Traffic Cameras, Voting Costs, and the 2015 United Kingdom General Election

• In Event: Advances in Deep Learning Methods for Text, Image, and Video Data

Abstract

As video data becomes more ubiquitous, social scientists must overcome several hurdles before inference is possible. One such hurdle is estimating the physical distance between objects in live footage. Since videos are a two-dimensional representation of a three-dimensional space, this task is not entirely straightforward, especially when the camera angle, zoom rate, and lens quality is unknown. Using 5 days of footage from over 150 traffic cameras before and after the 2015 United Kingdom election, this paper introduces an algorithm which uses the latitude and longitude from Google Maps to estimate the speed and distance between cars near polling locations. This information is then combined into a measure of traffic conditions in an effort to test whether poor traffic decreased turnout and disadvantaged incumbents on Election Day. Although our paper aims to help scholars better use traffic camera and other video data in their own research, the substantive contributions of our study are also noteworthy. Prior studies have demonstrated the importance of voting costs (Burden et al 2014; Leighley and Nagler 2015), especially as they relate to physical hurdles, like polling locations (Dyck and Gimpel 2005; Gibson et al 2013; Gimpel and Schuknecht 2003; Cantoni 2015; Brady and McNulty 2011; Newman, Johnson and Lown 2014), long lines (Barreto et al. 2009; Stein et al. 2019), or weather conditions (Gomez, Hansford, Krause 2007). The present study uses a similar argument in a new application that has yet to be explored in the literature. Ultimately, we find incumbents did significantly worse in polling locations with poorer traffic conditions on Election Day as compared to the day before. Not only do our results demonstrate how subtle changes in voting costs can carry profound electoral implications, but we also introduce a general methodology which will help scholars track objects in a way that is meaningful for social science research.

Authors

• Bryce Dietrich, University of Iowa

• Hyein Ko, University of Iowa

• Payel Payel Sen Sen, Stony Brook University