Friend-to-friend texting may be the most effective voter mobilization tactic during 2020 election

Newswise — Friend-to-friend text messaging may be the new door-to-door canvassing leading up to the 2020 election.

Columbia researchers have found that texting may be an effective alternative to conventional, face-to-face voter mobilization tactics during the COVID-19 pandemic. The team, which was led by Data Science Institute postdoctoral research fellow Aaron Schein, found that receiving a message from a friend led to increased voter turnout during the 2018 midterm elections. The estimated effect was equivalent to door-to-door canvassing, an important finding in the social distancing era.

Schein and his collaborators, including political scientist Donald Green, DSI member David Blei, DSI postdoctoral fellow Dhanya Sridhar, et al., partnered with Outvote, a get-out-the-vote tool, to conduct the study.

Outvote randomized an aspect of their system prior to the 2018 midterm elections to unobtrusively assess the causal effect of their users' messages on voter turnout. The Columbia team developed a statistical methodology to address the challenges of such data, and analyzed Outvote's previous findings. Their analysis reveals evidence of large and statistically significant treatment effects from friend-to-friend mobilization efforts. The statistical methodology may be used to study other friend-to-friend messaging efforts.

Schein uses statistical models to understand and predict factors that drive new voter turnout in U.S. political elections. He acknowledges that studying friend-to-friend appeals experimentally can be challenging. "If the study is too obtrusive, such that it hampers natural interaction, we may end up measuring the effect of friend-to-friend appeals that are stilted and inauthentic. On the other hand, if the experimental intervention is too weak, we may fail to estimate effects with precision," he said.

The study employed a "light-touch" scheme, Schein said, while injecting enough randomness into the system to identify causal effects. "The statistical challenges we overcome to estimate these effects are likely inherent to any study of authentic friend-to-friend appeals and are perhaps why there have been so few studies of this kind to date."

These research findings were presented to the 2020 International Conference on Computational Social Science (IC2S2 2020). Schein received the 2020 Best Presentation award during the conference.



The Data Science Institute (DSI) at Columbia University was established in 2012 to advance the state-of-the-art in data science; transform all fields, professions, and sectors through the application of data science; and ensure the responsible use of data to benefit society. We train the next generation of data scientists, develop innovative technology, foster collaborations in advancing techniques to interpret data and address pressing societal problems, and work closely with industry to bring promising ideas to market. Discover more at

Filters close

Showing results

110 of 5609
Newswise: UA Little Rock researchers investigate information campaigns designed to influence 2019 Canadian election
Released: 30-Oct-2020 5:25 PM EDT
UA Little Rock researchers investigate information campaigns designed to influence 2019 Canadian election
University of Arkansas at Little Rock

A group of researchers from the University of Arkansas at Little Rock have published an article that examined the possible use of online media campaigns orchestrated to influence the 2019 Canadian federal election. The article, “The Role of YouTube during the 2019 Canadian Federal Election: A Multi-Method Analysis of Online Discourse and Information Actors,” was published in the Journal of Future Conflict in September.

Released: 30-Oct-2020 12:00 PM EDT
5 Big Questions on Health Care and COVID-19
University of Virginia Darden School of Business

The coronavirus pandemic has once again thrust the unusual state of American health care into the spotlight. With a presidential election that could have a dramatic impact on the state of health care for millions on 3 November, Professor Vivian Riefberg considers the state of the industry.

Newswise:Video Embedded newswise-expert-panels-on-covid-19-pandemic-notable-excerpts-quotes-and-videos-available
Released: 30-Oct-2020 8:25 AM EDT
Newswise Expert Panels on COVID-19 Pandemic: Notable excerpts, quotes and videos available

Newswise is hosting a series of Expert Panels discussion on unique aspects of the COVID-19 pandemic. This tip sheet includes some notable quotes from the panelists.

Released: 30-Oct-2020 8:20 AM EDT
Newswise Live Event for Nov 2nd, 2PM EDT: the 2020 Presidential Elections

Experts from institutions including George Washington University and Cornell University will participate in an expert panel covering a wide variety of topics on the U.S. Elections, with questions prepared by Newswise editors and submissions from media attendees.

Released: 29-Oct-2020 1:30 PM EDT
Facing up to the reality of politicians’ Instagram posts
University of Georgia

A University of Georgia researcher used computer vision to analyze thousands of images from over 100 Instagram accounts of United States politicians and discovered posts that showed politicians’ faces in nonpolitical settings increased audience engagement over traditional posts such as politicians in professional or political settings.

Newswise: 247321_web.jpg
Released: 29-Oct-2020 11:35 AM EDT
How Twitter takes votes away from Trump but not from Republicans

A popular narrative holds that social media network Twitter influenced the outcome of the 2016 presidential elections by helping Republican candidate Donald Trump spread partisan content and misinformation. In a recent interview with CBS News, Trump himself stated he "would not be here without social media."

Released: 29-Oct-2020 10:05 AM EDT
Rutgers Expert Explains QAnon
Rutgers University-New Brunswick

The author of Conspiracy Panics: Political Rationality and Popular Culture describes QAnon, why it’s well-known and why we should not treat this as a misinformation problem.

Showing results

110 of 5609