Periodic Reporting for period 2 - DiCED (Digital Campaigning and Electoral Democracy)
Berichtszeitraum: 2021-08-01 bis 2023-01-31
The main objective of DiCED is to develop a more precise understanding of the nature, extent and impact of ‘data-driven’ campaigning (DDC) for voters, political parties and elections in established and newer democracies. The project is important because DDC is linked directly and indirectly, with a range of anti-democratic outcomes. Observing the growth in use of digital data and computational methods in campaigning, scholars have raised concerns that this is increasing the power of larger well-resourced organizations to run more opaque and unaccountable campaigns of persuasion and even social engineering. (Tufecki, 2014). Whether this is happening in practice, particularly outside the U.S. has not been subject to systematic investigation. This project seeks to fill that gap by evaluating these claims against the current reality of digital campaigning in five major democracies - the USA, Germany, France, Poland and the UK. It does so by distilling four inter-related objectives from the broader project goal. First, DiCED will define what a data-driven campaign is in conceptual terms. What are its essential features and does it constitute a new and distinct era of electioneering? Second DiCED will map the spread of DDC within democracies and examine how and why some parties and countries are more advanced in adopting DDC? Third, DiCED will investigate the consequences – negative and positive – of DDC for voters and parties. Are fears about spread of these new techniques really warranted? Finally, we examine the policy implications of DDC. How far is it currently regulated at the European and national level and are more controls necessary? From the citizen side, what digital skills are required to navigate the new campaign environment, and how might those be enhanced?
The project has completed a range of core tasks to meet these objectives. Specifically we have developed a comparative research design and set of data collection instruments adaptable for each election study and gained ethical approval. We have conducted fieldwork in four election studies – the 2020 U.S. Presidential election, the 2021 German Bundestag elections, the 2022 French Presidential and 2022 National Assembly elections. We have collected data at three levels of analysis. At the micro-level, we have measured citizen attitudes and online behaviours in each country via survey and social media feeds. At the meso-level, we have conducted interviews and surveys of campaign organization staff measuring parties’ perceptions and use of data-driven campaign techniques and collected content from their accounts on external platforms (Twitter, Facebook, and YouTube). Finally, at the macro-level we have designed a lexicon of key words and hashtags relevant for each election and collected aggregate-level data to track and monitor the campaign events and related public discussion. We have also produced a related series of qualitative event guides documenting the history of each election under study. These data and extensive desk-based research including a systematic literature review of DDC have been used to answer our four main research objectives and related questions.
The main results achieved to date are that we have developed a conceptual and operational definition of DDC that distils its core features and located it within the wider literature on campaigns historical development to show it can be interpreted as a continuation of past practice (the ‘modernization’ thesis) and transformative (the ‘disruption’ thesis). These theories will guide analysis of the empirical data collected in our election studies and be used to draw conclusions about the impact of DDC (Objective 1 and 3). Analysis of the U.S. and German micro and meso-level data has shown that views of DDC and adoption among parties system varies significantly by country and party with the U.S. being the most advanced or ‘driven’, and the major European parties following a more ‘data informed’ model. A key factor DiCED has identified to explain rates of adoption across countries is the nature of the regulatory environment, and particularly the power of data protection agencies. At the meso-level, a new role for political ‘influencers’ appears to be emerging where opportunities for paid individualised micro-targeted adverts are restricted (Objective 2). In terms of the effect of DDC, analysis of our U.S. micro-level data suggests the negative results of targeted political advertising maybe moderated if voters receive lection information that is personalised and relevant for them, particularly for younger voters. For parties, there appears to be a divide and power imbalance emerging between larger and smaller players in accessing DDC tools (Objective 3). Finally, concerning objective 4), our results show that the policy environment surrounding DDC is fragmented across several state, national and international agencies. While EU member countries i.e. France, Germany and Poland appear to be moving toward a more unified and strengthened approach under EU regulations, the direction for the UK since Brexit, and the U.S. which has no national data protection agency or policy, is less clear. However, some policy diffusion of EU GDPR principles appears to be emerging within U.S. states.
The main results achieved to date are that we have developed a conceptual and operational definition of DDC that distils its core features and located it within the wider literature on campaigns historical development to show it can be interpreted as a continuation of past practice (the ‘modernization’ thesis) and transformative (the ‘disruption’ thesis). These theories will guide analysis of the empirical data collected in our election studies and be used to draw conclusions about the impact of DDC (Objective 1 and 3). Analysis of the U.S. and German micro and meso-level data has shown that views of DDC and adoption among parties system varies significantly by country and party with the U.S. being the most advanced or ‘driven’, and the major European parties following a more ‘data informed’ model. A key factor DiCED has identified to explain rates of adoption across countries is the nature of the regulatory environment, and particularly the power of data protection agencies. At the meso-level, a new role for political ‘influencers’ appears to be emerging where opportunities for paid individualised micro-targeted adverts are restricted (Objective 2). In terms of the effect of DDC, analysis of our U.S. micro-level data suggests the negative results of targeted political advertising maybe moderated if voters receive lection information that is personalised and relevant for them, particularly for younger voters. For parties, there appears to be a divide and power imbalance emerging between larger and smaller players in accessing DDC tools (Objective 3). Finally, concerning objective 4), our results show that the policy environment surrounding DDC is fragmented across several state, national and international agencies. While EU member countries i.e. France, Germany and Poland appear to be moving toward a more unified and strengthened approach under EU regulations, the direction for the UK since Brexit, and the U.S. which has no national data protection agency or policy, is less clear. However, some policy diffusion of EU GDPR principles appears to be emerging within U.S. states.
To date we have identified five main areas – methodological, empirical, theoretical, conceptual and technical – that we anticipate going beyond the state of the art in the study of digital campaigns and political science research more generally:
(1) Methodologically by linking individual opinion data with new unstructured ‘big data’ in our
YouGov social media analysis panel we are helping pioneer the shift of survey research into a ‘fourth era’. The augmentation of survey data with new forms of observational trace data allows researchers to more accurately understand individuals’ exposure and responses to social and political online and offline stimuli. For DiCED, combining attitudinal measures with observational data on respondents’ digital campaign exposure provides an ideal ‘window’ to understanding the effects of DDC on voters. In addition to bringing more precision to our analyses, this work has required design of a rigorous ethical framework that we anticipate will serve as ‘best practice’ to guide other projects through this sensitive process. As a team, we regularly engage in reflexive thinking about the strengths and limitations of social media data, and have added a fifth research objective/question five to DiCED to accommodate this new area of enquiry. To promote this initiative the PI is co-editor for a Special Issue of the open access GESIS journal, methods,data analyses on this topic by the end of the project. This will encourage scholarship that openly dissects the challenges of working with social media data, and draws out explicit lessons to be learned for future scholarship.
(2) Empirically, our data will provide a unique, directly comparative perspective on citizens
experience of this new mode of campaigning and parties adoption of it. This will enable ambitious original research that will extend what has been so far largely single nation and U.S. centric research.
(3) Theoretically we are expanding the study of campaign change and effects to include insights
and theories drawn from a wider range of social science disciplines, notably law, policy and governance as well as consumer and social psychology. Our work to date has exposed the role of two important new ‘actors’ at the macro, meso and micro level in relation to the growth and impact of DDC. These are the regulatory bodies responsible for data protection policies at the national and international level and the online political influencers who ‘escape’ regulatory scrutiny and exercise ‘soft’ power on voters during elections. We are arguing for more attention to be given to psychological variables in accounting for voter responses to DDC. Specifically, we are developing hypotheses to account for the effect of personality traits and what we have labelled the ‘privacy-personalisation calculus’.
(4) Conceptually, we are working with researchers in computer and data science to better
understand the nature of algorithmic bias which is a topic of relevance to political targeting. Specifically we are engaged in a project run by North West Partnership for Security and Trust (NWPST), a partnership between the UK’s Government Communications Headquarters (GCHQ) and four Universities. The project is designed to test if it is possible to manipulate operational AI. We anticipate the results may be useful to address DiCED questions about the use of algorithmic biases in political targeting.
(5) Technically, and initially in response to the pandemic, we have established a new virtual
platform for research collaboration that forms a model for distributed researchers working with confidential sources of ‘big’ data as is the case for DiCED. Through a bespoke server, we combine an ‘open’ data environment for approved researchers to work with our data while maintaining compliance with strict ethical standards. This constitutes a first for the University of Manchester, and among our international project team. Although it forms a ‘prototype’ at this stage, we will seek to enhance its functionality as a workspace for the team, and we plan to seek continued support for the platform after the project formally ends.
(1) Methodologically by linking individual opinion data with new unstructured ‘big data’ in our
YouGov social media analysis panel we are helping pioneer the shift of survey research into a ‘fourth era’. The augmentation of survey data with new forms of observational trace data allows researchers to more accurately understand individuals’ exposure and responses to social and political online and offline stimuli. For DiCED, combining attitudinal measures with observational data on respondents’ digital campaign exposure provides an ideal ‘window’ to understanding the effects of DDC on voters. In addition to bringing more precision to our analyses, this work has required design of a rigorous ethical framework that we anticipate will serve as ‘best practice’ to guide other projects through this sensitive process. As a team, we regularly engage in reflexive thinking about the strengths and limitations of social media data, and have added a fifth research objective/question five to DiCED to accommodate this new area of enquiry. To promote this initiative the PI is co-editor for a Special Issue of the open access GESIS journal, methods,data analyses on this topic by the end of the project. This will encourage scholarship that openly dissects the challenges of working with social media data, and draws out explicit lessons to be learned for future scholarship.
(2) Empirically, our data will provide a unique, directly comparative perspective on citizens
experience of this new mode of campaigning and parties adoption of it. This will enable ambitious original research that will extend what has been so far largely single nation and U.S. centric research.
(3) Theoretically we are expanding the study of campaign change and effects to include insights
and theories drawn from a wider range of social science disciplines, notably law, policy and governance as well as consumer and social psychology. Our work to date has exposed the role of two important new ‘actors’ at the macro, meso and micro level in relation to the growth and impact of DDC. These are the regulatory bodies responsible for data protection policies at the national and international level and the online political influencers who ‘escape’ regulatory scrutiny and exercise ‘soft’ power on voters during elections. We are arguing for more attention to be given to psychological variables in accounting for voter responses to DDC. Specifically, we are developing hypotheses to account for the effect of personality traits and what we have labelled the ‘privacy-personalisation calculus’.
(4) Conceptually, we are working with researchers in computer and data science to better
understand the nature of algorithmic bias which is a topic of relevance to political targeting. Specifically we are engaged in a project run by North West Partnership for Security and Trust (NWPST), a partnership between the UK’s Government Communications Headquarters (GCHQ) and four Universities. The project is designed to test if it is possible to manipulate operational AI. We anticipate the results may be useful to address DiCED questions about the use of algorithmic biases in political targeting.
(5) Technically, and initially in response to the pandemic, we have established a new virtual
platform for research collaboration that forms a model for distributed researchers working with confidential sources of ‘big’ data as is the case for DiCED. Through a bespoke server, we combine an ‘open’ data environment for approved researchers to work with our data while maintaining compliance with strict ethical standards. This constitutes a first for the University of Manchester, and among our international project team. Although it forms a ‘prototype’ at this stage, we will seek to enhance its functionality as a workspace for the team, and we plan to seek continued support for the platform after the project formally ends.