The Effect of Billboards on Turnout in 4 States: A Randomized Experiment

• In Event: Billboards, Ballot Boxes and Voting Booths

Abstract

This paper will present the results of a large-scale randomized voter mobilization billboard experiment conducted during the 2019 election in conjunction with the voter turnout organization vote.org. To test the effect of billboards with non-partisan election day information on turnout, maps of parts of four states (Virginia, Kentucky, Louisiana, and Mississippi) were overlaid with a grid (or “fishnet”) containing 3-mile by 3-mile cells. Cells were randomly selected to be in the treatment or control group. Working with vote.org, all available billboards in the treatment cells that met certain locational and cost criteria were purchased and fitted with a reminder to vote, the appropriate election date, and a small vote.org logo. The hypothesis is that those who live in treatment cells will have turned out at a higher rate than those who live in the control cells. Based on previous work with vote.org, one of the key criteria placed on billboard purchases was that they could not be located on major highways (as many billboards are). Because billboards can theoretical spur turnout for people that live both near and far from the billboard location there is a need to utilize billboards that are more likely to have a spatially concentrated effect. Thus, the experiment operates on the assumption that the only billboards for which ordinary individual-level turnout data can reveal an effect are those located nearby to more residential areas. We term these billboards as having a "potential spatially proximate observable effect." Across the four states, 244 billboards were treated and 196 served as control billboards for the weeks leading up to Election Day.

Authors

• Scott Louis Minkoff, SUNY, New Paltz

• Christopher B Mann, Skidmore College