Data-driven campaigning and democratic disruption: Evidence from six advanced democracies

Abstract

Data-driven campaigning has become one of the key foci for academic and non-academic audiences interested in political communication. Widely seen to have transformed political practice, it is often argued that data-driven campaigning is a force of significant democratic disruption because it contributes to a fragmentation of political discourse, undermines prevailing systems of electoral accountability and subverts ‘free’ and ‘fair’ elections. In this article, we present one of the very first cross-national analyses of data-driven campaigning by political parties. Drawing on empirical research conducted by experts in six advanced democracies, we show that the data-driven campaign practices seen to threaten democracy are often not manifest in party campaigns. Instead, we see a set of practices that build on pre-existing techniques and which are far less sophisticated than is often assumed. Indeed, we present evidence that most political parties lack the capacity to execute the hyper-intensive practices often associated with data-driven campaigning. Hence, while there is reason to remain alert to the challenges data-driven campaigning produces for democratic norms, we argue that this practice is not inherently disruptive, but rather exemplifies the evolving nature of political campaigning in the 21st century.

Full text

Original Research Article
Party Politics
2022, Vol. 0(0) 1–15
© The Author(s) 2022
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/13540688221084039
journals.sagepub.com/home/ppq
Data-driven campaigning and democratic
disruption: Evidence from six advanced
democracies
Glenn Kefford
School of Political Science and International Studies, University of Queensland, Brisbane, Australia
Katharine Dommett
Department of Politics, University of Sheffield, UK
Jessica Baldwin-Philippi
Department of Communication and Media Studie, Fordham University, Bronx, NY, USA
Sara Bannerman
Faculty of Social Sciences, McMaster University, Hamilton, ON, Canada
Tom Dobber
Amsterdam School of Communications Research, University of Amsterdam, Netherlands
Simon Kruschinski
Department of Communication, Johannes Gutenberg University Mainz, Germany
Sanne Kruikemeier
Amsterdam School of Communications Research, University of Amsterdam, Netherlands
Erica Rzepecki
Faculty of Social Sciences, McMaster University, Hamilton, ON, Canada
Abstract
Data-driven campaigning has become one of the key foci for academic and non-academic audiences interested in political
communication. Widely seen to have transformed political practice, it is often argued that data-driven campaigning is a
force of significant democratic disruption because it contributes to a fragmentation of political discourse, undermines
prevailing systems of electoral accountability and subverts ‘free’and ‘fair’elections. In this article, we present one of the
very first cross-national analyses of data-driven campaigning by political parties. Drawing on empirical research conducted
by experts in six advanced democracies, we show that the data-driven campaign practices seen to threaten democracy are
often not manifest in party campaigns. Instead, we see a set of practices that build on pre-existing techniques and which are
far less sophisticated than is often assumed. Indeed, we present evidence that most political parties lack the capacity to
execute the hyper-intensive practices often associated with data-driven campaigning. Hence, while there is reason to
remain alert to the challenges data-driven campaigning produces for democratic norms, we argue that this practice is not
inherently disruptive, but rather exemplifies the evolving nature of political campaigning in the 21st century.
Paper submitted 20 October 2021; accepted for publication 10 February 2022
Corresponding author:
Glenn Kefford, School of Political Science and International Studies, University of Queensland, Brisbane, QLD 4072, Australia.
Email: g.kefford@uq.edu.au

Keywords
Political parties, campaigning, digital, data
Data-Driven Campaigning (DDC) has become a key con-
cept for those seeking to understand elections and cam-
paigns in the last two and a half decades. Whilst the practice
of collecting and mobilising information about citizens
within campaigns has long antecedents (Hersh, 2015), in
recent years it has been argued that we have entered a new
era of DDC facilitated by developments in digital tech-
nology, media and broader society (Roemmele and Gibson,
2020). Characterised by the collection, analysis and use of
increasingly personalised information online and offline,
data is widely seen to have transformed modern cam-
paigning (Nickerson and Rogers, 2014). However, DDC
has also been seen to be a force of democratic disruption,
with the collection, analysis and use of data in election
campaigns by political parties and other campaigners seen
to be challenging democratic norms and practices (Gorton,
2016). Focussing on political parties as the key campaign
actors in most advanced democracies, we argue it is im-
portant to know how political parties are using data in their
campaigning practices so that we can determine the likely
extent of disruption to these parties, as well as to our de-
mocracies. Hence, in this article, we answer the question:
‘how, if at all, do political parties in advanced democracies
undertake DDC?’. This question allows us to determine
whether DDC is the existential threat to democratic norms it
is often assumed to be, or whether DDC is a product of
broader socio-political forces which encourage and in-
centivise campaign actors such as political parties to
campaign in such a way.
Our approach departs from the prevailing tendency to
detail the theoretical, normative and legal aspects of DDC
(Dobber et al., 2019;Zuiderveen Borgesius et al., 2018).
Instead we provide one of the very first cross-national
empirical analyses of DDC practices, something severely
lacking in the literature thus far. While increasingly there are
case studies of DDC practices across the globe (Anstead,
2017;Kefford, 2021), scholarship lacks comparative ana-
lyses and is United States (US)-centric. This US focus is
particularly problematic because prevailing accounts of
DDC have become associated with highly resourced
presidential campaigns that are an outliner compared to
most election campaigns.
Offering an important corrective to these tendencies, we
bring together and comparatively analyse empirical data on
the DDC practices of political parties in six advanced de-
mocracies. Specifically, we reveal the variety of ways in
which data is collected, used and curated by campaigns,
showing that while there are a relatively uniform set of
practices employed by parties across advanced democracies,
the take-up and implementation of these practices differ
significantly across and within country contexts. Hence,
while there is evidence of DDC having a disruptive impact on
campaigning practices and parties, these effects are by no
means uniform, inherently new or necessarily democratically
problematic.
The remainder of our article is structured as follows: we
begin by surveying the literature on DDC and outline the
threat it is perceived to pose to democracy. We then move on
to our findings, setting out developments in our six ad-
vanced democracies. We start with how data is collected,
turn to how data is used and then move on to discuss data
infrastructure. We conclude by reflecting on the limitations
and logical extensions of our study and discuss the im-
plications of our findings in light of DDC’s alleged impact
on democratic norms and practices.
The rise of data-driven campaigning
DDC has gained widespread interest amongst academic and
non-academic audiences, coming to prominence as media,
technological and social-political transformations have led
to a fragmentation of media landscapes, ongoing data-
fication of society, increasing individualisation, and broader
electoral volatility across the democratic world. Some
scholars see DDC as a new ‘science of campaigning’(Pons,
2016: 36), suggesting it has opened the possibility of ‘direct
approaches in which political actors target personalized
messages to individual voters by applying predictive mod-
elling techniques to massive troves of voter data’(Rubinstein,
2014: 882). For Baldwin-Philippi, DDC has two main fea-
tures: ‘targeting, or deciding which messages go to what
potential voters at what time during the campaign, and
testing, or empirically measuring how well messages perform
against one another and using that information to drive
content production and further targeting’(2017:628).
Studies examining the effect of DDC on organisations
such as parties have been sparse. Some scholars focus on
data as a resource that has organisational consequences for
campaigning (Munroe and Munroe, 2018:8–9). Whilst by
no means novel (Hersh, 2015;Kreiss and Howard, 2010),
new forms of data are seen to allow parties to make cost
efficient decisions, (Kreiss, 2016), to improve communi-
cation attempts and to support ‘the organisation and eval-
uation of a campaign’(Dobber et al., 2017;Kruschinski and
Haller, 2017). Frequently, DDC is depicted as the latest
manifestation of longstanding trends of professionalisation
and modernisation within political campaigning (Plasser
and Plasser, 2002;Chester and Montgomery, 2017).
2Party Politics 0(0)

More prominent have been concerns about the demo-
cratic implications of DDC. Many scholars have made
connections between DDC and voter surveillance or pro-
filing, leading to coverage of electoral manipulation and
subterfuge. Accounts highlight how ‘[b]y analysing specific
datasets, political parties can achieve a highly detailed
understanding of the behaviour, opinions and feelings of
voters’(IDEA, 2018: 6). Or that ‘it even is possible to
predict a person’s beliefs, even before they have formed
them themselves. And, subsequently, it is possible to subtly
steer those beliefs, while leaving the person thinking they
made their decision all by themselves’(in ’t Veld, 2017:2–
3). Such narratives suggest DDC has democratic implica-
tions for individuals (Zuiderveen Borgesius et al., 2018),
raising questions about voters’privacy and capacity to
freely exercise choice free from manipulation (Burkell and
Regan, 2019). At a societal level, scholars have also
spotlighted the negative democratic implications of DDC,
arguing that it encourages campaigns to focus on individual
interests rather than interest aggregation (Kusche, 2020),
contributes to a fragmentation of political discourse (Pons,
2016;Harker, 2020), and undermines prevailing systems of
electoral accountability (Jamieson, 2013;in ’t Veld, 2017).
DDC is therefore currently seen to threaten established
democratic principles about individual and societal prac-
tices, trends that are seen likely to only intensify as cam-
paigns gain access to ever more personalised data and
technology adapts to enable more individual level targeting.
Whilst these democratic implications are widely dis-
cussed within scholarship and wider society, much existing
work exhibits significant limitations. First, empirical ob-
servations are scarce since access to campaigns is difficult to
obtain and there is little transparency around DDC practices.
Second, extant studies have adopted a ‘media-centric’
theoretical approach, focussing on the growth of digital
technology and the data insights these developments make
available (Jungherr et al., 2020). Third, most of the studies
focus on single country cases with a special focus on the US
context (Baldwin-Philippi, 2017;Kreiss, 2016).
Data and methods
We expand knowledge of DDC by comparing practices in
Australia, Canada, Germany, the Netherlands, the United
Kingdom (UK) and the US. In selecting our cases, we chose
countries with established history of DDC and utilised a
Most Different Systems Design (MDSD) to compare key
dimensions which theoretically should be significant in
shaping campaign practice in advanced democracies.
Hence, we included countries with different electoral,
regulatory, and institutional structures, allowing us to ob-
serve differing financial regimes, party systems, unitary and
federated structures, voluntary and compulsory voting, as
well as different regulatory and legislative environments. In
doing so, our aim was not to produce a set of findings that
are generalisable, or that explain the variation across ad-
vanced democracies, but rather to deepen understanding of
the varied nature of DDC and its impact on democracy. We
focus our attention particularly on political parties in rec-
ognition of the central role they play within election
campaigns and their significance as essential components of
the democratic framework.
Our approach, which is inductive and qualitative, has
proven insightful for the study of new campaigning tools in
our field (Kreiss, 2016;McKelvey and Piebiak, 2018), and
builds on emerging work that has shown variation in the
practices and organisational capacities of parties to un-
dertake DDC (Dommett, 2019;Kefford, 2021;Kruschinski
and Haller, 2017). Our analysis therefore allows us to reflect
more concretely on the way that changes to media, tech-
nology and political participation are –or are not –shaping
how political parties campaign in the 21st century.
To conduct our analysis, we brought together a group of
scholars with detailed knowledge of DDC practices of
political parties in the six democracies. Each featuring as
authors of this piece, we set out to pool our empirical in-
sights of single case studies to tackle the prevailing ten-
dency for isolated, or empirically impoverished studies of
data-driven campaigning. Whilst it may have been optimal
to conduct new, entirely comparative work simultaneously
in our six countries, this approach was not possible for a
number of reasons. In addition to resource constraints,
access was a pre-eminent concern. DDC is a highly sensitive
topic and it is common to encounter non-disclosure
agreements or extreme reticence about taking part in re-
search. Securing access, or re-access, is therefore far from
straightforward.
Our chosen approach, of course, raises a number of
challenges. Most immediately, it means that our data is not
directly comparable, either in terms of the type of empirical
insight collected, the time period covered, or the number of
parties (see Table A1 in the appendix). In the majority of our
cases, researchers have conducted extensive interviews,
scrutinised internal party documents, media coverage and
other documents within the last 2–3 years. The relative
propensity of each type of data reflects the dynamics of each
case. In Australia, for example, interview data far exceeds
any of our other cases, reflecting the limited insights which
are publicly available about the practices of political parties
in that country as a result of a regulatory regime that requires
almost no disclosure and transparency around party ex-
penditure. In contrast, in Canada or the Netherlands, for
example, experts were able to rely on other sources to gain
understanding, often drawing on party documents or con-
tent analysis. While acknowledging these limitations, we
argue that the merits of providing the first detailed cross-
national study of DDC outweigh these limitations as our
data allow us to provide a necessary corrective to the often
Kefford et al. 3

simplistic and uniform coverage of DDC by political parties
in scholarly and popular commentary.
In order to draw insights from this data, we asked all
authors to provide responses to a set of standardised re-
search questions inspired by Dommett’s (2019) theoretical
data-driven framework on data collection, data use, who is
using data, data regulation and recent election campaign
practice. Our analysis focusses on DDC practices in the run
up to national election campaigns for the sake of increasing
comparability.
1
However, given the organisational structure
of many political parties, interviews (when utilised) were
conducted not only with candidates, party officials and
campaign consultants working in national campaign
headquarters, but also with actors at sub-national and local
levels. Due to inconsistencies in data collection, we were
unable to offer comparative analysis on all topics, and hence
refined our focus to concentrate on three aspects of DDC:
Data collection, data use and data infrastructure. We discuss
each theme in turn and spotlight variance between major
and minor parties in each country. We also reveal areas in
which our data is less comprehensive, showing where
further empirical investigations can fruitfully build on our
work.
Findings
Data collection
Central to many of the concerns that scholars, policymakers
and commentators have about DDC is that it incentivises
parties –or other campaigners –to surveil citizens (Zuboff,
2019) instead of engaging them through other democratic
means. In accordance with many prevailing accounts, DDC
prompts parties to collect tens of thousands of data points
about voters’movements, engagement and behaviour in
online and offline environments (Schechner et al., 2019).
This data, and especially personalised forms of digital trace
data, are seen to make it possible for parties to influence and
even manipulate citizens to promote their electoral goals
(Madsen, 2019). Data can therefore be used to demobilise
certain groups of voters (Bodó et al., 2017) or to develop
subversive forms of persuasive influence (Burkell and
Regan, 2019), reshaping political practice and trans-
forming the logic of the democratic process.
To better understand and contextualise data collection
processes and the logic underpinning this activity, we, first,
asked team members to describe in detail whether and how
parties collected data. In particular, we wanted to reflect on
how data was gathered, asking whether this was freely
available, required party members and supporters to un-
dertake canvassing work, was purchased from external
brokers, or gathered through party polling (see Table A2).
We found that data is widely used and is collected in rel-
atively consistent ways across our cases, with limited
evidence that new media and social change has transformed
established data collection practices. There are important
differences in the practices evident in each country, and in
the capacities of major and minor parties. Challenging the
idea that all parties are able to collect ‘extraordinarily de-
tailed political dossiers’composed of ‘hundreds of millions
of individual records, each of which has hundreds to
thousands of data points’(Rubinstein, 2014: 863–864), our
analysis raises questions about the sophistication and
novelty of contemporary data collection activity.
First, looking at variation across our countries, we found
a longstanding tradition of data collection. Despite the
institutional and regulatory differences, there is relatively
little variation in the type of data collection methods uti-
lised. A combination of state information, canvassing data,
online tools (such as email sign up lists, cookies or social
media ‘matching’data)
2
, polling, and the purchasing of data
are commonly found, suggesting a high degree of com-
monality. However, we do find variation in the type of
information parties are able to collect in each country. Take,
for example, state provided information. In the UK, Aus-
tralia, US and Canada, information is available year-round
but comes in different forms. In the UK, parties can access
the electoral roll and the marked register, offering them
insight into who is registered, who cast their ballot at
previous elections, who has a postal vote and who is a first-
time voter. However, in Germany, whilst some data is
available,
3
certain information can only be purchased
6 months before an election including names, addresses and
educational qualifications if a party asks for data on a clearly
defined population group of a certain age. Moreover, in the
Netherlands, an electoral roll has not been provided since
1951, vastly affecting the information parties can access.
These variations are not only evident when comparing
different countries. In the US, there is little internal uni-
formity, with some states providing information freely and
publicly, some providing it freely but requiring people to go
through a request system or credentialing, and others re-
quiring payment. There are even differences in the available
data in different states, with some offering information on
partisan registration or ethnicity, and others not. In part these
variations reflect alternative regulatory frameworks, with
data protection in European countries in particular curtailing
what information can be shared and utilised, but the vari-
ations within single countries suggest that local rules and
norms are also significant to understanding differences.
Similarly, we found that canvassing was a long-
established practice in many countries, but once again
there was variation as this has longer antecedents in Australia,
Canada, the Netherlands, UK and US, where door-knocking
or phone canvassing are common (Bale et al., 2019;Dobber
et al., 2017;Kefford, 2021;Nielsen, 2012). In contrast, in
Germany, until 2019 it was only the two major parties who
canvassed strategically (Kruschinski and Haller, 2017).
4Party Politics 0(0)

Whilst parties within and between our countries gather
different types of data when canvassing,
4
parties usually
seek to gather data on vote intention and in many countries,
issue positions. Parties in all our cases are also adopting
digital canvassing tools in the form of mobile canvassing
applications, and yet rather than transforming canvassing
activity we found that this practice streamlined established
canvassing activities by removing the need for time-
consuming manual data input.
One area of particular interest is the collection of data
online and its potential to transform campaigning activity.
Often presented as offering a raft of new, more granular
forms of information to political parties (Dobber et al.,
2017: 12), we found that parties were indeed beginning
to gather more data online, but these activities were not
inherently disruptive. In large part online efforts reflect
established offline methods focused on getting individuals
to disclose their own information to campaigns, with parties
using tools such as the UK Labour Party’s‘NHS Baby
Number’which invited people to input their personal in-
formation and email address in order to find out what
number baby they were under the NHS system –a technique
which reportedly harvested over a million email addresses
that the party were able to use for targeted campaign
messaging (Culzac, 2014).
5
In a different vein, parties also
gathered information online without individuals’knowl-
edge or express consent. In Germany, for example, parties
reported themselves to be using the Facebook pixel function
to trace the online activity of voters. Whilst these practices
have the potential to raise privacy concerns, they mirror
established offline activity whereby parties record insights
about individuals that they have not disclosed themselves.
Indeed, there is evidence in Canada and the UK of parties
making inferences about ethnicity or gender from details of
voters’names (McEvoy, 2019), leading them, for example,
to send Eid cards to those believed to be Muslim voters.
Whilst some of the data gathered online was different than
previously available, this type of data collection did not
represent a radical departure from previous data collection
practices.
One possibility is that parties are using online data
sources to build detailed profiles of voters within their own
databases. To this point, we found that online data was often
being used to facilitate specific forms of online commu-
nication but was not being integrated into unified party data
sets. For example, parties in all our countries used the data
access services provided by companies such as Facebook
and Google to gather new insights, but these companies do
not allow parties to buy the underlying data they possess,
curtailing data collection possibilities. Whilst our cases
showed instances in which some data collected online –
such as emails –were paired with parties’existing voter
information data, online sources often appeared to
supplement rather than transform existing data-collection
activity.
Noting these trends, it is important to highlight important
differences in the data collection practices of major and
minor parties. This is especially the case in relation to the
collection of data via canvassing and parties’ability to
purchase data. The reasons for these trends reflect existing
financial disparities between parties. First, canvassing is
labour and capital intensive and some minor parties in the
Netherlands, Germany and Australia did not have either the
labour or capital resources to undertake these activities.
Second, smaller parties in the Netherlands, Germany,
Canada, UK and Australia were less able to finance the
purchase of data either from the state (for example in
Germany), from external companies, or from polling or-
ganisations. Even in the US, campaigns must pay the party
for use of their voter file during primary elections, sug-
gesting that available finance can limit a smaller campaigns’
access to data. Finally, we also found some examples of
smaller parties in the Netherlands and Germany (but not in
our other cases) who limited their data use and acquisition at
an ideological level. An example of this comes from the
Netherlands and the Democrats 66 (D66), whose use of data
was shaped by their principled stance on issues such as data
privacy. Whilst it is theoretically possible for major parties
to adopt such principled positions, we found no evidence of
this occurring within our cases.
Thinking through the consequences of these findings, the
evidence suggests data collection is commonplace, but by
no means uniform. We find little evidence that there has
been a marked shift in data collection practices prompted by
developments in digital technology, and rather, it appears
parties have adapted well-established data collection pro-
cesses to integrate new insights. This suggests that the
democratic concerns raised about ‘new’and revolutionary
practices are overstated. Moreover, while there is reason to
be concerned about parties collecting tracking and other
data via pixels and the subsequent effects on citizen privacy
–and parties in many of our cases were doing this –our
evidence suggests this tracking data was less influential than
is often assumed. In drawing these conclusions it appears
that data acquisition mirrors a trend found elsewhere in
campaigning and party organisation where rather than
providing opportunities for new or emerging minor parties,
we see data –as a significant resource that parties need to
collect to campaign in the contemporary political and media
environment –reinforcing existing hierarchies in party
systems and favouring established major parties (Gibson,
2015;Gibson and McAllister, 2015).
Data use
One of the central claims associated with prevailing de-
pictions of DDC is that campaigns use data to model voter
Kefford et al. 5

behaviour and can send targeted messages proven to be
effective on specific audiences. Christopher Wylie, the
Cambridge Analytica whistle-blower, famously told the
Guardian (cited in Cadwalladr and Graham-Harrison,
2018), ‘We exploited Facebook to harvest millions of
people’s profiles. And built models to exploit what we knew
about them and target their inner demons’. Such capacities,
if widely utilised and effective, would indeed represent a
disruptive force in campaigning practice and democracy,
and yet questions have been raised about how widespread
and new these practices are (Baldwin-Philipi, 2017; 2020).
To provide some much-needed comparative evidence of
how exactly data features in contemporary election cam-
paigns, we asked team members to describe how data was
used by political parties. In particular, we asked them to
outline to what extent: parties identify groups of voters with
certain demographic or attitudinal characteristics to target,
whether parties create models that profile voters, and to
what extent parties use data and analytics techniques to
either create scores about how likely voters are to be
supporters or to be persuadable (see Table A3).
First, across our six cases, in regards to the claim that
parties are developing detailed profiles of citizens, we found
it was common for parties in each country to create scores
on a citizens’likelihood to be a supporter and/or their
persuadability. The sophistication of the practices under-
pinning these scores and the granularity of the data was,
however, exceedingly difficult to assess. This was partially a
product of a lack of transparency and because parties often
delegate these processes to companies who are unwilling to
disclose their processes. This makes it challenging to un-
derstand precisely how models are constructed, but also to
determine how frequently they are used to underpin cam-
paign interventions. The exceptions to this are the US
(Hersh, 2015;Nickerson and Rogers, 2014) and Australia
(Kefford, 2021), where there have now been detailed dis-
cussions of the analytics process campaign operatives
undertake.
There was, however, some evidence that the idea of
modelling –broadly defined –has been widely embraced by
the major parties in the six case studies. When it comes to
identifying supporters, we found major and many minor
parties attempting to identify the attributes of likely sup-
porters to target voters. In the Netherlands, for example,
Groen Links used profiling techniques to identify likely
voters (e.g. showing highly educated women, living in a city
to be more likely to be a GroenLinks voter). Similar ap-
proaches can be observed in the personas identified by
campaigns as groups of target voters, with a search in the
Australian Liberal Party for ‘tradies’and similar techniques
also used by the German Christian Democrats to identify
groups of like-minded voters. These techniques are not,
however, particularly new, as polling and focus groups have
been used for decades to identify target audiences and this
has also been a common approach used by data brokers and
commercial marketing operations for decades (Kusche,
2020).
Our analysis also explored the extent to which parties
were using data to identify fine-grained target audiences. We
found that whilst occurring to some extent in our six cases,
these practices were not uniform. In the UK, Germany and
the Netherlands, for example, parties engaged in a form of
‘narrowcasting’(Kefford, 2021), communicating specific
policy pledges to particular groups such as students or
pensioners. Many of these appeals did not involve so-
phisticated modelling or attempts to determine the most
effective forms of messaging, but rather used simple geo-
graphic or demographic information to identify a target
group. Such practices occurred both offline and online and
were a longstanding feature of party communication
(Fulgoni et al., 2016).
It is, however, important to note that across our cases,
there was evidence of parties utilising the services of
technology companies to aid targeting. In all our cases
parties use Facebook tools such as ‘core’,‘lookalike’or
‘custom’audiences to identify specific types of voters or
those who had certain attributes in common, such as lo-
cation, age, language or gender. Previous research within
platform studies has highlighted how the affordances of
digital technologies can affect how campaigns are organised
(Nielsen and Ganter, 2018), and our findings mirror this. We
did, however, find differing relationships between parties
and such companies. Whilst in Germany we found evidence
of the CDU working closely with Facebook to construct
target audiences for specific topics and campaign times,
reflecting practices previously found in the US (Kreiss and
McGregor, 2018), this kind of practice was not found
widely elsewhere. Indeed, even looking in more detail
within the US, while the two major parties do this, and make
it easy enough that even congressional campaigns can pull
lists of targets to canvass and call, most House and even
some Senate races –especially those that were not contested
or which were poorly financed –do not engage in their own
sophisticated modelling or have access to direct relation-
ships with social media companies.
Beyond platform affordances, the precise form of tar-
geting evident in our cases was often heavily informed by
the institutional, behavioural and electoral dynamics of each
case. In Australia, for example, where there is compulsory
voting and where turnout is high, there was less incentive
for parties to focus on identifying infrequent voters and
deploying mobilising messages. In contrast, in the US,
where voting is not compulsory, where turnout in some
areas is historically low, and where specific districts are
electorally important, such targeting efforts were central.
The incentive to target different groups therefore varied
dependent on the particular context, resulting in inconsistent
practice across our cases.
6Party Politics 0(0)

Finally, we also explore the extent to which parties used
data to test message effectiveness. Financial and time
limitations have historically limited parties’ability to de-
velop highly differentiated, or multiple iterations of, cam-
paign messaging. Digital technology has, however, made it
easier for parties to test alternative messages, and to deliver
these easily at low cost.
6
In the Netherlands, for example, at
the 2021 election, the Facebook advertising archive showed
the CDA party to be running 40 different versions of one ad.
Similar experimentation was also evident at recent elections
in the UK and US. In other countries, such experimentation
was less evident despite platforms’efforts to make it easier
for campaigns to execute such a strategy. Evidence from
Germany and the Netherlands suggests that parties often
lack the capacity (discussed further below) to experiment
and test campaign interventions within an election period,
meaning that little testing occurs. Even in places where
testing is common, our analysis suggested that parties’
capacity to utilise these insights was limited, with evidence
from Canada showing that parties frequently failed to re-
view this data to monitor or evaluate campaign interven-
tions (Munroe and Munroe, 2018). Whilst parties are
therefore often interested in exploring the potential of
testing the effectiveness of messages, to date these practices
are not being as widely employed as is often assumed.
Our analysis therefore suggests data and analytics are an
important aspect of campaigning, and that in many of our
countries data is being used to develop models, to deliver
targeted and tested interventions, and to evaluate campaign
activity. However, the uptake and use of these tools is not
universal or as sophisticated as many prevailing accounts
imply. Whilst many parties are using the affordances pro-
vided by organisations such as Facebook to gain infor-
mation about voters and to tailor messaging, political parties
are largely using digital media to promote mundane and
well-established campaign targeting strategies within a
hybrid media system, and they are constrained in their
ability to micro-target and message test.
Data infrastructure
One often implicit implication of prevailing depictions of
DDC relates to the disruptive impact of data on campaign
(and party) organisation and the subsequent effects this has
on how these organisations engage with citizens. While the
work of Kreiss (2016) shows that such organisational
change has been a much longer term and uneven process,
contemporary arguments that data ‘drives’campaign in-
terventions imply that modern campaigns invest resources
in data personnel and infrastructure, and position data and
analytics teams at the centre of campaign decision-making.
To investigate this, we asked team members whether parties
had paid staff to deal with data and analytics processes,
whether they paid external companies to undertake data and
analytics work on their behalf, and whether they had be-
spoke or generic data management systems (see Table A4).
7
We found investment in data and analytics staff within
parties across national contexts, but there was a clear divide
between major and minor parties and important variations
over the electoral cycle. In each of the cases major parties
were investing in paid staff to do data management work.
Whilst it is not possible to get precise details on staffing
from all parties due to a lack of transparency, US parties
appear to contain the largest data teams, with dozens of staff
devoted to this work within both the Democratic and Re-
publican parties. In contrast, parties in other countries
contain far smaller data teams, often composed of a handful
of individuals. Election dynamics directly affect parties’
ability to do this work, as it is common in Australia,
Germany, the UK, Netherlands and US for major parties to
expand the size of their data teams in the run up to an
election, but outside the US this expanded capacity rarely
took data teams above 5–10 team members, and in minor
parties it was common to find only 1-2 devoted employees,
if there were any. Staff understanding of data analytics
within parties’wider campaign organisation is, however,
often limited. In Germany, for example, individuals rarely
have a background in data analysis, a point that became
apparent when, during one interview, a local political
campaign strategist asked what ‘predictive modelling’was.
One possible alternative to developing in-house capac-
ities is the potential for parties to use external companies to
analyse data and to create models, and we found significant
evidence of this. Developing the idea that parties are
drawing on a broader ecosystem of service providers
(Dommett et al., 2020), some major and minor parties paid
external companies to undertake such work. Companies
such as CrosbyTextor, Blue State Digital and Harris Media
worked with parties in multiple countries (i.e. UK, Aus-
tralia, Canada, Germany and the US), often (but not always)
working with parties with common ideological agendas.
There were also many specialist agencies who supported
and contributed to parties’data activities. In the Nether-
lands, parties have collaborated with made2matter, SUE,
and Roundabout. In the US, both parties have purchased
data from vendors (e.g. i360 and Catalist). Looking across
the parties within our cases, we found that by no means all
parties used these services, with the financial implications of
such relationships often acting as a key inhibitor on minor
parties.
Whilst data collection and analysis has become an im-
portant aspect of party campaigning, within our cases we
found campaign strategy was often not determined by data,
with other factors such as party leader/s preference or local
discretion instead determining how campaigning occurred.
Within the UK, for example, following the 2017 General
Election it was claimed that Theresa May elected not to
implement the ‘data-driven’strategy developed by the
Kefford et al. 7

consultancy firm CrosbyTextor and pursued her own
messaging (Valent Projects, 2020:3–4). Meanwhile in
Canada, Bennett (2015) has highlighted divisions within
parties between ‘traditionalists –relying on face-to-face
methods of canvassing –and the new breed of high-tech
party workers’, indicating that support for data-driven
techniques is not uniform within parties. We also found
evidence that party organisation and structure can affect the
extent of data influence (Kefford, 2018). In-depth analysis
of organisational structures in the German SPD party and
organograms of the UK Labour Party, for example, have
shown data and analytics teams to be cut off from other
teams, limiting their influence on parties’decision-making.
Elsewhere, we found evidence that decentralised parties did
not always draw on data when making campaign inter-
ventions. In Germany, for example, the SPD lack a top-
down organisational structure to allow them to implement
and communicate a data-driven strategy in their different
local chapters, resulting in local activists making autono-
mous decisions often uninformed by data.
In terms of investment in data infrastructure, we found
many parties across the six advanced democracies were
using a bespoke data management system. The evidence
suggests there were a wide range of large and often so-
phisticated databases that allow parties to upload, store and
analyse information. In Germany, Canada, the UK and the
US, all the major parties were using bespoke systems,
8
while in the Netherlands and Australia this was only true of
some of the major parties. We also found evidence that
many parties adopted systems found in other countries,
often purchasing and adapting systems from the US for their
own needs. In the UK, for example, the Conservative Party
has purchased ‘Voter Vault’which was developed for the
Republican Party, whilst in Canada the Liberal Party system
‘Liberalist’is modelled on the VoteBuilder software utilised
by the Obama campaigns (Bennett and Bayley, 2018: 14). In
addition, we found evidence of parties using external
companies to store and manage data. NationBuilder, for
example, was used in Australia, Canada, Germany and the
UK, with parties often running activist management op-
erations and creating websites via this platform. Whilst
these ‘off-the-shelf’systems are widely used by major and
minor parties in the Netherlands, Germany, Canada and the
UK, these systems were routinely described as glitchy and
difficult to use. Whilst many major parties had (often
limited) funds to invest in the maintenance and improve-
ment of these systems, or to supplement these services,
many minor parties did not have the resources to do so.
Cumulatively, data from our six countries suggests
parties almost universally recognise the importance of in-
vesting resources in data personnel and infrastructure,
however, many lacked the financial resources to fund full-
time staff or expert advice, and found it challenging to
maintain often unwieldy databases. These challenges were
made more substantive by the ‘boom and bust’cycles of the
electoral calendar –in major election years support comes
but dries up between elections. As such, even within major
parties, levels of investment were often low. Moreover, data
teams often remain peripheral parts of campaign organi-
sation and their influence is not guaranteed, suggesting that
the disruptive impact of data may not be as extensive as
often claimed.
9
These findings suggest that, as with data
collection and data use, evidence of how parties are in-
vesting in data infrastructure, does not reflect the threat to
democratic norms and principles outlined in much existing
scholarship. Instead, we see evidence –especially outside
the US –of organisations struggling to keep pace with
broader changes in media, technology and political par-
ticipation. DDC is expensive, and many political parties do
not have the resources to invest in bespoke systems or large
numbers of data and analytics personnel.
Conclusions
In setting out to study DDC across our six cases, we found
sustained evidence of parties collecting data, identifying
particular audiences for messaging, creating models and
investing in infrastructure either within or beyond their
organisation. However, in contrast to prevailing accounts
that have offered a fairly uniform depiction of the role data
plays in campaigns and the disruptive influence it is having
on parties, our analysis demonstrates variations in how data
is collected, used and resourced both between and within
cases. Recognising this diversity, our conclusions have
implications for current debates around the impact of DDC.
It is often claimed that changes in the technological, media
and social landscape have led parties to engage in hyper-
intensive data practices including subversive and invasive
forms of online data collection, sophisticated profiling,
highly personalised targeting and real-time campaign
evaluation. Whilst these practices can be found within the
US, there is variation even within this case. Even more
starkly, our evidence suggests these practices are not
common in the other five advanced democracies examined.
Indeed, we show that many aspects of parties’data col-
lection and analysis are long-standing and largely mundane,
whilst their data infrastructure is often curtailed and fre-
quently does not ‘drive’decision-making. There is therefore
little evidence of parties conducing sophisticated data
scraping operations, instead they tend to rely on basic, state
provided information complemented with simple voter
canvassing and information about voting behaviour and
issue positions. Whilst there is evidence that parties are
supplementing this information with new forms of data
gathered online and are utilising the affordances provided
by platforms to identify target audiences, much day-to-day
data use builds on longstanding principles of audience
identification and engagement, focussing on broad appeals
8Party Politics 0(0)

rather than targeting at the individual level. It may therefore
be technically possible for parties to collect fine-grained
data, and to personalise political messages, but in practice
we find that many parties engage in what Kefford (2021)
describes as narrowcasting. Our findings therefore suggest
that developments in digital technology, media and broader
society are not transforming data practices, but are
prompting them to evolve.
Reaching these conclusions, our analysis suggests there
is value in moving away from a ‘media-centric’account of
DDC that focuses on the transformational and disruptive
impact of new media developments. Instead, it encourages a
focus on campaign organisations, inviting us to explore how
in different countries these organisations are adapting to the
technological and societal developments they confront. In
utilising a ‘party-centric’approach, our analysis suggests
that rather than data being a disruptive force that is trans-
forming contemporary election campaigning, we see long-
standing power differentials maintained, and in some cases
reified because of the labour, capital, and skill needed to
conduct campaigning. DDC, we argue, is therefore not
inherently problematic or deterministic, but is a diverse set
of practices that reflect a new era of democracy in which
technology giants exercise significant power, the traditional
media landscape is fragmenting and changing modes of
citizen participation are creating new expectations about
politics.
There is also little evidence to suggest that these practices
are inherently a threat to democracy, while also recognising
the dangers for individual privacy. While not the primary
focus of our analysis, these findings are worth discussing in
terms of how they affect the competition between parties in
each of our countries. It is certainly true that DDC is labour
and capital intensive, and this has the potential to contribute
to the dominance of the major parties which are often highly
resourced at the expense of new or emerging parties.
However, we would suggest that while DDC is a feature of
party campaigning in many advanced democracies, the
efficacy of these campaigns remains a source of debate
(Broockman and Kalla, 2020;Kalla and Broockman, 2018).
Likewise, there is little to suggest that these practices are
inherently strengthening the linkage role that parties are
theoretically meant to play. While DDC may assist parties in
mobilising members and supporters and often manifests in
offline practices such as an increased emphasis on direct
voter contact, there is insufficient evidence to conclude that
this is assisting parties in placing themselves as the central
node between citizens and the state or even that political
parties wish for this to be the case. We therefore argue that it
is not the case that DDC is disrupting democracy, but
democratic developments –such as a fragmenting media
landscape, changes in political participation and techno-
logical advances –are disrupting campaign practices.
This study certainly has limitations. We acknowledge the
inconsistency of data, and particularly highlight the com-
parably limited data our experts were able to gather on the
Canadian case. These variations mean that certain con-
clusions, particularly in relation to Canada, should be
caveated (Table A4). Our data also focuses on six countries
where DDC use is already established. These cases should
therefore not be used to generalise to other countries whose
adoption of these techniques may be less advanced. Despite
these limitations, we contend that this article has made a
significant contribution to our understanding of DDC,
providing an important corrective to arguments DDC is
disrupting democracy. Instead, we argue that democratic
developments –such as a fragmenting media landscape,
changes in political participation and technological ad-
vances –are disrupting campaign practices.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with re-
spect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support
for the research, authorship, and/or publication of this article: This
work was supported by the Australian Research Council
(190100210), the Economic and Social Research Council (ES/
N01667X/1), the Canada Research Chairs program (950-231159),
McMaster University Faculty of Humanities, the NORFACE Joint
Research Programme on Democratic Governance in a Turbulent
Age and co-funded by ESRC, NWO, FWF and the European
Commission through Horizon 2020 under grant agreement No
822166.”
ORCID iDs
Glenn Kefford https://orcid.org/0000-0002-6733-3323
Katharine Dommett https://orcid.org/0000-0003-0624-6610
Jessica Baldwin-Philippi https://orcid.org/0000-0002-8664-
6827
Sara Bannerman https://orcid.org/0000-0002-2295-3254
Tom Dobber https://orcid.org/0000-0002-6657-4037
Simon Kruschinski https://orcid.org/0000-0002-3185-5656
Notes
1. Our focus on the run up to election campaigns should be noted
as our data suggests that the use of DDC fluctuates throughout
the electoral cycle. We found, for example, that increased levels
of staffing and resource are devoted to DDC in the run up to
elections, and that activity is often sparse after an election
occurs. This suggests the importance of further analysis of the
use of DDC in non-electoral periods, but such consideration is
beyond the scope of this particular article.
Kefford et al. 9

2. For example, Facebook’s‘lookalike’audience feature or Na-
tionBuilder’s‘social media matching’feature.
3. Accumulated voter data from statistical offices of the state
(residential districts, age, education, household size, proportion
of foreigners, religious affiliation, unemployment rate) and
accumulated voter data from the Federal Election Commis-
sioner can be accessed for free throughout the year (past
election results, voter turnout).
4. We found examples of variation in the scripts that were used by
different parties that showed some parties, for example, to
gather data on voters’interests, concerns or willingness to
display a poster at election time.
5. This inspired similar data harvesting operations in their ‘sister’
party in Australia –the Australian Labor Party, and Australia’s
other major party, the Liberal Party, were also employing
similar data harvesting techniques (Kefford, 2021:77–79).
6. As an example, Facebook’s Dynamic Creative tool, which
automates design variations of adverts, allows campaigns with
few resources to easily create and test a variety of adverts,
allowing them to run significantly more message variations in
that platform. This hints to the fact that –especially legally
restricted European parties –can utilise companies such as
Facebook to facilitate targeting.
7. By bespoke we mean that the data management system has been
built specifically for that party rather than the party uses a
system from an ‘off-the-shelf’provider such as NationBuilder.
8. Interesting variations at the federal and provincial level in
Canada have been revealed by McKelvey and Piebiak (2018).
9. Indeed, even in the US the centrality of data teams is debatable
and Baldwin-Philippi (2019), has noted how digital and data
teams often support all mobilisation and persuasion campaign
efforts and ‘often hold equal power in campaigns’.
References
Anstead N (2017) Data-driven campaigning in the 2015 United
Kingdom general election. The International Journal of
Press/Politics 22: 294–313.
Baldwin-Philippi J (2017) The myths of data-driven campaigning.
Political Communication 34: 627–633.
Baldwin-Philippi J (2019) Data campaigning: between empirics
and assumptions. Internet Policy Review 8: 1–18.
Bale T, Webb P and Poletti M (2019) Footsoldiers: Political Party
Membership in the 21st Century. Abingdon, UK: Routledge.
Bennett CJ. (2015) Trends in voter surveillance in Western so-
cieties: privacy intrusions and democratic implications.
Surveillance and Society 13(3/4): 370–384.
Bennett C and Bayley R (2018) The Influence Industry in Canada.
Berlin, Germany: Tactical Tech. Available at: https://cdn.ttc.
io/s/ourdataourselves.tacticaltech.org/ttc-influence-industry-
canada.pdf.(accessed on 19 May 2021).
Bodó B, Helberger N and and de Vreese CH (2017) Political
micro-targeting: a Manchurian candidate or just a dark horse?
Internet Policy Review 6. Available at: https://doi.org/10.
14763/2017.4.776
Broockman D and and Kalla J (2020) When and Why Are
Campaigns’Persuasive Effects Small? Evidence from the
2020 US Presidential Election. OSF Preprints. Available at:
https://osf.io/m7326/
Burkell J and Regan PM (2019) Voter preferences, voter manip-
ulation, voter analytics: policy options for less surveillance
and more autonomy. Internet Policy Review 8: 1–24.
Cadwalladr C and Graham-Harrison E (2018) Revealed: 50 Million
Facebook Profiles Harvested for Cambridge Analytica in
major data breach. London, UK: The Guardian. Available at:
https://www.theguardian.com/news/2018/mar/17/cambridge-
analytica-facebook-influence-us-election.(accessed on 19
March 2018).
Chester J and Montgomery KC (2017) The role of digital mar-
keting in political campaigns. Internet Policy Review 6: 1–20.
Culzac N (2014) Labour Launches Clever Tool to Find Out Your
NHS Baby Number and Harvests Thousands of Emails in the
Process. Independent. Available at: https://www.independent.
co.uk/news/uk/home-news/labour-launches-clever-tool-find-
out-your-nhs-baby-number-and-harvests-thousands-emails-
process-9587636.html.(accessed on 7 July 2019).
Dobber T, Trilling D, Helberger N, et al. (2019) Spiraling
downward: the reciprocal relation between attitude toward
political behavioral targeting and privacy concerns. New
Media & Society 21: 1212–1231.
Dobber T, Trilling D, Helberger N, et al. (2017) Two crates of beer
and 40 pizzas: the adoption of innovative political behav-
ioural targeting techniques. Internet Policy Review 6.
Available at: https://doi.org/10.14763/2017.4.777
Dommett K (2019) Data-driven political campaigns in practice:
understanding and regulating diverse data-driven campaigns.
Internet Policy Review 8: 1–18.
Dommett K, Kefford G and Power S (2020) The digital ecosystem:
the new politics of party organization in parliamentary de-
mocracies. Party Politics 27(5): 847–857.
Fulgoni GM, Lipsman A and Davidsen C (2016) The power of
political advertising: lessons for practitioners: how data an-
alytics, social media, and creative strategies shape US
presidential election campaigns. Journal of Advertising Re-
search 56: 239–244.
Gibson RK (2015) Party change, social media and the rise of
‘citizen-initiated’campaigning. Party Politics 21: 183–197.
Gibson RK and McAllister I (2015) Normalising or equalising
party competition? Assessing the impact of the web on
election campaigning. Political Studies 63: 529–547.
Gorton WA (2016) Manipulating citizens: how political cam-
paigns’use of behavioral social science harms democracy.
New Political Science 38: 61–80.
Harker M (2020) Political advertising revisited: digital cam-
paigning and protecting democratic discourse. Legal Studies
40: 151–171.
10 Party Politics 0(0)

Hersh ED (2015) Hacking the Electorate: How Campaigns
Perceive Voters. , Cambridge, UK: Cambridge University
Press.
in ’t Veld S (2017) On democracy. Internet Policy Review 6.
Available at: https://doi.org/10.14763/2017.4.779
International Institute for Democracy and Electoral Assistance
(2018) Digital Microtargeting. Available at: https://www.
idea.int/sites/default/files/publications/digital-microtargeting.
pdf (accessed on 4 May 2020).
Jamieson KH (2013) Messages, micro-targeting, and new media
technologies. The Forum 11: 429–435.
Jungherr A, Rivero G and Gayo-Avello D (2020) Retooling
Politics: How Digital Media are Shaping Democracy.
Cambridge, UK: Cambridge University Press.
Kalla JL and Broockman DE (2018) The minimal persuasive
effects of campaign contact in general elections: evidence
from 49 field experiments. American Political Science Review
112: 148–166.
Kefford G (2018) Digital media, ground wars and party organi-
zation: does stratarchy explain how parties organize election
campaigns? Parliamentary Affairs 71(3): 656–673.
Kefford G (2021) Political Parties and Campaigning in Australia:
Data, Digital and Field. London, UK: Palgrave Macmillan.
Kreiss D (2016) Prototype Politics: Technology-Intensive Cam-
paigning and the Data of Democracy. New York, NY: Oxford
University Press.
Kreiss D and Howard PN (2010) New challenges to political
privacy: lessons from the first US presidential race in the Web
2.0 era. International Journal of Communication 4: 19.
Kreiss D and McGregor SC (2018) Technology firms shape po-
litical communication: the work of Microsoft, Facebook,
Twitter, and Google with campaigns during the 2016 US
presidential cycle. Political Communication 35: 155–177.
Kruschinski S and and Haller A (2017) Restrictions on data-
driven political micro-targeting in Germany. Internet Policy
Review 6. Available at: https://doi.org/10.14763/2017.4.780
Kusche I (2020) The old in the new: voter surveillance in political
clientelism and datafied campaigning. Big Data & Society 7:
2053951720908290.
Madsen JK (2019) The Psychology of Micro-Targeted Election
Campaigns. Cham, Switzerland: Springer.
McEvoy M (2019) Political parties, campaign data, and voter consent.
OIPC. Available at: https://www.oipc.bc.ca/investigation-reports/
2278 (accessed on 7 July 2020).
McKelvey F and Piebiak J (2018) Porting the political campaign:
the NationBuilder platform and the global flows of political
technology. New Media & Society 20: 901–918.
Munroe KB and Munroe H (2018) Constituency campaigning in
the age of data. Canadian Journal of Political Science/Revue
canadienne de science politique 51: 135–154.
Nickerson DW and Rogers T (2014) Political campaigns and big
data. Journal of Economic Perspectives 28: 51–74.
Nielsen RK (2012) Ground Wars: Personalized Communication in
Political Campaigns. Princeton, NJ: Princeton University Press.
Nielsen RK and Ganter SA (2018) Dealing with digital
intermediaries: a case study of the relations between pub-
lishers and platforms. New Media & Society 20: 1600–1617.
Plasser F and Plasser G (2002) Global Political Campaigning: A
Worldwide analysis of Campaign Professionals and Their
Practices. Westport, CT: Greenwood Publishing Group.
Pons V (2016) Has Social science taken over electoral campaigns and
should we regret it? French Politics, Culture & Society 34: 34–47.
Roemmele A and Gibson R (2020) Scientific and subversive: the
two faces of the fourth era of political campaigning. New
Media & Society 22: 595–610.
Rubinstein IS (2014) Voter privacy in the age of big data. Wis-
consin Law Review 861–936.
Schechner S, Glazer E and Haggin P (2019) Political Campaigns
Know Where You’ve Been. They’re Tracking Your Phone.
New York, NY: WSJ. Available at: https://www.wsj.com/
articles/political-campaigns-track-cellphones-to-identify-
and-target-individual-voters-11570718889. (accessed on
14 October 2019).
Valent Projects (2020) Conservative party’s election 2019 digital
strategy. Available at: https://static1.squarespace.com/static/
5dfccc440f740335173284c3/t/5ef3473737cfa455b6efcfdc/
1593001786992/Valent-Conservative_ElectionComms.
pdf (accessed on 7 September 2020).
Zuboff S (2019) The Age of Surveillance Capitalism: The Fight for
a Human Future at the New Frontier of Power. London, UK:
Profile books.
Zuiderveen Borgesius F, M¨
oller J, Kruikemeier S, et al. (2018)
Online political microtargeting: promises and threats for
democracy. Utrecht Law Review 14: 82–96.
Kefford et al. 11

Appendix 1
Table A1 Comparing Data Across Our 6 Cases.
Type of data
collected
Number of
interviews
and type
Type of document
analysed Period of data collection Number of parties
Australia Interviews,
documentary
analysis,
participant
observation
167 including
candidates, party
officials and
campaign
consultants
Internal party documents
Public party documents
Party websites
Media coverage
2015–2020 6 (2 major, 4 minor)
Australian Labor Party
Liberal Party of Australia
Australian Greens
The Nationals
Pauline Hanson’s One Nation
The Nick Xenophon Team
Canada Documentary
analysis
0 Government documents
(Elections Canada;
Office of the Privacy
Commissioner of
Canada); political
parties’privacy policies;
secondary sources
(academic and news
sources)
2012–2020 5 (2 major, 3 minor)
Liberal Party of Canada
Conservative Party of Canada
New Democratic Party
Green Party
Bloc Quebecois
Germany Interviews,
Documentary
analysis and
content analysis
of Facebook,
Google and
Snapchat ad
libraries
43 (with 53
interviewees)
including
candidates, party
officials, campaign
strategists; and
consultants
Internal party documents
(Political parties’
privacy policies; ethical
guidelines for DDC)
Public party documents
Party websites
Media coverage
Government documents
(Office of the Privacy
Commission)
Memoires and first-hand
accounts
Parliamentary Reports
Regulatory reports or
investigations
Non-profit institution
reports
2014-2018 state elections
2017–2021 federal elections
7 (2 major, 5 minor)
Social Democratic Party;
Christian Democratic Union;
Christian Social Union;
The Greens;
Free Democratic Party;
The Left;
Alternative for Germany
Netherlands Interviews,
documentary
analysis and
content analysis
of Facebook,
Google and
Snapchat ad
libraries
8 Public party documents,
financial reports
2016–2021 (8 parties were
interviewed in 2016
[Labour Party, D66, CDA,
Green Left, Christian
Union, Senior’s Party,
Reformed Political Party,
Socialist Party]. Four
parties that were
interviewed in 2016, were
also interviewed in 2021
[D66, GreenLeft,
ChristianUnion, Labour
Party]. Content analysis
was conducted in 2019 and
2021. Analysis of
documents occurred
between 2016 and 2021)
16 (6 major, 10 minor)
People’s Party for Freedom and Democracy
Democrats 66
Party for Freedom
Christian Democratic Appeal
Labour Party
Green Left
Socialist Party
Christian Union
Forum for Democracy
Volt Netherlands
Party for the Animals
Right Answer 2021
Reformed Political Party
Think
Farmers Party
Bij1 (2gether)
UK Interviews,
documentary
analysis and
content analysis
of Facebook and
Google
advertising
archive
57
Including with party
staff, campaign
consultants, and
activists
Internal party documents
Public party documents
Party websites
Media coverage
Memoires and first-hand
accounts
Parliamentary Reports
Regulatory reports or
investigations
Financial spending returns
2015–2018; 2021 6 (2 major, 4 minor)
Labour Party
Conservative Party
Liberal Democrats
Greens
Scottish National Party
United Kingdom Independence Party
US Interviews and
documentary
analysis
13, including party
staff, and campaign
staff/consultants
(some of whom
have worked in
both positions)
Public party documents
Party websites
Media coverage
Financial spending returns
2018–2020 2 (2 major)
Democratic Party
Republican Party
12 Party Politics 0(0)

Table A2 Data Collection in Six Advanced Democracies
a
.
None Some minor All minor Some major All major Other
To what extent do parties in your country collect data through canvassing activity?
Netherlands X X
Germany X X
Canada X X
UK X X
Australia X X
US X
To what extent do parties in your country collect data through online activity?
Netherlands X
Germany X X
Canada X X
UK X X
Australia X X
US X
To what extent do parties in your country have access to data from the state?
Netherlands X
Germany X X
Canada X X
UK X X
Australia X X
US X
To what extent do parties in your country purchase data from companies such as data brokers?
Netherlands X
Germany X X
Canada X X
UK X X
Australia X X
US X
To what extent do parties in your country conduct private polling to gather data?
Nation X
Nation X X
Nation X X
Nation X X
Nation X X
Nation X
a
* means response comes with caveats based on limited data available to the researcher
Table A3 Data Use in Six Advanced Democracies.
None Some minor All minor Some major All major Other
To what extent do parties identify groups of voters with certain characteristics to target campaign interventions at?
Netherlands X X
Germany X X
Canada X* X*
UK X X
Australia X X
US X
To what extent do parties create models that profile voters?
Netherlands X X
Germany X X
(continued)
Kefford et al. 13

Table A3 (continued)
None Some minor All minor Some major All major Other
To what extent do parties identify groups of voters with certain characteristics to target campaign interventions at?
Canada X*
UK X X
Australia X X
US X
To what extent do parties use data and analytics techniques to create scores about how likely voters are to be
supporters?
Netherlands X X
Germany X X
Canada X X
UK X* X*
Australia X X
US X
To what extent do parties use data and analytics techniques to create scores about how persuadable voters are likely to
be?
Netherlands X X
Germany X
Canada X*
UK X* X*
Australia X X
US X
Table A4 Data Use in Six Advanced Democracies.
None Some minor All minor Some major All major Other
To what extent do parties have paid staff devoted to work on data management and analysis?
Netherlands X* X*
Germany X X
Canada X*
UK X* X
Australia X X
US X
To what extent do parties pay external companies to analyse data and create?
Netherlands X X
Germany X X
Canada X X
UK X* X
Australia X
US X
To what extent do parties have a data management system that is bespoke to the party, meaning it was either designed
or adapted for a specific parties’purpose and is not a generic system used by many others, for example, Voter vault,
Excaliber?
Netherlands X* X*
Germany X*X
Canada X X
UK X X
Australia X
US X
(continued)
14 Party Politics 0(0)

Author Biographies
Glenn Kefford is a Senior Lecturer in Political Science in
the School of Political Science and International Studies
at the University of Queensland and the author of Po-
litical Parties and Campaigning in Australia: Data,
Digital and Field.
Katharine Dommett is a Senior Lecturer in the Public
Understanding of Politics in the Department of Politics and
International Relations at the University of Sheffield and the
author of The Reimagined Party: Democracy, Change and
the Public.
Jessica Baldwin-Philippi is an Associate Professor at Ford-
ham University and the author of Using technology,
building democracy: Digital campaigning and the con-
struction of citizenship.
Sara Bannerman is the Canada Research Chair in Commu-
nication Policy and Governance at McMaster University and
the author of International Copyright and Access to Knowledge.
Tom Dobber is an assistant professor in the department of
Political Communication & Journalism of the University of
Amsterdam.
Simon Kruschinski is a Research Higher Degree student in
the Department of Communication, Johannes Gutenberg
University Mainz, Germany.
Sanne Kruikemeier is an Associate Professor in Political
Communication and Journalism in the Communication
Science department of the University of Amsterdam.
Erica Rzepecki is a Research Higher Degree student at
McMaster University.
Table A4 (continued)
None Some minor All minor Some major All major Other
To what extent do parties have paid staff devoted to work on data management and analysis?
To what extent do parties use a data management system that is not bespoke but is rather used in a number of different
countries or parties, for example, NationBuilder or ecanvasser?
Netherlands X
Germany X
Canada X X
UK X X
Australia X X
US X*
Kefford et al. 15