Authors:
Iain Weaver | University of Exeter | United Kingdom
Hywel Williams | University of Exeter | United Kingdom
Elected politicians do not represent their constituencies in a social vacuum, but inevitably and necessarily form social connections with their colleagues. It is therefore important to describe the social networks of politicians, in order to understand how network structures and processes might affect the actions taken on behalf of the electorate.
This paper uses time-resolved interactions between UK politicians in social media to analyse their interactions during an eventful period spanning the 2015 General Election and the 2016 Referendum on membership of the EU (Brexit). We develop a novel ‘multiplex community affiliation clustering’ (MCAC) method to track the evolution of community structure amongst UK politicians (Members of Parliament, MPs, and British Members of the European Parliament, MEPs). To ensure sufficient data for creation of robust networks, we augment the direct first-order interactions (where one politician retweets another) with indirect second-order interactions (where a two-step retweet path connects two politicians via a non-politician intermediary). This extension captures the wider UK political landscape, including partisan and non-partisan media outlets, journalists, party members, and the politically engaged public. Social networks of politicians derived from interactions in social media show coherent communities with strong linkage within political parties. Comparison of networks formed from direct retweet interactions between politicians and the extended networks including two-step paths via a non-politician intermediary shows that both are qualitatively similar.
Despite the inherent dynamism of social media, network structure typically falls into one of four distinct network states, while the topics of discussion typically fall into one of six distinct content states. Both network states and content states are each strikingly persistent and recurrent over time, and reflect ongoing political events and debates. For example, the politically divisive referendum on the UK’s memberships of the EU produced a network state which dominated from announcement of the referendum date in mid-February to the end of June when the referendum took place. This period saw MEPs affiliated with the United Kingdom Independence Party, who largely championed the vote to withdraw from the EU, forming consistent communities with the majority of MPs from other parties who expressed support for Britain’s exit from the EU. Politicians campaigning to Remain were well-linked, including an unusual linkage between left-wing Labour MEPs and right-wing Conservative politicians.
The MCAC network clustering method alone recovered several key features of UK politics during the study period. Firstly, the temporal sequence of weekly network snapshots was recovered based on their cluster similarity, without any temporal information being provided to the clustering method. Secondly, the clustering of individual politicians was able to accurately identify the ideological position of MPs and MEPs regarding the EU referendum. This ability of social network analysis to predict the ideological positions of individual politicians suggests a route towards predictive political science.
Reference: Weaver, Williams (H), Cioroianu, Coan, Williams (M), Banducci (submitted) Evolving affiliations amongst UK politicians on social media. In review.