Who Likes It More? Using Response Times To Reveal Group Preferences in Surveys.

Abstract:

Surveys are essential to measure societal preferences and uncover differences between socieconomic or demographic groups. However, survey data is noisy and biased, limiting the reliability of standard analyses. We provide a non-parametric preference revelation method to improve inference in surveys, panels, and large online experiments. Applications include eliciting social attitudes, support for policies, political trends, and marketing research. Using recent developments in neuroeconomics, the method leverages choices and response time distributions to uncover group preferences and unambiguously rank the relative strength of preference between groups of people, defined, e.g., by gender, age cohort, socio-economic status, political orientation, etc. The analysis relies on analytical results that deliver nonparametric criteria to reveal whether a group prefers an option over its alternative, and whether it does so more than another group, for a general class of behavioral models and without any assumptions on the underlying noise. We validate the techniques and apply them to important socio-economic topics in two representative surveys. The method is an inexpensive, complementary tool to existing approaches and it further indicates when standard analyses are unwarranted or when we can learn more from the data using response times. It also provides a quantification of biases in responses to surveys.

Carlos Alos-Ferrer, Lancaster University