Key points are not available for this paper at this time.
Does the fact that only a third of the respondents answering a mail questionnaire answer all items reduce substantially the reliability of the results This is a study of 14,600 usable questionnaires from a 40,000 mailing. It analyzes item nonresponses by characteristics of respondents and by type, position, and repetition of questions.
Robert Ferber (Sat,) studied this question.