You are here

Tracking the 2012 U.S. Presidential Election on Twitter

The next speaker at Social Media and Society is Christopher Mascaro, whose interest is in 'big data' on political communication online. Political discourse studies have traditionally been restrained by geographic and social access, and 'big data' from online activities can overcome some of these barriers; it also introduces some new limitations that must be considered, however.

Christopher's focus is on the 2012 U.S. Presidential Election: how do networks form around the political issues being discussed on Twitter during the campaign? The dataset was generated using TwitterZombie running on Amazon Web Services, tracking an extensible range of hashtags, user handles, and keywords to construct a broad dataset that extends well beyond single hashtags only. Some of these were event-specific; some of them were 'official' or commercially promoted hashtags, too. Data were collected over 85 days from 20 August to 13 November 2012. This captured some 52 million unique tweets.

It is crucial that such gathering methods for longitudinal collection are documented fully, for review and future use. The infrastructure needs to be flexible and extensible – and the complex and diverse dataset is valuable as it provides insights into a broader range of activities than simple hashtag datasets do. The data contain a somewhat lower percentage of hashtagged tweets as a result, which is important and desirable and documents that hashtags cover only a part of the full discussion on Twitter.

The context shifts markedly during acute events, such as televised debates: hashtag use often spikes, but the hashtags are barely used outside of the debate timeframe at all. Hashtag co-occurrence can also point to the injection of internal issues (e.g. political crises and natural disasters). A tracking of names and mentions also shows that Vice-Presidential candidates are highly inconsequential, outside of their own televised debate.

What still needs to be theorised more is whether people are actually taking an interest, or whether there is just a low-level conversation about the election. Technology might serve to fragment conversations here, and Twitter remains a one-way broadcast medium for candidates for much of the election period. Specific tracking approaches can also affect the shape of the data: people frequently @mentioned @BarackObama, for instance, but largely preferred to refer to Mitt Romney by name rather than Twitter handle.