Analyzing Questionnaire Editing to Improve Data Quality: Results from the 1996 Agricultural Resource Management Study

The Agricultural Resource Management Study (ARMS) collects information on production practices, costs, revenues, and assets for farm and ranch operations in three data collection phases. Phase I of the study consists of screening for target commodities. This report focuses on manual editing and imputation in the Phase II and III. Concern about the resources being used on manual editing and imputation led us to study the amount of editing and its potential causes. We began by examining the frequency with which each item on the questionnaire was edited. Next, for each frequently edited question, we examined each questionnaire on which that question had been edited to determine the reason. Finally, we sought patterns in the causes for edits for each question, and made recommendations as to how editing for the question could be reduced. We conclude that, by analyzing how and where edits are made, we can identify areas where questionnaire design, editing procedures and enumerator training can be improved. We also conclude that, while statistically sound automated imputation methods were appropriate in some cases, manual editing and imputation seemed to be appropriate in others. By making these changes to questionnaires, editing procedures, and enumerator training, editing may be reduced and the quality of data collected increased. Finally, we conclude that similar methods could be employed in a broader range of surveys, particularly the Census of Agriculture.