On Tuesday, July 9, I discussed the “10 Best Practices to Utilize When Designing an Effective Survey.” To complement this post, I will outline a few simple practices to follow when designing survey questions. These practices are overall meant to increase response rates and ensure validity is maintained.
1. Short as Possible: Questions should be as short as possible to ensure clarity and keep the respondent engaged. If you have to take a breath while reading, then the question is too long.
2. Shared, Simple Vocabulary: It is important to keep questions simple enough for respondents to understand the terms. Especially watch out for density.
3. Unbiased Language/Premises: When constructing questions, be aware of potential biases in the question wording. Different responses are elicited from terms like welfare versus assistance to poor.
4. Unambiguous Answers: Answer choices should be extremely clear and easy for the respondent to understand.
5. Confined to One Issue: Questions should be confined to a single issue to ensure accurate responses. Avoid double-barreled questions.
6. Exhaustive/Exclusive Categories: Respondents should fit under only one category.
7. Positive Construction: Questions should be constructed in a positive tone instead of a negative tone. Double negatives should be completely eliminated, as they will often cause respondent confusion.
8. Rotate Items in a Series: This will increase the survey’s validity as you are minimizing the effect previous questions have on the next question.
9. Use Likert Items - These are questions written as strongly agree, agree, disagree, and strongly disagree. They are easy for respondents to understand and answer quickly.
10. Filter Random Responses: This will increase the overall survey validity. Ways to filter random responses include adding don’t know option, screening using knowledge (objective questions) or asserted interest (subjective questions), and avoiding response set to try and minimize agreement response bias, such as varying whether the liberal answer is agree or disagree.
What Do You Think?
What are some other good practices to use when constructing survey questions?