Survey recruitment messages and reported turnout – an experimental study

Abstract

Voter turnout rates among survey respondents tend to exceed actual turnout by a significant margin. This note reports the results of an experiment whereby the indicated topic of a political survey was varied at the recruitment stage, with the goal of producing a sample with reported turnout rates that more closely mirror reality. Three experimental groups were informed that the survey was about politics, while a fourth received no indication of the nature of the study. Data from nearly 5000 respondents in Ontario, Canada suggest that this manipulation had no appreciable effect upon survey completion rates, but the level of reported turnout is lower when no topic is indicated. We theorize that this manipulation affects self-selection into (out of) surveys by voters (non-voters). By not disclosing the political nature of a survey, samples can be made to more closely reflect populations they are meant to represent in this important dimension.