A breakdown of the 2019 General Election targeted advertising campaign across all platforms – The election saw the parties utilise an array of advertising tools, 2019 was not just a Facebook election as in 2017. Instead, Google was seen on equal footing with Facebook by the Conservatives and of importance by Labour. The social networks were however not used equally be each party, there were major nuances between the parties use of the tools.

Summary

The election saw the parties utilise an array of advertising tools, 2019 was not just a Facebook election as in 2017. Instead, Google was seen on equal footing with Facebook by the Conservatives and of importance by Labour. The social networks were however not used equally be each party, there were major nuances between the parties use of the tools.

The Conservatives effective campaign was in sharp contrast to the Labour campaign. The Conservative Party’s simple messages, use of colour and humour was mixed with negative personalised attack advertising. The party knew their audience, where they were, how to reach them and how to speak to them. In contrast, Labour tried to recreate the organic success of 2017 but failed, content was lacklustre with many ads used to boost traditional organic content. Labour tried to activate a similar group to 2017 but this clearly didn’t occur. On Facebook, the Conservatives were also helped by the Brexit Party’s campaign whilst Labour was hindered by the Liberal Democrats.

Overall the 2019 targeted campaign was enormous, with massive spend and the rise of other social networks. Facebook no longer has total supremacy in UK elections instead audiences are more fractured leading to parties needing to be smarter in how they can reach and influence voters. Across the parties an estimated £8,553,343 was spent on the three platforms with spending seen especially on Facebook and Google, it is important to note that this figure is achieved from using the mean between low/high estimates the networks provide. There is a large margin of error caused by the platforms ineffective and weak tools.

The campaign was however much more than just the political parties. From the 1^st to the 12^th December we categorised 2481 political pages that sent adverts (excluding non-party-political pages such as those that promote voting). Overall 29153 adverts were sent from just 18 party political pages across Facebook/Instagram, Google/YouTube and Snapchat. From these 18 party pages 608,409,178 impressions were achieved across the three platforms. A huge number of actors are part of a system that is not capable of being held to account. Third party pages (that is non-party party-political actors or pressure groups) made up around a 3^rd of the spend and reach seen via all pages. Third party actors are of major importance, little is done to hold them to account with there being a serious lack of transparency. Overall 135,881,994 impressions were seen from just the top 19 third party Facebook pages via an estimated spend of £1,728,835. The biggest group in terms of spend and reach of these third party pages were anti-Conservative anti-Brexit pages while pro-Conservative/Brexit pages were much less active.

The problems

1. Current access to data is too obtuse and limits the media, academia and third sectors abilities to hold the social networks and advertisers to account. The limited tools for examination are too simplistic, they feature inaccurate spending levels, reach and targeting data that means political parties and third parties are given relative free reign in the use of targeting.
2. The scale of advertising and ease to advertise from untransparent third party actors has led to a huge scale of advertising especially via Facebook. These new threats from obscure groups are too numerous and clandestine to be held to account effectively during an election.
3. Some parties (for example the Liberal Democrats) engage in huge numbers of advertisements using high frequency and low spend, the huge number of adverts makes it impossible to be effectively held to account.
4. Election day and the days before saw many third party pages suddenly appear, actors shouldn’t just be able to pop up and advertise.

The UK 2019 General Election – A unique social media landscape

On the 29^th October 2019, Parliament voted for a general election to take place on December 12^th, the first December election since 1923. A near century of history existed between these two dates, with this not the only century wide shift to occur, as on election night we saw Labour’s share of seats in Parliament fall to the lowest level since 1935 and the largest Conservative majority since 1987. The results of the election shocked many pundits and commentators as polling was correct and Twitter wrong. Labour had fallen back, red wall smashed, Corbynmania absent. The Liberal Democrats had failed to make an impact piling up worthless votes in remain seats while losing their leader. All the while the Conservatives remained steady, boosted by The Brexit Party standing down and hurting Labour. Given the grand scale of the shifts, the tides of history appeared to turn on a wide axis.

This election saw the parties engage in advertising campaigns with nuance and skill. Although advertising has remained a constant for the last century, recently the advertising and campaigning landscape has changed rapidly due to social media and microtargeting. The parties have far more tools at their disposal than in years past. The 2019 election saw a very different landscape to the 2010 General Election, let alone 1987 or 1935. This is because over the last decade, we have seen the rise of social media and targeted advertising as a phenomenon that now lies at the very heart of election campaigning. The centrality of big data, microtargeted precision and a lack of transparency has meant it has become harder to hold the parties to account while they have gained more powerful campaign tools.

The black box of social media-based campaigning has through self-regulation in the wake of scandal, seen small changes made by the social media giants over the last two years. This meant that finally researchers were slightly better prepared to understand this election and better able to examine targeted advertising. This report uses all the sources available, examining what occurred during the 2019 General Election and presenting research on what should change to improve the situation. The UK General Election is an effective case study for the West as we saw rules being bent, new forces come into play, the radical scale of targeted campaigning and a fragmented social media landscape. These are key lessons for other nations to appreciate within their targeted and social media campaigns, with it being sensible to heed our calls for change given how elections currently play out in the UK. It is however vital to appreciate that we still have huge gaps in data access and understanding, with much more needing to be done to create a healthy environment where targeted advertising and democracy can coexist.

Unlike the more regulated TV era of 1950-2000, the new zeitgeist of the digital era is currently more akin to a political wild west. Given this situation and the untenable position of regulatory stasis. The report also examines what must be done to understand and keep in check how parties are campaigning. Balancing freedom of speech with the limitation of threats, to thus promote transparency through a realizable implementable framework. This report sets out a pragmatic vision for regulation that is implementable and will push social media organisations to work for democracy and not undermine it. Our vision is for targeted advertising to be moulded into a tool that does not undermine democracy but that can operate within acceptable parameters of risk and threat.

The new landscape – data, advertising and new tools.

In the UK, more than 200,000 issue-based and political ads have been shown to Facebook users since October 2018. As we have begun to live ever more internet mediated lives due to increased internet access and new systems such as smartphones; parties have taken on the opportunities (and pitfalls) the internet offers with gusto. Billboard posters and newspaper spreads have slowly been losing out to online advertising, with the changing nature of where we the voters spend our time, leading to a change in how we are being campaigned to. For example, campaigners spent an estimated 43% of their total advertisement spending on digital platforms during the 2017 election, this was almost double 2015 (24%)[i]. Although it seems so far that this number will not likely have increased by much in 2019, the fact remains that what was once a novelty is now a primary campaign technique. Half of campaign funds spent goes on digital, and most of this on only a handful of platforms, especially Facebook and Google. 2019 saw the parties spending millions of pounds buying countless thousands of ads across both the short one-month campaign, but also even more widely across the whole year.

This election saw the mainstream use of social media as a core part of campaigning strategy, no longer was social media merely used to reach the young and mobilize existing support as has been seen previously. Instead it was approached by all the parties as a medium that could really make a difference, from reaching ‘precarious pensioners’ and ‘recalcitrant Remainers’ to ‘Labour leavers’. All the parties had differing understandings of how social media could make a difference; thus, the parties approached the many social media platforms differently.

So what is new about social media?

For the people –

• Increased role and number of opinion leaders (political science terminology) or influencers (marketing terminology).
• Change in flow of communications – from one step flow to two step – then to multiple step flows of communication.
• Capacity for the creation of echo chambers as users congregate around those with similar interests/backgrounds.
• Personal data used to target us in detailed ways.
• Greater scope of information – whether alt news sites, fake news or legitimate highly partisan news sources.
• Bots and trolls – can influence what we see and how we see it.

For campaigners –

• Greater audio-visual capacities that are developing all the time.
• Capacity for users to create and help a campaign – Corbyn meme’s, personal posts.
• High levels of interactivity creating a feedback loop – allows for the improvement of content via analytics.
• Data on people – Personal data used to target important voters/influencers in detailed ways.
• Bots and trolls – can influence what voters see, allow for overseas control of information spread, e.g. Russian bot farms. Maybe overhyped?
• Satellite campaigns – interlinked supportive campaigns specified towards specific groups (e.g. Momentum).
• Permanent campaigning.

Overall the parties love these new social media tools as they allow them to reach out to new voters, organise new supporter groups and influence hard to reach demographics.

Targeted advertising and its benefits

Advertising online

Social networks as well as search browsers and websites give political campaigners a range of new tools to get political messages out to voters. Thus, over the last decade we have seen the online recreation of offline ‘bread and butter’ campaign materials. Street billboard posters are now seen on the side of news articles online, leaflet content is seen via local or national Facebook advertisements, while TV party election broadcasts are repurposed for online content streams. More than ever parties have the capacity to recreate their offline campaigns online, with the abilities of these forms improved in pertinent ways.

• Firstly, political advertising has become harder to ignore, if you want to use social media or the internet you will see advertising, especially during the peak of the one-month campaign. You cannot just turn off the TV election broadcast or put down the newspaper; targeted advertising is now ubiquitous and impossible to avoid.
• Secondarily data and control is the new lifeblood, with the capacity for parties’ campaigns to be led by data far increased from the past. Rather than random passers-by seeing the billboard at the end of the street, or a random assortment viewing your TV broadcast, your poster or video can now be digitally targeted to those electorally important voters who are likely to be interested. Digital content is less likely to be wasted upon safe voters, instead the messages can be targeted at key electoral demographics. All the while, parties still retain the benefits of broadcast traditional tools such as posters, TV and broadcast social media content.

These capacities represent a real ramping up of what parties can do, with this all driven by a combination of a party’s own data, bought data such as that Experian provides, and the data social networks offer. This situation means that although the audio-visual content seen on the internet is not revolutionary, the potential use of content has so massively changed that political campaign capacities are beyond those available even a decade ago.

Alongside this shift, digital technologies have also developed meaning that the audio-visual forms that were most difficult to create, such as video, are now easily used. Video is known to engage people more, with today’s barriers to video production very low, parties have in effect their own direct cable TV networks that can be forced into user’s news feeds or social media homepages. In both medium and message, the nature of how parties can campaign has changed for their benefit, as they have been given new mediums far beyond their wildest expectations.

Micro-targeting

The Cambridge Analytica scandal made people aware of what has been occurring for years, including many practices made normal by the Obama campaigns. The scandal caused shock due to the illegal political use of personal data and highlighted the lack of public education that has occurred. Today, parties are using lots of data sources to target us, and they know a lot about us. Campaigners no longer just focus on Mondeo man, but can target John of number 34, Pearlier Lane, who has a 4×4, is worried about the local day centre, has 2 kids, a sky subscription, likes fishing and whose home is worth £300,000. Online advertising is allowing parties deliver the right messages to the right voters. Nevertheless, there is nuance within online advertising as parties do not act as a monolithic block, indeed modern campaign best practice dictates that success is tied to parties understanding their audiences which are inherently different. This election saw the Conservatives heed these calls, and Labour lose sight of them.

Personalised data targeting as a communications approach offers a radical improvement in detail, however despite possibility, we see the parties generally engaging in wider levels of targeting. Given a lack of access for researchers alongside failures in new research initiatives, we still know relatively little about how parties are micro targeting voters in specifics or what political effects this has. Nevertheless, we know that targeted advertising offers parties a radical scale of options to reach broad and micro-scale groups of voters, even if evidence of effect is more limited.

It is more expensive to micro-target, thus the conservatism seen in approach may be because there are as many potential pitfalls as benefits. Thus, targeted advertising (at the moment) appears evolutionary rather than revolutionary. Nevertheless, we are at the thin end of the wedge in how the parties can target voters, with it important we steer the development of these tools now while we can.

Feedback loops

Running ads online itself creates large amounts of data. Information about messages, advertisers and payments are all saved digitally, as are precise data about who saw ads and how they engaged with them. This is valuable for the platforms running the ads, as they can refine their content. We extensively saw this via party’s A/B testing content, for example with the Liberal Democrats testing different images of Jo Swinson. Through this process, parties can hone and improve the effectiveness of their messages in real time. No longer do they have to rely on expensive and small-scale focus groups, instead parties can reactively improve their content. This testing can occur across many years as parties use A/B testing to prepare for the short campaign, this is something we saw extensively in the run up to the election.

What did the platforms offer the parties?

Parties appreciate that the electorate are fragmenting online, no longer is every social media user generally on Facebook like from 2014-2016. Instead todays social media landscape sees a wide divergence in user’s choice of network by age, location and political position. Although the online space is diverging as younger people flock to newer social networks, big beasts such as Facebook continue to be of central importance for political campaigning. It is however important to note that the UK social media space is saturated both in the platforms available and in userbases. Those who want to use social media already do so, with general increases seen on some networks such as Facebook more due to the growth in older users rather than the few new young people that enter into the market each year. As such for political parties the opportunities and pitfalls of reaching electorally important voters has never been so great. The platforms are all different and as such the complexity of how parties can use social media has increased over the last few years as people have segmented themselves onto different networks.

Facebook

Capacities – Organic campaigning & micro-targeted/broadcast advertising

Central positives – Huge broad userbase, massive targeting abilities and integration with party datasets

FACEBOOK owns the Facebook social network, Facebook messenger, Instagram and WhatsApp. Given data and privacy issues on WhatsApp and Messenger, only the Facebook platform and Instagram allow for clear analysis of targeted advertising. This is frustrating for many campaign scholars as WhatsApp has been growing in importance and has also been a location for viral misinformation and sectarian politics. Given the lack of access these tools are not further examined in this report, even though both these private messaging services are of potentially major importance.

Beyond basic categories, Facebook offers the best solutions to micro-targeting. They allow advertisers to create “Custom Audiences”, based on information they supply, such as likely voters or visitors to the campaigns’ website. Facebook also offers “Lookalike Audiences”, these target users based on a custom audience as above, by finding additional people who are like that custom audience, but not yet part of it. It thus constructs a new list of people to contact who are ‘similar’ to those who already like your content. Finally, Facebook allows parties to target people based upon their interests such as them following or engaging with a certain page.

Facebook is by far the largest social media platform in the UK, combined with its photo focussed subsidiary Instagram, these platforms offer the parties a gateway to the general public via both organic popularity and micro-targeted advertising. According to Facebook Audience Insights, there are 35-40 million monthly active users on the Facebook platform, and 24 million on Instagram. Facebook gender is relatively representative of the UK population compared to other social networks, this is seen in the gender skew of 52% women and 48% men (based on users aged 18 and older). The platform is generally more rightward skewed than other social networks, more rural and featuring an older userbase. In contrast Instagram has a younger userbase, more millennial, less political and less evenly spread across the country than Facebook. The Facebook/Instagram advertising platform is by far the most complex campaign tool offered to the parties, it also offers the widest and most representative access to voters.

Twitter

Capacities – Organic campaigning

Central positives – Ability to influence and prime media commentators and journalists, source content and a place to activate partisans

Twitter is not Britain; this is something that those who heavily use the platform often fail to appreciate. Looking at 2019 via Twitter one would expect the Labour Party to have won a supermajority. Although the platform is well used in the UK, featuring around 14 million users in the UK. The sorts of people who truly actively use Twitter are electorally abnormal. Users are more urban and far more left wing that the UK voting population. Equally, as well as being a younger audience, gender is skewed towards men at 60% men to 40% women.

Twitter is a useful tool for political campaigners, but its central importance is not direct communications from party to userbase. It instead has vitally important campaign uses to prime journalists, push out stories and to inform partisans. The use of Twitter for targeted advertising is even more limited, the platform until recently allowed for geo-targeted advertisements, but use was limited. During the 2017 General Election, only £56,000 was spent by all the major parties[iii]. This lack of revenue is one part of why, just before the election campaign kicked off, Twitter decided to ban all political advertisements on the platform. This was however more in response to the looming 2020 US Presidential elections rather than the UK 2019 General Election. Twitter has led the way in making a clear position on targeted political advertising which has cheered some[iv], while others find the choice short-sighted[v].

Google

Capacities – Organic/targeted campaigning via YouTube & micro-targeted/broadcast advertising via search

Central positives – huge reach, creative opportunities for location and implementation of ads, powerful targeting capacities

Google controls the ubiquitous search engine as well as YouTube, it’s estimated that there are 35.6 million UK visitors to YouTube each month, making it the second biggest social network in the UK after Facebook. However, this number is dwarfed by Google search, which has 87.5% percent of the desktop, tablet and console search engine market in the UK. Given such high internet usage rates in the UK, Google are likely to be reaching nearly every British voter. Although electorally very important Google search has not received very much attention or analysis despite potentially huge influence. This is in part due to limited tools for researchers, although the group did start releasing spending reports and an ad library during the election. YouTube is also a newly important campaign tool. YouTube is hugely popular to its varied visitors, with the platform able to offer parties reach into a different sort of internet user in a different location. YouTube is different to other social networks and offers reach to younger audiences that Facebook is losing. We saw the Conservatives spending heavily on the platform via front page banner ads, as such YouTube was clearly a target for party campaigners. However, there is a lack of access for researchers making specific analysis difficult.

In terms of use for political advertising, Google also took steps before the election to limit the capacities of targeting on their web browser. The group said it would take action against ‘demonstrably false claims’ as well as limiting the granularity of targeting to age, gender, and general location (postal code level). Overall Google is rising in importance as a campaign tool.

Snapchat

Capacities – Organic campaigning & limited micro-targeted/broadcast advertising

Central positives – Younger audience, content cannot be skipped, medium targeting capacities

The number of Snapchat users in the UK reached 16.2 million in late 2018. The platform has a young userbase and does allow for political advertisements by geo-targeting. We saw parties advertise on the platform far more than in 2017, however use was still limited compared to other social network tools. Given the changing demographics of the social networks, if parties are intent on reaching a younger audience, it is becoming more important to advertise via Snapchat, YouTube and TikTok.

TikTok

Capacities – Organic campaigning

Central positives – Young audience, viral capacities, ability to look ‘cool’

TikTok has banned political advertising, however the platform is growing as an organic political tool to reach a centrally Generation Z userbase, the app was downloaded 6.7 million times in the UK in 2018 alone. Given a lack of advertisement options, parties approaches focus on attempts at viral content and grassroots outreach.

Reddit

Reddit has not banned political advertising, as campaigners can advertise to certain sub-reddits. The website is widely used but can be quite specific depending on community. As such it may have a political use to activate partisans, however in reality usage in elections is limited. The platform can help drive huge amounts of organic traffic and thus is a potentially useful tool, however it its capacities for targeting and advertising are limited.

What was the regulatory context?

The 2019 General Election showed the digital campaigning landscape to still be akin to the ‘wild west’. Electoral law has not adapted nor kept up with advances in advertising, with governments only recently showing an interest in engaging with regulation. However, the major difference between this election and previous elections was a new pressure on the activities of these corporations and parties. There was far more public, media and governmental awareness.

Rules

Major changes that have occurred are mostly related to social networks self-regulating and changing their own rules and practices, or via major semi-related legislation such as the General Data Protection Regulation (GDPR). Nevertheless, although essentially unregulated, a series of laws and regulations do cover the use of advertising online as a campaign tool. The general regulatory landscape is outlined below across five key aspects of practicality; data, reach, spending, content and disclosure.

Data

Sophisticated targeting requires large amounts of high-quality data. Currently the rules covering the collection, storage and processing of data for political campaigns are geared towards traditional advertising and media channels. Changes for data access have been made and are being proposed so the landscape may change further. Centrally the General Data Protection Regulation (GDPR) 2018 has significantly tightened restrictions on the collection and use of personal data. Political campaigners must comply with its provisions although punishment of potential breaks with this legislation are unclear. For example, the Coalition for Reform in Political Advertising have asserted some 2019 election advertising may have breached the regulations, but nothing has yet occurred[i] . Brexit also throws into question whether the UK will continue following the GDPR, however any market access will likely demand it, making it of continued relevance in post-Brexit Britain.

The Information Commissioner’s Office (ICO) of the UK is in the process of developing a code of practice for the use of personal data in political campaigning. It includes advice on micro-targeting, and the use of lookalike audiences. It looks unlikely to become legal, instead becoming further ‘best practice’ guidelines for campaigners. A draft was released earlier this year with responses now being examined.

Reach

Under the Communications Act 2003, the UK does not allow political advertising to be broadcast on TV to avoid giving an advantage to better financed parties. Parties are instead offered airtime for party political broadcasts. Non-broadcast media are not subject to this regulation. Political advertising on non-broadcast media is exempt from the Advertising Code and therefore not subject to regulation by the Advertising Standards Authority, leaving it essentially unregulated. Parties are free to use as many social networks or send as many ads as they like, the only control is spending.

Spending

Regulation of campaign finance in the UK focuses on the expenditure of parties and candidates, rather than on the donations received. The Electoral Commission publishes specific guidance for political parties and candidates for the elections taking place in each year covering:

• the spending limits that apply for each party and candidate for each election
• the periods for which these limits apply
• the reporting deadlines for each election

After each election or referendum, the Electoral Commission gathers and publishes the reported spending by each party. Since December 2010, the Electoral Commission has had powers to investigate potential breaches of the rules, and to issue sanctions if breaches are found to have occurred, including fines up to a maximum of £20,000.

Spending rules are however highly problematic as there are differences between local constituency spends and national party spends. The ability to target specific people within a geographic area gives parties the opportunity to focus their attention on marginal constituencies with national funds via national social media pages. This undermines the principle of a level playing field at a local level and is a current legal grey area that needs improvement.

Content

Unlike commercial advertising, political advertising in the UK isn’t regulated by the Advertising Standards Authority (ASA). There are currently no laws on fake news or misinformation outside those that cover legal areas such as slander or misrepresentation which sanctions wilfully untrue or defamatory claims. This situation is unsatisfactory to voters: according to a YouGov poll commissioned by Coalition for Reform in Political Advertising 87% of voters believe ‘there should be a legal requirement for factual claims in political advertisements to be accurate’.

Disclosure / Transparency

Transparency is a key principle of electoral law; the Electoral Commission argue that voters should be able to understand ‘who is behind the campaign and who created it’.

Recently the UK Government committed to implementing an imprints regime for digital election campaign material so that the public would be able to see who was paying for ads. However, there are no rules yet compelling all political actors to disclose who they are online. The only requirement so far is for parties and ‘non-party’ campaigners to register with the Electoral Commission if they plan to spend over £20,000 in England or £10,000 in the rest of the UK.

Norms

Currently rather than laws and regulation, a legal culture of self-regulation has been in force. So far, norms have been as ineffective as governments in changing the culture surrounding targeted advertising, however this is largely because governments have not yet tried to engage with altering regulation or approach in any clear-cut way. International guidelines have led on best practice, while it seems that network on network pressure spurred by public opinion has created the most powerful force for change.

International guidelines

An array of international guidance on campaign financing exists to set out best practice regulations and guides for how advertising should operate to ensure that elections are fair, clean and free. Organisations such as the OSCE, IDEA and Venice Commission have laid out guidelines.

Public, governmental, 3^rd sector and network on network pressure

From Netflix documentaries like The Great Hack to newspaper reports, in the wake of the Cambridge Analytica scandal the public were made much more aware of targeted advertising. This was supported by greater governmental, media and 3^rd sector awareness and thus greater pressure on these digital platforms. There has consequently been a new norm growing amongst the social networks, that something must change and that they cannot continue as normal. As such we have seen the networks altering their rules and practices to maximise transparency and minimise harm. However, each network is taking a different stance, and some are going farther than others. This norm however can potentially be harnessed to promote further positive change as all the networks clearly agree that the current situation is not good for public trust. If government could work with these organisations, it is possible we could see a reduced need for future draconian legislation and regulation.

Nevertheless, although progress has been made by norm building, these groups have elected to make smaller steps than those necessary to challenge the threats posed by targeted advertising. As such, it appears government is needed to step in to enforce change.

How did the parties use social media during the 2019 General Election

This election campaign saw huge distinctions in how each party campaigned, with it clear that any changes made to regulation must appreciate that political parties appear to be diverging in how they campaign. This means regulation cannot be one size fits all, it must try and address all networks together.

Labour and Conservatives

The Conservatives won the election big, with their social media campaign a clear part of their overall success. Firstly, in terms of the organic campaign the Conservative Party ‘went a bit wild’, using lots of video content that was bright and colourful, memes and personalised content. Messaging was clear and narrowly focussed on Brexit, however the messages certainly had the capacity to cut through. Boris was well used to campaign on Brexit while the party page focussed on policy. In contrast, Labour used a trident approach. The party used the Labour Party page to deliver core policy, Corbyn’s page to promote the leader and energise younger people and Momentum to create viral humorous video content. The party however failed to recreate the spirit of 2017, there was a reduction in exciting novel content, some content from 2017 was even re-used. The party not only failed in its organic campaign but also poorly utilised targeted advertising.

Theme	Conservative Party	Labour Party
Networks and how used	Facebook, Google, YouTube, Snapchat, Instagram.	Facebook, Google, YouTube, Snapchat, Instagram.
Engagement	Lots of activity especially on Facebook with engagement on posts high for the Conservative Party, but nevertheless still behind the Labour Party. Low organic popularity outside of Boris Johnson. Content attracted a lot of negative engagement; many comments were negative.	Lots of activity across all networks with different pages receiving different levels of engagement. Corbyn’s page achieved the lions share. Corbyn less popular in engagement than 2017. Overall engagement seems to have dropped off somewhat compared to 2017. Weak share power. Huge popularity on Twitter. Passion of core party activists was channelled well via social media. Content attracted some negative engagement, some comments were negative, with this far more apparent than 2017.
Tactics	Engagement at all costs, led to some quirky creations and some visually awful content. Experimental, less staid than in 2017, clear the new digital team made a clear impact.	Lack of mixture of offline action with online content, content was unlike in 2017 where it often pushed Corbyn’s interactions and events. Less optimistic than 2017, more combative in a bad way that could put centrist voters off. Effective deployment of organic social media support, Momentum and the party managed to get the party faithful out to door knock in winter. Lack of defensive pivot especially towards northern heartlands.
Content	More positive content, alongside very negative, personalised content. Trolling / s***posting content. Misrepresentation of people and events – Keir Starmer	Reused content from 2017, made them look tired. Labour Party page’s focus didn’t join policy with people. Messaging was not clear, as core policy wasn’t clear, muddied the waters. Bread and butter issues centred on the use of the NHS, prescriptions and working conditions. Video content often dark in colour and without sound.
Post that summarises approach	Our NHS will never be for sale[a]	Matt Delaney NHS video[b]

[a] https://www.facebook.com/borisjohnson/videos/415478332669224/

[b] https://www.facebook.com/watch/?v=2157391067888058&external_log_id=f6a7c943cd88f762a1ee751b013db506&q=rob%20delaney%20nhs

The targeted campaign

Theme	Conservative Party	Labour Party
Networks and how used	Facebook, Google, YouTube, Snapchat, Instagram.	Facebook, Google, YouTube, Snapchat, Instagram.
Targeting	Leave voters Disillusioned Lab voters Tory core vote Those fearing a Corbyn government Former 2017 voters	Young People, BAME people Former 2017 voters Second vote Remainers WASPI women
Tactics	Catch people’s attention at all costs, led to some quirky creations and some visually awful content. Much more optimistic than 2017. Also, content very negative, personalised against Corbyn. Use of impartial journalistic sources as mouthpieces, altering the narrative of truth. Spend late	Lack of defensive pivot especially towards northern heartlands. Led social media spend throughout the campaign. Reached millions of younger voters Snapchat investing 75.9% of the total spend of major parties SnapChat spend. High spend on a limited number of ads. (6th of November – 5th of December) Spent a lot more than 2017.
Content	More positive content, alongside very negative, personalised content. Misrepresentation of people and events – Keir Starmer Content is not people focussed, use of lots of interactive poster style content. Clear messaging. 3 areas; Brexit, leadership and taxation/economy. Lots of video content, bright and colourful.	Labour Party page’s focus didn’t join policy with people. Focus was on policy not Corbyn. Messaging was not clear, as core policy wasn’t clear, muddied the waters. Bread and butter issues centred on the use of the NHS, prescriptions and working conditions. Video content often dark in colour and without sound. High policy focus via the Labour Party page on prescriptions, Green New Deal, NHS and Brexit. Labour Party page’s content not interactive, used dull colours, lack of video.
Post that summarises approach	Upside down Brexit advert[c]	Matt Delaney NHS video[d]

[c] https://www.youtube.com/watch?v=N7OwmsDxkTY

[d]https://www.facebook.com/watch/?v=2157391067888058&external_log_id=f6a7c943cd88f762a1ee751b013db506&q=rob%20delaney%20nhs

On Facebook the Conservatives prioritised lookalike audiences, while Labour focussed on Custom audiences. Labour attempted to recreate their 2017 campaign but with a greater focus on Facebook advertising, the party decided to spend much more on Facebook than in 2017, pivoting towards the platform in targeted advertising while trying to recapture their viral success seen in 2017. The party attempted to reach out to those doing less well, pushing their radical manifesto as a solution for the millions of ‘just about managing’ voters. The party continued to push for supremacy on Twitter as seen in 2017, with the party also using younger platforms including Snapchat and TikTok. Snapchat was used for organic virality and for some targeted advertising, while TikTok saw limited political content. Overall the party wanted to use social media to energise their supporters to campaign online and offline, using targeted advertising as a combined tool to get out the vote, activate partisans and reach out to target voters. This clearly did not occur as it had in 2017.

In contrast, the Conservatives were surprisingly nuanced in their approach, with the party page not mentioning Brexit as much as one would expect. Instead, the party pushed Brexit content via Boris Johnson’s page. The party was also talking about other things – the leadership of Jeremy Corbyn, but also the NHS and policing – and grasped that while Brexit was central to this General Election, it was not the only issue going. The Conservatives seized that the best tool they had at their disposal to reach Labour Leavers, the former face of Vote Leave, Boris Johnson. Equally, the party appreciated his ‘Marmite’ character was a turnoff for Remain-voting areas. As such the campaign appreciated this and used Johnson selectively depending on seat. As Image 1 overleaf shows, in remain voting locations Corbyn was used, while in leave seats Boris was used. It is this appreciation of audience that gave the Conservatives victory. Similar nuance is not seen in the Labour campaign. Finally, it is important to highlight the Conservatives use of mistruths and edited news footage, including the use of content that pushes the edges of acceptability. This election saw the parties push norms to their limits in search of votes.

Labour also avoided Brexit, a decision that was instrumental to their loss. Only 15% (122) of their party page adverts covered Brexit, and Corbyn avoided it completely. This repeated their approach in 2017, but this time it failed as they had since then turned to support a second referendum. The limited content about Brexit was stagnant and unchanging, and it was clear the party would have preferred to avoid the issue altogether. The Labour Party was trying to walk a very difficult tightrope between covering and discussing their Brexit position, and ignoring it completely. Labour was split between an electorate who wanted Brexit and those who saw it as an anathema, and its attempts to walk a tightrope between the two failed abjectly. The Conservatives, on the other hand, weaponised Brexit alongside other topics, through a strategy that appreciated nuance, colourful and changing content and a determination not to get bogged down in a single issue. Finally, Labours policy offerings were not effectively communicated or were absent, focus was on healthcare not on radical policy. There was therefore a huge disparity between Labour’s online and offline campaigns, this is in great contrast to the Conservatives who seemed to communicate in a way designed for social media offline. For example, Boris Johnson in the ITV debate repeated ‘Get Brexit Done’ consistently.

To understand the content of the adverts, Graph 1 shows the topic of the adverts sent across the first two weeks of the campaign. Overall for the Conservatives, Brexit, party action (getting people activated to register to vote, make voting plans and organise), the economy and leadership were core topics, while for Labour, party action and healthcare were important. One can clearly see that the Conservatives were using a wider array of topics speaking to a broader base in their messaging.

Smaller parties

The Liberal Democrats, Green Party and Brexit Party’s organic campaigns were more limited, although the Brexit Party and Green Party were popular in their issue-based campaign approach. The Liberal Democrats and Brexit Party were more active in targeted advertising than the Green Party, with parties spending big on Facebook in proportion to their size.

The Liberal Democrats and Brexit Party helped and hindered Labour and the Conservatives through their efforts. The Liberal Democrats focussed on using ‘interested audiences’ via Facebook, targeting voters who liked Barack Obama or Kamala Harris for example. In contrast the Brexit Party went regional, targeting leave voting Labour seats across the country. It appears likely the party helped the Conservatives given the pact they parties had, as well as the impact the Brexit Party had on the Labour Party vote in certain seats. The Green Party engaged in limited targeting, centrally pushing the Green agenda via issue campaigning.

The Liberal Democrats and Brexit Party as expected focussed heavily on Brexit (Graph 2). However, the Liberal Democrats most heavily focused on leadership. The use of leadership is interesting as they moved away from using Jo Swinson after they realised their/her personalised campaign was not working instead focussing more on Brexit content. As this data shows the first two weeks of the campaign, therefore this shift is not visible. It clearly shows that huge spend and thousands of adverts can be ineffective without the right message. The Green Party put the environment front and centre; the party also heavily tied the economy in with green issues. Overall the smaller parties’ specific interests and goals are more readily apparent in their approaches, with the wider approach of the major parties signalling the difference between the parties of government and issue-based parties.

What happened – not just the medium but the message

On election night it became clear very early that things had not worked well for Labour, this was even though they spent more on Facebook and were more ‘popular’ online. The story of 2019 does not admonish targeted advertising but highlights that the tool is just one of many within the campaign toolkit. Targeted advertising is not everything, but the parties view it as a central part of their campaign toolkit. All the parties spent big with the Brexit Party and Conservatives operating in relative unison to demolish Labour in key seats. Using both organic communications and adverts, and with the help of last-minute third-party groups, the red wall crumbled.

Labour was more popular online, however given demographics and activism levels that is forged into the system. Compared to 2017 the Labour Party fell back. In contrast the Conservatives surged forwards, more popular online with clearer messaging. The election saw the Conservative mix the right messages with the correct mediums, while Labour sold weaker messages through an ineffective mix of mediums. Targeted advertising was only a part of this, as similar issues were visible through other campaign tools. Given the lack of transparency we can only partially uncover how targeted advertising was used, but it was clearly deemed of great importance to the parties, but not a golden bullet. As such, analysis requires a full holistic appreciation of the 2019 campaign and how the parties acted.

Unfortunately given its snap election nature, it is likely there will be a lack of academic study of the potential effectiveness of targeted advertising during the campaign. Thus, although we cannot categorically say targeting made a big difference in 2019, we know the parties spent big and targeted heavily. However, as ever having all the tools does not offer success, it is a mixture between medium and message. As such, the parties targeted advertising and social media popularity are glimpses of wider truths and symbolic of their campaign performance as well as their individual success. This is important to bear in mind, as the parties used considerably different approaches this election. The 2019 campaign shows that more must be done to regulate social media targeted advertising, from third parties to edited news content, some unacceptable things occurred during the campaign. Had the social networks not altered their practice beforehand it could have been far worse, but clearly more needs to be done. Overall, government must step in to push the platforms further in order to promote a more harmonious campaign environment.

---

Scale of the campaign

Actors

On Facebook from the 1^st to the 12^th December we categorised 2481 political pages that sent adverts (excluding non-party-political pages such as those that promote voting). This huge number of advertisers signals the massive problem we have in holding to account those who advertise. This also goes far beyond party associated pages, as there is a hidden iceberg of political advertisers that are 3^rd party actors. Facebook is the centre of activity for this huge number of actors, in contrast Snapchat and Google saw more limited numbers. Over the same period, only 9 political pages were advertising on Snapchat and only 12 political pages on Google/YouTube. Only some of these actors are seen across all three of these social networks with it clear that the use of multiple pages by parties is a purely Facebook phenomenon.

The types of actors seen online using the example of Facebook are:

Main party pages = These are the official UK wide online instances of the parties, generally this is where most the adverts occur from. Example = Labour Party page.

National party pages = This is the official national instances of the parties, generally these pages are not that widely used except in Scotland and by Labour. Example = Welsh Labour.

Local/candidate party pages = These are the official representations of a party’s local group, candidate or regional pages. Example = Labour Bristol, Labour South West, Thangam Debbonaire.

Satellite party pages = These are quasi-independent pages that push a certain political narrative or goal, they are satellite to the central pages but are often closely linked. For example, Momentum is a satellite of the Labour Party, but is centrally supportive of Jeremy Corbyn and his politics. Example = Momentum.

Third party pages = These are supposedly independent pages, although they vary in connections to the parties. They push a certain political angle, such as pro-remain or anti-Corbyn and are generally run by members of the public. The whole third-party system is a massive grey area as will be shown later. Example = Capitalist Worker, Hope not Hate, Led by Donkeys.

Government actor pages = these are apolitical get out the vote pages intended on increasing electoral turnout, they are not examined in this report as they are not political. Example = Yes I will vote.

Overall use of targeting and data by the parties

You may expect that there was a set tactical approach to targeted advertising seen during the election, a best use form exhibited by all the parties. However, given the newness of targeted campaigning alongside differing electoral goals, funding and electoral geographies, the reality is the parties approached through a variety of methods. There are huge variations in the scale, spend and messaging the parties used. The wide array of approaches seen alongside the huge scale and spend of the different parties, highlights the need for clearer tools to analyse targeted advertising. Approach is not currently easily measured, as we are currently only scratching the surface of how parties and others are campaigning using targeted advertising.

Across the different social media platforms, we saw the parties use different pages and different approaches. Thus, how the parties used targeted advertising the spending and reach of the parties’ central pages must be examined. Local/candidate pages were excluded with only central party pages examined because these are generally the most important to the parties both in spend and reach.

The calculations were based on the following pages for each platform;

Snapchat – Boris Johnson page, Liberal Democrats page and Labour Party page.

Google + YouTube – Labour Party + Welsh Labour, Liberal Democrats, The Brexit Party, The Conservative & Unionist Party.

Facebook + Instagram – Labour party + Scottish Labour Party + Welsh Labour + Jeremy Corbyn + Labour Unions + Momentum, Boris Johnson + Conservatives + Scottish Conservatives, Jo Swinson + Liberal Democrats, Plaid Cymru, Nigel Farage + Richard Tice + Brexit Party, UK Independence Party (UKIP), Green Party of England and Wales, Scottish National Party (SNP).

The number of adverts from the parties

A huge number of adverts were seen across the platforms, with Facebook as expected seeing the most. The sheer scale indicates the difficultly for the media or 3^rd sector to hold the parties to account effectively.

Facebook	Snapchat	Google
Boris Johnson	Liberal Democrats	Labour Party
Conservatives	The Labour Party	Liberal Democrats
Green Party of England and Wales	CCHQ (Boris Johnson’s page)	The Brexit Party
Jeremy Corbyn		The Conservative & Unionist Party
Jo Swinson		Welsh Labour
Labour Unions
Liberal Democrats
Momentum
Nigel Farage
Plaid Cymru
Richard Tice
Scottish Conservatives
Scottish Labour Party
Scottish National Party (SNP)
The Brexit Party
UK Independence Party (UKIP)
Labour party
Welsh Labour

Across the pages seen in Table 4 – via Google, Snapchat and Facebook 29153 adverts were logged from the 1^st November to the 12^th December.

This huge number of adverts was seen centrally via leader and party pages; approach however was heavily party dependent. Alongside party pages a limited number of satellite and national pages also saw large scale numbers of adverts. The colossal scale of advertising seen signals the serious problems we have in ‘keeping a tab’ on political advertising during the campaign. This problem is exacerbated further when including third-party pages and local/candidate pages, the number of adverts is even higher. This figure is revealed in the later analysis of third parties. Nevertheless, as one would expect, the mainstay of content and spend is from party pages.

Facebook, as expected from the number of advertising pages is far ahead of the other platforms in number of party adverts seen. Nearly 30,000 adverts from parties alone across a one-month period averages at 1000 adverts a day, a number that is impossible to keep a track of.

	Total number of party adverts seen
Snapchat	30
Google + YouTube	514
Facebook + Instagram	28609
Sum	29153

The overall number of adverts sent by the parties is seen in Graph 4. A huge number of advertisements were sent, especially from the Conservatives (11150 adverts) and Liberal Democrats (11459 adverts). It is important to highlight that the number of adverts is different to spend, as different tactics between the parties was clearly apparent. Overall the Conservatives and Liberal Democrats focussed on greater levels of A/B testing, sending greater numbers of lower spend adverts. In contrast, Labour (2177) and the Brexit Party (3925) sent fewer adverts but that featured a higher average spend. Finally, the use of targeted advertising is not ubiquitous as can be seen from limited use by parties such as the SNP, UKIP and Greens. Collectively these parties only sent 442 adverts across the one-month short campaign (Graph 4).

Reach of the party adverts

Overall 608,409,178 impressions were logged via the 29153 adverts sent. This figure was arrived at using the mean average of the minimum and maximum ranges supplied via Facebook and Google, while exact numbers were supplied by Snapchat.

The scale is enormous; millions saw party political adverts across only a one-month period. Nevertheless, it is important to appreciate results will have a wide margin of error due to the social networks providing a low level of clarity in their reporting. Even where supposedly correct numbers are given, as is the case with Snapchat, the platform has history in providing incorrect data[1]. Given the wide ranges of the statistics given by Facebook and Google, there is a large margin of error. The numbers could be higher, or more likely lower.

Overall, every UK Facebook user (estimated by Facebook at ~45 million people) would have each ‘seen’ 8 adverts each across the platform. However, the capacity for targeting important voter groups means millions of impressions will largely be saturated within a few marginal seats. Many of these seats were also in the crosshairs of multiple parties, such as the Conservatives and the Brexit Party meaning potentially even higher saturation of certain seats.

As previously shown, by far the greatest quantity of adverts was seen via Facebook. However, in terms of spend and reach, the field is much more balanced. This is due to the way the parties used each tool. Facebook’s greater number of adverts is largely because the parties use Facebook as a playground to test messages through A/B testing, alongside utilising its capacities for microtargeted communications that Snapchat or Google are less capable of. In contrast, although advertising via Google and YouTube was popular, little A/B testing occurred and most adverts saw increased spend. Thus, in spend and reach Google received a near equal level of spend compared to Facebook.

Examining impressions by party, one can appreciate the role the different networks played. Unlike the 2017 General election this was not a solely Facebook election, other platforms have begun to play a key role in the campaign, especially Google/YouTube. This is evidently clear from the Conservatives and Labour numbers. According to the figures the Conservatives received more impressions from Google than from Facebook, and similarly for Labour. Labour received more impressions via Snapchat and Google than from Facebook.

Spend of the party adverts

In terms of spend, numbers were again taken for Google and Facebook from the mean of the low and high spend estimates, while Snapchat offered accurate spend. Across the parties £8,553,343 was spent on the three platforms with spending seen especially via Facebook and Google. It is very important to appreciate the wide margin of error in the spending data, as the wide ranges offered mean that total spend may be far higher. This is especially so given the reported spend here using the data is higher than that reported by Facebook. The issue is in determining the spend seen via the numerous low-spend adverts sent especially by the Conservatives and Liberal Democrats. This report uses the mean value between low and high estimate. For most posts this is effective, however for posts that spend £0-100, £50 may be too high given many are test posts. Nevertheless, although imperfect this spending data is currently the best available as the Electoral Commission’s spending reports are not yet released.

	Spend (~)
Snapchat	£92541
Google	£2939975
Facebook	£5520827

When breaking spend down by party (Graph 7) interesting trends are visible.

The Conservatives were clearly far ahead, spending an estimated £3,540,511.17, far ahead of Labour, the Liberal Democrats and the Brexit Party. The other UK parties spent very little, with the SNP not engaging in targeted advertising, a fascinating tactical approach. Even the Green Party, spent a relatively large amount of money for their size (£182,573). It is important to note, that for the Liberal Democrats and Brexit Party, the spend we are seeing will have been across only a few seats. However, in the Liberal Democrats case, it is understood they were overly optimistic in their targeting. In contrast, the Brexit Party were clearly trying to target around ~100 leave voting seats. As such their spend per seat on targeted advertising was very high, meaning that they were heavily saturating certain seats.

Google was clearly of major importance to the big two, with Snapchat an alternative avenue used most heavily by Labour. The use of these other networks speaks to a diverging online audience. Not everyone is on Facebook anymore and the parties are beginning to appreciate the tools that Google offers.  The ability to steer online traffic via sponsored positions in Google search was seen especially by the Conservatives and Labour. Alongside search advertising the parties also used YouTube, most notably the Conservatives achieved huge reach via a YouTube banner ad. This tactical shift speaks to the Conservatives approaching 2019 in a very different way to 2017.

Spend vs reach for the parties

One would expect a totally linear relationship between spend and impressions, this is generally the case, however some interesting nuances are seen within the data.

Facebook shows a closely linear relationship between spend and reach especially for the Conservative Party, whereas for Labour and the Liberal Democrats the relationship is less close. The data is likely due to people sharing adverts, with this predominantly seen via the Green Party and Labour Party. Indeed, many viral Jeremy Corbyn posts were also used as adverts increasing reach. The Brexit Party and to a lesser extent the Conservatives show an inverse correlation with these parties’ content likely not organically shared as much, this then leads to spend achieving less reach. This trend is also shown in terms of the cost per 1000 impressions (CPM) with Labour and Liberal Democrats benefitting from a greater reach from a lower spend. However, it appears that those parties who engaged with less complex advertising achieved more economical rates of impressions. It could be that the Conservatives and Brexit Party were just going after certain electorally important voters in certain seats rather than a generalised approach.

The story is different on Google/YouTube, it appears that for all the parties excluding Labour, spend and reach are very closely related. Labour in contrast achieved far greater reach from a lower spend in comparison to the Conservative Party. This may be because of the Conservatives tactics to buy up the most expensive slots closer to voting day, a method that would increase spend rate.

The trend is clearly seen in the cost per thousand impressions, with Labour reaching more for a lower price. This again suggests they used less complicated targeting parameters in comparison to the other parties.

Snapchat supposedly offers accurate data, with the reach and spend data matching very closely.

Examining the Snapchat CPM relationship, it appears that the Conservatives achieved a lot of reach for little spend. This is because the party did not utilise complex targeting, with the adverts run via Boris Johnsons page targeted generally across the country excluding Uxbridge, Boris Johnson’s seat. This was done to avoid local spending limits. In contrast, Labour were clearly trying to target certain people as they spent more to reach voters, with the Liberal Democrats in the middle. It is fascinating to note that of the networks Snapchat features the lowest average CPM impressions (£1.60) compared to Facebook (£11.52) and Google (£15.29). The lack of competition on the platform makes advertising cheap.

Overall, all the social networks are not equal. For the parties that advertised heavily, Facebook cost the most followed closely by Google. However, if one looks at the party approach, the Conservatives were spending nearly as much on Google as on Facebook, this is also seen in the Brexit Party. These parties either were utilising targeting parameters more or were buying up certain expensive locations and advertising spaces on Facebook and Google. In contrast, Labour went with a cheaper and likely less targeted approach.

CPM per 1000	Labour Party	Conservative Party	Liberal Democrats	Brexit Party	Average
Facebook	£14.74	£17.22	£14.18	£21.87	£17.00
Google	£10.75	£16.12	£15.16	£19.12	£15.29
Snapchat	£2.07	£1.05	£1.68	.	£1.60
Average	£9.19	£11.47	£10.34	£20.49

Party associated pages across the networks

To get a greater understanding of the nuances behind this data, the available pages seen on the networks are now examined, all non-apolitical pages that advertised on Google and Snapchat are examined, with party political pages shown for Facebook. Because of the large number of third parties on Facebook, third-party influence on Facebook is separately examined.

Facebook + Instagram

The following party pages were seen advertising (Table 8), the individual spend of the parties varied depending on page. For example, the Labour party heavily used the main party page as did the Conservatives. Other parties avoided their leader pages, such as Jo Swinson.

	Number of ads	Impressions	Spend
Boris Johnson	1192	17236806	228006
Conservatives	9667	84740202.5	1520349.5
Green Party of England and Wales	341	25073774.5	182573.5
Jeremy Corbyn	57	16254902.5	295850
Jo Swinson	2	32999	349
Labour party	1164	38973422	674768
Labour Unions	84	2363952.5	18802.5
Liberal Democrats	11268	96022970.5	1361267
Momentum	634	17889677	142727
Nigel Farage	754	1208623	79423
Plaid Cymru	66	1833464	12964
Richard Tice	2	21999	199
Scottish Conservatives	10	3130489.5	61739.5
Scottish Labour Party	43	6358946.5	73496.5
Scottish National Party (SNP)	32	3865979.5	24929.5
The Brexit Party	3144	40045652	823200.5
UK Independence Party (UKIP)	3	74497.5	247.5
Welsh Labour	146	1333435	19935
	28609	356461792.5	5520827

Google + YouTube

Google saw the main parties alongside a group of smaller pages advertise. The spend and reach of many of these smaller groups was very low, with the Conservatives spending by far the most.

Advertiser	Number of adverts	Spend	Reach	Reach per £	Spend per advert
Alliance Party of NI	20	1500	2480000	1653	75
Bridgwater and West Somerset Liberal Democrats	6	1150	130000	113	192
Concerned Citizens Limited	26	1650	130000	79	63
Hackney Liberal Democrats	5	375	620000	1653	75
Labour Party	44	696650	70480000	101	15,833
Liberal Democrats	170	241325	15915000	66	1,420
Robin Michael Horsley	27	14650	1675000	114	543
Shropshire Party	2	300	10000	33	150
The Brexit Party	25	166225	8695000	52	6,649
The Conservative & Unionist Party	275	1727975	107165000	62	6,284
Welsh Labour	80	107800	4320000	40	1,348
Westminster North Conservative Association	7	175	35000	200	25
	687	2,959,775	211,655,000	4166	326,55

The huge number of adverts sent by the Conservatives especially shows how seriously the party took the platform, with the party spending a large amount of money on each advert (£6282). This figure was only surpassed by the Brexit Party (£6649) and Labour (£15833). Third party activity seen was limited. Only Concerned Citizens Limited, a listed company created on the 25^th November 2019 that created anti-Conservative party links was seen. Their content approach is seen in the image below.

Snapchat

A limited group of pages used the platform to advertise, UK politicians, parties and third parties that did are shown in Table 10.

Advertiser	Spend	Reach	Reach per USD	Number of Ads
Avaaz Campaigns UK	13968	8486764	608	4
CCHQ	3172	2325795	733	6
HOPE not hate Charitable Trust	3637	1979567	544	173
Jo Dale for Congleton	3	1686	562	1
Liberal Democrats	19706	9050458	459	19
Mid Suffolk Green Party	954	356615	374	19
Midlothian SNP	111	19396	175	4
The Labour Party	97368	36119130	371	5
Vikki Slade GE Campaign 2019	0	405	405	2
	138919	58339816	4231	233

Snapchat was clearly of importance to labour as well as of limited use to third party Avaaz. The third-party activity on the platform was very limited, but still shows the large reach of these pages, 10,466,331 impressions were seen from just Hope not Hate and Avaaz campaigns across the one-month period. This contrasts with 47,495,383 impressions from the Conservative, Labour and Liberal Democrat pages. Third party influence on Snapchat is clearly an important phenomenon.

Third party advertising

To properly get a handle on the third-party problem we must examine where the phenomenon is occurring; Facebook.

	Platform	Number of ads	Spend	Reach	Reach per £	Reach per Ad
Third party anti Conservative	Snapchat	177	17605	10466331	594.51	59132
	Google	26	1650	130000	78.79	5000
Third party pro Conservative	Snapchat	n/a	n/a	n/a	n/a	n/a
	Google	n/a	n/a	n/a	n/a	n/a

Alongside the party pages previously examined (a total of 18 pages), 19 of the most active third-party Facebook pages were examined. The third-party pages on Facebook were sampled because of the huge number of pages that advertised on Facebook over the one-month campaign period (n = 2481). 2481 pages advertised on Facebook over the one-month campaign period, this is clearly unmanageable to examine. The page’s advertisements were examined across the one-month campaign from the 1^st November to the 12^th December. Apolitical third parties and government GOTV operations were excluded. There were far more remain third pages than leave and as such, there is a bias towards that position in these pages. Some of the third parties chosen (such as Capitalist Worker) were selected because they were particularly active in the final week of the election. As such a wide sample of types of third-party group were gathered. It is important to note that this group of pages is not totally representative of all the pages seen because it does not include local, regional or many other third-party pages. The sample is not designed to be representative, but instead give insight into the role third parties had.

A Facebook phenomenon?

Facebook third party political advertisers examined – Advance Together, Avaaz UK, Best for Britain, Campaign Against Corbynism, Capitalist Worker, City Action, Fair Tax Campaign, HOPE not hate, In Facts, Led by Donkeys, NHS for a People’s Vote, Parents’ Choice, Remain United, Right to Rent Right to Buy Right to Own, Save Brexit, Scientists for EU, Stop School Cuts, Vote for a Final Say, Working4UK.

	Number of ads	Impressions	Spend
AdvanceTogether	1444	2090402	65502
Avaaz UK	163	1795932.5	26382.5
Best For Britain	283	70538914.5	755058.5
Campaign Against Corbynism	60	462996.5	8246.5
Capitalist Worker	39	1424493	23643
City Action	21	1279994	19744
Fair Tax Campaign	95	1994981.5	30581.5
HOPE not hate	3471	16544195.5	249945.5
InFacts	49	2392975	22975
Led By Donkeys	2168	10763867	155016
NHS for a People’s Vote	78	3452979	34279
Parents’ Choice	32	1494993	18843
Remain United	1861	1722838	26038
Right To Rent, Right To Buy, Right To Own	64	2251987.5	30887.5
Save Brexit	107	822472.5	22122.5
Scientists for EU	4	1007497	10446.5
Stop School Cuts	649	11887674	81274
Vote for a Final Say	1482	2565305.5	100255.5
Working4UK	39	1387496	47594
Sum	12109	135881994	1728834.5
Average	637	7151684	90991

This small collection of the biggest third-party advertisers shows the huge scale of advertising these groups are generally engaged in. Some pages see enormous activity such as Led by Donkeys, Vote for a final say, Hope not Hate and Advance Together. It is interesting to see the number of adverts so high and yet spend so low, this tactical approach is strange as one would expect greater spend on a smaller number of adverts, however it appears that many of these third-party groups were explicitly targeting certain seats multiple times.

As such, a large quantity of adverts was needed. As Graph 14 shows, comparing directly these 19 third party pages with the 18 party associated pages

When examining the third-party pages by political position (Anti-Conservative and/or Anti-Brexit or Pro-Conservative and/or Pro-Brexit) there is a huge difference between political wings. Pages were – Pro Conservative/Brexit = 8 (Campaign Against Corbynism, Capitalist Worker. City Action, Fair Tax Campaign, Parents’ Choice, Save Brexit, Right To Rent Right To Buy Right To Own, Working4UK). Anti-Conservative/Brexit = 11 (AdvanceTogether, Avaaz UK, Best For Britain, HOPE not hate, InFacts, Led By Donkeys, NHS for a People’s Vote, Remain United, Scientists for EU, Stop School Cuts, Vote for a Final Say).

	Platform	Number of ads	Spend	Reach	Reach per £	Reach per Ad
Third party anti Conservative/Brexit	Facebook	10170	1,426,917	1221972745	86	12015
Third party pro Conservative/Brexit	Facebook	457	201,662	11119414	55.1	24331

There was clearly a concerted effort by remain and anti-Conservative pages to reach the public, using lots of smaller spend adverts to reach a colossal number of impressions. Nearly £1.5 million was spent by anti-Conservative third-party pages across 10k adverts, in contrast, pages supporting the Conservatives and Brexit featured a much smaller spend of £200,000 and 450 adverts. It is important to note here however that many of these pro-Brexit/Conservative pages were only advertising towards the end of the campaign meaning they were spending more money across a smaller period. As such the reach of pro remain pages was astoundingly large compared to pro-Conservative pages.

Nevertheless, it is apparent that the pro-Brexit/Conservative pages were reaching more people per ad using a different tactical approach.

But what about the reach of these third parties compared to official UK party pages? When comparing the 19 third party pages with the 18 party pages seen earlier it is clear that official parties still make up the lion’s share of impressions during the General Election. Nevertheless, third parties are nowhere near negligible, but instead a clear player in the online realm.

Official pages are clearly in the lead, however, the fact that so much reach is achieved by third party pages offers new questions for the operation of our democracy. Third party pages have enormous power, they had 38% of the impressions of official party pages, the central pages that advertise were seen by shadier third-party actors. These actors are less transparent and harder to hold to account, featuring unknown control or aims.

Party tactics are changing in how they are using social media; this is intersected by the role third party pages are now playing. This force has grown out of nowhere, with the role of unknown forces now a part of the online electoral map. Alongside the diversification of parties approaches, especially seen in the rising use of Google, we have seen the rise in difficulty in holding to account the forces that use targeted advertising. The current diverse tactics exhibited by the parties and third groups requires easier access for the media and academics to examine the advertisements being bought.

---

Self-regulated transparency – issues of research and accountability

The self-regulation epoch has led to some change on the part of social network platforms, albeit in the face of enormous pressure. However, the current landscape is not working as efficiently as it should. We are failing to hold the powerful to account, limit negative effects or understand effectively how parties are campaigning. Across research tools, data access, advertiser transparency and the media, the current self-regulation environment is not providing an effective solution to the problems micro-targeted communications and organic social media present.

Ineffective research tools and data access

Despite the growing influence of targeted advertising and social media on our politics, access for researchers to the data that drives these platforms is still hard to come by. Detail is lacking on the advert’s placed on parties’ pages, the videos that go viral or the targeted advertisements sent. Even though elections and referendums from the USA to India appear to show the importance of targeted advertising and social media, researchers still have mixed and limited levels of access. This is not just access to targeted advertising data either, access to even basic organic content data is also heavily restricted. In some cases, we have even seen a reverse of access over the last decade, as social media organisations have been shutting down avenues for collecting and analysing organic data. In the wake of the Cambridge Analytica scandal the platforms went into protective overdrive, in the process smashing systems such as Netvizz that researchers have been using for half a decade.

Across targeted and organic data, research access to the public and academia is either restricted or limited. This situation is effectively engendering a censorship war on knowledge as research areas from psychology to economics are denied access to organic data. Even within the study of political campaigning (the area given the most access) targeted advertising data access is patchy, inconsistent and very limited, while organic data is placed in the hands of a select number of journalists and think tanks. Instead of the platforms pursuing transparency in data access and operation, social networks are using transparency as a cover to keep certain systems and data obscure. This situation is untenable, and is bad for the public, the platforms and researchers. How can we examine or question posts that are “fake news” or appreciate whether a party has fulfilled the promises they made in the past, if the promises they have made are hidden from view?

What is required is a system of regulation that moves beyond transparency ‘rhetoric’ towards pushing social media platforms to have accessible data resources pertinent to public and researcher’s eyes. The public have a right to view organic and targeted posts they have been historically sent, while for researchers’ careful access to more complex targeted data must be made possible. This system would allow groups external to the social networks to hold both the platforms and the powerful who use them to account.

Current access to organic and targeted content is patchy and not comparable across the different networks. Given the lack of cohesive structure, the two core areas of targeted advertising and organic data are examined across the networks.

Facebook and Instagram

Targeted advertising

Facebook has gone the furthest of the social networks in opening data access, with rapid progress made over the last two years through the creation of an Ad Archive and Ad report. However, akin to the failure of their research collaboration Social Science One, Facebook have also failed to make adequate arrangements for data accessibility and as seen later have closed organic access points as they have opened targeted data.

Ad Archive/Library

Facebook offers access to targeted ads sent and some data associated with them via the Facebook Ad Archive launched in the UK from May 2018. The database originally only showed political adverts, however later in March 2019 the system was upgraded into the Facebook Ad Library. This resource allows people to see all the ads running in a country and offers tools to examine and search the adverts by sender.

During the 2019 General Election, the Ad Library was a useful tool that allowed for the rough analysis of how parties were using Facebook advertising as a campaign tool. More complex analysis required the data to be scraped from the library, this is because Facebook do not provide an easy way to download the advertisements. The database provides only rough integers of vital data points, this hampers academic research considerably as any results from analysis are muddied via the wide estimates Facebook provides. Advert spend, reach and the basic gender and age demographics targeted are the central data points offered, they do not provide effective clarity on the deeper aspects of how users are targeted or where these adverts are being sent.

Facebook themselves appreciated the lack of data points, as soon after the 2019 General Election they announced further improvements to the system[i]. Users will be able to view audience size in the Ad Library, alongside a better Ad Library search and improved filtering e.g. audience size, dates and regions reached. The implementation of regions and audience size is welcome; however, it is likely that akin to the other data points, Facebook will only provide the most limited data possible.

These changes are probably all we will see before the 2020 Presidential Elections; they represent a step in the right direction but there are still real issues with Facebook’s Ad Library as it is not an effective tool for either the public or researchers.

Positives of the Ad Archive

• Searchable and some level of accessibility for the public.
• High level of coverage of political advertisements.
• Generally, works well despite limited data points.
• Core elements are there including reach, gender, age and spend.
• Offers clear lines of who paid for the advert and where advert was from.

Problems with the Ad Archive

• Location, or postcodes not visible.
• Data not downloadable or workable.
• More complex elements of targeting such as via interests or other demographic characteristics are not visible.
• Regular delays of up to a full day between when ads start running and when they show up in the archive.
• Bugs and errors, for example on December 10^th, just two days before the United Kingdom went to the polls, Tristan Hotham with Roland Manthorpe of Sky News broke the story that 40% of the UK’s political advertisements vanished from Facebook’s Ad Library[ii].
• Bugs have been reported, during the EU parliamentary elections in 2019, the Ad Library API (application programming interface) faced various bugs, including an infinite loop — in which the library returned the same results repeatedly. The API system allows the creation of applications that access the features or data of an application.

Facebook Ad Library report

Facebook alongside the Ad Library offers a report that summarises the key data in an easy to use system. During the General Election many journalists used Facebook’s report to write on total spending and other large-scale metrics because the system offered quick headlines for those who did not have proper time to do more full analyses. The system compliments the Ad library well, but suffers from the same limitations as the library, as well as problems during the election with updates being very slow. Equally, when the Ad Library lost 40% of the UK’s adverts, all the report data was also incorrect.

Snapchat

Snapchat Political Ads Library

Following Facebook’s lead Snapchat began offering a csv. data set (data that can be used via excel and other tools easily) of targeted ads sent via the platform in September 2019. Each year is downloadable from their website, and although not as wide ranging as Facebook’s Ad Library, the data set is easy to use and work with.

Positives of Snapchat Ad data

• The data also includes some more complete data points including the postcodes of locations targeted that Facebook’s Ad Library does not offer.
• Downloadable as a csv. data file making it useful for academic researchers or journalists.
• Specific numbers on spend, impressions and region ID unlike Facebook’s wide ranges.

Problems with Snapchat Ad data

• The data does not offer workable data points on gender, age, spend or reach, the data is messy and needs improvement.
• During the election Boris Johnsons adverts geotargeting postcode data was wrong leading us to report a story incorrectly[iii]. Snapchat fixed this quickly, however it is unacceptable to release incorrect data at such an important time.

Google

Google Transparency report

Google have tried to be more accessible to researchers, again following Facebook’s lead they offer a library of adverts seen across Google search and YouTube. Given how important Google and YouTube have become for parties targeted advertising campaigns their creation of a transparency report is welcome, although again as with the other database offerings are very limited.

Positives of the Google Transparency report

• Offers data on when and where the adverts ran, including spend and reach.
• Can download individual ads or all ads via csv.

Problems with the Google Transparency report

• Lack of data points, less detailed than Snapchat or Facebook.
• Data was very slow to update during the 2019 General Election.

Organic social media

Issues of access plagues targeted advertising, but also is a major feature for organic content.

Facebook and Instagram

Social Science One was intended to allow university researchers access to organic data via Crowdtangle, but since Facebook’s failure to work with the partnership, there is currently no easy access for researchers to access organic Facebook content. This has led to academics fearing a return to a wild west of scraping data, a situation that will put users’ personal data in greater risk[iv]. Experts such as Deen Freelon expect data scraping to rise in practice again[v], an act that breaks the Facebook terms of service, but as the last avenue available it is researchers only course of action.

Currently, for organic content, Facebook only offers select partners access such as ITV or BBC. This is operated via Crowdtangle a tool that Facebook owns, that allows researchers to gather the posts and engagement a registered page has sent. Crowdtangle is an efficient service that collates organic posts from pages logged by users, data is complete anonymised and downloadable.

However, unless you are the BBC or a thinktank, no clear avenue for organic Facebook data is available for researchers as other API accessing tools such as Netvizz or Facepager have been curtailed. There is equally no access for the public, the posts that promised them things during the election are hidden from them. Crowdtangle currently only offers access to a restricted cadre of news organisations and think tanks. It does not offer the public the political posts party pages send, and access is institutional with almost no universities having access to the platform.

Facebook has closed data access to organic data while opening access to targeted ads. Why are the public allowed to see the ads they have been targeted with but not the posts of a page they follow? We are seeing fields of study interested in Facebook’s millions of public pages shut down. We are witnessing the implementation of a hierarchical and restrictive set of parameters that are stifling of research. This was not always the situation, previously Facebook had a more open API for organic content allowing several tools that could extract organic posts, their content and the engagement. A new balance needs to be struck.

Twitter

Twitter banned advertising[vi], and as such has no onus to provide any tool for researchers’ access for understanding political advertising. In terms of organic content Twitter is mixed in terms of accessibility, the API is still relatively open, but the platform has been getting more restrictive in recent years as its API is not exhaustive[vii]. Given this situation a lot of academic research has been undertaken on Twitter as it is easier to get post data to analyse, this has skewed the research body away from more important networks such as Facebook.

Snapchat

Snapchat does not provide an open API, thus although open in terms of targeted advertising the organic political posts sent by parties are very hard to extract and analyse. The only solution is scraping data.

Google

As a location centrally for advertising and not a social network, Google itself understandably do not place great effort into making organic data accessible, with YouTube access a more mixed affair. As such it is understandable that Google have been very careful to not allow access to researchers or the public to much organic data. YouTube’s API is still relatively open allowing for some interesting analysis of video channel links such as within Alt-right communities[viii], however Google data is far more closed. Other than a handful of tools such as Google Trends where you can examine search rankings, or some data maps or search rankings created around election time[ix], the platform offers relatively little access to data. Available data can however be downloaded as csv but given how important these platforms are becoming more should be done to open access to the public and researchers to understand political parties organic outreach activities via these two sources.

TikTok

TikTok appears to have no avenue for the access of organic content, the only solution is scraping.

Ineffectual accountability – The media and third estate being held back

As mentioned, the media and third estate have the best access to tools possible. Despite this, some organisations decided to collect their own advertising data due to the problems associated with the tools available. For example, Rory Cellan-Jones of the BBC called for screenshots of advertisements before the election. Only resources such as Who Targets Me provided complex data on who the parties were targeting.

Facebook/Instagram’s ad library and spending reports, Snapchat’s advertising data and Google’s spending reports were all used to produce reports from the Guardian, First Draft and BBC teams. However, these official tools were not enough to truly elucidate the campaign. The reports created were understandably limited by the data they were reliant upon, focussing on overall spend and a few cherry-picked examples of targeting, the news coverage seen across the election symbolise the difficult situation researchers find themselves in when analysing political advertising. The most effective studies scraped the Facebook Ad library or used their own data collected themselves, even then their ability to go behind the curtain of how the parties were targeting was heavily constrained. The inaction of social networks in providing effective data sources to examine the advertising sent during an election hampered journalists’ abilities to spot misinformation, fake news nefarious actors and to truly appreciate what happened with the election online.

The 2019 General Election saw the media engaging in analysis of the targeted campaign more than ever, but other than a few stories surrounding spending, the misuse of new content and organic engagement, the media was barely able to scratch the surface. There was simply too much advertising and organic content, ineffective tools and too few researchers to properly hold the parties and platforms feet to the fire. The failure of the media to make an impressive impact in appreciating the party’s online campaigns signals the need for superior tools and data access.

Legitimate parties and nefarious third parties

As seen in the rise of third-party pages advertising, platforms such as Facebook must make advertiser clarity stronger, currently people can create pages and start advertising instantly. Pages must disclose their funders but often individuals are hard to trace and mysterious as we saw in the case of ‘3^rd Party Limited’ a page that ran many ads during the 2019 General Election campaign from a non-descript position. Similarly, although Snapchat makes clear who the advertiser is, as the platform gains in popularity, it is clear they will need to do more to address the current issues surrounding advertiser information.

Alongside current efforts in advertiser transparency not addressing harm, the current self-regulation zeitgeist has meant that social media companies have found a new solution to transparency woes, the ‘control for users’ escape hatch has allowed the platforms to wash their hands of bad actors, deferring activity and protection onto the populace.

Facebook will be rolling out controls which will allow users on their platforms to choose how an advertiser can reach them with a Custom Audience approach or allow the user to choose to see fewer political and social issue ads. This feature builds on other controls in Ad Preferences they have released in the past, like allowing people to see fewer ads about certain topics or remove interests. Google chrome have improved cookie controls, but again akin to Facebook’s solution this is not necessarily a solution as users are not generally aware of how to alter settings and exclude themselves from cookie schemes. The user control approach is in part because Facebook and Google have decided to defer the protection of their users onto individual activity. The problem with this is awareness and action, many users will not be made aware of the choices and millions more will not engage with the changes, this in turn means they are in effect offering a status quo solution that will fail to alleviate concerns.

Conclusion

The 2016 Brexit Referendum, 2017 and 2019 General Elections all occurred within a regulatory environment that clearly is not working. The clandestine way the parties campaigned, the rise of third-party pages, the content that bordered on misinformation and our inability to examine the huge number of adverts sent, all highlight a system that is being taken advantage of. The problems we are currently seeing are potentially worse than we can appreciate as we do not have the access to data needed, nor the tools required to properly hold those using targeted advertising to account. Parties are bumping into the edges of legality and acceptability, with this a state of affairs that cannot be allowed to continue.

---

[i]https://about.fb.com/news/2020/01/political-ads/

[ii]https://news.sky.com/story/researchers-fear-catastrophe-as-political-ads-disappear-from-facebook-library-11882988

[iii]https://medium.com/@WhoTargetsMe/is-johnson-going-under-the-uxbridge-decapitation-campaign-37283c237df0

[iv]https://www.researchgate.net/publication/328525888_Computational_Research_in_the_Post-API_Age

[v]http://hussman.unc.edu/directory/faculty/deen-freelon

[vi] https://ads.twitter.com/transparency

[vii] https://developer.twitter.com/en/docs/tweets/search/api-reference/get-search-tweets

[viii] https://arxiv.org/abs/1908.08313

[ix] https://trends.google.co.uk/trends/story/GB_cu_wp70Nm4BAABYKM_en