A Sampling Procedure for Mailed Questionnaires

T HE mailed questionnaire is desirable in surveys because it is economical of money and effort, free from the interviewer's effect on the respondent, and able to reach individuals not accessible by visit or telephone. However, it has several drawbacks which can seriously limit its efficiency as a means of gathering reliable data. The mailed answers are often exposed to two kinds of bias: (1) bias arising from the respondents' misunderstanding the questions, resentment of interference in their personal affairs, or falsification for reasons connected with the subject of the survey; (2) bias of non-response, which arises from differences in the characteristic under investigation between respondents and non-respondents. The latter bias is often more pronounced in mailed questionnaires than in personal interviews and its reduction to acceptable limits will always necessitate extra effort, time and skill. The problem of non-response has already been tackled from various angles. Hansen and Hurwitz [4] studied a survey which combines the advantages of mailed questionnaires and personal interviews. The plan first utilizes the economies involved in the use of questionnaires by mailing them to a sample of the population under investigation. A follow-up is then carried out by interviewing a subsample of the non-respondents to the mail canvass, thus eliminating substantially the bias of non-response in the first stage. The optimum allocation of the mail and field samples is obtained by requiring minimum cost for an assigned precision. The sizes of the two samples naturally depend on the rate of non-response to the mailed questionnaire and the variances of the characteristic under investigation both among the whole population and among the non-respondents. These parameters are assumed to be known from previous experience. Politz and Simmons [5] devised a technique for circumventing the need for call-backs in interviewing surveys. Each person in the sample is visited onlce, at a time determined at random. From each person interviewed information is obtained as to whether he was at home at certain times, which leads to an estimate of the proportion of the time he is at home during the interviewing hours. The sample estimate is produced by weighting the results for each group that spends the same proportion of time at home during the interviewing hours by the reciprocal of the estimated percentage of time that the members of this group are at home.