Mind-reading a friend: A better way to ask the polling question?

Abstract

This article reports on differences observed when asking a simple polling question in a traditional way—that is, asking respondents for predictions about their own voting behavior versus asking respondents for predictions about a friend’s voting behavior. My intuition, based on a reading of the behavioral science literature, was that a polling question based on “reading the mind” of a friend might yield more accurate results than a traditional self-declared polling question. Four independent studies among UK adults were fielded using Google Surveys, two for each approach at the start of the electoral campaign and two for each approach at the end of the campaign. Analysis of survey results suggested that the question asking for predictions about a friend was more accurate, and more sensitive to campaign shifts, than the self-declared question. Furthermore, it was the only approach that seemed to predict a decline in turnout among older voters—a subtlety that was not more widely recognized until the numbers had been crunched post-election. We conclude by suggesting that the authenticity of the approach could be enhanced when combined with pollsters’ weighting models and that as a technique it deserves greater consideration within quantitative market research generally.