Peter Cihon, Taha Yasseri, Scott Hale, and Helen Margetts, University of Oxford
The Internet provides the means to radically alter citizen participation in government. One such means is popular in a number of democracies, including the United Kingdom and United States, where governments have established websites for citizens to petition their representatives online. Petitions are not new, but these platforms—together with social media—offer a means of circulating and tallying signatures like never before. This paper analyzes, for the first time, the relationship between sharing activity on social media and petition signature outcomes. This analysis uses some 1 million tweets over a 20 month time span that shared more than 11 thousand petitions on the UK petition website.
Quantitative analysis of online petition campaigns is a recent focus in the literature. Twitter and other social media platforms offer researchers ‘trace data’ of individuals’ interactions, permitting large-scale analysis after the fact for the first time. Researchers have widely studied these trace data in varied contexts across the world. Indeed, much research has focused on collective action on social media surrounding political events (See, e.g., Cihon and Yasseri forthcoming; Margetts, et al. 2016). Among social media, Twitter is particularly well situated for the study of collective action and political engagement. The Twitter platform facilitates the spread of particular messages through re-tweets, which can cascade through many individual users’ networks (See, e.g., Gonzalez-Bailon 2011). Use of hashtags links conversations across networks, and popular hashtags are further publicized by the platform itself. Moreover, Twitter provides researchers with data on public tweets through its API. This work makes use of that data in conjunction with UK Petition website outcomes.
The UK government established the No. 10 Downing Street petitions website in 2006, and garnered some eight million signatures from five million unique email addresses between its founding and 2010. In 2010, the No. 10 Downing Street website was replaced by the current UK Government and Parliament Petitions website. In 2011, U.S. President Obama launched We the People, a petition platform on the White House website. Structure and procedure for petition websites vary, though the broad concept is as follows. Individuals may create a petition on a particular topic of concern and then make the petition viewable for others to sign. If the petition garners sufficient signatures, then it will receive a government response. In the case the current UK website, only British citizens or residents can write and sign petitions, and petitions with 10,000 signatures will receive a government response and those with 100,000 or more signatures will be considered for debate in Parliament.
Since their launch, government e-petition platforms have been the subject of research. Dumas, et al (2015) examine a small sample of petitions’ signature communities on the We the People website. Hagen, et al (2015) use language processing on the content of We the People petitions to produce topic communities. Hale, et al (2013) analyze signature growth of over 8,000 petitions submitted to the UK website over a two year period. Only 6% of petitions received enough signatures to cross the response threshold, and of those, 43% crossed the threshold on the same day they were posted. The importance of early signatures brings into question the methods of publicizing petitions. As of yet, no research has addressed the crucial relationship between social media and petition signatures; we seek to address the gap in the literature with this paper.
This work is informed by a preliminary analysis conducted on three months of Twitter activity and petition signatures. This analysis followed the performance of 860 UK e-petitions from April 1 to June 1, 2013. During that time, 36,621 users tweeted 83,099 posts that linked to the subject petitions. Regression analysis indicates that number of tweets was positively associated with petition growth, but an even greater predictor of growth was unique users who authored tweets: a 10 percent increase in unique users tweeting about a particular petition was associated with a 14.52 percent increase in signatures, holding number of tweets constant. Yet this regression model does not account for over 40% of variance in the data.
These observations raise the significance of particular sharing activity on Twitter. Using Twitter data, we create two network projections: (1) individual users are linked if both tweeted the same petition and (2) petitions are linked if the same user had tweeted both. Analysis of these networks yield interest communities of users and petitions. Users passionate about animal rights issues and soccer were two particularly tightly knit communities. This analysis too raised the importance of brokers in spanning interest communities; central brokers were mainly individuals, not organizations or Twitter ‘bots’. Brokers were not simply those users who tweeted the most, raising the importance of tweet content. Communities among petitions revealed similarities in theme, with the largest community addressing general governance issues, the second largest addressing animal rights, the third about fiscal issues, and a fourth about the reburial of King Richard III. Petitions that saw the highest growth in signatures tended to be centrally located in the second network projection, indicating that successful petitions are those that enjoy widespread appeal across different issue groups. The centrality of petitions was not a simple function of tweet count, again raising the importance of network sharing activity over simple frequency.
This preliminary analysis informs our current study. New data was collected that spans approximately 20 months, from July 2013 to March 2015. During that time, over 1 million tweets linked to 11,706 petitions. Each tweet is associated with considerable information about the post and its posting user. Tweet content, time posted, whether the tweet was a reply (and to whom), as well as the favorite and retweet count are available in the data. This offers rich insight into the content of the tweet as well as its reception within the user’s network. Posting user information includes all profile information, including images and bio, total number of tweets, number of followers, number of users the account is in turn following, whether or not the account is verified and the date the account was initially created. This information offers deep insight into the beliefs of the person as well as their network. Our analysis will seek, first, to reproduce results found in the preliminary study, second, to quantify the significance of those results against a random network, and, third, to interrogate the results with further data on qualities of individual users and petitions that lead to high signature growth.
Our results will offer the first insight into the important publicity mechanism for government petitions online today: social media. The relationship between network sharing activity on Twitter and petition outcomes will inform the budding e-petition literature that, until now, has largely focused within the petition platform itself.
Cihon, Peter and Taha Yasseri. “A Biased Review of Biases in Twitter Studies on Political Action.” Frontiers in Physics. Research Topic: “At the crossroads: lessons and challenges in Computational Social Science.” Forthcoming.
Dumas, Catherine, Daniel LaManna, Teresa M. Harrison, S.S. Ravi, Loni Hagen,
Christopher Kotfila, and Feng Chen. “E-petitioning as Collective Political Action in We the People.” iConference (2015): 1-20.
González-Bailón, Sandra, Javier Borge-Holthoefer, Alejandro Rivero, and Yamir Moreno. 2011. “The dynamics of protest recruitment through an outline network.” Scientific Reports 1: 197. DOI: 10.1038/srep00197.
Hagen, Loni, Özlem Uzuner, Christopher Kotfila, Teresa M. Harrison, and Dan Lamanna. “Understanding Citizens’ Direct Policy Suggestions to the Federal Government: A Natural Language Processing and Topic Modeling Approach.” 48th Hawaii International Conference on System Sciences (HICSS) (2015): 2134-2143.
Hale, Scott A., Helen Margetts, and Taha Yasseri. “Petition Growth and Success Rates on the UK No. 10 Downing Street Website.” WebSci ’13 Proceedings of the 5th Annual ACM Web Science Conference (2013): 132-8.
Margetts, Helen, Peter John, Scott Hale, and Taha Yasseri. Political Turbulence: How Social Media Shape Collective Action. Princeton, N.J.: Princeton University Press (2016).
 The king’s bones were found in a parking lot shortly before the beginning of the sample time period: http://www.theguardian.com/science/2013/feb/04/richard-iii-dna-bones-king