Social desirability and the fear of sanctions can deter survey respondents from responding truthfully to sensitive questions. Self-reports on norm breaking behavior such as shoplifting, non-voting, or tax evasion may thus be subject to considerable misreporting. To mitigate such response bias, various indirect question techniques, such as the randomized response technique (RRT), have been proposed. We evaluate the viability of several popular variants of the RRT, including the recently proposed crosswise-model RRT, by comparing respondents' self-reports on cheating in dice games to actual cheating behavior, thereby distinguishing between false negatives (underreporting) and false positives (overreporting). The study has been implemented as an online survey on Amazon Mechanical Turk (N = 6, 505). Our results from two validation designs indicate that the forced-response RRT and the unrelated-question RRT, as implemented in our survey, fail to reduce the level of misreporting compared to conventional direct questioning. For the crosswise-model RRT we do observe a reduction of false negatives. At the same time, however, there is a non-ignorable increase in false positives; a flaw that previous evaluation studies relying on comparative or aggregate-level validation could not detect. Overall, none of the evaluated indirect techniques outperformed conventional direct questioning. Furthermore, our study demonstrates the importance of identifying false negatives as well as false positives to avoid false conclusions about the validity of indirect sensitive question techniques.