Introducing the Visual Conjoint, with an Application to Candidate Evaluation on Social Media

Abstract

Conjoint experiments have enabled scholars to understand the preferences of citizens in a variety of political contexts. We propose a method to modify the standard text-only “box conjoint” to make the treatment higher in external validity with respect to a common target context. Citizens frequently encounter political information encoded as images and in particular in the form of politicians’ social media posts and profiles. We deploy “visual conjoint” experiments where subjects select between two images that encode the same explicit information as is standard in the box conjoint. We conduct an experiment in which we randomize the modality of a conjoint experiment where subjects evaluate the Twitter profiles of hypothetical candidates. We demonstrate that the visual conjoint more effectively encodes image-based information and social endorsement information. The visual conjoint also allows the salience of different attributes to vary naturally the way they do on social media, in contrast to the artificially enforced uniformity of the box conjoint.