey (JAS), which is conducted annually in June, was used as the independent survey in the capture-recapture approach. For capture-recapture, a matched dataset consisting of all matches of a COA record to a JAS record is formed. This dataset is the foundation for modeling the probabilities of coverage, response, and correct classification of farms/non-farms for the COA. These probabilities are estimated through a series of weighted logistic regression models. Demographic characteristics are essential covariates considered in the models’ variable selection process. In 2017, NASS redesigned the demographics section of the COA questionnaire to allow up to four principal operators per farm operation. The JAS questionnaire collects information on only one principal operator. Multivariate imputation was used to address this missing-data problem. The effects of the redesigned COA demographics section and the JAS imputation effort on the capture-recapture methodology are assessed through a series of simulations and data analyses. This paper presents the results of these analyses.