Can ATMs get out the vote?

November 19, 2019

This paper reports on a large-scale (randomized) field experiment designed to assess ATMs’ (automatic teller machines) capacity to “get out the vote”, an unexploited method of voter mobilization. The experimental design used the full universe of functioning ATMs in Portugal, which benefits from a sophisticated world class system with wide national coverage. The experiment randomly selected a set of treatment civil parishes, where a civic message took over the totality of ad time in ATMs, and compared with a set of control civil parishes where advertisements ran as usual. The ATM campaign was active for three days immediately before and during the 2017 local elections in Portugal.

Considering the intensity of treatment for both the entire campaign period and the weekend, the results show a statistically significant increase in the likelihood of voting arising from the ads. All three proxies for treatment intensity employed, i.e., the number of cards used, the number of ATM operations, and the number of withdrawals, deliver robust positive coefficients. Placebo tests using turnout rates in previous elections do not show significant coefficients, strengthening the interpretation of the experimental results. The paper also ran a post-treatment survey around ATMs located in two neighboring civil parishes in Lisbon, one treated, the other not, one week after the local elections. This survey found a sizable difference in recall between the treatment and control groups.

Click here to go to the paper by João Pereira Santos, José Tavares, and Pedro C. Vicente.