Towards a General Theory of Survey Response: Likert Scales vs. Quadratic Voting for Attitudinal Research

"Likert scales" are the most standard and widespread instrument in survey research when measuring public opinion on political and economic issues. In this simple approach, respondents are given the opportunity to voice their agreement or disagreement on a set of issues by placing their attitudes on a scale that runs from "strongly disagree" to "strongly agree." One assumption commonly made by social scientists using such scales is that they provide faithful - if noisy - measures of respondents' views. We challenge this assumption, highlighting several reasons why respondents may be expected to systematically exaggerate their views in political surveys using Likert scales. We propose a simple decision-theoretic model of survey answers to discuss whether Quadratic Voting might overcome these pathologies. We provide conditions under which one might expect Quadratic Voting to outperform Likert scales.