Robust Emotion Manipulation for Surveys: Evidence from Three Experiments

A large number of experiments investigate the effects of emotions on critical political outcomes---including policy attitudes, support for authoritarians, tolerance, political participation, and more. Using causal graphs, we show that in such experiments, emotions are intermediate outcomes of randomly assigned emotion manipulation instruments---not randomized treatments themselves. Understanding emotions as lying causally downstream of emotion inductions underlines the necessity of using emotion manipulations that are strong with respect to their effects on target emotions and specific with respect to priming the target emotion while minimally priming other emotions. In response to these imperatives, we present evidence from three experiments (total N=6,649) on the effectiveness of vignettes, autobiographical emotional memory tasks, images, and more for inducing anger, gratitude, fear, political anger, political gratitude, and political cynicism. We show that vignettes are broadly reliable manipulations in terms of both strength and specificity. We also investigate predictors treatment compliance in emotion manipulation experiments and find that pre-treatment attitudes toward research may moderate treatment effects.