The length of the survey is important in determining response rates. The longer and more complex the survey, the less likely respondents are to complete it. Each question should be scrutinized for its importance to the final analysis. The wording of the questions is also critical in obtaining an accurate result. Questions should clearly get at the issues of interest, and the language and reading level must be appropriate for the audience. Questions should be free of ambiguity, and each question should address a single issue. (For example, "Do you currently have a satisfying job?" addresses two issues¾ whether the respondent has a job and whether the job is satisfying). Questions should be reviewed for response bias (i.e., the tendency to give socially acceptable answers).
Response formats should be appropriate to the question. Questions may be asked in open-ended, multiple choice, ratings, or other formats. Questions should cover the entire range of possible answers. Answers may be complex and may not be captured by simple answer choices (e.g., "Why are you no longer receiving benefits?"). If interviewers are used, it should be clear whether they should read the questions and whether they should provide interpretations.
Sensitive material should be elicited without stigmatizing the respondent. It should be placed after more general questions and should seem a natural part of the questionnaire.
Pretesting the questionnaire on a small group of respondents will often reveal problems in interpreting questions. These problems can then be corrected before the survey is distributed.