Data-Driven Politics: Using Networks to Predict Political Outcomes
Did you know network science can be used to predict voting behavior on political issues?
2016 marked a momentous time in the United Kingdom’s political history, most notably with the Brexit referendum voting. Responses to contentious political issues often allow people to be grouped into various opinion camps, but Brexit had divided people across the political spectrum like never seen before. This blurring of party lines signified a shift in the relationship between ideology and opinion and represented an anomaly in the UK’s orderly political system.
Researchers Carla Intal and Taha Yasseri explored the possibility of using network science to predict political outcomes by investigating the voting patterns of British Members of Parliament (MPs). The UK party system is historically known for its discipline, with MPs mostly voting along their party lines. However, Brexit’s shakeup presented an opportunity to reveal the challenging forces exerted on the UK party system.
This was done by analyzing the voting records of each MP in the 57th parliament following the 2016 referendum. The result was a complex network visualization, where each node is an MP, and the edges between them represent quantification of how similarly they voted across legislative issues. MP’s similarity scores could be negative if they tend to vote in opposite directions, which was often the case with MPs of different parties.
Interesting cases were where two MPs of the same party have a large dissimilarity or MPs of different parties have high similarity. These MPs were classified as “rebels” based on the forces of “within-party repulsion” and “cross-party attraction” calculated by the researchers. The result allows us to visualize a network that can identify these rebels with 90% accuracy!
This study represents a fascinating direct application of graph theory and structural balance theory to analyze a real-world scenario. Each MP is a node in a complete graph, where the links between them can be classified as positive or negative. Similar to how we discussed in class examples of positive and negative connections, such as allies and enemies, the links in this network study quantify similarity in voting behavior. This opens up an opportunity to apply structural balance theory to this political network, helping us determine whether or not this complete graph is balanced and allowing us to predict future voting behavior accordingly.
Source: https://blogs.biomedcentral.com/on-physicalsciences/2021/06/23/brexit-shenanigans/