The prevalence of extradyadic sex (EDS) tends to be underestimated due to underreporting in national surveys, which use direct questioning. Self-administered questionnaires can reduce this response bias but may cause the anxiety of privacy exposure. Randomized Response Technique (RRT) can ensure participants' confidentiality under the randomized design of indirect questions to and was found to yield more valid estimates of sexual or other sensitive behaviors than direct questions. This study estimated the EDS rate among Taiwanese aged 18 years and over, using a hybrid of Randomized Response Technique and Crosswise Design (RRTCD) and the Weighted Conditional Likelihood (WCL) estimator. The data analyzed were from the 2012 Taiwan Social Change Survey, in which the answer to the innocuous question from the unrelated-question RRT of Greenberg, Abul-Ela, Simmons, and Horvitz (1969) was obtained indirectly from a demographic question related to the innocuous question. This RRTCD provided more information on the innocuous question to effectively improve the efficiency of the unrelated-question RRT of Greenberg et al. The WCL estimator was found to be more efficient than the Greenberg et al. estimator for estimating the EDS rate in terms of smaller standard errors and smaller differences in the levels of EDS across sociodemographics and extramarital-sex attitudes. Similar to those suggested in the literature, the estimated rates of EDS in two subsamples were higher among men, homosexuals, those who have or had wages, and those who accepted extramarital sex. The levels of EDS varying with sociodemographics were different between the married and the unmarried.
Keywords: Extradyadic sex; Random Response Technique; Randomized Response Technique with Crosswise Design; Weighted Conditional Likelihood estimator.