Sensitive Questions and Social Desirability - Theory and Methods

点击翻译按钮获取中文摘要

The project is situated within the research area of survey methodology and deals with measurement error in survey interviews. We focus on response bias resulting from misreporting by survey respondents (evincing of "incorrect" instead of "true" values) when answering survey questions. What the project is mainly interested in with respect to this are, on the one hand, opinion expressions on fictitious issues, i. e., expressing an opinion or attitude related to issues that actually do not exist. On the other hand, we focus on so-called sensitive questions (for instance on substance abuse or deviant behavior), where respondents often deny socially undesirable characteristics and overreport socially desirable ones.The overall objective of the project consists in providing theoretical and methodical basic research to the topic of response bias in surveys. In so doing, we wish to contribute to the theoretical understanding of the reasons and mechanisms causing response bias, and to gain methodical knowledge about which "best practice" and which questioning techniques yield valid survey data. The central point of the theoretical research agenda is an elaboration and application of frame-selection-theory, as a general theory of action, to response behavior in surveys. An important question with this regard is, for example, whether and when misreporting result from a cognitively reflective-calculating process or instead occurs automatically-spontaneously. The methodical research interest consists, first, in assessing what general design aspects (interviewer instructions, spontaneous or reflective answering of survey questions) have a positive or negative impact on data validity. Second, we are interested in the performance of new variants of the item-count-technique (ICT), an anonymizing questioning technique especially developed for posing sensitive questions in survey interviews. Recent further developments of ICT are being tested and empirically evaluated with regard to the question whether the complete anonymization of the response process reduces misreporting as compared to standard questioning techniques.Our research methods consist in carrying out a large-scale telephone survey (N=3000) and in conducting a meta-analysis on the performance of ICT. In the survey, we employ, first, classic methods of research on response bias like measuring respondents' need for approval; second, response latencies as a measure for the respondents' cognitive information processing mode are recorded and analyzed; third, the survey will contain experimental method splits in order to investigate the effects of different question designs on response behavior. We also experimentally investigate the performance of new variants of ICT. The meta-analysis will pool preferably all published and unpublished research on the efficacy of ICT in avoiding response bias with sensitive questions.