Demobilizing, Dividing and Disinforming? Unpacking the effects of social media and web use on voter behaviour and attitudes in the 2020 U.S. Presidential election

Authors: Rachel Gibson, Jonathan Nagler, Esmeralda Bon, Jahander Musayev and Peter Smyth

Social media platforms and websites are assuming an ever more prominent role in elections as campaigners flood the internet with online ads, mobilizing hashtags and misinformation, and voters’ increasingly consume, share and comment on, this content. This prominence took on added relevance for elections held during 2020, given the shift of much of daily life to the online sphere in response to the Covid-19 pandemic. Given these developments building an accurate understanding of how this more immediate, immersive and interactive mode of communication is affecting voter behaviour and attitudes is both a pressing and very timely undertaking for researchers. To date, research on the topic to date has, by and large, supported the view that internet use and particularly social media has had a positive effect on individual’s political engagement, particularly within the electoral context (Boulianne, 2020). However, more recent studies have begun to suggest that a range of more negative behavioural, attitudinal and cognitive effects now need to be considered, with Facebook and other platforms seen as tools for demobilizing, dividing and disinforming electorates, fuelling distrust in mainstream politicians and the support for populist and extremist parties (Freelon and Lokot, 2020; Howard, 2020; Schroeder, 2019; Postill, 2018). More systematic investigation is currently needed to assess the extent to which the negative effects of social media and web use are now starting to outweigh their positive impact. In particular, models of internet use and political engagement need to be extended to specify and test for a range of explicitly anti-democratic outcomes, in addition to any democratic gains.

This paper seeks to address this challenge in by exploiting an innovative new research methodology that links individuals’ web activity with survey responses during the U.S. 2020 Presidential election. Specifically, we collected the Twitter use and web browsing habits of up to 1,500 respondents in a YouGov panel in the final two months of the campaign These data are combined with pre and post-election survey data measuring a range of political variables including vote intention, issue and ideological positions, levels of political interest, trust in the candidates and electoral system, and democratic satisfaction, as well as perceived exposure to misinformation during the election and knowledge about the current economic and health situation facing the U.S. Through the combination of these two types of data in a panel study design we are able to more accurately understand what voters saw, heard and did online during the election, and how this affected their likelihood of voting, sentiment toward the candidates, trust in key political actors and understanding of the key issues in the campaign. As such, we are able draw fresh and powerful insights into the causal impact of internet use on political engagement and particularly on this question of whether higher exposure to, and/or interaction with, online content is now leading to a net loss rather than net gain in electoral engagement? As well as looking at the impact of internet use on voter behaviour our dataset will also allow us to measure attitudinal and cognitive change among voters over the course of the campaign, and specifically whether increased reliance on digital media is linked with more negative perceptions and distrust of political elites, adoption or endorsement of more extremist views, as well as declines in political knowledge and the quality of voter decision-making. While our findings are not seen to provide a definitive answer as to whether the internet is now contributing to trends toward democratic backsliding, they are anticipated to offer fresh and important insight into this debate.

Measuring and Mapping Data-driven campaigning – A comparative study of the U.S. and Germany

Authors: Esmeralda Bon, Rachel Gibson and Andrea Rommele

This paper proposes a more explicit empirical definition and quantitative measurement of an emergent mode of electioneering – data-driven campaigning (DDC) – that is then applied and ‘road-tested’ in two major party systems, Germany and the U.S. following the ‘most different systems’ logic. The key aim of the paper is to contribute to the development of a more systematic and replicable measurement of campaign change that can be used in comparative research, and to derive a deeper understanding of the factors that drive innovation and adoption of these new techniques at the party and systemic level. To do this, the paper is divided into four main sections. Part I undertakes a systematic review of the literature on the concept of DDC in order to arrive at shared or ‘essentialist’ definition of the practice. In part II this conceptual definition is operationalised by deriving observable indicators that measure the extent to which parties or campaigns are deploying DDC practices. Potential indicators include the use of large scale integrated databases, analytics software and message testing. These indicators will be converted into binary and ordinal variables and form the basis for development of a multi-dimensional DDC index that can be applied to individual parties. The resultant scores can then be used to measure and compare adoption of the new mode of campaigning within and across party systems. The operationalisation and index construction draws on prior experience of the authors in the design and application of a party professionalization campaign index (Gibson and Rommele, 2000, 2009). Part III of the paper will attempt to validate the index by applying it in two divergent party systems – the U.S. and Germany. Scores will be assigned to parties based on a combination of elite interview data and desk based research. The results will be evaluated in terms of the extent to which they are successful in capturing variance across parties and party systems and how far they have face validity. Where do we see highest and lowest rates of adoption among parties? How far does size, ideology, structure, electoral performance or party goal matter in explaining the results? Given we have a MDS design is there a marked difference in scores at the country or party system level? The final section of the paper will interpret and contextualise the findings to reflect on what they tell us about the drivers and barriers to the uptake of DDC at the organizational and systemic level, and more generally on the value of the new index for cross-national research. In taking this more reductionist empirical approach to measuring DDC we inevitably lose nuance and theoretical richness. However, by making its features explicit and quantifiable it becomes possible to benchmark and measure variance in campaign style and practice within and across party systems, and also change over time. This is arguably an area of party activity that has to date proven resistant to efforts at empirical precision.

I always feel like somebody’s watching me….”: How much do Americans know about who is targeting them in elections and do they really care?

Authors: Rachel Gibson, Kate Dommett and Esmeralda Bon

This paper will measure and analyse U.S. citizens understanding of, and reactions to, political micro-targeting during the 2020 Presidential election. Specifically, we will examine levels of awareness and concern among voters toward campaigns’ use of their online and offline personal data during elections, and their behavioural reactions and responses to such usage. Such an investigation is both timely and important. The 2016 Presidential campaign saw concerns about the misuse of citizens’ personal data and manipulation of online advertising rise to the fore with the Cambridge Analytica scandal and Senate Intelligence Committee report confirming foreign interference on Facebook and Twitter. At the same time campaigns show no sign of slowing down in their efforts to collect, link and analyze the increasing volumes of voter data that digital platforms increasingly generate. Studies that systematically seek to understand how the public perceive and respond to these new ‘data-driven’ techniques in the U.S. and elsewhere are increasing, but are still quite limited. The evidence that exists indicates that a majority of the public typically disapprove of political micro-targeting, although Americans appear to be somewhat more accepting of it than their European counterparts (Kozyreva et al. 2020). In addition, it appears that despite having concerns, the vast majority of citizens are unaware of the range of data that platforms such as Facebook hold on them and many do not take simple steps to protect their privacy (Pew Report ‘Facebook Algorithms and Personal Data’ 2019). Such findings suggest a discrepancy exists between the concerns individuals have about data-driven campaign practices, and the extent to which these concerns inform their behaviour, or increase their understanding about how their own data is used online.

This paper seeks to unpack and investigate these findings further, to provide a richer and more detailed profile of U.S. citizens’ relationship with political micro-targeting. We do so using data from a specially commissioned YouGov survey conducted in the final weeks of the U.S. Presidential campaign. The survey measured citizens’ orientations to campaigns’ use of voters’ personal data and political micro-targeting in three dimensions: knowledge, attitudes (supportive or critical) and behavioural responses. The goals of the analysis are three fold: first to provide an in-depth profile of U.S. voters’ relationship with this more data-intensive mode of campaigning; second, to use statistical techniques to understand what predicts varying levels of awareness, concern and behavioural responses of micro-targeting at the individual level and how far this variance has political consequences in terms of levels of trust in institutions, democratic satisfaction, vote intention and also positions on key issues or support for conspiracy theories? In a third and final step results of our analysis will be used to construct a series of generic voter profiles or ‘types’ in relation to micro-targeting from the ‘know-nothings’ to ‘fully aware’. By empirically defining and conceptualising these types we aim to provide a clearer aggregate picture of current levels of voter literacy with regard to micro-targeting in the U.S., which can then form a benchmark for over time analysis and applied in other countries, thereby enabling more direct comparative research on this topic.

Measuring the uptake and implications of data-driven campaigning for party organizations – A comparative study of the U.S. and Germany

Authors: Esmeralda Bon, Rachel Gibson and Andrea Rommele

As parties rely increasingly on digital tools in response to the Covid-19 pandemic this is accelerating their use of so-called ‘data-driven’ techniques as part of their campaign strategy. This new approach brings an expansion in the scale and reach of the extent of personalised contacting or micro-targeting parties can carry out during elections. To date much of the attention on the growing use of data-driven campaigning or DDC has examined the impact of these more individualized targeting activities on parties’ ability and success to mobilize voters. Less emphasis has been placed on the internal drivers to adoption of DDC and the impact of the new infrastructure and expertise required to engage in this more data intensive mode of electioneering on party organization and internal decision-making. This paper aims to fill that gap by providing a clearer understanding of what this new mode of electioneering actually requires in terms of party organizational resources and comparing the extent to which parties in two major systems Germany and the U.S. – are actually engaging in data-driven campaigning. Specifically we aim to design a more systematic, transparent and replicable measure or index of DDC at the party level which we can use to better compare parties’ take-up, and also the key organizational changes it introduces to their internal workings. What are the new data and resources required for DDC, what type of personnel are required to deliver it, and does it change the nature and structure of campaign management, use of conventional or offline media, and who now controls strategic decision-making.

To address these questions the paper is divided into two main parts. In part I we undertake a systematic review of the literature on the concept of DDC in order to arrive at a shared or ‘essentialist’ definition of the practice that is then operationalised by deriving observable indicators that measure the extent to which parties or campaigns are deploying DDC practices. Potential indicators include the use of large scale integrated databases, A/B message testing, predictive modelling, and hiring of data analysts and research software engineers. These indicators will be converted into binary and ordinal variables and form the basis for development of a multi-dimensional DDC index that can be applied to individual parties. The operationalisation and index construction draws on prior experience of the authors in the design and application of a party professionalization campaign index (Gibson and Rommele, 2000, 2009). In part II of the paper we apply the index across two divergent party systems – the U.S. and Germany – and assign scores to parties using a combination of elite interview data and desk based research. The scores will be used to identify the key internal party organizational drivers behind adoption of the new techniques, and particularly how far features such as size, resources, ideology, internal structure and hierarchy, and primary goals determine adoption rates? In addition we will use the rankings to draw some preliminary conclusions about the impact of a more intensive use of the new techniques on other aspects of party campaign practice. In particular, how far is the use of DDC displacing use of other more conventional campaign techniques such as mass media advertising? Is it leading to the rise of a new internal digital elite and reliance on computers and AI for decision-making in place of traditional field operatives? What does this mean for the future of intra-party democracy? While our findings are clearly speculative at this point, we argue that the introduction of the DDC index is a key step toward systematizing study of the implications of these more automated campaign methods for the internal workings of political parties.