One of the more common mistakes survey writers make is including a “don’t know” or “no opinion” response. Doing so provides an easy way for respondents to avoid engaging with the subject of the question.

 

Some researches have provided “don’t know” response options because they hoped it would help get rid of inaccuracies. Research indicates, however, that including “don’t know” options actually produces more problems than it solves.

 

How “don’t know” is hurting your data

 

In the past, some researchers thought that if they did not include “don’t know” options, they were forcing respondents to make choices about something they had no opinion about. The conventional wisdom was that this led to low reliability between responses to survey questions over time.

Additional research supporting this notion found that respondents will often provide opinions about fictitious issues or events.[1]

 

So it’s not hard to see how including “don’t know” options in surveys was an appealing choice, it provides those without attitudes or knowledge an opportunity to express that directly.

 

However, subsequent research has indicated that offering a “don’t know” option actually creates a worse problem than it solves by allowing respondents to disengage from the survey. Given the option, some people tend to just select “don’t know” to avoid having to think about the answer.[2]

 

Why you should ditch “don’t know”

 

First, while some respondents may believe they don’t know the answer, research indicates that when they’re asked to guess they’re better at it than you or they would think. In fact, they’re able to provide correct answers to factual questions more frequently than if they were just choosing randomly.[3] When people really don’t know, it’s okay to ask them to provide their best guess. The best guess will still be more beneficial than a “don’t know” response.

 

Second, providing a DK response option is likely to produce less valid data. For example, researchers have found that voting behavior is predicted more accurately by other political variables when DK responses are discouraged.[4] As a general rule, it’s best to avoid using “don’t know” or “no opinion” response options.

 

    [2] Oppenheim, 1992
    [3] Mondak & Davis, 2001
    [4] Visser et al, 2002