Introduction
Political advertising in the UK is undergoing radical change, challenging a longstanding regulatory framework. Restrictions on political advertising have a long provenance, dating back to the nineteenth century, but in more recent times emphasis has shifted to the control of mass media campaigning tools. In terms of the normative basis of regulation, the system seeks, on the one hand, to protect freedom of expression, while on the other, it ensures the democratic system is not distorted by powerful interest groups. The ban on political advertising on broadcast media was for many years core to the regulatory framework, as a means of ensuring that there was no electoral ‘arms race’ in advertising expenditure.Footnote 1 A restriction on party spending has been another flank of regulation; the growth of national media campaigns tended to eschew strict limits on constituency-level campaign expenditure. This was remedied by the introduction of legislation in 2000 imposing national campaign spending caps supervised and enforced by a new regulator, the Electoral Commission (ELC).
While for a period relatively stable, in recent years it has been suggested that electoral law is in need of substantial reform to deal principally with the shift to digital advertising, particularly the use of data to target individual voters with tailored political messages (so-called microtargeting).Footnote 2 The case for reform is gaining traction, with the ELC publishing a raft of proposals, albeit modest in nature.Footnote 3 The controversies surrounding microtargeting are part of the broader debate on online disinformation (aka ‘fake news’), and it is often analysed in that context. In that regard, there have been a plethora of other reports and investigations which have touched on the issue of microtargeting, notably the House of Commons Digital, Culture, Media and Sport Committee's (DCMSC) inquiryFootnote 4 and the EU Commission's initiatives,Footnote 5 including the adoption of a code of practice.Footnote 6
Microtargeting also raises serious issues concerning individuals’ privacy and data protection rights.Footnote 7 Even the most casual observer cannot have escaped the press coverage, parliamentary scrutiny and regulatory action taken with respect to the alleged misuse of data in recent elections and referendums.Footnote 8 While of the upmost importance, for the purposes of this paper, the question of the legality of the procuring and use of personal data will not be discussed directly.Footnote 9
Rather, the central focus of this paper is on whether such techniques have the potential to distort the public discourse and subvert the democratic process, and if so, how should the law respond. The analysis here is on the regulation of political parties and campaign groups who seek to influence electoral outcomes directly, rather than broader forms of political expression online. Such a distortion may result from the unmediated nature of the messages. First, targeted advertisements can escape the attention of the media and the public such that political campaigns may make contradictory claims to different segments of the electorate. Second, messages may be designed to play to some voters’ prejudices and biases without any exposure to countervailing viewpoints. This may lead to a greater polarisation and fragmentation of society, and ultimately undermine the ability of voters to engage in meaningful democratic deliberation.Footnote 10 The evidence which exists is considered in more detail in the next section. One of the key questions is whether intervention can be justified on the basis of risk rather than proof of actual harm. A further problem relates to the need to avoid regulation which limits political discourse online. There are also inherent dangers in seeking to restrict or control the content of political speech.Footnote 11 Measures which seek to maximise transparency for voters, while exposing political messages to public scrutiny, should always be preferred over more restrictive controls. The policy dilemma arises when such measures prove to be ineffective.
In terms of the dimensions of the problem and possible solutions, there a number of options. First, there are measures to increase the transparency of digital messages, enabling voters to judge the credibility of targeted messages on the basis of their provenance and whether they have been paid for. The use of online repositories may also lead to greater scrutiny of the messages targeted. Second, reforms could be made to the reporting requirements political campaigners are required to follow, and tighter restrictions on advertising from outside the UK. There remains a third approach, which is to place substantive restrictions on online advertising. In designing any interventions, great care would be needed to minimise any limits on political expression. The case for substantive restrictions should not be ruled out, however, especially if other less interventionist measures prove to be unworkable or are subverted. The problems also raise classic concerns of regulatory theory, in particular: regulatory fragmentation, since responsibility for regulation is likely to be dispersed across different actors (government, agencies, and crucially the industry);Footnote 12 information asymmetries, given the techniques used are characterised by high levels of innovation; and the costs of regulatory failure, which are particularly high here given the potential harm to the democratic process.
1. A primer on microtargeting
This section explains the ‘mechanics’ of microtargeting, identifies the key actors, how it is used, and its possible effects, both negative and positive.Footnote 13 As a cautionary note, the negative effects of microtargeting should neither be assumed nor exaggerated. The literature overall tends to frame new innovations in political campaigning in pessimistic terms. The intention here is to expose the risks to enable a discussion of if and how to design appropriate legal responses.
Targeting of voters is by no means a new phenomenon. Political parties have for many years collected data on voters, dividing them into segments based on their characteristics (in particular, geographic, demographic and previous voting behaviour), with a view to designing specific political content for each segment.Footnote 14 The difference with microtargeting is the potential: to bring together data from various sources (or ‘datapoints’), to draw inferences about political beliefs and biases from seemingly unrelated data (such as internet browsing and purchases), to target messages to voters at a high level of granularity, to test the efficacy of campaign messages in real time, and to differentiate messages to particular voters. We also know, as elections become ‘datafied’, political campaigners are increasingly reliant on digital intermediaries, whose influence over the democratic process may be problematic.Footnote 15
(a) The key techniques of microtargeting
What follows is a brief sketch of the key methods of microtargeting, the different types of intermediaries involved, and their respective roles.Footnote 16
(i) Data collection and aggregation
Political parties have traditionally maintained information on individual voters, gathering information from public sources (principally the electoral register), together with their own canvassing data. Other data is gathered from various other commercial sources, including on: geographic location, basic demographic data (age, gender, income, education etc); search and online activity (which can be used to identify a voter's specific political interests or leanings, or seemingly less relevant information, such as online purchasing behaviour).Footnote 17 As has been the case for a number of years, growing out of commercial online advertising, there are a number of ‘data brokers’ supplying data for political advertising.Footnote 18 The platforms themselves also collect and aggregate data on their users, facilitating targeting. Data exchanges, which provide for the automated buying and placing of digital advertisements through real-time ‘auctions’ (so-called ‘programmatic advertising’), also exploit large amounts of data held on individuals.Footnote 19
(ii) Voter segmentation and profiling
Voter segmentation means ‘dividing the electorate into smaller blocks, and using different campaign methods for each segment’.Footnote 20 While not a new practice, it is becoming increasingly efficient due, first, to the high level of granularity of the data which is now collected and, secondly, the sophisticated methods of profiling which can be employed.
There are two principal means of profiling. Personal data can be used to predict political views and party allegiances, the probability of voting, and which issues are likely to have particular resonance with individuals (‘predictive’ profiling). More controversially, what is referred to as ‘psychographic’ profiling is increasingly employed, a technique whereby data analytics firms profile voters according to personality traits.Footnote 21 In terms of data collection, one method is to incentivise people to undertake personality tests, and then cross-reference results with other online or personal data.Footnote 22 Psychographic data can also be inferred from online behaviour more generally. Although the veracity and effectiveness of such techniques is currently a matter of debate,Footnote 23 its use in political campaigning is well documented, especially its potential to achieve greater precision of salient and persuasive messages.Footnote 24
While many of the techniques involved in microtargeting are controversial, ‘lookalike modelling’ has attracted particular opprobrium. This technique enables campaigners to reach new audiences on platforms (referred to as ‘peer groups’ or ‘persuadables’) due to their sharing similar characteristics or profiles to existing audiences (‘cloning’);Footnote 25 Facebook offers a ‘lookalike’ service which has been used by political campaigns.Footnote 26
(iii) Creating and testing personalised messages
As was noted above, a key advantage of microtargeting is the increased precision and efficiency of targeting. One of the reasons for this is the testing of messages on an iterative basis through the use of algorithms (so-called ‘A/B testing’). While campaign messages have long been subject to testing to maximise their resonance with voters (especially through focus groups), the use of social media and commercial advertising methodologies has resulted in a ‘step change in pace and scale’.Footnote 27 One of the disturbing effects of such technology is said to be the dislocation of the politicians from any effective engagement with, or understanding of, the electorate.Footnote 28
(b) The potential benefits and harms of microtargetingFootnote 29
Microtargeting may lead to a levelling of the political playing field and a greater diversity of viewpoints. Political parties benefit from more precise targeting, reducing costs and possibly lowering barriers for smaller parties to reach the electorate.Footnote 30 With a shift away from mass media, political campaigns may also become more diverse, as parties seek to engage voters on issues which previously had been regarded as of marginal relevance. Indeed, one of the lauded benefits of social media platforms (SMPs) is their facilitating the greater participation of individuals in political campaigns, and a wider prevalence of ‘satellite campaigns’ advocating particular electoral outcomes while being outside the control of the political parties themselves.Footnote 31 Voters in turn may become more engaged and likely to vote as campaign messages are tailored more effectively to their interests and concerns. Taken together, such effects may lead to greater electoral participation.
According to Gorton, given the ability of microtargeting to alter voter behaviour, citizens become ‘potential objects of control rather than autonomous political actors’.Footnote 32 Even if this is true, of more concern is the effect this could have on public discourse; the ability to segment voters and target individualised messages may lead to a fragmentation of political debate, and a polarisation of the electorate.Footnote 33
For a number of years now, partly because of their commercial models, concern has been expressed that the use of online intermediaries can lead to the related phenomena of ‘filter bubbles’ and ‘echo chambers’.Footnote 34 It is true to say that the evidence is ambiguous.Footnote 35 Some have found some modest effects,Footnote 36 while others have found vulnerable users may be particularly susceptible.Footnote 37 In any event, filter bubbles are materially different from political microtargeting. The former is a product of the platforms’ algorithms, optimising the salience of messages in order to maximise commercial gains from advertising. The latter is concerned explicitly with selecting the relevant audience, according to their political opinions, and tailoring campaign messages in order to influence their voting intentions.
Traditionally, the electorate have tended to gather information via the mass media, whereas microtargeting offers the ability to tailor (unmediated) messages to voters according to their pre-existing political preferences and biases.Footnote 38 This has the potential to fragment the public sphere by reducing the opportunities for voters to engage in a shared public discourse.Footnote 39 Instead of promoting a common conversation, campaigns tend to focus on ‘wedge issues’ – the promotion of emotive and superficial matters – which may distort public debate and lead to greater insularity.Footnote 40
Furthermore, as targeting becomes more efficient, there is an increased tendency to focus on those voters who are most useful to the campaign, excluding others. For this reason, microtargeting is particularly effective in first-past-the-post electoral systems where resources tend to be focused on key marginal seats.Footnote 41 ‘Political redlining’ is a technique employed by campaigns where they concentrate on ‘swing’ voters in target constituencies, ignoring others who are likely to support opposing parties, or are least likely to vote.Footnote 42 The danger is this exacerbates pre-existing political inequalities by ignoring voters who are least likely to be engaged.Footnote 43 Microtargeting may also lead to the suppression of voter turnout for opposing campaigns,Footnote 44 a practice referred to as ‘digital gerrymandering’.Footnote 45 While negative campaigning is not new, the increased efficiency of microtargeting increases the adverse consequences in terms of voter disengagement.
Another issue relates to the power of the intermediaries. In recent years, as politicians have realised the advantages of digital marketing, there has been a significant growth in big data experts and consultancies upon whom parties are reliant. Platforms themselves have developed services which are closely integrated with political parties’ campaign teams,Footnote 46 controlling access to the electorate with which political parties wish to communicate, their data, and targeting campaign tools.Footnote 47 Facebook, in particular, appears to be emerging as a ‘one-stop-shop’ for services related to microtargeting.Footnote 48 There are a number of possible dangers as intermediaries assume an increasingly powerful role as ‘gatekeepers’ over parties’ abilities to communicate with voters. In contrast to (some of) the traditional media, there is little incentive on intermediaries to maintain a healthy public debate or a level playing field for political parties; ‘political audiences are ultimately sold to the highest bidder’.Footnote 49 Another area for caution is the ability an intermediary has to refuse to deal with political parties whose viewpoints run counter to its own ideological standpoint.Footnote 50 One related area of disquiet is ‘political capture’. As politicians decide upon the regulatory environment in which intermediaries operate, the latter's increasing potential to shape electoral outcomes, in a setting which lacks transparency and public scrutiny, may result in the emergence of more favourable rules across all the spheres of their commercial activity, not simply those relating to political advertising.Footnote 51 Conversely, a party which advocates greater regulation that is likely to damage the commercial interests of intermediaries may be discriminated against.Footnote 52
Of all the possible risks from microtargeting, the most serious is the distortion of the democratic process by undermining public discourse. It may do this in two key ways. First, the collection and use of data may have a ‘chilling effect’ on voters’ behaviour online, deterring them from expressing their opinions freely.Footnote 53 Secondly, microtargeting may bypass the traditional media, with political messages reaching voters without being mediated in any meaningful way,Footnote 54 whereas previously a campaign might have been deterred from propagating false or misleading information due to the potential adverse reputational effects, as messages become more targeted and differentiated, misleading information is less likely to be exposed. As Gorton cautions ‘microtargeting produces messages that can more easily fly under the radar of the press and the broader public, markedly increasing their power to mislead and misinform viewers with impunity’.Footnote 55
The lack of transparency relating to the origin of an advertisement is likely to compound this problem. A ‘dark ad’ is a paid-for political message where only the publisher of the advertisement and the individual(s) targeted can see the message (unless she chooses to share the message with others), thus making it difficult to hold campaigners to account for any false or misleading information, and facilitating contradictory messages or policy claims. The term is also used to describe the situation where the target audience is unable to determine the provenance of the message and whether or not is it paid for. Conversely, technologies such as ‘bots’ can be used to artificially increase the apparent popularity of SMP accounts from which messages are sent, and fake accounts and paid trolls may also be used to deceive voters about the true level of a campaign's support (so-called ‘astroturfing’Footnote 56).
2. The current legal regime for the regulation of political advertising
This section gives a brief overview of the current rules which seek to control political advertising. First, there are controls on campaign expenditure which limit advertising spending by political parties and others. Secondly, the UK has maintained a longstanding prohibition of political advertising on television and radio. Thirdly, the ‘imprint’ requirement, which currently applies only to printed materials, aims to increase the transparency of political advertising.
(a) Campaign expenditure rulesFootnote 57
Electoral law sets limits on the amount of money campaigners can spend on campaigns. There are two key pieces of legislation. Individual candidates’ expenditure is governed by the Representation of the People Act 1983 (RPA). The spending limits at constituency level are relatively low.Footnote 58 For this reason, parties would normally prefer to report advertising spending under the more generous national spending limits, even where a limited number of marginal seats are targeted.
National campaign expenditure, the focus here, is governed by the Political Parties, Elections and Referendums Act 2000 (PPERA), which also set up the ELC. Before PPERA there were no restrictions on national spending. The ELC has civil sanctions for unlawful behaviour, including fines (up to a maximum of £20,000). However, criminal offences can only be pursued by the police.Footnote 59 The ELC's principal role is to enforce the reporting requirements for campaigners and ensure spending limits are not exceeded. Spending limits apply to expenditure during the ‘regulated period’ preceding the vote (eg 12 months for UK parliamentary elections, four months in respect of the 2016 EU referendum). The spending limit for a party contesting all UK parliamentary seats would be £19.5 million.Footnote 60
In terms of national spending for the 2017 general election, of a total expenditure by registered parties of £39,147,932, around a quarter (£10,052,016) was directly attributable to ‘advertising’,Footnote 61 though this may be an understatement as advertising expenditure (including digital campaign spending) may fall within other reportable categories.Footnote 62
Third party spending is also regulated by PPERA.Footnote 63 Where a non-party campaigner spends over a prescribed limit within the ‘regulated period’ (£20,000 in England), with the purpose of promoting particular candidates and/or policies aimed at influencing the outcome of an election, they must register with the ELC and comply with its reporting requirements on spending.Footnote 64 Non-party campaigners are only entitled to register if they are based in the UK; those not entitled to register must not spend above the prescribed spending limits.
Referendum spending under the EU referendum was also governed under PPERA.Footnote 65 Here the ELC designated two lead campaign groups which were each subject to a spending limit of £7 million. Other registered campaigners (those spending over £10,000) were permitted to spend up to £700,000. Political parties were also entitled to register with the ELC provided they specified the outcome they were campaigning for; spending limits were based on the proportion of the vote in the preceding general election.Footnote 66
In addition to spending, PPERA also put in place restrictions on donations and loans to political organisations (in both elections and referendums), which generally must originate with a person or organisation from the UK (where the amount exceeds £500). The ELC referred two individuals to the National Crime Agency concerning alleged offences relating to illegal sources of funding and loans in the EU referendum.Footnote 67
In terms of spending on digital advertising, the ELC noted that social media had been used extensively in the EU referendum campaign, following the trend in the 2015 general election.Footnote 68 Digital advertising spend (as a share of total advertising spend) has increased rapidly over the last decade: from 0.3 per cent in 2011, to at least 42.8 per cent in 2017.Footnote 69
(b) The prohibition on broadcast political advertising
The regulation of political advertising in the UK is the subject of a strict dichotomy: a complete ban with respect to broadcasting, and little or no regulation of everything else.
The statutory prohibition (now contained in the Communications Act 2003)Footnote 70 has been in place since 1954 (the beginning of commercial television). It applies not only to organisations which have political objectives, but all advertisements which have a ‘political end’, which has a very broad definition including influencing public opinion on a matter ‘public controversy’.Footnote 71 Broadcasters are, however, required to offer some campaigners free party political broadcasts.Footnote 72 The UK's prohibition has been challenged, most notably in Animal Defenders International (ADI) v UK (the ADI case),Footnote 73 where the Grand Chamber of the European Court of Human Rights (the Court) upheld the ban (by a majority of nine to eight).Footnote 74 This case is particularly pertinent for the future regulation of digital political advertising, as the Court held a general prohibition may be permissible where less restrictive means of controlling advertising could lead to uncertainty or abuse.Footnote 75 This point is elaborated upon below.
(c) The current ‘imprint’ requirement
At present, there is a requirement for an ‘imprint’ to appear on all printed election materials, ranging from unsolicited mail to newspaper advertisements.Footnote 76 The imprint requirement serves two key purposes. First and foremost, it enables voters to judge the provenance and credibility of campaign materials. Secondly, it assists the ELC in monitoring and enforcing electoral spending rules.
‘Election material’ has a very broad meaning, defined as ‘material which can reasonably be regarded as intended to promote or procure electoral success’ of particular political parties or their candidates.Footnote 77 It may include the advocacy of particular policies without mention of a particular candidate/party,Footnote 78 or material ‘prejudicing the electoral prospects at the election of other parties or candidates’.Footnote 79 The ‘imprint’ must specify ‘the name and address of the promoter of the material’, and (where not the promoter) the name and address of any person on behalf of whom the material is being published.Footnote 80 It applies to all materials, not merely those of campaigners registered with the ELC, and there are no minimum spending thresholds. The rule's application is not limited to the registered period.Footnote 81 Printed materials in referendum campaigns are subject to similar requirements.Footnote 82 The ELC has used its civil enforcement powers to enforce the imprint requirement,Footnote 83 although failure to comply also attracts criminal liability.Footnote 84
In summary, the current electoral rules which govern political advertising are in need of reform to reflect the challenges of the digital environment. First, digital advertising expenditure has increased rapidly, but it is difficult to monitor directly given the opaque reporting requirements. Secondly, the ban on broadcast advertising is increasingly difficult to justify given the shift of campaigning to the online environment. Thirdly, the imprint requirement does not apply to digital election materials, leaving a serious gap in the electoral rules. The following section now considers how these problems and others can and should be dealt with.
3. The options for reform
This section explains the potential reforms to electoral law to deal with the challenges of digital campaigning, especially those arising from the use of microtargeting. First considered are relatively modest changes to the reporting of campaign expenditure, to enable the ELC to monitor and capture all of the attendant costs of digital campaigning. One live controversy is foreign spending on campaigns, and the apparent gaps in the present regime to deal with it effectively. Secondly, are two measures which are aimed at increasing transparency: a digital imprint requirement, and online repositories for campaign messages. Thirdly, and more radical, is the possibility of placing substantive restrictions on microtargeting. These range from modest interventions (eg stopping lookalike profiling) to a complete prohibition of targeting. As is discussed below, placing limits on political advertising will need to be carefully calibrated with Article 10 of the European Convention on Human Rights (ECHR). Another possible means of mitigating some of the harms of microtargeting is to regulate the ‘truthfulness’ of content. Finally, some of the issues around regulatory design need to be confronted.
(a) Reforming spending rules
The ELC has made a number of proposals to both increase the transparency of spending on digital campaigning, and to ensure all associated costs are captured.Footnote 85 First, it argues for greater specificity in reporting. One particular problem is securing a correct division of spending between local and national campaigns. Clearly there is an incentive to report expenditure against the national spending limit, which is considerably higher, but digital communications are often focused on swing voters in marginal seats. To address this issue, the ELC proposes more specific information including on the messages used, the amount spent on each campaign, and which parts of the country the message was targeted at.Footnote 86 Spending categories would also need to be revised, as PPERA currently contains no separate category for digital spend.Footnote 87 A second reform is to revise the rules to capture all digital campaigning costs. One seeming anomaly is money spent by political parties (and lead referendum campaigns) on direct staff costs not counting towards spending limits.Footnote 88 It is clear parties have used permanent employees, and have taken on additional staff or consultants, at significant cost, for their digital campaigns.Footnote 89 The ELC also stress that the spending limits do currently include all the costs of developing and using databases; this applies even if the original purchase of hardware and software falls outside the regulated period.Footnote 90 The ELC notes that resources are needed to make and manage databases of supporters which can in turn lead to organic reach; these costs should count towards spending limits.Footnote 91 So too the costs of bots and paid trolls.Footnote 92
One set of reforms the ELC has argued for relates to information and reporting requirements. Given the immediacy and ubiquity of digital campaigning, it stresses that it needs to be able to investigate campaigns in ‘real time’, not just after a vote as at present.Footnote 93 The reach of its powers to compel the provision of evidence also needs to be extended to third parties, including social media platforms.Footnote 94 Another of the major problems is the lack of effective sanctions the ELC has. At present, the maximum fine it can impose is £20,000 for each offence, which the ELC considers ‘risks becoming a cost of doing business’, with no effective deterrent value.Footnote 95 This is particularly the case for referendums, where reputational damage is far less likely to be an issue for non-party campaigners.Footnote 96
Another particular area of concern identified by the ELC is foreign spending in UK political campaigns. The only individuals or groups who can give money to parties or campaigns (or register as non-party campaigners) are people who live or are registered to vote in the UK and organisations based here.Footnote 97 At the time PPERA was passed, there was limited potential for foreign actors to purchase campaign advertising.Footnote 98 The internet has certainly made cross-border communication easier and cheaper for non-UK organisations.Footnote 99 There exists a surprising lacuna in the political finance regime: while there is a ‘general principle’ against foreign spending, there is no ‘explicit’ ban.Footnote 100 Where a non-UK individual or organisation spends below the minimum levels requiring registration, there is no specific infraction of the UK electoral laws.Footnote 101 Even were this apparent loophole to be removed, there would remain serious scope for abuse.Footnote 102 While monitoring foreign interference in political campaigns will engage the intelligence and security services,Footnote 103 in the view of the ELC, social media platforms must put in place mechanisms to ensure those paying for political advertising in the UK are actually based, or registered to vote, here.Footnote 104 As the ELC noted, such arrangements had been put in place by Google, Facebook and Twitter in advance of the US mid-term elections in 2018.Footnote 105 While there were reported problems with the schemes in the US,Footnote 106 Facebook has subsequently announced a new enhanced registration procedure which requires those placing political advertisements on its platforms to prove they are based in the UK.Footnote 107 All this points to the ELC's increased reliance on intermediaries to ensure UK spending restrictions are enforced.
(b) Extending the imprint requirement to digital media
The apparently simple proposal to extend the ‘imprint’ rule to digital media has gained increasing traction.Footnote 108 It represents a longstanding policy of the ELC, was recommended by the Law Commission in 2016,Footnote 109 received the broad support of the DCMSC,Footnote 110 and the Committee on Standards in Public Life endorsed the reform.Footnote 111 Arising out of the Government's response to the latter's report, the Cabinet Office launched a consultation on the issue,Footnote 112 and a response to the consultation in May 2019.Footnote 113
The key challenge is to create a requirement which promotes transparency, while not having a chilling effect on online political discourse by ordinary citizens.Footnote 114 Previous experience has demonstrated the difficulties in striking the appropriate balance. The Scottish Parliament had introduced a digital imprint rule for the independence referendum in 2014.Footnote 115 While the intention was for the rule to apply only to campaigners, not individuals expressing personal opinions over social media, there was considerable confusion over its scope.Footnote 116 The problem appeared to stem from the broad definition given to qualifying materials (those ‘relating to’ the referendum), rather than the narrower formulation used for print (those ‘reasonably regarded’ to intend to promote or procure a particular outcome).Footnote 117 While the Law Commission advocates the latter approach, in addition it suggests a ‘reasonably practicable’ defence for digital media.Footnote 118
The Government's consultation raised a number of other issues. Recognising the dangers of over-inclusiveness, it floated the option of a spending thresholdFootnote 119 and suggested the digital imprint requirement should not apply outside of the regulated period, as is the case for the print equivalent.Footnote 120 The Government now appears to favour parity with the imprint rule in these two respects.Footnote 121 It also posed the question of whether certain types of digital messaging ought to be prioritised over others.Footnote 122 No firm decision has been taken on this, though it observes that a differentiated approach may result in confusion.Footnote 123
The issue of who should bear the responsibility for including an imprint, and liability for failing to do so, also needs to be settled.Footnote 124 The potential addressees of a digital imprint rule could include, in the context of a ‘tweet’: the creator of the material, the advertising agency, the original tweeter, or an individual retweeting the message.Footnote 125 Again the Government appears to be equivocating on this issue, noting the tension between ensuring that any rule promotes transparency for voters while not stifling the debate.Footnote 126
The consultation also asked for views on the form an imprint should take, especially whether the imprint should appear as a ‘pop up and hover’ text, on the message itself, or whether inclusion of the author of the material on their social media ‘bio’ would be sufficient.Footnote 127 The Government again appears to be postponing any real engagement with these difficult questions, stating only that any rule would need to be platform neutral.Footnote 128
Clearly the issue of enforceability needs to be considered, given the ubiquity of social media. In response to the Law Commission's consultation, it is noteworthy that the Association of Chief Police Officers entered serious reservations over the feasibility of a digital imprint rule.Footnote 129 The ELC suggested criminal liability ought not to attach to non-compliance, though the Law Commission demurred.Footnote 130 The Government did not engage meaningfully with the question of enforcement, beyond asking whether current enforcement powers are sufficient.Footnote 131 In its response to the consultation, the Government merely noted that substantial fines could be a means of ensuring compliance.Footnote 132 Perhaps the major impediment to effective implementation of any digital imprint rule will be the enforcement resources available to the ELC (and the police).
There are clearly a number of difficulties in applying an imprint rule to digital communications and any regime would need to be ‘future-proofed in expectation of developments in media and technology’.Footnote 133 Whilst the Government appears to be asking (some of) the right questions, providing clear answers on appropriate rule design is proving problematic, and it is obvious a simple read across of the print rule is not appropriate, especially if priority is given to the need to protect political discourse online. Furthermore, while compliance among those who are registered with the ELC may well be high, there remain serious enforcement issues where organisations outside of the UK seek to influence campaigns.
The debate over digital imprints is not limited to the UK. Similar proposals have been considered in the US, with the introduction of bipartisan legislation (the ‘Honest Ads Act’) into Congress.Footnote 134 The US Federal Election Commission has also launched a consultation on extending imprint requirement to online advertisements.Footnote 135 The EU's Code of Practice likewise advocates increasing transparency, requiring ‘public disclosure of political advertising’ online,Footnote 136 though it also notes the need to avoid restrictions on political discussion or the publication of public opinion.Footnote 137 What is significant is the response of the social media platforms. It appears the threat of legislation has resulted in both Facebook and Twitter implementing an imprint requirement in the US,Footnote 138 and this has been extended by Facebook to the UK (using the US term ‘disclaimer’, rather than imprint).Footnote 139 Perhaps this may yield more evidence on how to design a digital imprint rule which does not have the paradoxical effect of suppressing democratic discourse online.
(c) Database(s) of microtargeted advertisements
The provision of an online repository is a further initiative aimed at increasing the transparency of digital advertising. According to the Independent Commission on Referendums (ICR), such a database would serve two purposes: first, it would enable the ELC to assess online spending and identify any discrepancies in spending returns; and, secondly, it would increase the accountability of campaigns, exposing instances where contradictory messages are being targeted at different groups or where adverts are using inappropriate or harmful messages.Footnote 140
Social media companies have already committed to doing this individually, as part of their initiatives to tackle ‘dark ads’.Footnote 141 For example, Facebook has put in place a repository for advertisements served over its platform which are ‘related to politics or issues of national importance and target the UK’;Footnote 142 those which do not carry ‘disclaimers’ and are reported by users to be of a political nature will also be added to the database.Footnote 143 While this may increase the transparency of online campaigning, the database will only carry limited information on targeting.Footnote 144 The EU code would appear to require more, stipulating that users should be given information ‘to understand why they have been targeted by a given advertisement’.Footnote 145
While the platforms’ moves to create repositories of political advertisements has been welcomed, a number of bodies and organisations have questioned whether this can be an effective substitute for a central public register. The ICR points to possible problems if each platform follows its own approach, with the resulting information being ‘confusing and disjointed’.Footnote 146 Instead, it stresses the need for coordination and cooperation between firms, government, parliament and the regulators.Footnote 147 The DCMSC, which supports calls for an independent public register,Footnote 148 also suggests the Government should impose ‘transparency requirements’ on SMPs, ensuring data on political advertising ‘are clear and easily searchable, and identify the source, explaining who uploaded it, who sponsored it, and its country of origin’.Footnote 149 The ELC welcomes the initiatives of the platforms, although it appears less committed to a central register.Footnote 150
(d) Restricting digital campaigning
This brief section, in the light of there being no case in point, discusses the principles enunciated by the Court in the ADI case and others, and the extent to which it is possible to place restrictions on microtargeting consistent with freedom of political expression under Article 10 of the ECHR.Footnote 151 It then goes on to reflect upon some of the restrictions which could be placed on digital campaigning, including microtargeting.
(i) Freedom of political expression and the Strasbourg jurisprudence
As the Court has confirmed on numerous occasions, political expression, including advertising, benefits from the highest level of protection under Article 10,Footnote 152 with States enjoying a narrow margin of appreciation in determining whether a restriction is necessary to address a pressing social need.Footnote 153 Clearly, any restrictions on targeted advertising is likely to engage Article 10.Footnote 154 Nevertheless, the Court has repeatedly stated that restrictions on political advertising are permissible to prevent a distortion of the democratic process.Footnote 155
One of the crucial issues in the ADI case concerned the availability of less restrictive means in the alternative to a general prohibition on political advertising over broadcast media. First, the Court was persuaded, in determining the proportionality of a measure, that it must ‘primarily assess the legislative choices underlying it’.Footnote 156 Secondly, in the context of prohibition, it was appropriate for a state to ‘take into account the risk of abuse if a general measure were to be relaxed’.Footnote 157 Further, where a relaxation would ‘give rise to a risk of significant uncertainty, of litigation, expense and delay, as well as of discrimination and arbitrariness’,Footnote 158 these were factors relevant to the UK's assessment of whether a general measure was preferable.Footnote 159
One of the arguments in favour of the prohibition turned on the particular potency of broadcasting advertising relative to other forms, including online. The Court accepted the differential treatment of broadcasting still to be logically cogent, recognising ‘the immediate and powerful effect of the broadcast media, an impact reinforced by the continuing function of radio and television as familiar sources of entertainment in the intimacy of the home’.Footnote 160 In contrast, ‘the choices inherent in the use of the internet and social media mean that the information emerging therefrom does not have the same synchronicity or impact as broadcasted information’.Footnote 161
Whether the Court's observations on the relative power of broadcasting advertising stand up to scrutiny, both then and now, is a matter of debate. As early as 1998, the Committee on Standards in Public Life (the Neill Committee) noted that the advent of the internet and digital communications could render the broadcasting ban out of date.Footnote 162 However, rather than justifying its removal, the Committee called for the legislation governing political advertising on broadcast media to be reviewed ‘to ensure that its reach is sufficiently wide to block attempts at evasion by new modes of communication’.Footnote 163 More recently the ICR suggested that social media may have overtaken broadcasting as the most influential form of media, it being ‘far from clear that social media lack “intimacy” or that exposure to advertising online is a matter of consumer choice’.Footnote 164 So if a ban can be justified when broadcasting was the most influential media, why cannot the same be true in principle for its successor?
The immediate problem with applying the reasoning of the Court in ADI to digital campaigning is its apparent inconsistency with previous case law, especially the TV Vest judgment.Footnote 165 Nevertheless, the Court has accepted that States may impose restrictions on political advertising to protect the integrity of the democratic process. While the rights to both freedom of expression and free elections, ‘form the bedrock of any democratic system’ and are ‘inter-related and operate to reinforce each other’,Footnote 166 the Court has acknowledged that there are certain circumstances where the ‘two rights may come into conflict and it may be considered necessary, in the period preceding or during an election, to place certain restrictions, of a type which would not usually be acceptable, on freedom of expression’.Footnote 167
Further, as the ADI case demonstrates, in judging the necessity of a restriction, the Court has had regard to the potential for abuse if a less restrictive mechanism were put in its place. So, in the context of microtargeting, if measures aimed to mitigate some of its harmful consequences proved to be unsuccessful, there is a stronger justification for more robust mechanisms (perhaps even a ban). The behaviour of those who bear the responsibility for making a system work – especially campaigners and intermediaries – and their willingness to ensure any regulatory solutions are effective and workable are clearly relevant to any evaluation of whether more restrictive means are necessary in the future. This raises the question of what types of further restrictions could be contemplated?
(ii) Possible restrictions on microtargeting
There a number of ways in which microtargeting could be subject to substantive restrictions in order to mitigate some of its more harmful potential, though there is as yet little consensus around any. Notably, as an industry body, the Institute of Practitioners in Advertising (IPA) has called for a temporary moratorium, with the prospect of an agreed minimum limit upon the number of voters sent an individual message.Footnote 168 Explaining its reasoning, the IPA opined that the ‘open, collective debate’, upon which politics and the ‘public sphere’ depends, is being circumvented by microtargeting.Footnote 169 The ICR considered the idea of a total ban on online political advertising, as an extension of the UK's approach to broadcast media.Footnote 170 While restricting political speech, it observed that campaigners could still communicate with voters through ‘organic’ content, with online reach being determined by their level of public support rather than advertising spend.Footnote 171 It cautioned against such a move, however, on the basis it would lead to disparity of treatment between print and digital advertising that would be difficult to justify.Footnote 172 Instead, it recommended a review of political advertising across all media.Footnote 173
The DCMSC was similarly equivocal, arguing the ELC should establish a code of practice on microtargeting, perhaps limiting its use, during the regulated period, to those registered with the ELC.Footnote 174 In addition, the Committee argued the ELC (and Information Commissioner's Office (ICO)) should consider the ‘ethics’ of SMPs selling lookalike audiences to advertisers during the regulated period (in particular, whether users should have the right to opt out from being included in lookalike profiling).Footnote 175 Elsewhere it rejected an outright ban, arguing that microtargeting ‘when carried out in a transparent manner, can be a useful political tool’.Footnote 176
At present, it appears there is no clear appetite for substantive restrictions on microtargeting, though the political consensus may shift, especially if other measures aimed at its control, particularly those focused at increasing transparency, do not prove to be successful. What is perhaps surprising at this stage is the lack of a clear signal to intermediaries that unless its harmful effects are mitigated, microtargeting will be curtailed. This may blunt the potential leverage government and the regulators have over the industry to achieve meaningful reforms.
(iii) Regulating for the ‘truthfulness’ of digital campaigning
One of the core objections to microtargeting is that it enables campaigners to escape public scrutiny of the content of their political messages. This raises the issue of whether campaign content should be regulated, particularly the factual basis for any underlying claims being made?Footnote 177 For a number of years, it has been the position that there is no direct regulation of the ‘truthfulness’ of campaign materials.Footnote 178 The ELC has consistently rejected any such role for itself.Footnote 179
The Advertising Standards Agency (ASA) likewise does not have a role in regulating election and referendum advertisingFootnote 180 (but it does oversee adverts relating to some non-party political subjectsFootnote 181). Before 1999, party political advertising did fall within its remit, but some of the general rules did not apply, specifically those prohibiting misleading advertisements, making unfair comparisons or unfairly attacking or discrediting other ‘products’. In the ASA's opinion this ‘part in, part out’ approach risked bringing ‘advertising in general into disrepute with the public’.Footnote 182 It was also concerned that, given the short campaign timeframes, it was unlikely any decision could be made before an election's conclusion.Footnote 183 More significant was the lack of a political consensus between the main parties to bring advertising wholly within the code.Footnote 184
The Neill Committee in 1998 ‘exhorted’ the political parties to formulate a bespoke code of practice, in consultation with the advertising industry, although it noted any such code would ‘have to be robust enough to be effective in the heat of a general election campaign’.Footnote 185 The government of the day was robust in its rejection of the (soon to be formed) Electoral Commission as a possible regulator:
The Electoral Commission has been canvassed as a possible alternative regulatory body. The Government, however, sees dangers in conferring such a role on the Electoral Commission. Adjudicating over complaints about political advertisements would inevitably draw the Electoral Commission into the party political arena in a way that could compromise its reputation for even-handedness and independence.Footnote 186
It did not provide any alternative candidates.Footnote 187
The ELC revisited the question of if and how to regulate political advertising in 2004.Footnote 188 It ruled out a statutory code, given the freedom of expression concerns, and the scope of any voluntary code could not, in its view, reach the truthfulness of campaign messages: ‘it would seem inappropriate and impractical to seek to control misleading or untruthful advertising, given the often subjective nature of political claims’.Footnote 189 It also rehearsed many of the practical problems with trying to enforce a code – the need for expeditious decisions, the potential for spurious complaints, the need for an independent adjudicatory body (rejecting itself as a candidate, together with the ASA model), and the lack of any effective sanctions. Again, the biggest obstacle to the adoption of a regulatory framework was the lack of consensus among the political parties.Footnote 190 This apparent political impasse has now persisted for over two decades.
While there currently exists no regulation of political advertising content, the ELC and ASA do nevertheless receive a substantial number of complaints on political advertising during political campaigns (eg the ELC received in excess of 1,000 complaints from the public concerning the accuracy of campaign literature during the EU referendum).Footnote 191 This appears to suggest that there is a public demand for regulation, as is clear from public opinion research.Footnote 192 Indeed, in evidence before the DCMSC, Guy Parker (CEO of the ASA) accepted there is a general public consensus that the content of political advertising should be regulated; the ‘difficulty is by who and how’.Footnote 193 Regulation would also need ‘buy-in’ which has not been forthcoming from political parties.Footnote 194
Another possible solution is to require social media firms to regulate the content of political messages on their platforms. As the ICR has observed, in the broader context of disinformation, this carries potentially ‘dangerous consequences for freedom of expression’ with SMPs, especially in the shadow of sanctions, having an incentive ‘to err on the side of caution’ rather than carefully balancing harm against freedom of expression.Footnote 195 Alternatively, the EU Commission envisages a greater role for ‘fact-checkers’, with the creation of an independent European network.Footnote 196 Some have gone further, calling for the creation of a ‘truth commission’ to intervene to expose misleading campaign claims.Footnote 197 Others have canvassed the idea of pre-clearance of factual claims made in political advertisements.Footnote 198 While fact-checking may have an important role to play, practical problems abound, not least how to reach the voters who were originally targeted?
The simple hard fact is that there is unlikely to be the political desire or will to put in place a robust regulatory regime for political advertising content while, as microtargeting increases in scale and ubiquity, the need and demand for it continues to grow. Regulation is probably always going to be an ineffective substitute for the glare of public and media scrutiny which microtargeting tends to obscure.
(e) The regulatory perspective: ‘speaking softly carrying big sticks’?Footnote 199
It has already been noted that social media companies have implemented a number of reforms to guard against the distortion of the political debate in the UK and beyond. In particular, progress has been made on the display of ‘disclaimers’ and the provision of online repositories. These initiatives have been broadly welcomed, though the ELC has stated that if the industry's ‘voluntary’ proposals do not work, the Government should consider direct regulation.Footnote 200 On both these reforms, the Government has committed to including them in a code of practice,Footnote 201 as part of its broader initiatives to deal with ‘online harms’.Footnote 202 The code, which will also include fact-checking requirements,Footnote 203 would be backed up by a regulator with significant enforcement powers (including the ability to impose substantial fines and, as a last resort, ‘ISP blocking’ powers for non-compliant platforms).Footnote 204
The EU Commission's efforts in implementing its Code of Practice have also become a focal point for reform. In anticipation of the European Parliament elections of May 2019, the Commission stepped-up its monitoring of the Code's implementation, with requirements on its signatories to report monthly on progress.Footnote 205 The Commission has been highly critical of the major platforms, for example, in failing to provide ‘hard data’.Footnote 206 Nevertheless, the industry has made significant moves to increase transparency.Footnote 207 Reporting after the conclusion of the elections, the Commission was able to announce progress, while leaving open the option to legislate in the future.Footnote 208
While not the focus of this paper, data protection and privacy laws may be other regulatory vehicles to deal with microtargeting.Footnote 209 Under the General Data Protection Regulation (GDPR)Footnote 210 the processing of data pertaining to ‘political opinions’ (a category of ‘sensitive data’) is prohibited subject to strictly limited exceptions.Footnote 211 It also contains a public interest exception, permitting the processing of data on political opinions ‘in the course of electoral activities’.Footnote 212 According to the ICO, where political campaigners make use of data analytics, aggregating data from SMPs and other sources, they must inform the data subjects, although a public statement on the organisation's website may suffice.Footnote 213 The EU Commission's guidance notes ‘when lifestyle data brokers or platforms collect data for commercial purposes, that data cannot be further processed in the electoral context’.Footnote 214 Similarly, the European Data Protection Board (EDPB) recently reiterated the requirement of explicit consent for automated profiling.Footnote 215
There are also restrictions on campaigning by political organisations under the ePrivacy Directive.Footnote 216 As a form of direct marketing, ‘unsolicited’ political communications may require consent.Footnote 217 Since 2009, that Directive also requires prior consent to the storage of cookies and other tracking technologies on end users’ devices.Footnote 218 There is currently a draft ePrivacy RegulationFootnote 219 which, if adopted, would significantly restrict the use of ‘metadata’, a matter of particular controversy.Footnote 220
Both the GDPR and ePrivacy rules offer the potential to substantially restrict microtargeting, although much depends upon whether data protection authorities have the resources and incentives to enter an inherently political arena.Footnote 221 Nevertheless, recent experience in the UK has shown an increasing willingness of the ICO to intervene, both in individual casesFootnote 222 and the broader policy debate, with the Information Commissioner arguing for an ‘ethical pause’ on microtargeting.Footnote 223 The critical question here is the extent to which it is desirable to use data protection and privacy laws as a means of regulating microtargeting. There is the possible danger that freedom of political expression will not be given appropriate weight when it comes to the careful balancing of rights necessary in designing appropriate restrictions on digital campaigning.
It appears then that there is an emerging consensus on the need to regulate online advertising, and the industry has made efforts to ensure that political campaigning on its platforms is more closely aligned to the underlying norms of electoral law, especially as regards transparency. Indeed, in this regard, it is significant that Mark Zuckerberg (the CEO of Facebook) has called for legislation to regulate digital campaigning.Footnote 224 The emerging solution is a system of co-regulation, whereby government and regulatory agencies frame principles and guidelines at a formal level, with implementation resting with the industry. The likelihood of such measures achieving success depends upon the credibility of threats to intervene with sanctions, and even direct regulation, should industry-led initiatives fail.
Conclusions
As one of the leading critics of microtargeting observes, ‘we have anxieties about micro-targeting because we have anxieties about democracy itself’.Footnote 225 So we should caution judging this practice against an idealised version of democracy, where voters are rational and seek to be informed, shaping their preferences in the general interest. Furthermore, we know that despite the apparent appetite of political parties to use political advertising generally, there is evidence that it has limited persuasive effects.Footnote 226 That said, the so-called ‘minimal effects thesis’ tends to underplay the ability of political advertising to increase the saliency of particular types of issues, especially those of an emotional nature, likely to invoke feelings of fear and anxiety.Footnote 227
While the evidence of the effects of digital campaigning generally may be ambiguous, there are specific techniques and consequences associated with microtargeting which give rise to particular unease. Principal among these is the danger that targeting allows politicians to communicate with voters in an unmediated way, thereby escaping the glare of public scrutiny, and undermining the integrity of the public sphere. Increasing the transparency of online advertising, while necessary, is only a partial solution to this problem. A digital imprint requirement will contribute to transparency, but enforcement brings with it a new set of problems, not least the potential ‘chilling effect’ of online political discourse. And as microtargeting becomes more efficient, online repositories are likely to be ‘overwhelmed’ by the ever-increasing volumes of messages generated.Footnote 228
A different way of framing the question is to ask what ‘value’ does microtargeting have, ensuring we continue to benefit from innovations in digital campaigning. It would certainly be useful if there was further evidence of the effects of microtargeting, both positive and negative.Footnote 229 Any case for reform is clearly easier to advance if regulation itself is targeted; seeking to mitigate some of the less palatable consequences of the practice, while not stultifying some of the potential benefits. To this end, attempts to regulate the content of political campaign messages, either by intermediaries or through direct regulation, are likely to be fraught with dangers, restricting political speech in a way which is inimical to democracy. The best solution is to increase the transparency of political campaigning through imprint requirements and the use of digital repositories. Nevertheless, restrictions on the means of communication can and should be considered in the future if less restrictive measures prove to be unsuccessful. The risk to democratic discourse is sufficient to justify restrictions on microtargeting, as a last resort.
As we have seen previously, it is very difficult to build a consensus among politicians and their parties over regulating political advertising. Politicians are alert to digital campaigning as a means of securing a comparative advantage over their adversaries. Perhaps the ‘arms race’ narrative, so successfully employed in the UK to justify the ban on broadcast advertising, is now being realised in the online environment. Is it likely that politicians will be able to remove from themselves the use of such a powerful and efficient campaigning tool? Even if the political class are persuaded of the need for regulation, are the regulators likely to enthusiastically assume the mantle? Mindful of the need to protect the perception of their independence, would any rational regulator want to adjudicate on a political party's marketing behaviour in the heat of an electoral campaign? What might be the possible implications for that regulator, and its personnel, after the polls have closed?
Freedom of political expression is a core value in any democratic state, and any restrictions upon it need to be justified carefully and fully. Nevertheless, online political advertising is changing the nature of our campaigns, our politics and, ultimately, our democracy; for this reason it must not escape the most intense scrutiny, now and in the future.
