Worth 1000 Words? Image-Based Conjoints and Respondent Experience

Public opinion researchers seeking to improve the ecological validity of conjoint experiments have incorporated visual media such as candidate photos, campaign posters, and videos to replace or complement text-based tables. While previous work has investigated visual delivery of racial or gender information, there has been little evaluation of how visual stimuli influence conjoint results more generally. This note identifies three areas where the introduction of images might affect candidate choice experiments: parameter estimation, conjoint task difficulty, and the introduction of additional features not intended by the research design. Analysis of a split-sample conjoint experiment where respondents were randomly assigned an image or table-based version of an otherwise identical candidate choice experiment show that there was little difference in parameter estimates between the table and image-based group, respondents demonstrated greater difficulty completing the image-based task, and unintended features introduced by candidate photos did not systematically affect respondent behavior. The main trade-off that researchers should consider when selecting the medium of their conjoint treatment is therefore between the realism of the task and its difficulty.