Frequently a question will elicit responses that can only be classified as “don’t know” or “no answers” responses. Such responses are even encountered when using multiple choice and dichotomous questions. Such responses must be properly coded, and researchers should be prepared to include them in the tabulation procedure.
Some projects may contain so few “don’t know” and “no answer” responses that they are included only to show complete statistical tables. In other studies, it might not be possible to state what the most prevalent or common answer was if a large number of ‘don’t know’ was received.
Example: In a study concerned with which type of retail outlet was most important in the purchase of a microwave oven, researchers asked the question: At which type of outlet did you purchase your microwave oven? (Check one of the following store types).
The results are shown below, where the percentage of “don’t know” is so large that researchers were unable to draw any sound conclusions about which type of outlet is most important. The high percentage of “don’t know” may have resulted from question wording. Many respondents may not have known what was meant by the different outlet terms and, therefore, may have replied that they did not know the type of outlet at which they purchased their microwave ovens.
Ways of Handling “Don’t Know” Answers: There are three ways of handling the “don’t know” problem – none of which represents a fully satisfactory solution.
Percent of respondents Purchasing Microwave Ovens at Types of Outlets:
Type of Outlet Percent of respondents
Appliance store 26
Furniture store 16
Mail order house 6
Department store 6
Discount store 3
All others 15
Don’t know 28
(Base = 231 respondents)
1. Show the “don’t knows” as a separate category. This is the best procedure for it does not mislead anyone about what happened.
2. Estimate answers from other data contained in the questionnaire. Occasionally the “don’t know” answer can be inferred by studying other information contained in the questionnaires. For example, family income might be estimated by referring to the number of individuals in the family who are working and the occupations of each.
3. Distribute the “don’t knows” proportionately among the other categories. This procedure assumes that the remainder of the sample (those who gave an answer other than “don’t know”) will be representative of the universe. This assumption may not be correct, and therefore if the extent of the “don’t knows” is not shown the reader may be misled.
Dichotomous or Multiple Choice Questions That allow only one answer:
Each of these questions types has predetermined response categories – usually pre-coded that have been established in accord with the overall objectives of the study. As a consequence, the tabulation of responses to such questions is predetermined and consists merely of counting the number of responses falling into each category. The two most common approaches to tabulating the responses to multiple-choice questions: (A) showing both quantities and percentages for each response category, and (B) showing percentages for each responses category, but only total quantity. The two approaches can be applied to both scale and dichotomous questions as well.
Open Questions that allow only one answer:
As mentioned above, categories must be established for responses to open questions and individual responses to such questions must be read and coded according to these pre-established categories. If an open question elicits only a single answer, tabulation procedures will be similar to those described in the previous paragraph. An example of such a question is, what one characteristics is most important in deciding which brand of after shave lotion you purchase? (If the question asked, what characteristics are important in deciding which brand of after shave lotion you purchase? More than one response would be possible.