In statistical surveys, data is often collected through questionnaires and sometimes by making observations. A well-designed questionnaire is essential for ensuring the quality of data gathered. Here are some points to consider when designing a questionnaire. First, questions should be relevant to the survey objectives. Lengthy questionnaires can put off respondents. Note that the inability of people to recall facts and events in the past may result in inaccurate data. Ensure that questions do not relate to past events. The second point is that the wording in a questionnaire needs to be appropriate, specific and precise. It needs to be simple and should mean the same thing to all respondents. If not, questions could be misunderstood and generate inaccurate responses. Vague questions could produce invalid answers. For instance, if someone were to say 'yes' to the question: 'Did you see the dean of students last week?', it might mean many things, such as (1) I saw him (with my eyes); (2) I consulted him, or (3) I met him. This could be avoided by rephrasing question to: 'Did you consult the dean of students last week?' if this is what is meant. The third point to note is avoid double-negative questions. The question 'Don't you think alternative X is not a good solution to problem Y?' is a typical of such a question. This could be confusing. It is better to ask: 'Do you think alternative X is a good solution to the problem Y?' If an opinion is being solicit ed, avoid leading, loaded and composite questions. Consider the questions: 'Please name the TV programmes you watched yesterday, say, programme A or B.' This is a leading question. Consider another one: 'This policy is undesirable, do you think the Government should keep this policy?' This is a loaded question which will elicit a negative response. These could affect the credibility of survey results. Steer clear of composite questions. A composite question combines two or more questions and is complex. Consider this: 'Do you possess an optimistic view about Hong Kong's political and economic prospects?' Ideally, the question should be in two parts. Another point to note is that questions must be arranged in a proper order. If people are forced to answer questions which they are unable to respond to or on which they do not have any opinions, then their answers will be invalid. Therefore alternative answers such as 'don't know' and 'no opinion' should be included as options. Also, add screening questions where necessary. For example, questions on salaries and earnings should only be answered by working people. Therefore these questions should be preceded by a screening question such as: 'Do you have a job at present?'. And finally, questionnaires should be tested on some prospective respondents before they are finalised for use in a survey. For more information on this series of articles, write to the General Statistics Branch (2) of the Census & Statistics Department at Wan Chai Tower, 12 Harbour Road, Wan Chai, Hong Kong, or call 2582-4732
