Retrieving True Preference under Authoritarianism

Abstract

Scholars of authoritarian politics identify preference falsification in public opinion surveys by measuring the difference between a respondent's answers to politically sensitive questions and non-sensitive questions. Yet, the selection of questions is not empirically tested but justified only by the researchers’ prior knowledge in the field. In this paper, we explain how latent profile analysis (LPA), a tool to analyze survey respondents based on their answer patterns, can provide observation-based evidence of the potential existence of preference falsification. We first provide a theoretical framework where we classify survey respondents under authoritarianism into true regime supporters, candid non-supporters, and preference-falsifying sub-populations. Then, we demonstrate the application of LPA to public opinion research through the use of data simulation, a quasi-experimental setting from Chinese General Social Survey data, and World Value Survey data.