The Effects of Prelisted Items in Business Survey Questionnaire Tables (303622)

Keywords: business, establishment survey, questionnaire design, respondent burden, prelisted items, data quality

Surveys conducted by the National Agricultural Statistics Service (NASS) ask respondents to report the type, production and sales of commodities they harvest by writing them into large response tables, one commodity per line. To ease respondent burden, some questionnaires prelist the names of certain commodities into the response tables, aiding recall and requiring the respondent to write less. Prelisted items can also help with data quality, as they reduce the chance of error during data entry. However, this can create space problems, as not all items can be prelisted, and an adequate number of blank lines need to be available for nonprelisted commodities. We investigate this issue by comparing three versions of these tables which are used in the Census of Agriculture and Organic Production Survey. One version prelisted some items in the response tables, with the remaining items printed in a listing at the bottom of the questionnaire pages, and in a separate instruction booklet for the survey. A second version had response tables with no commodities prelisted within the table, and commodities printed at the bottom of the questionnaire pages, and in the separate instruction booklet. A third version only had commodities printed in the separate instruction booklet, with no items prelisted on the questionnaire. Findings from 105 cognitive interviews will be discussed, including how respondents report their commodities on these different forms, the errors introduced by each, and impacts on the data quality.