The proximity effect: The role of inter-item distance on reverse-item bias

This paper introduces the proximity effect model. The proximity effect model explains the correlation between two items as a function of (1) items' conceptual relationship (i.e., nonreversed items measuring the same construct; reversed items measuring the same construct; or items measuring unrelated constructs), and (2) items' proximity in a questionnaire. The proximity effect model draws from the belief-sampling literature. We first conduct cognitive interviews to better understand the ways that respondents process items in a questionnaire depending on items' conceptual relationships and proximity. Next, we quantitatively test the resultant model using primary data (N=3114). Finally, we use simulated correlation matrices to assess the impact of the proximity effect on factor structure and reliability. The results indicate that the (positive) correlation between a nonreversed item pair decreases with increasing inter-item distance. In contrast, the (negative) correlation between reversed item pairs decreases (i.e., becomes stronger) with increasing inter-item distance. This is partly due to the fact that respondents tend to minimize retrieval of additional information when answering nearby nonreversed items, and they tend to maximize retrieval of new and different information when answering nearby reversed items. The proximity effect model leads to a set of key recommendations pertaining to questionnaire construction (i.e., the use of reversed items dispersed throughout the questionnaire) and data analysis (i.e., the use of a confirmatory factor analysis model specification including a response style factor).