Changing the Lens: The Added Value of Analyzing Marginal Means and Using the Rating Outcome in the Analysis of Conjoint Experiments

Abstract

Conjoint experiments have gained considerable popularity in political research. Until recently, the majority of the studies utilizing this design have focused on the analysis of the causal quantities - Average Marginal Component Effects (AMCEs), and used mainly the forced-choice outcome. However, Leeper et al. highlighted the importance of also considering marginal means (MMs), which offer descriptive quantities of preferences. Expanding upon their work, this study demonstrates how examining both MMs and AMCEs and utilizing the rating outcome can lead to different substantive conclusions and policy implications. Using an original conjoint experiment conducted in Israel and the US to study the determinants of public attitudes toward paternalistic policies, I show that although the causal effects of different policy attributes are similar in both countries, the policy implications of these effects differ substantially in each country when levels of support are examined through MMs. To further exemplify the added value of MMs, I replicate two recent studies that focused on AMCEs and demonstrate how including MMs in the analysis could have enriched or even altered the conclusions drawn. Finally, the article provides practical recommendations for analyzing conjoint data.

Keywords: conjoint experiment, marginal means, AMCEs, methods