Publication Type: Conference Paper/Unpublished Manuscript
Review Method: Peer Reviewed
Abstract: Web surveys often include an explicit “Decline to Answer” response option in closed-ended questions. This explicit “Decline to Answer” option can take a variety of forms such as “Decline to Answer,” “Refuse to Answer,’ “Refuse/Don’t Know,” and “Prefer not to Answer.” Hypotheses about the impact of explicit “Decline to Answer” options relate both to item response rates and mode effects in mixed mode surveys. At the item level, one hypothesis is that an explicit “Decline to Answer” option will decrease item nonresponse by eliminating accidental skips; another is that it will increase item nonresponse by making prominent the option to refuse. An important mode effect hypothesis is that the explicit “Decline to Answer” can be a source of mode effects for surveys employing multiple data collection modes since neither paper-based SAQs nor CATI questionnaires typically include explicit “Decline to Answer” options as a response category.

The National Opinion Research Center (NORC) programmed the 2003 Survey of Doctorate Recipients (SDR) Web questionnaire without offering a “Decline to Answer” option as a response category for closed-ended questions. This setup allowed the user to continue past a question without answering by clicking the “Continue” button without selecting an option category. We compare the item non-response on selected items in this questionnaire with item non-response from similar questions on other NORC web-based studies where an explicit “Decline to Answer” option was provided. We also compare item non-response to selected items on the paper-based SAQ version of the SDR survey. We evaluate the impact of the explicit “Decline to Answer” option on overall item non-response in the survey and we discuss the value of additional programming that may be required to implement a Web instrument without an explicit “Decline to Answer” option.