Abstract
Conjoint analysis has become one of the most commonly used statistical tools in social science, particularly political science. In its standard form, respondents compare two profiles (e.g., hypothetical candidates, immigrants, or policies) with randomly juxtaposed attributes and select one of the profiles. Drawing on the social psychology literature, we show that the presence of a comparison profile systematically influences how respondents evaluate a given profile. As a result, stated preferences differ between single-profile designs without comparison and paired-profile designs with comparison. Our results highlight the importance of explicitly justifying design choices in conjoint analysis and aligning them with the underlying decision process of interest. Two-profile tasks are most appropriate when the research goal is to examine how respondents evaluate options in comparative contexts, in which case quantities of interest should be defined at the level of choices rather than profiles.
Citation
Horiuchi, Yusaku, and Martha C. Johnson. n.d. “To Compare or Not to Compare: Testing the Effects of Profile Comparison in Conjoint Analysis.” Working paper. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5957676