Effect of branching middle responses in dichotomous scales on web surveys

Abstract

Previous research shows that in telephone surveys, 11–49% of respondents select a middle alternative when it is offered on dichotomous bipolar questions, although they do not volunteer a middle option when it is not mentioned. Furthermore, offering a middle option leads to different systematic response effects, including social desirability bias and satisficing effects, which are related to respondent characteristics. Research has shown that in telephone surveys, a question form with a branching middle response has a lower response validity compared to that of a non-branching middle response. However, compared to the non-branching form, the branching form could potentially motivate respondents to spend more time and effort in their responses in the absence of an interviewer. Therefore, validity and reliability differences between branching-question form responses and non-branching-question form responses in general population web surveys have become an area of interest for research. Using a randomized experiment, this study tests the impact of a branching middle alternative in a general population survey on the validity and reliability of responses. The branching middle alternative did not change the validity and reliability of responses, and compared to the alternative, it reduced satisficing behavior based on the proxies.