Despite its wide usage in explaining political dynamics of non-democracies, preference falsification remains an empirical myth for students of authoritarian politics due to the challenge of measurement. We offer the first quantitative study of this phenomenon in a non-democratic setting by exploiting a rare coincidence between a major political purge in Shanghai, China, and the administration of a nationwide survey in 2006. We construct two synthetic measures for expressed and actual political support and track their changes before and after the purge. We find that the purge caused a dramatic increase in expressed support among Shanghai respondents, yet the increase was paralleled by an equally evident decline in actual support. We interpret this divergence as evidence for preference falsification and conduct a number of robustness checks to rule out alternative explanations. We also show that falsification was most intense among groups that had access to alternative information but were vulnerable to political sanctions.