On the logical structure of census and survey questionnaires

Abstract

Skip instructions in census and survey questionnaires result in different groups of persons being eligible for different questions—the logical structure of the questionnaire. Logical structure is important for analyzing census and survey data, for processing raw questionnaire data, and for questionnaire design and testing. We define a simple, general, computer-readable representation of a questionnaire that incorporates information on skip instructions. Algorithms for automatic calculation of questionnaire paths and sets of eligible persons are presented and applied to the Malawi 2008 census questionnaire and IPUMS International sample dataset. The R code provided may be used to reproduce our results and to apply the algorithms to other questionnnaires and datasets. We show that dividing questionnaires into sections is a fundamental tool for understanding and limiting logical complexity. We model questionnaires as programs intended for execution by an interviewer or respondent, rather than by a computer, and represent them by a software design diagram that presents information on questions, responses, and skip instructions in a form that is as readable as a traditional paper questionnaire, while simultaneously providing a precise specification that programmers may use to create an electronic version of the questionnaire.