The Center for Social Data Analytics Brownbag
Speakers: Giancarlo Visconti and Yang Yang
Title: "Unrealistic Comparisons: Why Forced-Choice Conjoint Experiments Fail to Mimic Voters' Electoral Decisions"
Abstract: Political scientists increasingly rely on conjoint experiments to simulate and comprehend
the intricate decision-making process of voters when confronted with candidates differing on
multiple dimensions. However, in implementing these experiments, researchers often use
forced choices between candidates to elicit responses from participants, neglecting the realworld option for voters to abstain or cast a blank/none vote. This mismatch can lead to significant measurement errors and bias. We empirically demonstrate and evaluate the bias induced
by forced-choice design in typical conjoint experiments, using existing conjoint data from
published articles, a series of simulations, and an original conjoint experiment that randomizes different choice scenarios. We then propose a practical guide that incorporates substantive
knowledge about specific contexts to help researchers avoid possible pitfalls and enhance the
precision of their estimations.
the intricate decision-making process of voters when confronted with candidates differing on
multiple dimensions. However, in implementing these experiments, researchers often use
forced choices between candidates to elicit responses from participants, neglecting the realworld option for voters to abstain or cast a blank/none vote. This mismatch can lead to significant measurement errors and bias. We empirically demonstrate and evaluate the bias induced
by forced-choice design in typical conjoint experiments, using existing conjoint data from
published articles, a series of simulations, and an original conjoint experiment that randomizes different choice scenarios. We then propose a practical guide that incorporates substantive
knowledge about specific contexts to help researchers avoid possible pitfalls and enhance the
precision of their estimations.