Measuring Second-Order Opinions

Expectations regarding what others believe (second-order opinions) influence people's own opinions, as well as their behavior. Voters’ opinions are also important for political elites. There has been an increasing interest in these questions among political scientists in recent years. Various methods are employed to measure and analyze these second-order opinions. Approaches include within-subject designs – asking respondents both about their own opinion and their estimate of public opinion, and between-subject designs – comparing aggregate results between randomized groups. Some ask respondents to estimate the share of fellow citizens or voters who support something, others ask respondents to place the weight of the majority on one side or the other. Drawing on survey experiments and panel data, this paper introduces novel evidence, testing different designs, response scales, and randomization procedures to document respondents' second-order opinions. The study incorporates data collected across survey samples of citizens and elites. Specific recommendations are provided to guide future research in this evolving domain.