Asking Sensitive Questions in Surveys

THE purpose of this class exercise is (a) to reinforce the student's grasp of tree diagrams, probability calculations, etc., and (b) to mention a topic of considerable practical importance, the randomised response technique. It comes from Hutchinson (1993). Suppose I wish to determine what proportion of the population drive faster than the speed limit, fail to pay taxes, go to church on Sunday, etc. I am not the least bit interested in whether a specific individual drives faster than the speed limit, but I do want to get an honest overall estimate for the whole population - and people are often not honest when answering questions like these. The randomised response method can be used here. It relies upon the respondent knowing that the interviewer does not know whether the respondent is answering the sensitive question, or is answering a perfectly harmless question. An example suitable for class use is the following.