Social media and the 2017 general election

Daily user interactions on Facebook pages of Labour Party and Conservative Party over the course of the campaign.

While much has been written about the role of social media in last week’s general election, to understand the part it played we must look more closely at the digital traces the campaigns left behind. In this post I analyze data on the Facebook activities of the two main parties — the Conservative Party and the Labour Party — to show how they used social media and to see how the campaigns compared.

I address a number of questions: Which party’s posts were more popular with social media users? Which reached more users? And which attracted more new users over the course of the campaign? Did the Conservatives run a more negative online campaign, as has been reported, and did social media users react differently to their messaging?

I find that Labour’s campaign seemed to be more successful in attracting user engagement and support, although the Conservatives’ posts received more attention on average. Analysis of the content and sentiment of the posts shows that around one-third of each party’s output focused on attacking its rival and that these attacks tended to be reciprocal. Perhaps most surprisingly, I find large differences in the way people reacted to the material shared by the two parties.

Social media and the 2017 election

In the past few years, British political parties have started putting more resources into online campaigning. Social media was first adopted by the major parties around the 2010 General Election but does not seem to have been used seriously by them until the 2015 election, when the Tories reportedly spent £1.2m on Facebook advertising in campaign to get David Cameron re-elected, more than seventy times the amount spent by Labour. It played a large role in the Brexit referendum, with campaign groups on either side having a large social media presence. Both Leave.EU — an unofficial campaign group funded by millionaire Arron Banks that ran in parallel to the official Vote Leave campaign — and the official remain group, Britain Stronger In Europe— which has re-branded itself as Open Britain — continue to operate on Facebook, with 850,000 and 570,000 Facebook supporters respectively. As Theresa May announced the snap election in April, these groups joined a cacophony of others as the election campaign took to social media. The Conservative Party spent over £1m on Facebook ads, many of which purportedly focused on attacking the Labour candidate, Jeremy Corbyn, while the Labour Party and other supporting groups produced slick-videos that were shared widely by young voters, reportedly reaching 9.8m Facebook users (22% of the total UK user base).

To understand these campaigns and how they compared, I used the Facebook Graph API to extract the content from the public pages of the two major parties, Labour and the Conservatives. To keep this analysis focused, I ignored the pages for other parties, the personal pages of politicians, other political groups, and the media, and only look at the period from the election announcement on April 18th through to election day on June 8th.

Which campaign was more popular?

It is difficult to precisely determine the popularity of the campaigns on social media (at least without access to the internal information viewable by page owners) but there are a few ways we can get a rough sense of how the campaigns compared. First, we can see how active the parties are by counting the number of posts they made. These posts include campaign materials like videos and images, as well as news articles and posts from other groups that they shared. Overall, we see that Labour was much more active, posting 545 times over the 52 days of campaigning, compared to 161 posts by the Tories. Labour continually ramped up its digital output as the election approached, although both parties halted their campaigns following the terrorist attacks in Manchester and London.

To assess how popular these posts were, we can look at how users interacted with them. The graphs below shows the average number of “likes” and “shares” that posts each day received. While Labour put out more content, the posts made by the Conservatives got significantly more likes, averaging around 9,000 per post compared to the approximately 3,000 per post that Labour received. If we look at shares, however — when a user shares the post on their own Facebook page and in their friends’ news-feeds — we see a different picture. Posts by Labour tended to get shared more for the majority of the campaign period, except for a video the Conservatives released attacking Corbyn, that was shared by over 90,000 people and viewed by over 8 million.

Comparison of the mean number of likes and shares per post.

To get a sense of the overall volume of interactions over the course of the campaign, we can look at the cumulative number of likes the posts received, shown in the graph below. For both parties the slopes get steeper the closer they get to the election, indicating increases in user activity. We see that the Labour party received many more engagements from Facebook users, almost 2 million likes by election day. The graph at the top of the article further illustrates how Labour’s posts also tended to get shared and commented on more than those of their counterparts. Although these numbers appear to be quite large, it is worth noting that the number of “impressions” made— people who saw material from either party — is likely an order of magnitude bigger, as indicated by the numbers of views some of their videos have received. Unfortunately these data are not publicly available.

A problem with the analysis so far is that we don’t know whether the same users are repeatedly interacting with the posts made by each party and inflating their popularity. The next graph shows the cumulative number of unique users who liked each party’s posts over the course of the campaign. Both groups started out with similar trajectories but there is a jump of almost 20,000 new likes on the Labour party’s posts a few days after the time their manifesto was leaked. Looking more closely it appears that they shared ten different videos on May 15th, including a profile of Corbyn directed by Ken Loach and explanations of their tax policies and vision for the NHS.

In sum, it appears that Labour was much more active online, tended to get more engagements, and attracted more unique users than the Conservatives — consistent with reporting that Labour membership has surged in recent weeks — , although the posts published by the latter tended to get more likes on average.

Who ran a more negative campaign?

The Conservatives have been accused of running a negative campaign, with more attention devoted to attacking their rivals than promoting their own policies. To assess this claim, I measure the extent to which each party’s materials mentioned their rivals and the overall sentiment of their posts. The graph below shows the proportion of posts made by each party that explicitly mentioned the other party or its leader (e.g. Conservatives using the words “Labour” or “Corbyn”). These attacks fluctuated over the seven-and-a-half week period, making up approximately one third of the posts made by each party. Based on this cursory analysis, it appears that the Conservatives did not engage in this tactic any more than Labour.

To gauge the overall mood I used a sentiment analysis algorithm to measure the emotional valence of the text used in the posts and took the average value each day. Scores below zero (below the dashed line in the graph) indicate more negative content, while values above zero and closer to one indicate more positive language.

Note that the higher variance of the trend for the Conservatives is likely an artefact of the relative sparsity of the data.

The graph shows that the sentiment of both parties tended to track fairly closely over the period studied, notwithstanding some fluctuations. The campaign seems to have started of relatively positively, but we see contrasting sentiments in late April when Labour began to attack the Conservatives’ manifesto and Theresa May’s ill-conceived promise to bring back fox hunting. In the first week of May, the Tories began posting more negative content, claiming that a Labour government would cause “chaos” and “wreck the economy” and a note by Defence Secretary Michael Fallon associating Corbyn with terrorism. These attacks were met with more negative sentiment from Labour, who focused on cuts to the NHS and police under the Conservative government. The tone seemed to gradually get more positive, bar the periods where campaigning paused (I set the value to zero, or neutral, when there were no posts). In sum, these two graphs show that both parties spent a lot of time attacking each other (in absolute terms Labour actually put out more content mentioning the Tories) and that the sentiment of their language tended to track each other fairly closely. While inconclusive, this appears to contradict the narrative that the Conservatives ran a more vitriolic campaign, at least in their social media postings.

How did users react?

Although there are many ways we could try to understand how users reacted to these posts, the reactions feature that Facebook implemented last spring provides a simple way to get a sense of their responses. It is worth noting that reactions should not necessarily be taken to indicate people’s attitudes towards the topic of the post or the party making it — for example a user might react angrily in agreement when Labour criticizes the Tories and also react angrily in opposition to a video posted by the Tories — but they nonetheless capture some information of interest. Facebook currently offers users six core “reactions”. Here I focus on five of them, ignoring likes, which I have already discussed above.

Facebook reactions: Like, Love, Haha, Wow, Sad, and Angry.

The graphs below shows the proportion of reactions that fell into each category for each party. We see very different patterns across the two parties. The most frequent reaction to the Conservatives posts is “haha”. Without closer analysis, it is difficult to tell whether people find their content more amusing, whether Tories are more easily amused, or whether these are people laughing at the Tories. I was unable to tell from scrolling through the Party’s page. They also seem to use angry reactions more than people who interact with Labour.

The most frequent reaction to posts on the Conservative Party page seems to be “HAHA”, while people appear to “LOVE” posts by the Labour Party.

By far the most common reaction to Labour’s posts is “love”. The relative proportion of love also seems fairly constant over the course of the campaign; in contrast it appears to decline for the Tories. The stark difference between these two distributions is quite surprising and highlights the importance of considering the relationship between emotions and politics. These findings offer some support claims that Labour’s positive ads and message of unity may have been a more successful strategy than attacking their opponents, a rebuke for the politics of anger and fear that have dominated the headlines over the past year.

Conclusions

These data have helped to shed light on some of the dynamics of the election campaign on social media. While the picture presented is at a high-level, abstracting from the nuances of the rich and variegated media shared and people’s responses to it, it illustrates some of the similarities and differences between the campaigns. The Labour Party was far more active on social media and managed to engage with many more voters over the duration of the campaign. While the both parties attacked each other, and the Tories undoubtedly attempted to smear Corbyn, the difference between their tactics, at least online, appears to have been over-stated. Most promisingly, the reactions of social media users suggest that a more positive politics may be possible, as love and laughter prevailed over sadness and anger, despite the uncertainties of Brexit and the vile terrorist attacks.

Thanks for reading! I’d appreciate your feedback and thoughts on this.
Tom Davidson

Find me @thomasrdavidson & t-davidson.github.io

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.