Improving Studies of Sensitive Topics Using Prior Evidence: A Unified Bayesian Framework for List Experiments

Abstract

Estimates of sensitive questions from list experiments are often much less precise than desired. We address this well-known inefficiency problem by presenting a unified Bayesian framework which combines indirect measures with prior in- formation. Specifying informed priors amounts to a principled combination of information which increases the efficiency of model estimates. This framework generalizes a whole range of different design and modeling approaches for list experiments, such as the inclusion of direct items, auxiliary information, the double list experiment and the combination of list experiments with other indirect questioning techniques. As we demonstrate in several real-world examples from political science, our Bayesian approach not only improves the efficiency and utility but also changes the substantive implications drawn from list experiments. This way, it contributes to a more accurate understanding of sensitive preferences and behaviors of political relevance.