This Insight is a guest article by Dr Roderick Mackenzie from Nottingham University. Dr MacKenzie is currently undertaking a Data Science Fellowship with the Parliamentary Office of Science and Technology (POST).
Social media platforms such as Twitter have become an important part of the political conversation. In this article, we use data mining techniques to explore how MPs engage with the public through Twitter with the aim of better understanding the role of social media in their Parliamentary work. The result is a broad picture of how MPs use Twitter to influence and participate in public debate.
Summary of findings
The Guardian is the most cited and shared online source in Westminster, this is followed by the BBC and Facebook. Paywalls seem to affect information uptake.
The majority of MPs are very active on Twitter, tweeting at least once or twice a day.
Only 50% of tweets issued by MPs are original tweets, the rest are retweets.
An MP’s tweet is most likely to be retweeted one hour after publication; after 24 hours it is effectively forgotten – 24 hours is a long time in politics.
The majority of tweets are about party political issues, with health being the most covered policy topic. Groups with particular policy interests will therefore have to fight to find their place in MPs’ Twitter feeds.
For information to be noticed by MPs on Twitter it should already be pre-packaged in a tweet.
Social media as a stream of consciousness
An active social media account can be thought of as a stream of consciousness, providing insight into a person’s interests and political focus. A model of information flow on Twitter between MPs and the public is shown in Figure 1. The dotted box represents all MPs who tweet. Within this box, original tweets are created and existing tweets are retweeted by MPs. In this diagram, tweets that do not originate from MPs enter the box from the left, and original information from newspaper articles and other web sources enter on the right.
By understanding the information flows within this diagram we aim to answer the following questions:
- What sources of information do MPs use to inform debate?
- How reliable are the information sources they use?
- Where should researchers place information so it reaches MPs most effectively?
- What topics do MPs debate on Twitter, and how are these related to policy?
- How does information flow between MPs and the public through Twitter?
- What role does Twitter play in informing political debate on the topics of the day?
Historical tweets from MPs were read from Twitter using the Tweepy library. This data was then analysed using Python and the Natural Language Toolkit to explore trends and relationships in the data. The data shown in this article covers the previous year (September 2017 – September 2018). At the start of this work, a total of 558 MPs were identified as having active Twitter accounts (86% uptake), of which Conservative MPs accounted for 248 accounts (79% uptake) and Labour MPs accounted for 245 accounts (95% uptake). In many of the smaller parties every MP has an active Twitter account.
What sources of information are used by MPs?
If an MP links to a website within a tweet, they draw attention to that source, whether their view of it is positive or negative. Figure 2 plots a histogram of the websites linked to by MPs in their tweets during the last year. The most popular website to link to was Twitter itself, so perhaps the best way to get research and information noticed by an MP on Twitter is to already have it pre-packaged within a tweet. The most frequently cited online source for MPs was the Guardian, closely followed by the BBC. Websites with paywalls tend to come further down the list than those that are freely accessible. Political blogs such as Order Order also tend to come further down the list. Each bar has been colour coded to show the number of times a website was mentioned by MPs in each party. The Guardian and the BBC were cited more often by Labour MPs than Conservative MPs. However gov.uk was cited more by Conservative MPs, perhaps because MPs in the governing party are more likely to link to Government statements and policy. It is also interesting to note that the Times appears to be the most popular Conservative leaning publication.
Facebook, Instagram and YouTube occupy places within the top ten sources, ahead of more traditional media sources. Out of the 1.7 million tweets issued by MPs in the last year, only around 19% link to any other source of information, therefore the majority of tweets can be thought of as comment, reaction or conjecture.
How often do MPs tweet and what are they tweeting?
Figure 3a plots how often MPs tweet. The red bars represent Labour MPs, the blue bars represent Conservative MPs, and the yellow bars represent MPs from the Scottish National Party. The graph shows the majority of MPs tweet between one and ten times a day, although some MPs tweet up to 60 times a day. In general, Labour MPs tweet more than Conservative MPs: this is shown by the long red tail extending to the right of the graph.
Figure 3b plots the number of retweets compared with original tweets from MPs. The majority of MPs combine tweets and retweets. The top five non-MP accounts which were retweeted by MPs were : @UKLabour(13,287), @Conservatives(10,554), @theSNP (5,200), @HouseofCommons (4,729), @labourpress (2,646), @CCHQPress (2,135), and @ParlyApp (2,120). The MP most retweeted by other MPs was the leader of the Labour Party @jeremycorbyn, with 9,538 retweets by MPs, while the leader of the Conservatives and Prime Minister @theresa_may was the second most retweeted among MPs with 7,014 retweets. Overall we found a weak correlation between Twitter usage and political party. However, one key difference is that 25% of Conservative MPs do not mention their party affiliation in their Twitter description, instead identifying themselves as ‘the MP for x constituency’, while for Labour the figure is lower at 7%.
What are MPs using Twitter to talk about?
Each tweet was automatically categorised into topical areas which included health, transport, education, science/environment and party political tweets. The number of tweets in each category over the last 100 days is plotted in Figure 4 (also on a log scale). The number of tweets relating to party political issues outnumbered tweets relating to specific policy topics by an order of magnitude. Policy topics seem to get more or less equal attention, with the exception of health, which sometimes reaches the same level of interest as party politics.
Figure 5a plots the hashtags used by MPs over the last year. Brexit and Prime Minister’s Questions Time rank highly, supporting the results in Figure 4, where party political issues dominate the tweets of MPs. Tweets about health were second. Figure 5b shows the top 20 nouns used by MPs over the last 100 days in their tweets. Once again, general political terms focusing on current events dominate.
How tweets reflect the mood among MPs
Using sentiment analysis, one can classify each tweet as negative or positive. By classifying every tweet issued by MPs one can examine the overall mood of the House. A score of 1.0 means a tweet was positive, a score of -1.0 means a tweet was negative. This was done for the last two years and plotted in Figure 6. Most notably it can be seen that the average mood of the House was around -0.4, i.e. negative. An analysis of the relationship between the two variables during the last year found a weak but statistically significant negative correlation between changes in these variables i.e. that when Conservative MPs became more positive in mood, Labour MPs became slightly less positive, and vice versa.
How long is a tweet relevant for?
One measure of how long a tweet is relevant for is how long MPs retweet it after its initial publication – this is shown in Figure 7. It should be noted that the scales are log scales, so there is an exponential decaying likelihood of a tweet being retweeted. The red line represents the 24 hour point after a tweet is written. After 24 hours, there is a sudden drop in the likelihood of a tweet being retweeted. Thus 24 hours, not a week, is a long time in politics on Twitter.
Most MPs are active Twitter users. Usage ranges from writing original tweets to simply retweeting those of others. The most frequently cited websites are those of newspapers and the BBC. The Guardian outperforms all other online sources, and can therefore be considered the most far-reaching on this measure. Social media platforms are the next most likely websites to be tweeted, followed by institutional websites such as hansard.parliament.uk. The fact that online sources without paywalls were the most frequently cited sources underlines the role of freely available information in digital democratic debate.
Overall it would seem that MPs tweet most about party political issues. However health sometimes competes with political tweets. For the past year the mood of MPs’ tweets was on the whole negative.
Twitter is the most cited website on Twitter by at least an order of magnitude, suggesting that if researchers want their work to play a part in political debate on Twitter, it helps to pre-package it in a tweet first with the aim of attracting MPs’ attention. We also note that given MPs’ focus on party politics, people with particular policy interests may need to look for opportunities to draw attention to their subject.
About the author and fellowship
Dr. Roderick MacKenzie (email@example.com) is an Assistant Professor in the Faculty of Engineering at The University of Nottingham. He is currently undertaking a fellowship with the POST, exploring how data science research methods can help the House of Commons Research Service. POST is an independent research arm of the Houses of Parliament, with the remit of providing independent evidence based advice to the members of the House. POST currently answers requests for information from MPs and updates briefing documents on an ad hoc basis. This small team currently has a limited global understanding of which topics are emerging in the political sphere, this makes it hard to predict demand for work and leads to an inevitable delay between the advice being sought and information provided to MPs. The aim of this fellowship is to apply large-scale data mining techniques to help POST predict which topics are becoming ‘hot’ in the House and thus increase the responsiveness of the services. The results will also be used to help POST better disseminate it’s research to MPs via Twitter. Dr. MacKenzie would like to thank Oliver Hawkins for his support during the fellowship, and EPSRC for funding the fellowship through Nottingham’s Impact Acceleration account.