Effects of Question Wording and Context: An Experiment with Religious Indicators

Abstract

When a question on change in interest in religion follows questions on religious beliefs and attitudes, more respondents report either an increase or a decrease in interest (instead of ‘same’) than when the antecedent questions concern religious identification and participation. The association between change in interest and change in attending religious services appears stronger when the two questions have similar response categories than when somewhat different categories are used. These conclusions emerge from multiway contingency analysis of data from a survey including a factorial experiment in which context and wording of questions are varied. The models of association for a 3 × 3 table used in the analysis extend earlier work on multiplicative models for contingency tables.