How to Measure Second-Order Opinions

I decided to prepare two shorter drafts (“letters” or “research notes”) for this conference instead of a full research article. Letter 1: Split-Sample Test of Second-Order Opinion Response Scales Word count: 2,700 In recent years, political scientists have shown a growing interest in political actors' ability to estimate public policy preferences. These estimations, or second-order opinions, play a crucial role in shaping how political actors advocate for their agendas. However, measuring second-order opinions presents unique methodological challenges. This study demonstrates that the method used to measure and operationalize second-order opinions can significantly affect the resulting conclusions. By employing a random split-sample design, this research compares the outcomes of two distinct methods for recording second-order opinions. One group was exposed to a slider scale ranging from 0 to 100, while the other group was given a binary response option: "most people in Norway agree/disagree." The primary research question is whether these two measurement methods yield different substantive conclusions. The findings indicate that a simplified binary response scale improves the accuracy of public opinion readings when there is a clear consensus but exacerbates biases when public opinion is divided. Letter 2: Projection, Conformity and Question-Order Effects in Second-Order Opinion Studies Word count: 1,576 This research letter addresses a methodological question that has been insufficiently explored in the now growing literature on second-order opinions: Can first- and second-order opinions be measured within subjects without introducing consistency bias? This inquiry is crucial because existing studies often show that second-order opinions are heavily influenced by first-order opinions. We present original experimental evidence demonstrating statistically significant within-subject effects when respondents are asked about both their own opinions and their assessments of others' opinions in the same survey. Notably, biases appear in both directions: first-order opinions can influence second-order opinions, and second-order opinions can influence first-order opinions.