Public Mood in UK Twitter during the Brexit Referendum
Public mood has been shown to play a role in how individuals process information and form political opinions, with affective experiences influencing political reasoning. However, assessing public mood on a large scale using traditional methods can be prone to error from a range of sources including the wording of questions, individuals using their momentary affective state to judge their overall state and even the characteristics and background of the interviewer.
To overcome many of these issues, use of social media as an alternative method for assessing the public mood has been demonstrated in previous studies. The causes behind changes in public mood can be difficult to explicitly capture, due in part to its diffuse nature.
Previous work has gone some way in explaining the changes in public mood as measured using social media, finding circadian and seasonal patterns of affect, along with public mood changing in response to specific real-world events.
In this study, we are interested in investigating changes in public mood through analysis of simultaneous and sudden changes in five affect components, and the events that triggered them. We make use of the fast group LARS algorithm, that detects shared change-points across several time series’ at once. These specific points in time, where many time series’ change together, have the potential of signalling specific realworld events that explain the variation in the public mood.
We find that there are three key times in the period leading up to and including the European Union (EU) referendum, coinciding with the football violence in Marseille between English and Russian fans and the Orlando nightclub shooting, the murder of Labour MP Jo Cox, and the results of the EU referendum itself. In each of these cases, the public mood is characterised by a decrease in positive affect, and an increase in negative affect, anger, anxiety and sadness, with the reaction corresponding to when the outcome of the referendum result became clear causing the largest negative change in public mood.
Furthermore, we find that analysing the affect components in different geographical regions of the United Kingdom shows a robust signal, with each region following a very similar trajectory over the period, and that the hour by hour evolution of public mood in the 48 hours starting on the day of the referendum significantly correlates with the GBP/EUR exchange rate for the five affect components used in this study.
Paper
- Change-point Analysis of the Public Mood in UK Twitter during the Brexit Referendum by Thomas Lansdall-Welfare, Fabon Dzogang, and Nello Cristianini in the Data Mining in Politics Workshop of the IEEE International Conference on Data Mining (ICDM) 2016.