Attribute Importance in Conjoint Analysis

• In Event: Advances in Survey Experiments

Abstract

In this research letter, we propose the use of a random forest model to study the importance of attributes in conjoint analysis. Traditional conjoint methods can be limited in their ability to accurately measure the relative importance of different attributes. We argue that the use of a random forest model can overcome these limitations and provide more accurate attribute importance. We present both simulation results and results from a replication of a recent eye-tracking experiment to validate our new method. Our results indicate that the random forest model is a promising approach for studying attribute importance in conjoint analysis and can provide valuable insights in political decision-making.

Author

• Lukas Frederik Stoetzer, University Witten/Herdecke