Imputing nonresponses to mail-back questionnaires.

Many mail-back questionnaires are expected at the outset to elicit poor response rates, perhaps as low as 15-30%. Corrections can be designed into such a survey by using either two or three mailouts of the questionnaire at regular intervals. Assuming a trend in responses as a function of the number of mailouts a person receives before filling out and mailing back the questionnaire, responses are imputed for those who do not mail back the questionnaire after the final mailout. Standard errors are derived, and an example is included. The imputation is easily programmed. A validation of this method is also included.