Nick Anstead, Richard Stupart, Damian Tambini and Joao Vieira-Magalhaes
Recent elections in various parts of the world have seen much discussion of the increasing role played by Facebook as a means of political communication. These debates have ranged widely, understanding Facebook as a platform for citizen self-organisation, news-consumption and deliberation. More recently, there has been an increasing focus on Facebook as an advertising platform.
It is in this latter guise that Facebook has caused most concern. While these anxieties have taken a number of forms - including discussion of Fake News, "demobilisation" adverts, hyper-targeting which allows political campaigns to offer different messages to different cohorts of voters, and foreign interference in elections - the root cause of all these fears is the relative non-transparency of the Facebook platform, which makes it hard for either electoral regulators or academics to understand the precise form that advertising on it takes. It is for this reason that we still have very little empirical data that allows us to understand Facebook as an electoral advertising platform, despite the wide-ranging discussion is provokes.
This paper offers a new methodology for tackling this empirical shortfall. Working with the social enterprise Who Targets Me, we employed a Chrome browser plugin installed by 11,421 users in England, Scotland and Wales to gather data on all the Facebook adverts they were exposed to. In total this allowed us to collect 783 unique adverts purchased by British political parties and identify the 16,109 times they appeared on our users' timelines.
Our paper is constructed in two parts. First, we analyse the dataset gathered during the UK election campaign of 2017, in order to understand the content and targeting of Facebook adverts harvested through the Who Targets Me plug-in. We do this in two ways. First, we examine the content of adverts, coding for topic, negativity, and leader-focus. Diverging from much of the contemporary discussion on Facebook advertising, our analysis of the dataset suggests that adverts are dominated by nationally defined party narratives, rather than narrowly defined targeted messages. Second, we look at where adverts are targeted. Here, the differences between parties suggest a complex interplay between political strategy, internal party politics and the incentive structure created by the UK's first-past-the-post election system.
Second, we reflect on the methodological limitations of our approach and seek to offer solutions as to how the approach might be improved in subsequent studies. Ultimately, we argue, while solutions like Who Targets Me can be improved (through, for example, recruiting panels who more accurately represent the broad electorate), the major challenge remains the non-transparency of the Facebook platform, which needs to be more closely integrated into the over-arching electoral regulatory logic of individual countries.
