Don't Knows (DKs)

“Don't Know” responses (DKs) occur when respondents report that they do not know or are uncertain about the answer to a survey question, whether it be about their behavior, attitudes, beliefs, perceptions, or a factual question. Don't Knows are often considered to be nonsubstantive responses and a form of item non-response. A high level of DKs is sometimes used as an indicator of poor data quality. DKs also create problems for statistical analysis. Researchers often treat DKs as missing data and often exclude cases in which respondents said “Don't know” from their analyses.

Respondents may say they don't know in response to a survey question for a number of reasons. They may genuinely not have an opinion or belief to report. Alternatively, satisficing theory suggests that respondents may report that they don't know because it is a strategy for providing acceptable (satisfactory) answers to survey questions without going through the mental processes necessary to provide a careful response. Finally, some respondents may say they don't know to avoid answering an uncomfortable, embarrassing, or politically charged question.

Don't Know responses are one example of a larger category of no-opinion responses, which reflect that respondents do not have a judgment to report. “Don't know” is sometimes included with the list of substantive response options that are offered to respondents in survey questions that use a closed-ended response format—for example, or don't you know? When presented in a survey question with a list of response options that does not include an explicit don't know option, respondents may volunteer “don't know” responses to interviewers in telephone and face-to-face surveys and may write them in on self-administered surveys.

Thus, respondents can report that they don't know the answer to a survey question regardless of whether such an option is explicitly offered to them. However, explicitly including such an option to respondents dramatically affects the proportion of respondents who say they don't know. The inclusion of an explicit [Page 209]Don't Know response option has been found to substantially increase (from 5 to as much as 30 percentage points) the proportion of respondents who report that they don't know, particularly for questions about issues with which respondents may not be familiar.

Because including an explicit Don't Know option can have a dramatic impact on responses, the decision about whether to explicitly offer such a response option is a very important one for researchers when creating a survey instrument. Two perspectives—nonattitude and satisficing—provide competing theoretical arguments about this decision.

The nonattitude perspective suggests that respondents who genuinely do not know an answer nevertheless may choose a substantive response option when no other option is available. The nonattitude perspective comes from Philip Converse's observation that survey interviews may exert implicit pressure on respondents to appear to have an opinion on a wide range of topics. When respondents are faced with a question to which they genuinely do not know the answer, many may be uncomfortable admitting that they know little about the topic or that they do not know the answer, and this may be particularly true when multiple questions for which they are uninformed are included in a survey interview. Respondents who do not have attitudes on an issue may respond to a question about the issue essentially by randomly selecting responses from among the choices offered. Including an explicit Don't Know response option would provide these respondents with a way to accurately report that they do not know how to answer the question.

...