Assessment of an Imputation Process Used in the 2017 Census of Agriculture

• Activity Number: 416 - Nonresponse Errors and Fixes
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Tara Murphy* and Habtamu Benecha and Denise A. Abreu and Darcy Miller
• Companies: USDA National Agricultural Statistics Service and NASS/USDA and USDA National Agricultural Statistics Service and National Agricultural Statistics Service
• Keywords: capture-recapture; imputation; list frame; area frame; logistic regression

Abstract

The National Agricultural Statistics Service (NASS) conducts a Census of Agriculture (COA) every five years using a list frame. The 2017 COA used capture-recapture methods to adjust the COA for undercoverage, nonresponse and misclassification of farms/non-farms. NASS's June Area Survey (JAS) was used as the independent survey in the capture-recapture approach. The JAS uses an area frame and the data are collected via in-person interviews. 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 crucial covariates considered in the models’ variable selection. In 2017, NASS redesigned the demographics section of the COA questionnaire to allow up to four principal operators per farm. The JAS questionnaire gathers information on only one principal operator. Multivariate imputation was used to address this missing-data problem. This paper evaluates the effectiveness of the imputation.