A Dual-Frame, Multi-Mode Sample Survey Design for Exit Polls in States With Election Day in-Person, Early, and By-Mail Voting*
Geoffrey Cannon, Howard B. Christensen, David B. Magleby, Joseph A. Olsen, Dan Williams · 2015
Adopting no-excuse absentee voting, vote-by-mail, and early voting centers will change how exit polling is done in the future. Conventional Election Day exit poll sample designs are incapable of reaching the early voter. Until 2010 the Utah Colleges Exit Poll relied on a sample based on a geographical frame consisting of strata/counties, and polling places within counties that ultimately connected the survey to the voters. With two large counties that have been part of the sampling frame for the exit poll adopting some form of vote-by-mail in 2014 we expanded our sample design adopting a dual frame, multi-mode approach. This paper describes changes in sample design for the Utah Colleges Exit Poll and our 2014 experiences in using a second sample design. We had four different modes of data collection: the traditional, Election Day paper ballot, an iPad-based ballot, a telephone ballot, and Internet ballot. The latter two modes utilized data provided by the state on who had voted, with the data updated daily. We found some significant differences in who votes early in-person, by mail, and on Election Day. We also comment on lessons learned for our 2016 exit poll. Our findings from this paper are valuable for exit pollsters both at the state and national level as we look at the likelihood of more and more states offering or requiring vote-by-mail or other early voting options.