One of the reasons for Labour’s surprising performance in this year’s general election has been said to be voter turnout – which, at 68.7%, reached its highest level since 1997. An analysis by LSE researchers Damian TambiniNick Anstead and João Carlos Magalhães suggests that a shift in the party’s Facebook advertising strategy in the last 48  hours of the campaign might be one of the factors involved in this phenomenon. In this period, Labour’s campaign on the platform not only mushroomed, but also zoomed in one main goal: prompting users to get out and vote.

This post is the fourth in a series that is examining data collected as part of a joint project recently launched by the LSE Media Policy Project and the “Who Targets Me” initiative. Previous analyses have considered how Liberal Democrats, Labour and Conservatives were using Facebook to court voters.

The data (extracted by a browser extension from the news feed of more than 11,000 initiative participants) give an insight into pieces of political advertising which were designed to be viewed by individual users and are not systematically disclosed by political parties. While huge, this dataset must be approached with caution, as it is uncertain whether it is a representative sample of all Facebook ads paid by political parties in this election.

The present analysis focused on 5,800 exposures to Labour ads (the party’s ads impressions) from May 3 to June 8 2017, with a focus on the last week of the campaign. It seems to indicate two patterns.

  • First, Labour increased its investment in Facebook advertising in the last three days of the campaign, easily outdoing the other two main parties (Lib Dems and Conservatives).
  • Second, this investment largely aimed at prompting users to vote. Given anecdotal evidence that Labour won the ‘organic’ battle (shares and likes) as well as this paid political advertising, this paid political advertising is likely to have enjoyed a significant online multiplier effect.

The advertisements circulated on Facebook by the party between June 6 and 8 (3,036) total 52% of all Labour ad exposures in our sample. While Conservatives also boosted their campaign (and Lib Dems deflated theirs), it barely compares to what Labour did, as the chart below shows.

The increase in the number of ads must take into account that, from the end of May to election day, the number of people using the “Who Targets Me” browser extension also increased by around 50%. This fact alone, however, doesn’t explain what happened. First, the evolution of ad impressions was not linear – with one day, June 4, having zero ad impressions. Second, in the same period, impressions increased by more than 1,000%.

An exploratory manual content analysis of Labour’s ads impressions indicates that as interesting as the size of this surge, is its nature.

From May to the first few days of June, Labour’s Facebook campaign focused on avoiding Brexit, exposing the alleged decay of UK’s public services and employing this deterioration to frame opponents (mostly Conservatives) as elitists (see our first post on Labour’s campaign for more detail).

As we coded each ad impression according to its main topic and assessed the evolution of topics’ prevalence over time, it became apparent that the last two days of their campaign saw a radical shift towards one main message: the elections itself. The following chart (which considers only June; there were virtually no mentions of the election itself in May) makes it easier to visualize how this movement occurred.

Until June 5, “elections” was a non-existent topic. But, beginning on June 6, it went on to become almost the only topic of the party’s Facebook campaign. In a sense, of course, all ads are at least indirectly about the election, and a great number of them would ask users to “vote Labour”. The difference with these ads is that they not only urged users to vote, but also gave them temporal details about the election itself – day of the month, day of the week and the period of the day polling stations would be open. The most circulated message in our sample, from June 8, for instance, reads:

“Today’s the day. You have the chance to change the future of this country – to one that works for the many, not just a privileged few. Find your polling station and vote Labour before 10pm.”

In short, we are not claiming that these messages indeed had a major effect on voter turnout – which is a complex matter, influenced by multiple variables. What our analysis does point to is the possibility that these messages might have had some form of impact – or, at the very least, that the Labour Party thought that it could have. If that is the case, they would be backed not only by common sense, but also by social science. In a study of 61 million Facebook users published in 2012, researchers showed that voters who were shown “mobilization messages” on Facebook were slightly more likely to vote in the 2010 US Congressional elections.

An important point to note is that our dataset captures just a fraction of all the Facebook messages about the election, as the browser extension ignores the so-called “organic” messages, i.e., messages which are not paid for to be circulated on the social media platform.

The importance of older voters

We also considered a key variable regarding the performance of Labour in this campaign: age. There is a widespread belief that Jeremy Corbyn’s success is linked to young people’ support, and data have suggested the rise of a sharp political divide based on age in the UK.

In a somewhat surprising result, our evidence indicates that the party made, particularly in the last days of the campaign, an effort to court older voters – maybe due to their confidence on how cemented youngsters’ support was.

A clear indicator of this manoeuvre is the number of ads talking about pensions and the so-called “dementia tax”. Almost 90% (364) of the ad impressions in our sample focusing on these topics were seen by users in June, of which roughly half were seen in the last two days of the election campaign. These ads attacked Conservatives’ supposed lack of concern with “pensioners”, and portrayed Tories as not deserving their trust.

Our data show a remarkable point indicative of targeting based on age: only a tiny fraction (2%) of the users who were exposed to ads about these topics in May and June were younger than 39 years old.

Labour did try to court younger voters, but in less intensely targeted ways. One tactic was through ads about Corbyn’s promise to abolish tuition fees. There were 149 ads impressions on this topic in June, which account, however, for less than half of the total of 305 in the sample.

In addition, while there’s some age segmentation, it is not so stark as with the impressions about “dementia tax” and pensions. A fifth of the users who were shown tuition fees ads in May and June were above 40 years old. On the other hand, this is not exactly unexpected, as tuition fees are also a concern of parents.

While again, the representativeness of our sample is still to be determined, the average age of users of the “Who Targets Me” browser extension (38.3 years old) is not so distant from the UK’s (40 years old).

A less negative campaign?

A last item we considered is the number of attacks in the last days of the campaign. What seems to have happened is a decrease in circulation of ads attacking opponents, and the emergence, especially on the last three days, of ads that did not attack anyone – they were mostly ads prompting people to vote. The decrease seems to have affected in particular the Lib Dems, but also the Tories. In an additional sign of a less negative campaign, Labour appears to have ceased, in the last two days before the election, to circulate ads which criticized Theresa May directly. Her name disappeared from Labour’s Facebook advertising on the last 48 hours of the campaign, according to our sample.

These findings are, to repeat, preliminary and have not gone through peer review. In addition, the content analysis results still are to be confirmed by inter-coder reliability tests. The authors are currently working on a longer analysis which will explore these themes further.

This post gives the views of the authors and does not represent the position of the LSE Media Policy Project blog, nor of the London School of Economics and Political Science.