The New ‘Data Science‘ of Election Campaigning in Comparative Perspective

Introduction and Summary The 2012 campaign by President Obama was seen as pioneering a new 'science' of electioneering, driven by big data and computational analytics (Issenberg, 2013). Insiders within the campaign revelaed that they had collected more than 1000 data points for every single voter and armed with this information could accurately predict how that person would decide - even before the individual themselves knew. Since then, interest in the use of these techniques in elections has increased and spread to other countries. Concerns over their power to shape and even manipulate voter choices have also been raised, with questions raised about how far data analytics companies have influenced the outcomes of recent major national elections and the 2016 UK Brexit referendum. Research is now at an early stage in terms of investigating these claims and the wider implications of these new techniques. This Workshop will build on these efforts by offering a forum for studies that map, model and examine the supply and demand for this new style of campaigining worldwide. More specifically it will seek to understand what are the key techniques and features of this new data-driven mode of electioneering?; How widely used are they in local, national and supra-national elections? What are the systemic conditions most conducive to its development?; and also what institutional and regulatory frameworks exist to control and monitor the take-up of 'data driven' campaigning. From the demand side it will focus on examining the effects of these new techniques on voters. Are they reinforcing current trends among electorates toward increasing levels of distrust or cynicism toward politicians and the system more generally. Or do they constitute a new and more efficient way to mobilize undecided and under-engaged voters in politics? Background and Literature Review Recent developments in campaigning have seen digital technology move to the centre stage of operations and particularly in regard to delivering on the core task of voter mobilization. This has major implications for both who runs campaigns and how they are run. From the voters’ perspective there are equally significant changes in the extent and type of contacting they experience. The micro-targeting methods and message specialisation that are now emerging mean that there is no longer a ‘one size fits all’ method in terms of campaign advertising and tracking and tracing exposure becomes an increasing challenge for academic study. To date there has not been a systematic attempt to comparatively audit where these new practices are emerging and analyze their effects on party systems, intra-organizational democracy, the quality of voter communication and its impact on voter turnout and choice. In addition, the aspect of legal limitation (data protection laws) is also one to zoom in to. This workshop aims to fill that gap. While the growth of the new data science techniques has progressed very rapidly in a short space of time, scholars of campaigns and elections had been speculating for some time on the voter targeting potential of the new medium (Norris, 2000; Plasser and Plasser, 2002; Farrell and Schmitt-Beck, 2003, Römmele 2001; Römmele/von Schneidemesser 2017, Gibson 2017). According to these authors, election campaigning had entered a new professionalized or postmodern era by the late twentieth century. While this shift preceded the arrival of the web, a key characteristic of the new era was its emphasis on careful message targeting through direct mail methods and multi-channel TV and a rejection of the ‘one size fits all’ mode that had dominated in the earlier ‘modern’ era. The arrival of the internet and particularly tools like email were thus seen as aligning very well with this wider tendency toward ‘narrowcasting’ and micro-messaging of voters.Initial attempts by parties to exploit the ‘narrowcasting’ capabilities of the web had included the development of content for particular audiences such as young people, female voters and journalists. Other more explicit vote-getting initiatives such as ‘vote swapping’ sites had been promoted, particularly by smaller parties to encourage the strategic exchange of ballots with ideological allies in a bid unseat or deny victory to a common ‘enemy’. More recently as social media networking tools emerged, digital campaigners’ made extensive efforts to develop a new two-step or indirect model of voter mobilization through online channels (Gibson, 2013; Stromer-Galley, 2014, Kreiss, 2013). While these initiatives were clearly designed to increase parties’ and candidates’ electoral support they also faced some significant challenges in regard to precision and inherent passivity as mobilization tools. Voters either needed to find the sites themselves or be connected in some way with the campaign through their online social networks in order to gain exposure to the mobilizing messages. Connecting with that all-important pool of undecided voters that could help swing an election outcome was thus a matter of luck and random exposure rather than coordinated intent. In the most recent elections witnessed in the U.S. and also the UK it seems that parties are now moving to a point of removing the uncertainties and inefficiencies in the process and setting up a new set of highly accurate ‘scientific’ methods for pinpointing and mobilizing these latent pockets of support. The shift in focus to more micro or even what some termed ‘nano’ targeting’ of the electorate is matched by an intensification of efforts by parties to ‘pull’ in more detailed information about individual voters in to the campaign. While websites had previously gathered personal details about visitors through newsletter sign-ups and feedback forms, the new methods mean that such requests become more frequent and intrusive. Innovations such as pop-up surveys and ‘landing pages’ that directly ask users’ about their vote intentions and geographic location are now common features on parties’ sites along requests for Facebook or Twitter accounts in order to access certain content. Once collected the new datum are immediately added in to the parties’ already vast voter files to be pored over and dissected by newly appointed teams of computer and behavioral scientists. The results of their analyses are then fed back into the micro-targeting process to produce yet more accurate online and offline voter contacts. The expertise and technical support required to deliver the new data intensive forms of campaigning necessitates a huge expansion in the budget and personnel allocated to digital. The growth in staff numbers is accompanied by an inevitable differentiation and specialization in the roles they perform. Long-standing ‘public’ facings jobs such as home page and email management continue as core activities while new teams focusing on the ‘back-end’ analytical and infrastructure support tasks also now come to fore. Outside any restructuring in the size and shape of the campaign organization that results, arguably the biggest internal change triggered is arguably among top level elites and the culture of decision-making. The importance of field experience and guru-like intuition as a guide to strategic planning is increasingly challenged by a new set of experts trained in cutting edge social scientific methods. More sophisticated ‘remote’ data-driven approaches to understanding and predicting voter behavior now form the basis for key decisions in field operations, media advertising and fund-raising efforts (Issenberg, 2013 Nickerson and Rogers, 2014; Anstead, 2017). The shift into data intensive mode of campaigning thus carry important consequences at the party system level in that competition is potentially even more imbalanced than ever. The bigger parties’ are the only organizations that are realistically capable of developing and running such a complex voter management system. At the intra-organizational level, national-local relations show a similar shift toward normalization as the new tech gurus take control over key decisions. Local level operatives essentially become delivery mechanisms for the strategic decisions over resource allocation taken nationally. That said, there is clearly some scope for local groups to exercise some autonomy over the particular tactics employed through access to a new suite of DIY-like resources for fund raising, meeting organization and direct canvassing. The consequences for voter communication are similarly profound in that it becomes more precisely controlled and targeted by central elites through specialist apps and smart software. Voters are now seen as essentially as the recipients of micro targeted messages and particularized content. Any interaction is conducted largely for purposes of extraction of information rather than the promotion of dialogue and participatory engagement. Empirical Evidence To date, data driven campaigning appears to be an emergent trend but one that is gaining adherents and ground as a more embedded practice (Hersh, 2015 Kreiss, 2016; Bennett, 2016; Anstead, 2017). Areas of notable growth include North America with Canada and the U.S. witnessing the most rapid advances (Bennett, 2016). Within the U.S. the Democratic National Committee (DNC) in particular has been particularly innovative in developing a new scientific approach to campaigning over the past two decades. It was, however, the Presidential election of 2012 and particularly the efforts of Barack Obama that first really put the new modus operandi of electioneering on the national and international electoral map. According to most observers the campaign of 2012 was a ‘game changer’ for digital, elevating it into a pivotal position within the wider decision-making hierarchy. Post-election media reports contained extensive coverage of the radical overhaul of campaign hardware and software that took place and the new laser-like focus on seeking out and mobilizing undecided voters. Forecasts of the vote split at the precinct level were reportedly accurate to within less than one percentage point down to the town hamlet level in New Hampshire. Post-election academic analysis did raise some questions as to the efficiency of Obama’s data crunching machine and how far it actually secured his victory (Sides and Vavreck, 2013). However, 2016 saw the techniques honed and developed still further by the Democrats as the Clinton campaign put them centre stage in the development of their voter mobilization. While the Trump campaign was publicly more skeptical about the power of the new scientific approach to campaign management, subsequent reports of extensive use of negative Facebook ads to discredit Clinton and allegations of collusion with foreign hackers to circulate fake news on social media indicated a more strategic and ‘below the radar’ use of the medium. Elsewhere, in advanced industrial democracies of Europe, the adoption of the new techniques has been much less evident. Some countries such as the UK have seen increasing interest among parties in data driven campaigning since the 2015 election, particularly among the right-wing Conservative party (Anstead, 2017). In Germany, parties have employed data analysts for the 2017 campaign, other major countries such as France and Italy have seen some ‘startup’ activity by tech companies that have offered data analytics services to parties to help them map and target new latent pockets of electoral support (Bennett, 2016). Attempts are patchy, however, and European legislation surrounding political communication, most notably the General Data Protection Regulation (GDPR) is seen as major restraint on the spread of such practices. For Anstead (2017) such variance underscores how the shape of ‘data driven campaigning’, and the uses it is put to, is “shaped by political context.” The goal of this workshop will be to document, explore and explain that variance and the likely shape of future direction for such regulation. To what extent will there be a softening of the rules as parties weigh up the potentially higher returns that the new micro-targeting offers to the investment of their increasingly limited resources. Summary and Next steps. The past decade has seen significant change in how campaigns are practiced. The rise of digital data and analytics is revolutionizing the ‘art’ of campaigning into a science. In addition it is introducing a whole set of new factors such as foreign influence and fake news as potential drivers of voter decision making. Where and how these changes are occurring beyond the case of the U.S. is a question of increasing importance for political science scholars to address. How are such developments changing the nature of competition and the internal organization of campaigns? Do they lead to an inevitable reinforcing of major parties and central elites? How far do they bring new players such as software engineers and data analysts into the process? While they may have a wealth of technical and computing and statistical expertise the clearly lack elections and campaign experience. Does this mean a de-politicizing of the elections and wider political process? Finally what does the rise of this new data elite within parties mean for the electorate? Are we now at the starting point of an increasing narrowing of campaign mobilization efforts where only those deemed to be persuadable through algorithms and computer modeling are considered worthy for contacting?