School of Social and Political Science

Using Twitter data to study protests



Someone using Twitter on a phone


Surveying protesters is challenging because protests are often unexpected, but new research has identified better ways to study protests, using Twitter data.

In a new article for the PLOS One journal, Christopher Barrie at the University of Edinburgh’s School of Social and Political Science and Arun Frey at the University of Oxford investigate the possibility of surveying protests using only Twitter data.

Estimating demographics through users’ online activity

Given that protest and dissent are increasingly visible online, the researchers propose a new method for identifying and examining the profile of on-the-ground protest participants.

Instead of only relying on the hashtag used in tweets to generate their sample, the authors use geolocation or image data to locate individual users to the actual routes of protest marches.

The study takes an innovative approach to identifying protestors on the basis of their digital activity, and estimates the demographic and ideological composition of protest crowds on the basis of users’ online activity alone.

The study was based on the case of the 2017 Women’s March in the United States.

Challenges of physically studying protests

The question of who participates in protest is a crucial one for academia, policy, and public debate. Studying protest participation, however, is difficult.

To survey protests, scholars often have to go to protests themselves and survey participants. However, as protests often erupt spontaneously, researchers rarely have the time to prepare survey questionnaires, gain clearance from institutional review boards, and hire interviewers.

Comparing digital and in-person data

Using multidimensional scaling and machine learning techniques, the authors estimate demographic and ideological ideal points on the basis of users’ Twitter profiles, and compare their results to two in-protest surveys of protest participants in Washington, DC.

While the article makes significant advances, the study does however, also point to persisting discrepancies between online and offline samples of protestors.

Urging caution

The researchers do urge caution, considering the amount of information now accessible on sometimes sensitive forms of political action, noting that this data is accessible not only to researchers, but also to governments.

Lead researcher Christopher Barrie said: “This is the first time that researchers have tried to ‘survey’ protestors from afar and the first comparison of online and offline protestor samples. Future work could automate this strategy to generate data on protestors cross-nationally and at scale.

“This research makes a major advance in demonstrating how close we are to being able to accurately build a picture of the demographics and ideological outlook of participants at political events, through what individuals say and do online.”

Read the article

Read: Faces in the crowd: Twitter as alternative to protest surveys

Find out more

Visit Dr Barrie’s SPS profile

Sociology at the School of Social and Political Science

Photo by Claudio Schwarz on Unsplash