Don't Know Responses in Survey Research
Open Access
- Author:
- Young, Rebekah Lynn
- Graduate Program:
- Sociology
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- February 10, 2012
- Committee Members:
- David R Johnson, Dissertation Advisor/Co-Advisor
David R Johnson, Committee Chair/Co-Chair
Paul Amato, Committee Member
Alan Booth, Committee Member
Melissa Hardy, Committee Member
Eric Plutzer, Committee Member - Keywords:
- don't know responses
missing data
item nonresponse
survey methods
questionnaire design
surveys - Abstract:
- Survey respondents are occasionally unable to generate the type of response researchers hope to record. Respondents may, instead, offer non-substantive answers such as saying they are unsure, cannot recall, or don’t know (DK). The first article of this dissertation asked which respondents were most likely to offer DK replies and under what circumstances. DK responses were frequently recorded for sensitive questions, to questions that occurred in the middle of a survey, and to questions which asked about politics, government, finances, and economics. Although women offered DK responses more often than men did, men were more likely than women to say DK to questions asking about family, friends, health, and health care. The second article of this dissertation explored the meaning of DK responses by examining response patterns in longitudinal data. An important finding was that forty percent of respondents offered DK replies to the same questions asked nine to ten years apart. This is strong inferential evidence that while some DKs may result from survey satisficing or passively refusing to answer, a sizable proportion also indicate genuine lack of knowledge. If a DK reply is, indeed, the best answer to a question, it seems prudent and perhaps necessary to allow these responses in data collection. At the same time, DK responses can pose a serious challenge to data analysts, a problem addressed in the third article. Using a simulation study, I explored six common methods for analyzing DK responses when different meanings were assumed to underlie them. In general, if researchers are uninterested in analyzing DK responses as a categorically meaningful response, and instead prefer to treat DK responses as missing values, modern missing data methods treated these responses with minimal bias. The exception to this finding occurred when DK responses represented “passive refusals,” where respondent’s intentionally withheld true answers from the researcher. Under this circumstance, no method for analysis emerged without limitations. An important theme that emerged from each article is that DK responses are much more nuanced, meaningful, and context-based than perhaps we recognized in the past.