Using Twitter to Estimate Crowd Size at the National Mall

Suzy Moat and Tobias Preis, scientists from the Data Science Lab at Warwick Business School and Alan Turing Institute Faculty Fellows have presented a new timely application on how data from Twitter and mobile phone networks may inform about the number of attendees recorded in certain restricted areas.

After President Donald Trump’s inauguration ceremony, last Friday and the subsequent Women’s March a day later, questions have arisen over which event attracted the larger crowd to the National Mall in Washington D.C.

To answer this question, scientists analysed the number of geotagged tweets sent by Twitter users from the National Mall on both days. They found that over 2.5 times as many people tweeted during the Women’s March than during Donald Trump’s inauguration.

TrumpCrowdTwitter_1st half

Tobias Preis, Associate Professor of Behavioural Science, Finance and Co-Director of the Data Science Lab and Faculty Fellow of the Alan Turing Institute said, “We were interested to see whether we could build on our earlier findings that Twitter usage can be related to the size of a crowd. We analysed the tweets we had collected from the National Mall in Washington D. C. on Friday and Saturday. We found that the maximum hourly Twitter user count during the Women’s March was around 2.5 times higher.”

In their original paper, Quantifying crowd size with mobile phone and Twitter data, published in Royal Society Open Science, Federico Botta, Research Fellow in Data Science at Warwick Business School, Moat and Preis analysed two months of both Twitter data and mobile phone data from Milan, from November 1 to December 31, 2013. They found that the size of spikes in mobile phone usage and Twitter usage during the matches strongly correlated with the official attendance figures for each match. In this study, the scientists exploited the access they had to data recorded directly by the mobile phone network provider and found that using data on mobile phone internet activity, they could generate estimates of the number of attendees which fell within 13% of the reported value.

TrumpCrowdTwitter_2nd half

Suzy Moat, Associate Professor of Behavioural Science, Co-Director of the Data Science Lab and Faculty Fellow of the Alan Turing Institute said, “There are various challenges in drawing on social media and mobile phone datasets to estimate crowd size, including questions about demographics and network accessibility. Nevertheless, our work to date suggests that these datasets can give us useful insights into the size of a crowd.” 

For more information please visit the Warwick Business School webpage and contact Martin Boonham, in the Warwick Business School PR and Press team at or on +44 (0)2476528307