What drives preference for the compromise option? Disentangling position-based versus attribute-based effects

Jungkeun Kim*, Mark T. Spence, Harmen Oppewal, Jongwon Park, Roger Marshall

*Corresponding author for this work

Research output: Contribution to journalArticleResearchpeer-review

2 Citations (Scopus)
55 Downloads (Pure)


The compromise effect exists where consumers show a disproportional preference for the option with intermediate attribute levels in a choice set. The effect has been widely observed in the literature. Other research has demonstrated the benefits of a product physically occupying a middle position, regardless of relative attribute levels. Both streams of research thus show that consumers' preferences are influenced by the display format of the choice set, with benefits accruing to the middle option. This research demonstrates that the compromise effect comprises both an attribute-based and a position-based component by disentangling these two effects. We demonstrate the coexistence of both components in a series of five experiments, including a field experiment, that systematically vary an option's position based on relative attribute levels and in terms of position. Existing literature on the compromise effect only to a limited extent accounts for this possibility, even though it has significant theoretical and practical relevance. While brands largely determine attribute levels, marketers/retailers can often determine a brand's display position. We also establish the moderating role of time pressure and find that the position component becomes more prominent when time is constrained while the attribute effect becomes more prominent when time is unconstrained.

Original languageEnglish
Pages (from-to)2153-2170
Number of pages18
JournalPsychology and Marketing
Issue number11
Publication statusPublished - Nov 2022


Dive into the research topics of 'What drives preference for the compromise option? Disentangling position-based versus attribute-based effects'. Together they form a unique fingerprint.

Cite this