Tuesday, September 04, 2018

Facebook Political Advertising in the 2017 UK Election: New Research

I was one of the people that took part in a research project Who Targets Me in 2017. I voluntarily installed a browser plug-in that allowed the project to track all adverts, including political adverts, that appeared in my Facebook feed. I was therefore very interested to read a first draft of the research findings from an LSE team led by Nick Anstead. The data set has limitations but the research provides some insights into UK political advertising on Facebook including:
  • the topics and messages of the adverts
  • the degree of focus on the leaders of the parties
  • the level of negativity in the adverts
  • the targeting of specific constituencies
  • how they used Facebook adverts to mobilise supporters

My main takeaways from the initial research are set out below.

Topic Focus of Facebook Adverts

The political parties focused on quite different issues in their Facebook adverts as we can see from the chart below. 65.5% of Conservative adverts focused on Brexit whereas none of the Labour adverts mentioned Brexit. This seems logical to me as Labour were trying to face both ways on Brexit. There appears to be no evidence they targeted leave voters with one message and Remain voters with another message. 

Topics of Facebook adverts for each party (n=754)

Focus on Political Leaders in Adverts

The Conservative Facebook adverts focused on Theresa May heavily (34.5% mentioned her explicitly). By contrast, Jeremy Corbyn was not mentioned in a single Labour Party Facebook advert. The Conservative position makes sense as the Conservatives designed their campaign (unwisely) around May as providing strong and stable leadership. Labour's decision not to feature Corbyn may reflect his personal desire not to personalise issues but also possibly a sense early on that Corbyn may not have been the electoral asset he became. 

Personality/leader focus of Facebook adverts for each party (n=754)

Negativity of Facebook Political Advertising

The researchers found the proportion of negative messages seems comparable to previous elections and campaigns on other platforms. That said, the majority of adverts placed by the four largest political parties were negative (which the researchers defined as naming a specific opponent politician or party). Overall, Labour had the highest proportion of negative adverts (64.1%), followed by the Liberal Democrats (61.6%). Research into UK parliamentary election broadcasts, has found the proportion of negative content to be somewhere around 40 to 50 percent (Walter, 2014: 52).

The Greens had by far the lowest proportion of negative adverts but this may due to their electoral position and desire to promote their policies more generally.

Proportion of negative adverts for each party (n=754)

Constituency Targeting of Facebook Adverts

The Conservatives were most likely to target traditional marginal seats but the researchers found that the Brexit referendum defined constituency targeting. The Conservatives and Labour focused on constituencies that voted to leave the EU. In Labour’s case, every single constituency in their top ten voted Leave. In the Conservatives case, eight of the top ten constituencies voted Leave. In contrast, the Liberal Democrats targeted areas that had supported Remain in the referendum to try to mobilise these voters.

The research suggests that Labour were more defensive, with four of their five most targeted constituencies represented by a Labour MP.  By contrast, Conservatives five most targeted seats were represented by Labour MPs. This makes sense in the context of the election as the Conservatives called it when they were well ahead in the opinion polls and anticipated taking Labour seats.

Using Facebook Adverts to Mobilise Supporters

It has previously been suggested that Facebook adverts are not that effective at persuading someone to change their mind but are effective in mobilising supporters. It was interesting therefore to see how the parties used adverts to mobilise their voters. The Conservative Party did not attempt to mobilise their supporters to do anything other than vote. By contrast, 19.8 percent of Liberal Democrat adverts ask users to sign a petition, presumably with the aim of getting details to target them more effectively.  32.1 percent of Green Party adverts asked people to donate to the party and to share their posts on Facebook. Some adverts were used to try to suppress voters in particular areas with messages such as a vote for the Liberal Democrats is a wasted vote.

Data Limitations

11,421 volunteers installed the Who Targets Me browser plug-in during the 2017 UK General Election campaign. This allowed the project to identify 783 unique Facebook political adverts that collectively appeared 16,109 times in users’ timelines.

This project data does have a number of significant limitations. The users were self-selecting such as myself. They were about 80% men with an average age around 35 (I would have increased the average age significantly). Participation varied by constituency, with higher numbers in urban, university areas. Data was only collected from Chrome desktop browsers so the data missed all adverts on mobiles and other desktop browsers. Thus the project was working with a very limited dataset. The dataset for example, did not include a single advert from UKIP which might reflect the nature of the audience that participated.

Access to social media data is an issue for political scientists. Facebook removed the free Pages API earlier this year and researchers need to apply for approval or purchase data from social media providers. Social media adverts are less transparent than traditional adverts in that they are generally only visible to the people targeted, which is why projects such as Who Targets Me are important.  Facebook did recently release political adverts used in the Brexit campaign to a parliamentary select committee.


Overall the study found that Facebook political adverts:
  • adhere closely to national campaign narratives, messages do not appear to vary significantly across the audience
  • do not appear to be much more negative than those on other traditional communication channels
  • were used by some parties to mobilise supporter actions such as signing petitions and joining campaigns
  • were targeted to specific constituencies but not necessarily marginal constituencies

These findings would suggest that Facebook targeting is not as sophisticated as is often discussed. However, it could be that the nature of the election, which was called at short notice, did not provide enough time for more sophisticated targeting. Clearly Facebook political advertising is an area which warrants a lot more research and larger, more representative datasets.