This article asks whether apparent attribute effects in candidate choice reflect unobserved masking (inference about missing information) or genuine lexicographic preferences (respondents treating one attribute as dominant). The authors propose an empirical test that uses conjoint experiments to distinguish between the two explanations and illustrate the approach with data on gender, race, and party; they also report a preregistered replication.
Bold testable distinction between masking and lexicographic preferences Use of conjoint experiments on candidate attributes Preregistered replication
Data and methods
The paper develops an empirical test implemented within conjoint experiments. The abstract states the authors illustrate the method using data on the effects of gender, race, and party in candidate choice and then replicate the results in a preregistered conjoint experiment. The presentation is methodological: it compares competing behavioral models using experimental attribute manipulations.
Main findings
According to the abstract, the method yields results distinguishing pure masking from lexicographic preferences in the context of candidate choice. The authors report an illustration and a preregistered replication; the abstract does not report numerical effect sizes or specific empirical estimates, so readers should consult the full article for detailed results and robustness checks.
- The paper frames masking as inference about unobserved attributes (for example, inferring party from gender) and offers lexicographic choice as an alternative explanation.
- It provides recommendations for researchers studying candidate and political choices based on the test and replication reported in the article.
