A Two-Stage Sampling Model for the Estimation of Population Proportion and Cheating with Randomized Response and Direct Questioning

• Activity Number: 689
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Survey Research Methods Section
• Author(s): Evrim Oral* and Husam I. Ardah and Edward J. Trapido
• Companies: LSUHSC School of Public Health and LSUHSC School of Public Health and LSUHSC School of Public Health
• Keywords: Randomized Response ; Direct Questioning ; Social Desirability Bias

Abstract

Randomized response techniques (RRTs) are proposed in survey sampling as a solution to the problem of social desirability bias (SDB) when dealing with sensitive questions. RRTs reduce the SDB by providing privacy protection for respondents. However, their variances are inflated with respect to the direct questioning (DQ), i.e., the RRTs provide unbiased estimators in exchange for less precision with respect to the DQ. The success of RRTs heavily depends on the assumption that both the interviewers and all respondents fully understand and correctly apply the RRT procedure. More importantly, the sensitive question of interest may not be considered as really sensitive by most of the respondents, in which case using an RRT instead of the DQ will inflate the variance of the estimates. In this study, we propose a two-stage sampling design where one can accurately estimate the prevalence of the sensitive characteristic under study without paying the penalty of the inflated variance by choosing between the proposed model and the DQ. With the proposed model one can also estimate the probability of cheating in the population. We support the theoretical results with various simulations.