Online Program

Addressing Non-response and Incomplete Coverage for Hard-to-Reach Populations in the 2007 Census of Agriculture

*Virginia Lee Harris, National Agricultural Statistics Service 
Bill Iwig, National Agricultural Statistics Service 
Matt Fetter, National Agricultural Statistics Service 

Keywords: coverage, weighting, dual-frame, nonresponse adjustment

The census of agriculture data covers all types and sizes of farms, including counts of farms operated by females and by individuals belonging to different racial and ethnic groups. Regardless of the effectiveness of outreach efforts, the census of agriculture requires non-response and coverage adjustment weighting of record-level data. In order to improve the quality of census data for these hard-to-reach populations, the National Agricultural Statistics Service utilized new statistical methodology for the 2007 Census of Agriculture that addressed their unique non-response and coverage characteristics. Non-response weighting groups for each state were created based on a number of different variables, including minority population indicators. Minority population records with significantly different response rates could be grouped separately for improved non-response weighting. State-level targets were established for specified census items, including farm counts for minority populations, which were used in calibration to provide coverage-adjusted record-level weights. The state-level targets were developed based on a dual-frame estimation procedure, consisting of the census mail list (CML) and an area frame component that covers the incompleteness of the CML. The area-frame sample was designed to provide additional samples in the hard-to-reach population areas. The paper will describe the non-response and coverage adjustment procedures used for the 2007 Census of Agriculture and some results.