Indirect Questioning in Sample Surveys

Collecting data on human populations by means of sample surveys is not an easy task. Survey practitioners often experience difficulties in collecting reliable data due to various sources of nonsampling error and in particular due to nonresponse. In case the issues under investigation are of sensitive nature, such as issues on sexual orientation, tax evasion, or involvement in criminal activities, people are reluctant to participate, and even if they agree to participate, false or misleading answers are given by many of them. Indirect questioning techniques offer a solution to this problem. These are techniques designed in such a way that the information provided by a participant is not incriminating and thus his/her privacy is protected. However, based on the information collected from all participants, the investigator is able to estimate parameters of interest related to the sensitive characteristic. In this chapter we make a case in favor of the use of indirect questioning techniques. We briefly discuss hypothetical as well as real examples where the methodology presented in this book can be implemented.

In this chapter the basic and rudimentary aspects of sample surveys for finite populations are presented in a compact way. The concepts of population, sample, sampling design, survey data, estimating finite population parameters of interest and consequent errors and their control will be explained and detailed illustrations will be provided. The theory to address general issues will be explained first. Then the need for modification to cover the case of sensitive issues and how to do that will be explained. It will be clearly shown how in a general situation one may handle indirectly procured observations to estimate parameters of interest and also derive estimated measures of accuracy. Sophisticated theoretical details will be presented only in brief. Finally in this chapter we put emphasis on the fact that any probability sampling design may be employed for the purpose of estimating parameters related to stigmatizing characteristics.

It will be first emphasized that no matter how a person is selected to be in the sample, an indirect procedure may be adopted to endeavor to gather sensitive data from him/her in deriving sensible estimators for population parameters along with estimated measures of their accuracies. This chapter offers a preview of the indirect questioning techniques that will be presented in subsequent chapters in more detail. The broad and rich class of Randomized Response Techniques and their rationale are presented first. Then three more techniques are discussed briefly, the Item Count Technique, the Nominative Technique, and The Three Card Method along with their rationales. The chapter ends with previews of two of the most recent indirect questioning techniques, the general class of Non Randomized Models which includes techniques where no randomization device is needed and the method of Surveying with Negative Questions where questions are asked in a negative way so that a respondent can provide one of many possible answers.

This chapter is devoted entirely to randomized response techniques which can be implemented to estimate certain parameters of a qualitative stigmatizing characteristic. Descriptions of the randomized response procedures of specific techniques are given. In particular details are provided for Warner’s, Simmon’s, Kuk’s, Christofides’, and the Forced Response randomized response techniques. For those techniques, explicit formulae are given for the various estimators of interest and measures of their accuracy, assuming that the sample is chosen according to a general sampling design. However, given that most practitioners are more familiar with simple random sampling without replacement, the formulae are explicitly stated for this particular sampling scheme as well. In addition to the numerous randomized response techniques reviewed, this chapter includes a recently developed randomized response technique which uses the Poisson distribution to estimate parameters related to a stigmatizing characteristic which is extremely rare. Furthermore, we discuss an approach using the geometric distribution to generate randomized responses. Also in this chapter, techniques dealing with multiple sensitive characteristics are described. Finally, some aspects of the Bayesian approach in analyzing randomized response data are presented along with a brief literature review on the topic.

A brief outline of the general theory of estimating finite population totals and means based on a sample selected with a suitable sampling design is given. Initially it is assumed that direct responses are available and then the theory is developed in the case when the sensitivity of the data on the quantitative characteristic makes it necessary to implement suitable devices to collect randomized response data. Two different randomized response devices are considered. The theory of estimation is illustrated in case the sample is selected employing the Rao-Hartley-Cochran sampling scheme as well as in the case of a general sampling scheme and when the data are collected using either of the two devices. Techniques which allow for direct responses by participants are presented. Such approaches are based on the idea that some people may consider the item in question not sensitive enough and therefore both options for providing a direct response or a randomized one are available. The main advantage of these optional randomized response techniques is the variance reduction of the produced estimators.

Numerous randomized response techniques have been developed to handle the case of stigmatizing characteristics. Warner’s (1965) pioneering technique was just the beginning. One of the main disadvantages of randomized response techniques is the fact that participants often are very skeptical about the whole process because, either they do not understand it or because they feel that their privacy is not really protected. In addition, in cases where a randomization device is being used, people think of randomized response as a trick, or as a process which is not really a serious scientific method. Because of these and other drawbacks, for example the fact that randomized response very rarely can be incorporated into survey questionnaires, other alternative methods have been devised. In this chapter, five of those techniques and their variations are presented along with the relevant theory. The most popular one, the Item Count Technique is discussed first, and various versions of it are given. Another technique included in this chapter is the Nominative Technique, which, as explained, can be thought of as an application of network sampling. The Three-Card Method, a simple and easily understood technique is also discussed in brief with theoretical details omitted. A special treatment is given to the recently developed class of Non-Randomized Models. Those are techniques which do not use any device. However, this does not mean that no randomizations are taking place. The last section of the chapter is devoted to the so-called Negative Surveys. Those are surveys where questions are phrased in a negative way so that all but one of the possible answers are true for each and everyone one of the participants.

The main motivation behind indirect questioning is to increase participation, reduce nonresponse and reduce untruthful responses in surveys dealing with stigmatizing characteristics. Indirect questioning techniques are often advertised as methods protecting the privacy of the participants in surveys dealing with stigmatizing or sensitive issues. This is easy for anyone to see. However, different techniques do not necessarily offer the same level of protection. Quantitative measures of the protection of privacy and measures of jeopardy have been devised, which can be used, among other things, to compare one indirect questioning technique to another. Those measures are discussed and special emphasis is put on their limitations. In this chapter a case is made for the need to develop quantitative measures of the protection of privacy as perceived by the participants. A person is willing to participate in indirect questioning sample surveys if he/she feels that his/her privacy is protected and that the answer provided is not sufficient for someone to determine whether he/she bears the stigmatizing characteristic. Real life examples are cited which prove that the issue of privacy protection from the participant’s point of view is indeed very important.