Content uploaded by Jean-Nicolas Bordeleau
Author content
All content in this area was uploaded by Jean-Nicolas Bordeleau on Aug 31, 2022
Content may be subject to copyright.
Do Unfounded Allegations of Election Fraud Influence the Likelihood of Voting?
1
Jean-Nicolas Bordeleau
2
School of Political Studies
University of Ottawa
Abstract: The legitimacy of the electoral process is often put into question by political candidates
and elites who seek to account for their loss. As a result, a significant portion of voters are
presented with unfounded allegations of widespread election fraud even though such fraud seldom
occurs in consolidated democracies. Previous research has determined that misleading claims
regarding the integrity of elections carry important implications for citizens’ perceptions of
electoral fairness. However, the literature has yet to systematically explore the impact of electoral
fraud allegations on voter participation. Using original survey data from the United Kingdom, this
research measures the impact of unfounded allegations of election fraud on the decision to vote or
not. The results of the survey experiment do not support the hypotheses according to which
exposure to unfounded allegations of fraud influences confidence in elections and voter
participation. However, results from supplementary analyses highlight a significant relationship
between perceptions of fraud and subsequent desire to cast a ballot. Explanations for these findings
are discussed.
Keywords: voter participation; election fraud; electoral trust; fact-checking; survey experiment.
Corresponding Author: Jean-Nicolas Bordeleau
School of Political Studies, University of Ottawa
120 University Private, Faculty of Social Sciences
Ottawa, Ontario K1N 6N5
1
Preprint version prepared for the 2022 Annual Meeting of the American Political Science Association
held in Montreal, Canada.
2
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 Centre for the Study of Democratic Citizenship
(CSDC). There are no known conflicts of interests to declare.
1
1. Introduction
Politicians and elites sometime use unfounded claims and political misinformation to promote their
interests and rally their supporters (Berlinski et al., 2021; Clayton et al., 2020; Davis & Ferrantino,
1996; Fogarty et al., 2015). Most notably, political candidates and elites in the United States have
made claims of widespread election fraud in an attempt to delegitimize the electoral process and
rationalize their election loss (Pennycook & Rand, 2021a). These claims, while unsubstantiated,
have gained tremendous attention among the electorate (Fahrenthold et al., 2020). Surveys
conducted following the 2020 U.S. presidential election found that 77 per cent of Trump voters
and 26 per cent of Biden voters believed the election to have been fraught with fraud (Pennycook
& Rand, 2021a). This phenomenon is not isolated to the American context: false allegations of
election fraud have been made by political actors and news media in several other democracies. In
Brazil, for instance, presidential candidate Jair Bolsonaro made claims that the election would be
rigged against him throughout the campaign period (Berlinski et al., 2021). Similarly, in the United
Kingdom, Boris Johnson’s government introduced election reform legislation under the pretext
that stricter measures were needed to combat widespread fraud (Smith, 2021).
Previous research on election fraud has focused on two distinct dimensions: objective cases
of election irregularities and subjective perceptions of fraud. There is an extensive literature which
examines the first dimension. Indeed, objective levels of electoral integrity have been at the centre
of the research agenda on electoral malpractice (Alvarez et al., 2009; Norris, 2014). With some
notable exceptions, researchers have paid significantly less interest to subjective beliefs regarding
election fraud. Recent research has begun examining the determinants of subjective perceptions of
fraud as well as their impact on democratic citizenship norms. In a recent experiment conducted
in the United States, Berlinski et al. (2021) show that unsubstantiated claims of voter fraud
undermine voters’ confidence in elections. Further research has also found election fraud rhetoric
to reduce citizens’ respect for important democratic norms (Albertson & Guiler, 2020; Clayton et
al., 2020). These studies focus almost exclusively on political attitudes, creating an important gap
in the literature when it comes to understanding the impact of subjective perceptions of election
fraud on voter behaviour and electoral participation.
This article therefore seeks to address the gap in the literature by asking the following
research question: do unfounded allegations of election fraud influence voter participation? More
specifically, this research has three objectives: 1) to evaluate the impact of election fraud claims
on the disposition to vote; 2) to understand the mechanisms that underlie the effects of fraud claims
on electoral participation; and 3) to examine whether corrective messages in the form of fact-
checks reduce the impact of allegations of election fraud on voting. Identifying and understanding
the effects of electoral misinformation on voters is crucial to increasing the resiliency of
democratic institutions.
To achieve these research objectives, I rely on the rational calculus of voting (Downs,
1957; Riker & Ordeshook, 1968) and Birch’s (2010) electoral competitiveness reasoning to
propose testable preregistered hypotheses. These hypotheses follow the logic that elections which
are perceived as less competitive have lower turnout (Birch, 2010). Several studies have indeed
demonstrated the sensitivity of voters to the competitive context of elections (Blais, 2000; Fortin-
Rittberger et al., 2017; Franklin, 2004; Norris, 2004). Building on this theoretical reasoning and
using a between-subject survey experiment, I assess my predictions. The online survey experiment
makes use of priming to simulate exposure to claims of election fraud and reduce participants’
perceptions of electoral fairness.
2
2. Election Fraud, Voters, and Voting
Concerns about the integrity of elections are not new. In fact, worries of manipulation in electoral
processes go back as far as electoral democracy itself. In Ancient Greece, certain social groups
exerted overwhelming influence in elections and oftentimes manipulated votes (Taylor, 2007).
Fear of electoral manipulation and fraud was also a major factor in the adoption of the electoral
college in the founding years of the United States (Levin, 2020). More recently, election fraud has
made the headlines around the world and has become a popular term used by populist political
elites and citizens alike (Pennycook & Rand, 2021a). Whether it is referred to as election fraud,
electoral manipulation, election rigging, voter fraud, or meddling; these terms all refer to the same
broad concept. This section focuses first and foremost on defining election fraud and then
exploring the scientific literature relevant to the topic.
2.1. Defining Election Fraud
Defining the concept of election fraud has proven challenging. Some scholars argue for a
more voter-specific approach to election fraud while others adopt an all-encompassing definition
(Birch, 2010; Fogarty et al., 2015; Fortin-Rittberger et al., 2017; Udani et al., 2018). For the
purpose of this research, we adopt the definition developed by the International Foundation for
Electoral Systems (IFES): “Electoral fraud can be defined as any purposeful action taken to tamper
with electoral activities and election‐related materials in order to affect the results of an election,
which may interfere with or thwart the will of the voters.” (López-Pintor, 2010, p. 9)
In modern electoral democracies, widespread electoral fraud can be prevented through
procedural certainty and predetermined electoral governance (Mozaffar & Schedler, 2002;
Przeworski, 1988). In other words, free and fair elections must be conducted according to a fixed
set of rules that does not leave room for flexible enforcement. These rules form part of large legal
frameworks which rely on international standards published by international organizations such as
the United Nations (see, for instance, Article 25 of the International Covenant for Civil and
Political Rights). Therefore, in consolidated democracies, it can be said that electoral fraud takes
place when “the outcome [of an election] is compromised by the politically motivated application
of electoral rules.” (Birch, 2010, p. 1603)
In practice, election fraud can take many forms and can occur at any stage of the electoral
process. For instance, the decision to count some ballots but not others may be considered as
election fraud if the rejection of ballots is done in a partisan manner that advantages or
disadvantages a candidate. Another example includes non-citizens registering and voting illegally
in elections (Fogarty et al., 2015). Other examples include voter impersonation, multiple voting,
and vote buying, to name but a few.
2.2. Election Fraud and Voters
Having defined the concept of election fraud, we now turn to the literature surrounding
electoral manipulation and perceptions thereof. First, it is important to understand that perceptions
of election fraud may differ from actual levels of fraud and electoral integrity. A clear example of
how perceptions can be different from real levels of fraud comes from the 2020 U.S. presidential
elections. Indeed, a survey conducted by Pennycook and Rand (2021a) shows that a majority of
Americans believed in widespread election irregularities. These beliefs were unsubstantiated and
shown to be false by several investigations (Berlinski et al., 2021). In that specific election, proven
3
instances of fraud represented less than 0.0000001% of ballots cast (17 cases out of the 159 million
votes cast). This clearly demonstrates how citizens often overestimate the prevalence of election
fraud and why, in established democracies, studying subjective perceptions of fraud is necessary.
Previous research had been primarily focused on examining the actors who conduct
election fraud, the tactics they use, and the policies that can be adopted to prevent election
irregularities (Alvarez et al., 2009; Lehoucq, 2003; Levin, 2020). Recent developments, however,
have examined perceptions of procedural fairness (Norris, 2013). This shift stems from what
researchers have identified as increased election-related rhetoric which has led to a gap between
real levels of fraud and citizens’ perceived levels of election fraud (Coffé, 2017). Politicians in
several well-established democracies have indeed begun to capitalize on electoral integrity as a
political issue. In doing so, they have politicized electoral management and cast doubt on the
integrity of the process. Even in countries where elections are conducted according to the highest
standards of integrity, some citizens now adhere to the belief that the electoral process is biased,
fraudulent, and unfair (Bordeleau, 2021b; Coffé, 2017; Pennycook & Rand, 2021a).
A large portion of the research on perceptions of election fraud explores the individual
differences that explain higher or lower belief in such fraudulent acts. In the United States, recent
survey evidence suggests that perceptions of fraud are highly polarized along partisan lines. In a
sample of over 1,500 American voters, Pennycook and Rand (2021a) find that 77 per cent of
Trump voters believed in widespread election fraud while less than 26 per cent of Biden voters
did. Perhaps more striking is their finding that over 65 per cent of Trump voters did not perceive
the results of the election as lawful and upheld the idea that Trump was the legitimate winner
(Pennycook & Rand, 2021a). The US partisan divide in perceptions of fraud can be explained by
motivated reasoning and elite discourse. The overwhelming fraud rhetoric coming from
Republican elites has made election fraud a Republican partisan belief. In the specific case of the
2020 US presidential election, attitudes towards fraud were also used as defence mechanisms to
justify defeat at the polls (Fortin-Rittberger et al., 2017). Feelings of electoral loss are hence
correlated with higher levels of fraud beliefs (Levy, 2020).
Moving away from partisanship, recent scholarship has also established that some political
attitudes are strong determinants of belief in fraud. Most notably, researchers have found populist
traits and belief in conspiracy theories to be related to adherence to election skepticism. Indeed,
studies show that populist attitudes and conspiratorial thinking predict a large portion of the
variance in perceptions of electoral integrity (Albertson & Guiler, 2020; Norris et al., 2020).
Sociopolitical worldviews are also related to beliefs in fraud, with individuals scoring high on
social dominance orientation and right-wing authoritarianism holding more negative attitudes
towards the electoral process (Bordeleau, 2021b). Altogether, these findings support the idea that
individuals who respect strong authority figures are more susceptible to adhere to false beliefs. In
this case, those with authoritarian tendencies who like populist leaders are exposed to rhetoric
undermining election legitimacy which leads them to believe in widespread election fraud
(Bordeleau, 2021b; Clayton et al., 2020; Norris et al., 2020).
Elite discourse has also been found to be a crucial element in the diffusion of negative
attitudes towards the electoral process. Clayton et al. (2020) find that exposure to elite statements
regarding election fraud (i.e., allegations of fraud) erodes trust in the process and confidence that
elections are run legitimately. Such false claims also lead to greater beliefs that elections are
rigged, specifically among individuals who approve of the political elite making those claims
(Clayton et al., 2020). In short, exposure to allegations of fraud leads to greater beliefs in fraud
and potentially influences key attitudes towards elections as well as democratic norms.
4
In an experimental survey study, Berlinski et al. (2021) provide an initial examination of
the consequences of exposure to fraud claims on citizens’ confidence in elections and satisfaction
with democracy. Their results suggest that “telling respondents that 'experts' believe that the 2016
election was vulnerable to manipulation and fraud increased perceptions of fraud, lowered
confidence in the electoral system, and reduced willingness to accept the outcome.” (p.3) Emphasis
is placed here on the causal mechanism described by the authors. Exposure to fraud claims causes
higher perceptions of fraud which in turn cause lower confidence in the process. This is key to the
framework which will be used in this research. The Berlinski et al. (2021) experiment also provides
support for the political congeniality hypothesis according to which fraud claims by partisans will
influence co-partisans but not counter-partisans. Similar experimental evidence has been presented
by Goodman (2022) as well as Justwan and Williamson (2022). In their survey experiment,
Justwan and Williamson (2022) find that claims of fraud lead to significantly lower satisfaction
with democracy. Similarly, Goodman (2022) finds that elite rhetoric regarding electoral
manipulation affects citizens’ trust in electoral institutions. Bordeleau (2021a) also reports that
media rhetoric surrounding foreign electoral interference influences voters’ confidence that
elections are free and fair. These studies provide clear empirical evidence that exposure to
allegations of election fraud and higher perceptions of fraud influence individuals’ attitudes
towards the electoral process and democracy.
2.3. Perceptions of Fraud and Participation
One variable has so far received very little attention in the literature on perceptions of
election fraud. Indeed, very little research discusses or engages with turnout or voting behaviour.
Berlinski and colleagues (2021) propose a clear direction for research on the topic: “Future
research could also test the effects of allegations in a pre-election context and possibly examine
effects on turnout or participation intentions.” (p. 13) While the effect of claims of fraud on
electoral participation has not been investigated through experimental research, Birch (2010)
examines this relationship using cross-sectional comparative data.
Using data from over 31 countries, Birch (2010) finds that perceptions of electoral fairness
predict turnout even when controlling for several micro and macro level control variables. A shift
from the bottom to the top of the fraud perception scale (less fraud to more fraud) leads to a turnout
decrease of almost 12 percentage points. This is the first empirical evidence available supporting
the argument that higher perceptions of fraud decrease voter participation. While these findings
are significant, they rely on aggregate-level data (turnout) which limits the ability to make
inferences regarding individual-level behaviours. The findings also rely on observational cross-
sectional data, implying that causality cannot be inferred. It also limits the ability to understand
the role of other factors, such as partisanship and ideology, in shaping the effect of fraud
perceptions on voter participation. With that being said, the Birch study constitutes a first step
upon which the present research will build on.
3. Fact-Checking Allegations of Fraud
Beyond understanding the impact of allegations of election fraud there is also a need to mitigate
their potential effects. A leading approach in dealing with false information regarding electoral
integrity is fact-checking. Fact-checking is defined as “the practice of systematically publishing
assessments of the validity of claims made by public officials and institutions with an explicit
attempt to identify whether a claim is factual.” (Walter et al., 2020, p. 351) In other words, fact-
5
check messages are used to set the record straight regarding a piece of information which has been
shown to be factually incorrect or exaggerated. Such messages can be issued by different actors,
from the original author of the message to third parties.
Recently, political scientists and social psychologists have focused on the development of
a new method to correct misperceptions: accuracy-nudge fact-checking. This type of corrective
messaging was used by Pennycook et al. (2020) and consists in “nudging” individuals in the
direction of thinking about the accuracy of what they are reading. Their study relies on previous
work by psychologists which found that susceptibility to fake news is often caused by lack of
reasoning rather than motivated reasoning (Pennycook & Rand, 2019; Pennycook & Rand, 2021b).
This implies that individuals do not readily consider the accuracy of the news information they are
presented with. As a result, Pennycook et al. (2021) present participants with what they call an
accuracy nudge. These nudges are simple reminders to consider the accuracy of the information
they read. In political journalism, for example, this could be achieved through a banner at the
header of an article cautiously warning individuals to be conscious of the accuracy of the
information they are consuming.
Alongside social psychologists, political communication scholars have taken keen interest
in accuracy-nudge interventions. Andı and Akesson (2020) find that nudging reduces susceptibility
to believe in false news, but also significantly reduces the likelihood that an individual will share
false information. Their experiment focuses on the group norm aspect of news consumption to
demonstrate that accuracy nudges not only activate reasoning mechanisms, but also inhibit
information sharing processes normally prevalent with political misinformation.
4. Present Research
This study seeks to answer two straightforward research questions: do allegations of election fraud
influence voters’ likelihood of voting? and: does fact-checking limit the impact of election fraud
claims on voter participation? Building on rational choice arguments and recent research
developments, we posit several hypotheses regarding the effect of electoral fraud claims on
citizen’s attitudes toward voting and their self-reported likelihood of voting. This specific section
describes the theoretical framework and the main arguments behind the hypotheses.
Previous research suggests voters consider the utility of voting when deciding to cast a
ballot or abstain. Existing rational choice models indeed argue that individuals rely on extrinsic
motivational factors such as the costs of voting (Blais & Daoust, 2020; Downs, 1957; Riker &
Ordeshook, 1968) and the benefits of casting a ballot (Birch, 2010; Karp & Banducci, 2008; Riker
& Ordeshook, 1968) when deciding to vote. In line with this, I argue that individuals’ perceptions
of fraud will factor into these rational motivations. To better illustrate this argument, I propose a
two-step framework which begins with exposure to claims of fraud and ends with the individual
decision to vote or abstain. Figure 1 presents a graphical representation of the two-step framework.
The first step of this model is exposure to claims of fraud. Such allegations influence
citizens’ perceptions of electoral integrity (Berlinski et al., 2021). Indeed, when exposed to these
unfounded claims, individuals develop beliefs that the electoral system is corrupt and that their
votes do not necessarily matter. Those who perceive election fraud to be widespread will have
beliefs according to which each vote does not have equal value or that the outcome of the electoral
process is predetermined (Berlinski et al., 2021; Birch, 2010). In fact, this is supported by the
politicized rhetoric surrounding electoral integrity. In a 2020 address, Donald Trump alleged that
the results of the election had been decided months before election day (Rev, 2020). Under steps
one and two of the model (see Figure 1), this rhetoric and these allegations will lead some electors
6
to develop perceptions that the electoral process is fraudulent. In turn, these negative perceptions
will influence citizens’ attitudes and decision-making, specifically their confidence in the electoral
process and the decision to turn out and vote.
In making their decision, citizens with higher perceptions of fraud (due to allegations) will
believe that the results of the elections may already be out of their hands because of fraudulent
activities or that their vote will matter less than that of fraudsters. Naturally, this would reduce
their perceived benefit of voting and their levels of trust in the system. Eventually, when election
period comes around, this would potentially lead to the decision to abstain. As Birch (2010)
mentions, “If voters fear that polls are corrupt, they have less incentive to bother casting a vote;
participating in a process in which they do not have confidence will be less attractive, and they
may well perceive the outcome of the election to be a foregone conclusion.” (p.1603) Of course,
the influence of perceptions of fraud on voting is not singular, and other factors previously found
to mobilize or demobilize voters remain important (see Blais & Daoust, 2020; Smets & van Ham,
2013). Nevertheless, electoral integrity ought to be a meaningful element in citizens’ evaluation
of the relative importance of their vote. In short, voters who perceive elections as fraudulent will
believe their vote to be inconsequential and therefore will be less likely to want to cast a ballot.
Figure 1. Theoretical framework
Note. This model relies on concepts developed by Berlinski et al. (2021) and Birch (2010).
4.1. Preregistered Hypotheses
The theoretical model presented above puts forth the process under which allegations of
fraud lead to higher perceptions of fraud and subsequently influences the electoral decision-
making. Based on this model, I posit three primary hypotheses. First, I expect there to be a negative
relationship between exposure to unfounded claims of election fraud and individual likelihood of
voting (Hypothesis 1). In addition to effects on behavioural intentions (voting), it is expected that
exposure to fraud claims will negatively affect overall confidence in the electoral system
(hypothesis 2). I further expect the relationship between exposure to fraud claims and voting to be
mediated by individuals’ confidence in elections (Hypothesis 3).
Step 2: Attitude and
behaviour change
Step 1: Exposure
Allegations of Fraud
Perceptions of
Fraud
Confidence in
elections
Likelihood of
voting
7
Hypothesis 1: Individuals exposed to unfounded claims of election fraud will report
a lower likelihood of voting in the upcoming election.
Hypothesis 2: Individuals exposed to unfounded claims of election fraud will report
lower confidence in elections.
Hypothesis 3: The relationship between unfounded claims of election fraud and
intention to vote will be mediated by individuals’ confidence in elections.
My hypotheses are consistent with the rational choice and motivational models of voting. Exposure
to unfounded claims of election fraud should affect the probability that an individual will perceive
their vote as decisive. If an individual believes there is fraud in the electoral process, they should
be more prone to believe that the outcome of the election is out of their hands and rather in the
hands of the fraudsters. They should therefore not have the same attitude towards the importance
of their vote. Indeed, if the election is manipulated, why should one vote? As such, the utility of
voting is decreased, and one’s likelihood of voting should be lower.
In addition to the hypotheses presented above, a fourth hypothesis concerning fact-
checking is posited. Accuracy-nudge interventions provide a quick, effective, and tested way to
incorporate fact-checking in this study. Allegations of election fraud are often unfounded,
unsubstantiated, and provide little details as to the exact nature of the supposed fraudulent
activities. Consequently, an accuracy nudge should lead individuals to consider the accuracy of
the allegations. The fourth hypothesis therefore focuses on the ability of corrective messages to
reduce the effect of unfounded claims of election fraud on confidence in elections and voting
intentions. Recent scholarly developments have found that fact-checks can indeed correct voters’
beliefs and support for democratic norms (Aird et al., 2018; Swire et al., 2017). In line with this
growing literature, I predict that presenting a corrective message (in the form of an accuracy-
nudge) after a claim of election fraud will significantly reduce the effect of claims of election fraud
on voting intentions and confidence in elections (Hypothesis 4).
Hypothesis 4: The effect of unfounded claims of election fraud on confidence in
elections and likelihood of voting will be lower among individuals presented with a
corrective message (fact-check).
5. Methodology
This survey experiment was approved by the relevant Institutional Review Board. The hypotheses,
study design, and material were preregistered on the Open Science Framework (available here:
https://doi.org/10.17605/OSF.IO/ZNHVK).
3
Overall, 3003 participants were recruited using the
online survey panel service Prolific (https://prolific.ac). Prolific is fast, reliable, and produces
samples that are more diverse and representative than alternative platforms (Palan & Schitter,
2018; Peer et al., 2017). Samples from online recruitment panels are also more demographically
representative than university laboratory samples (Peer et al., 2017). However, Prolific sampling
can introduce response bias should respondents go through the survey as fast as possible to obtain
3
All hypotheses, procedures, and material were preregistered before data collection. In addition, the data has been
made available on the preregistration site. The R syntax used to compute the models presented herein have also been
published online to facilitate replication and independent verification of the analyses.
8
their financial rewards (Peer et al., 2017). In order to account for these biases, this study included
‘attention’ questions. Seventeen participants who failed to answer this item correctly were
removed from the analyses.
5.1. Research Design
In order to understand the effects of exposure to claims of election fraud, this study relies
on a survey experiment. Three condition groups were elaborated in order to test the effects of both
claims of fraud and accuracy-nudge interventions on participants’ attitudes and behavioural
intentions. The first treatment was the fraud allegation group. Participants in this group were
exposed to a real unfounded claim of election fraud which circulated in a prominent UK
newspaper. The second treatment group included the same allegation of fraud but added an
accuracy-nudge intervention to test the fact-checking hypotheses. Lastly, the third group served as
the control group and only had participants answer the survey questionnaire. Table 2 presents the
two treatment conditions. The first question asked to respondents directly after the various
treatments was the perception of election fraud. This question, which is presented under Table 2,
serves as a manipulation check.
Table 1. Vignettes for the treatment groups
Treatment Group
Vignette
Election Fraud
Some say that there is a lot of fraud in U.K. elections.
See, for instance, this recent newspaper headline:
Election Fraud
+ Accuracy nudge
Some say that there is a lot of fraud in U.K. elections.
See, for instance, this recent newspaper headline:
However, several independent investigations have cast doubt on
the accuracy of these allegations.
Note. The control group will not be presented with a vignette, they will proceed directly to manipulation
check question: “In your opinion, how frequently does election fraud occur?” (1 = never occurs; 7 = occurs
a lot)
9
Once participants completed their respective conditions, they were directed to a survey
questionnaire which measured various attitudes and behavioural intentions. This survey included
measures for the dependent variables, several potential control variables, and sociodemographic
characteristics. The full questionnaire can be found in Appendix B and C.
4
To measure respondents’ confidence in the electoral process, this study relied on a
modified three-item scale with items from Berlinski et al. (2021) and Bordeleau (2021a). The items
are “all in all, how well do you think elections work in the UK?”; “to what extent do you trust UK
elections?”; and “how well do you think that the results of UK elections represent the people’s
will?” Responses were captured on eleven-point scales (0 = not [well] at all; 10 = very well/A lot).
These items measure various aspects of electoral trust which, when scaled together, offer a general
measurement of individuals’ confidence in the electoral process. For analytical purposes, the three
items will be recoded into a single confidence index ranging from 0 = no confidence to 10 = a lot
of confidence.
In order to measure individuals’ likelihood of voting at the next election, this study
included a self-reported turnout intention question. This item asked participants: “how likely are
you to vote in the next UK general election?” Participants answered on an eleven-point scale (0 =
very unlikely; 10 = very likely). While self-report items are susceptible to respondent bias, they
offer the most adequate and efficient way to measure behavioural intentions among individuals.
Lastly, sociodemographic questions were asked to gather information on the sample and
control for certain individual-level factors (i.e., gender, age, education level, etc.).
5
Political
interest was measured on an eleven-point scale (“How interested are you in politics? 0 = Not at
all; 10 = A great deal). Political ideology was also be measured using an eleven-point scale (In
politics people sometimes talk of left and right. Where would you place yourself on the following
scale? 0 = Left; 10 = Right). Appendix B provides a full list of control and sociodemographic items.
6. Results
In total, 3003 individuals took part in the study of which 2986 were kept for analyses (17
failed the attention check item). Participants were spread almost evenly among the condition
groups. Prolific pre-screened participants in order to ensure a representative sample with regards
to gender, age, and education. As such, the sample used in this study is representative of the UK
population for these sociodemographic characteristics according to the latest available census data
at the time of study. Further, balance tests presented in Table 6 (Appendix E) show that, when it
comes to sociodemographics and key political attitudes, all groups are balanced and do not
significantly differ from one another.
The first set of preregistered hypotheses posited that exposure to allegations of election
fraud would lead to lower levels of confidence in elections as well as lower individual likelihood
of voting. In addition, the fourth hypothesis predicted that fact-checking interventions in the form
of an accuracy-nudge would reduce the negative effect of allegations of fraud on both confidence
4
The sociodemographic questionnaire (Appendix B) was filled out by participants before their respective treatment
conditions. The attitude survey (Appendix C) was presented after the treatments. Participants in the control group
answered both questionnaires one after the other (no treatment).
5
For a control variable to be included in the model, it needs to meet the two following criteria: 1) be significantly
correlated with at least one of the predictors; and 2) have theoretical/literature support for inclusion in the model.
Variables that meet only one of the criteria will not be included. For instance, a variable that is correlated with the
predictors but has not previously been found—in the literature or theory—to be relevant will not be included.
10
in elections and likelihood of voting. Table 3 report the OLS regression models which test these
hypotheses.
As we can see from Model 2, being presented with an allegation of fraud did not
significantly decrease respondents’ confidence in the electoral process. Moreover, exposure to an
accuracy nudge also did not influence overall confidence in the process. Model 2 offers similar
results for likelihood of voting. Being presented with a fraud claim or an accuracy nudge did not
significantly influence participants’ desire to vote. In both models, the variance explained by the
treatments is basically 0, indicating that treatment effects are inexistant.
Table 2. Regression coefficients for main treatment models
Model 1
Likelihood of Voting
Model 2
Confidence in Elections
(Intercept)
8.343***
5.749***
(0.087)
(0.072)
Fact-Check
0.058
0.089
(0.123)
(0.102)
Fraud Allegation
0.132
0.091
(0.123)
(0.102)
N
2986
2986
R2
0.000
0.000
F
0.581
0.526
* p < 0.05, ** p < 0.01, *** p < 0.001
Note. Simple OLS regression coefficients. Control group is the baseline.
Standard errors in parentheses. DVs range from 0 to 10. Table generated
using the <modelsummary> package by Arel-Bundock (2022).
These results do not support our hypotheses or the proposed mediation model (hypotheses
1, 2, and 4). Our allegation of fraud and fact-checking treatments had no effect on both trust in
elections and likelihood of voting. In fact, the effects observed for the fraud allegation group—
which are non-significant and negligible—are in the opposite direction of what was predicted.
According to these coefficients, being in the fraud claim treatment group is associated with higher
confidence and likelihood of voting than in the control group.
Having not found any significant relationships between exposure to allegations of election
fraud and confidence in elections, hypothesis 3 was disregarded from the analyses. The mediation
model depended on a relationship between allegations of election fraud and confidence in
elections. Since none of the coefficients for confidence in elections are significant, it implies that
there is no potential for a causal mediation path towards likelihood of voting. In short, our
preregistered hypotheses are not supported by the survey experiment.
11
6.1. Manipulation Check
In order to make sense of the null results of the experiment, I examine the mean scores on the
manipulation check item by condition group. As a reminder, this manipulation check measured
perceptions of fraud. As such, there should be clear differences between the three groups on this
measure of fraud beliefs. Based on the results of previous studies and the theoretical framework
developed in previous chapters, participants exposed to allegations of fraud (group 1) should report
higher perceptions of fraud than those in the control group (group 0) and those in the fact-check
condition (group 2).
Table 3. Regression coefficients for manipulation checks
Model 1 (all groups)
Model 2 (treated groups)
(Intercept)
3.395***
3.218***
(0.074)
(0.071)
Fact-Check
-0.364***
-0.187
(0.105)
(0.101)
Fraud Allegation
-0.177
(0.105)
N
2988
1991
R2
0.004
0.002
F
6.070
3.455
* p < 0.05, ** p < 0.01, *** p < 0.001
Note. Simple OLS regression coefficients for perceptions of fraud. Control group is the
baseline for Model 1. Fraud allegation group is the baseline for Model 2. DV ranges from 0 to
1. Standard errors in parentheses. Table generated using the <modelsummary> package by
Arel-Bundock (2022).
An OLS regression model was produced with the categorical variable of condition group
as an independent variable and the perceptions of fraud item as the dependent variable. The results
are presented in Table 4 (see Model 1). The coefficient for the allegation of fraud treatment is not
significant, indicating that the treatment used did not affect participants’ perceptions of electoral
integrity. This is contrary to expectations set forth in the theoretical model and contradicts the
findings of Berlinski et al. (2021). With that said, the coefficient for the fact-check treatment
indicates that being presented with an accuracy nudge did result in lower perceptions of fraud
among respondents. However, the size of this effect is small.
It should be noted that Model 1 compares the two treatment groups to the control and not
to each other. In other words, the results highlight that exposure to an accuracy nudge leads to a
decrease in perceptions of fraud in comparison to the control group (not the fraud allegation
treatment). Since hypothesis 4 stipulated that exposure to a fact-check would reduce the negative
12
effects of allegations of fraud, it is necessary to compare between treated groups. Model 2 presents
the OLS regression coefficient for the fact-check with the allegation of fraud group as a baseline.
As we can see, the coefficient is not statistically significant, indicating that participants in the fact-
check group did not have perceptions of fraud that differed significantly from those in the
allegation of fraud treatment group. Overall, these results indicate that being exposed to allegations
of election fraud does not influence individuals’ perceptions of fraud. However, being presented
with a fact-check slightly reduces perceptions of fraud. This is true only when comparing to
individuals who do not receive any information (control group).
6.2. Supplementary Analyses
In order to better understand the relationship between claims of fraud, beliefs about election
fraud, confidence in elections, and likelihood of voting, this section provides supplementary
analyses which were not included in the preregistration. These analyses serve to examine
alternative explanations for our findings.
The large sample size used in this study allows us to conduct an analysis of cross-sectional
data from non-treated participants (control group). Using the manipulation check as the
independent variable, Table 5 presents the results of regression analyses for the relationship
between perceptions of fraud and participants’ self-reported likelihood of voting. The same models
with confidence in elections as the dependent variable are available in Appendix F. It should be
taken into consideration that these models stem from cross-sectional survey data and that
participants answered all questions (including IV and DV) in the same setting. As a result, causal
inference cannot be drawn directly from the results of these models. Nevertheless, they can guide
our interpretations of the experimental findings as well as provide direction for future research on
the topic.
As we can see, individuals’ subjective perceptions of fraud are significantly related to their
self-reported likelihood of voting. This is true even when controlling for several other variables.
Table 5 presents four models, each more conservative than the previous one. Model 1 includes
only our independent variable. In this model, a one-point increase on the perception of fraud
measure leads to a decrease of 0.353 on the likelihood of voting scale. Models 2 to 4 contain
different sets of control variables which have previously been shown to influence individuals’
likelihood of voting.
Our second model includes standard socioeconomic controls (gender, age, and education),
all previously found to influence turnout (Blais & Daoust, 2020; Smets & van Ham, 2013). In this
second model, a one-point increase on the fraud perception scale leads to a decrease of 0.281 on
the likelihood of voting. This is a slightly smaller coefficient than the one from Model 1. Moreover,
age and education (but not gender) are found to be significant predictors of the desire to vote. The
third model focuses on controlling for political attitudes (political interest and self-reported
ideology). This third model shows that political interest and to a lesser extent political ideology
influence the decision to vote. This is consistent with previous research on the motivation to vote
(Blais & Daoust, 2020), as well as individual-level predictors of turnout (Smets & van Ham, 2013).
The coefficient for fraud perceptions is still significant, although to a lesser extent than Models 1
and 2 (B = -0.225, p < .001). Lastly, the fourth model includes all control variables within a single
regression equation (illustrated in Figure 3). Even when controlling for socioeconomic variables
and political attitudes, perceptions of election fraud remains significantly related to participants’
likelihood of voting. In this full model, a one-point increase on the perception of fraud scale leads
to a decrease of 0.189 points on the likelihood of voting.
13
Table 4. Relationship between perceptions of fraud and likelihood of voting
Model 1
Model 2
Model 3
Model 4
(Intercept)
9.544***
6.540***
7.034***
5.522***
(0.143)
(0.487)
(0.289)
(0.489)
Fraud Perceptions
-0.353***
-0.281***
-0.225***
-0.189***
(0.034)
(0.035)
(0.032)
(0.033)
Age
0.029***
0.021***
(0.006)
(0.006)
Education
0.449***
0.212**
(0.085)
(0.080)
Gender
-0.005
-0.011
(0.018)
(0.017)
Ideology
-0.083*
-0.094*
(0.036)
(0.037)
Political Interest
0.432***
0.404***
(0.030)
(0.031)
N
996
995
994
993
R2
0.098
0.133
0.260
0.270
F
108.086
37.846
116.015
60.768
* p < 0.05, ** p < 0.01, *** p < 0.001
Note. Coefficients for OLS linear regressions with likelihood of voting as the dependent
variable. Both variables measured on eleven-point scales from 0 to 10 (low to high).
Standard error in parentheses. Models computed with cross-sectional data from non-treated
participants (control group). DV and fraud perceptions range from 0 to 10. Table generated
using the <modelsummary> package by Arel-Bundock (2022).
The full model accounts for 27% of the variance in likelihood of voting. Fraud perceptions
alone account for roughly 10% of the variance in the dependent variable. This would suggest that
beliefs in election fraud are considerably correlated with an individuals’ self-reported likelihood
of voting. These results are consistent with previous studies and support the aggregate-level
findings of Birch (2010) and to some extent the experimental findings of Berlinski et al. (2021).
The results for confidence in elections are similar, with stronger coefficients than
likelihood of voting and some different patterns for control variables (see Table 7 and Figure 4 in
14
Appendix F). This suggests that perceptions of fraud are more related to individuals’ confidence
in the process than their self-reported likelihood of voting. This fits with the expectations put forth
in the theoretical framework.
Figure 2. Regression plot for simple predictive model of likelihood of voting
Note. Regression line for Model 4 from Table 5. Perceptions of fraud measured on 11-
point scale from 0 = no fraud at all to 10 = a lot of fraud. Likelihood of voting measured
on 11-point scale from 0 = I will not vote to 10 = I will vote for sure. Simple OLS linear
regression equation line. Graph created using the <ggplot2> R package by Wickham
(2016).
7. Discussion
The results of the survey experiment do not support that of previous studies. With that said,
there are several potential explanations for why the results do not align with that of previous
research or with my hypotheses. In this chapter, I discuss the findings, explore alternative
explanations for the results, and examine the implications of my analyses.
7.1. Quality of Treatment
The first potential explanation for the failure of observing significant changes in
respondents’ perceptions of fraud is the weakness of the treatment. The theoretical model
suggested that allegations of fraud would influence individual levels of fraud perceptions which
would in turn lead to attitudinal and behavioural changes (i.e., confidence in election and the
decision to vote). However, as we have seen from the manipulation check analyses, the main
treatment (allegation of fraud) had a weak and statistically non-significant effect on participants’
15
beliefs in fraud. Similarly, the fact-check treatment did not have the desired effect when comparing
with the main treatment group. This suggests that the treatments presented to participants did not
have the intended impact. This section explores how our treatment differed from that of previous
survey experiments in order to identify why it did not function according to expectations.
One of the main differences between the treatment used in this study and that of previous
research on allegations of election fraud is the non-partisan and non-authoritative nature of the
treatment condition (Berlinski et al., 2021; Clayton et al., 2020). In their experimental study,
Berlinski et al. (2021) also rely on real allegations of fraud, however they rely on claims from
partisan elite sources rather than news media. For example, one of the treatment conditions in their
survey experiment included tweets from senior Republican politicians who claimed fraudulent
activities undermined the integrity of the 2016 US election. The presence of a partisan source for
the treatment is in stark contrast to the neutral and non-partisan treatment used herein. This contrast
may explain the inability of our treatment to influence beliefs in fraud. It is indeed possible that,
as suggested by Clayton et al. (2020), the development of beliefs around electoral integrity depends
on the partisan dispositions of citizens and the discourse of co-partisan elites.
The absence of a partisan component in this study could therefore explain why our
experiment did not yield significant results. This implies that allegations of fraud are consequential
only if they stem from partisan sources and are congruent with consumers’ ideological (or partisan)
preferences. Norms regarding what is “good” citizenship in times of democratic crises are
significantly shaped by patterns of partisanship (Goodman, 2022). Accordingly, it is logical to
expect beliefs regarding electoral fraud to be shaped by similar partisan patterns. In the case of the
US, for instance, these patterns of partisanship on the issue of election fraud are characterized by
the strong fraud rhetoric expressed by Republican elites and the counter-rhetoric on the Democrat
side. In sum, the influence of allegations of fraud on citizens’ perceptions appears to be
significantly shaped by partisan patterns of elite discourse (Berlinski et al., 2021; Clayton et al.,
2020), which would explain why the neutral treatment used in this experiment did not influence
such perceptions.
7.2. Pre-existing Beliefs and Partisanship
It is also possible that exposure to claims of fraud and fraud-related fact checks may act as
an activation mechanism for partisan beliefs. In this case, some individuals who do not hold strong
beliefs regarding election fraud were primed by the treatments to retrieve their partisanship-based
belief. As a result, rather than having the effect of changing people’s perceptions of electoral fraud,
the treatments simply led participants to retrieve pre-existing beliefs regarding elections. These
pre-existing beliefs may have been developed based on experiences at the polls, prior exposure to
elite rhetoric and fraud claims, or other sources of information. What is important is that the
treatments did not lead to the creation of new beliefs (or changes in perceptions) but rather led to
the retrieval of already existing ones.
When presented with a supplemental allegation of fraud (as was the case in this study),
individuals do not appear affected by the information regardless of that information’s congruence
with their own beliefs. That is because the allegation acted as a priming mechanism for respondents
to retrieve the partisan pattern for this specific issue. The literature provides support for this
alternative explanation. Results from our manipulation check also collaborate this narrative. Voters
in the control group had the highest levels of fraud perceptions when compared to the treated
groups. This implies that when exposed to nothing, participants did not have cues to recall their
partisan beliefs (whether positive or negative) regarding fraud. On the other hand, when exposed
16
to allegations of fraud or fact-checks, respondents exhibited lower levels of beliefs in fraud. Under
this alternative explanation, this would be due to the fact that the treatments acted like a heuristic,
leading participants to recall that they do not (or do) adhere to the partisan narratives put forth by
these unfounded allegations of fraud. The overall effect of the manipulation check was negative
because most participants held very little beliefs in fraud. So, for the large majority of respondents,
exposure to the treatments led to a more salient expression of their disbelief in election fraud.
These findings imply that voters have already had the opportunity to be exposed to claims
of fraud and build beliefs regarding such fraud. Building on previously elaborated alternatives, it
can be said that being presented with additional information regarding election fraud—regardless
of the nature of this information—makes salient previously developed beliefs instead of generating
new ones. This would explain why the experiment presented herein did not lead to significant
changes in citizens’ perceptions of electoral integrity and consequently did not affect their
confidence in elections and their self-reported likelihood of voting.
If the failure of this experiment to cause changes in participants’ beliefs in fraud does
indeed stem from the deep-rooted nature of pre-existing partisan beliefs, then it is appropriate to
modify our original research question to the following: do beliefs in election fraud influence the
likelihood of voting? Fortunately, we were able to test this new research question through cross-
sectional data from non-treated participants (control group). These supplementary analyses
provided support for the idea that perceptions of election fraud are related to individuals’ self-
reported likelihood of voting. However, these results are only suggestive and depend greatly on
the quality of the control variables included. If under different circumstances (i.e., partisan cues or
greater dosage) allegations of fraud were to lead to changes in citizens’ beliefs in fraud, then such
changes might also potentially affect their attitudes towards the electoral process as well as their
likelihood of voting.
The alternative explanations are not mutually exclusive. It is probable that the more citizens
are exposed to claims of election fraud, the more claims they need to be exposed to change their
beliefs. This would explain why a single claim of fraud would not influence beliefs today, but may,
for example, have been influential pre-2016. Berlinski et al. (2021) and Clayton et al. (2020) both
conducted their experiments in the immediate aftermath of the 2016 US election. At the time, elite
rhetoric regarding election fraud were in their early stages and partisan patterns were only
beginning to form. At the time the present experiment was conducted in 2022, allegations of fraud
had been circulating for several years and partisan patterns had been clearly established.
7.3. Limitations
This study has several shortcomings. First and foremost, the reliance on self-report scales
for likelihood of voting limits our ability to understand how perceptions of election fraud influence
real voting behaviour (that is, whether someone actually votes or not). Solving this limitation will
require researchers to use validated voting and abstention data.
Another important limitation of this research is the inability to measure a causal effect
between fraud beliefs and voting. The use of cross-sectional data in the supplementary analyses
limits our ability to make such causal inferences since participants answered questions in one
sitting and there is therefore no temporal differences between when the dependent and independent
variable were measured. This also raises another concern: endogeneity. Since the survey collects
data in a single short period of time, it is impossible to clearly identify the directionality of the
results. While it is not expected by theory, it is possible that higher likelihood of voting leads to
17
lower levels of fraud beliefs. With that being said, the theoretical framework developed in this
research argues that the directionality of the effect is in the hypothesized direction.
8. Conclusion
The results of this study provide a first look into the potential effects of perceptions of
election fraud on voter participation. While this study has several limitations, it provides
individual-level cross-sectional evidence for the hypothesis that fraud perceptions are related to a
demobilization of voters. This complements findings from Birch (2010) who relied on aggregate-
level comparative data to examine this same hypothesis.
These results bear important implications for democratic governance. First, voter
participation is the single most important method of participation in most consolidated
democracies. The potential for harmful rhetoric and false beliefs to disrupt civic engagement at the
polls should be troubling to all who study democracy or practice policymaking. Second, the results
of this study highlight the importance of independent and transparent election management. In
maintaining high standards of election administration, consolidated democracies can ensure that
the gap between actual levels of fraud and perceived levels of fraud remains just that: a gap.
In line with the concerns of neutrality of the treatment, future studies should consider the
implication of partisanship and elite discourse on voters’ attitudes towards the electoral system in
a non-US setting. Thus far, Berlinski et al. (2021) have successfully shown that partisan messaging
influences attitudes towards democracy and elections in a highly polarized two-party system. Are
the negative effects of elite election fraud rhetoric as consequential in less polarized multi-party
and systems? Is it possible that more fractionalized party systems act as safeguards for the spread
of negative attitudes towards electoral processes? Future research certainly ought to consider these
questions.
Overall, my findings suggest that unfounded allegations of election fraud do not influence
citizens’ beliefs regarding election fraud, nor do they affect their intention to vote in future
elections. With that being said, supplementary analyses are suggestive of the important role of
individuals’ perceptions of fraud on their likelihood of voting. Future research should therefore be
concerned with two distinct phenomena: 1) the consequences of false beliefs regarding election
fraud on a variety of political behaviours; and 2) the processes which lead to individuals’ holding
such beliefs. It is also in the interest of scholars to consider how to engage with false beliefs in
election fraud. This is true especially since traditional forms of fact-checking are not effective
against strongly rooted partisan beliefs.
The null results of the survey experiment do not suggest scholars should refrain from
studying allegations of fraud in the future. On the contrary, the alternative explanations explored
in the discussion provide ample support for the idea that more research needs to be done on this
topic. Under the right dosage and quality, allegations of fraud have been shown to influence key
attitudes before (Berlinski et al., 2021; Justwan & Williamson, 2022). It is therefore crucial to
continue building the research agenda on unfounded allegations of election fraud, since such
rhetoric poses great threat to the stability and legitimacy of democratic institutions.
18
References
Aird, M. J., Ecker, U. K. H., Swire, B., Berinsky, A. J., & Lewandowsky, S. (2018). Does truth
matter to voters? The effects of correcting political misinformation in an Australian
sample. Royal Society Open Science, 5(12), 180593.
https://doi.org/doi:10.1098/rsos.180593
Albertson, B., & Guiler, K. (2020). Conspiracy theories, election rigging, and support for
democratic norms. Research & Politics, 7(3), 2053168020959859.
Alvarez, R. M., Hall, T. E., & Hyde, S. D. (2009). Election fraud: detecting and deterring
electoral manipulation. Brookings Institution Press.
Andı, S., & Akesson, J. (2020). Nudging away false news: Evidence from a social norms
experiment. Digital journalism, 9(1), 106-125.
Arel-Bundock, V. (2022). modelsummary: Summary Tables and Plots for Statistical Models and
Data: Beautiful, Customizable, and Publication-Ready. In R Package (Version
0.10.0.9000) https://vincentarelbundock.github.io/modelsummary/
Berlinski, N., Doyle, M., Guess, A. M., Levy, G., Lyons, B., Montgomery, J. M., Nyhan, B., &
Reifler, J. (2021). The Effects of Unsubstantiated Claims of Voter Fraud on Confidence
in Elections. Journal of Experimental Political Science, 1-16.
https://doi.org/10.1017/XPS.2021.18
Birch, S. (2010). Perceptions of Electoral Fairness and Voter Turnout. Comparative Political
Studies, 43(12), 1601-1622. https://doi.org/10.1177/0010414010374021
Blais, A. (2000). To vote or not to vote?: The merits and limits of rational choice theory.
University of Pittsburgh Pre.
Blais, A., & Daoust, J.-F. (2020). The motivation to vote: Explaining electoral participation.
UBC Press.
Bordeleau, J.-N. (2021a). Does Election Interference Rhetoric Influence Voters’ Behaviours and
Attitudes? Empirical Evidence from an Experimental Survey Study
Bordeleau, J.-N. (2021b, July 2021). The Effects of Social Dominance Orientation and Right-
Wing Authoritarianism on Perceptions of Election Fraud and Support for Aggressive
Fraud Control Measures Electoral Integrity Project Conference,
Clayton, K., Davis, N. T., Nyhan, B., Porter, E., Ryan, T. J., & Wood, T. J. (2020). Elite rhetoric
can undermine democratic norms. Proceedings of the national academy of sciences,
118(23). https://doi.org/10.1073/pnas.2024125118
Coffé, H. (2017). Citizens’ media use and the accuracy of their perceptions of electoral integrity.
International Political Science Review, 38(3), 281-297.
19
Davis, M. L., & Ferrantino, M. (1996). Towards a positive theory of political rhetoric: Why do
politicians lie? Public Choice, 88(1), 1-13.
Downs, A. (1957). An Economic Theory of Political Action in a Democracy. Journal of Political
Economy, 65(2), 135-150. http://www.jstor.org/stable/1827369
Fahrenthold, D. A., Viebeck, E., Brown, E., & Helderman, R. S. (2020, 10 November 2020).
Here are the GOP and Trump campaign’s allegations of election irregularities. So far,
none has been proved. The Washington Post.
https://www.washingtonpost.com/politics/trump-election-irregularities-
claims/2020/11/08/8f704e6c-2141-11eb-ba21-f2f001f0554b_story.html
Fogarty, B. J., Curtis, J., Gouzien, P. F., Kimball, D. C., & Vorst, E. C. (2015). News attention to
voter fraud in the 2008 and 2012 US elections. Research & Politics, 2(2),
2053168015587156.
Fortin-Rittberger, J., Harfst, P., & Dingler, S. C. (2017). The costs of electoral fraud: establishing
the link between electoral integrity, winning an election, and satisfaction with democracy.
Journal of elections, public opinion and parties, 27(3), 350-368.
Franklin, M. N. (2004). Voter turnout and the dynamics of electoral competition in established
democracies since 1945. Cambridge University Press.
Goodman, S. W. (2022). Citizenship in Hard Times: How Ordinary Citizens Respond to
Democratic Threat. Cambridge University Place.
Justwan, F., & Williamson, R. D. (2022). Trump and Trust: Examining the Relationship between
Claims of Fraud and Citizen Attitudes. PS: Political Science & Politics, 1-8.
Karp, J. A., & Banducci, S. A. (2008). Political efficacy and participation in twenty-seven
democracies: How electoral systems shape political behaviour. British Journal of
Political Science, 38(2), 311-334.
Lehoucq, F. (2003). ELECTORAL FRAUD: Causes, Types, and Consequences. Annual Review
of Political Science, 6(1), 233-256.
https://doi.org/10.1146/annurev.polisci.6.121901.085655
Levin, D. H. (2020). Meddling in the Ballot Box: The Causes and Effects of Partisan Electoral
Interventions. Oxford University Press, USA.
Levy, M. (2020). Winning cures everything? Beliefs about voter fraud, voter confidence, and the
2016 election. Electoral Studies, 102156.
López-Pintor, R. (2010). Assessing Electoral Fraud in New Democracies: A Basic Conceptual
Framework. IFES White Paper.
20
Mozaffar, S., & Schedler, A. (2002). The Comparative Study of Electoral Governance—
Introduction. International Political Science Review, 23(1), 5-27.
https://doi.org/10.1177/0192512102023001001
Norris, P. (2004). Electoral engineering: Voting rules and political behavior. Cambridge
university press.
Norris, P. (2013). The new research agenda studying electoral integrity. Electoral Studies, 32(4),
563-575.
Norris, P. (2014). Why electoral integrity matters. Cambridge University Press.
Norris, P., Garnett, H. A., & Grömping, M. (2020). The paranoid style of American elections:
explaining perceptions of electoral integrity in an age of populism. Journal of elections,
public opinion and parties, 30(1), 105-125.
Palan, S., & Schitter, C. (2018). Prolific.ac—A subject pool for online experiments. Journal of
Behavioral and Experimental Finance, 17, 22-27.
https://doi.org/https://doi.org/10.1016/j.jbef.2017.12.004
Peer, E., Brandimarte, L., Samat, S., & Acquisti, A. (2017). Beyond the Turk: Alternative
platforms for crowdsourcing behavioral research. Journal of Experimental Social
Psychology, 70, 153-163. https://doi.org/https://doi.org/10.1016/j.jesp.2017.01.006
Pennycook, G., Epstein, Z., Mosleh, M., Arechar, A. A., Eckles, D., & Rand, D. G. (2021).
Shifting attention to accuracy can reduce misinformation online. Nature, 592(7855), 590-
595.
Pennycook, G., McPhetres, J., Zhang, Y., Lu, J. G., & Rand, D. G. (2020). Fighting COVID-19
misinformation on social media: Experimental evidence for a scalable accuracy-nudge
intervention. Psychological science, 31(7), 770-780.
Pennycook, G., & Rand, D. G. (2019). Lazy, not biased: Susceptibility to partisan fake news is
better explained by lack of reasoning than by motivated reasoning. Cognition, 188, 39-50.
https://doi.org/https://doi.org/10.1016/j.cognition.2018.06.011
Pennycook, G., & Rand, D. G. (2021a). Examining false beliefs about voter fraud in the wake of
the 2020 Presidential Election. The Harvard Kennedy School Misinformation Review.
Pennycook, G., & Rand, D. G. (2021b). The psychology of fake news. Trends in cognitive
sciences, 25(5), 388-402.
Przeworski, A. (1988). Democracy as a contingent outcome of conflicts. In Constitutionalism
and democracy. Cambridge University Press.
Rev. (2020). Donald Trump Speech on Election Fraud Claims Transcript December 2. Rev.
Retrieved 25 June from https://www.rev.com/blog/transcripts/donald-trump-speech-on-
election-fraud-claims-transcript-december-2
21
Riker, W. H., & Ordeshook, P. C. (1968). A Theory of the Calculus of Voting. American
political science review, 62(1), 25-42.
Smets, K., & van Ham, C. (2013). The embarrassment of riches? A meta-analysis of individual-
level research on voter turnout. Electoral Studies, 32(2), 344-359.
https://doi.org/https://doi.org/10.1016/j.electstud.2012.12.006
Smith, C. C. O. (2021). Voter fraud measures announced in the Queen’s speech. Government of
the United Kingdom. https://www.gov.uk/government/news/voter-fraud-measures-
announced-in-the-queens-speech
Swire, B., Berinsky, A. J., Lewandowsky, S., & Ecker, U. K. H. (2017). Processing political
misinformation: comprehending the Trump phenomenon. Royal Society Open Science,
4(3), 160802. https://doi.org/doi:10.1098/rsos.160802
Taylor, C. (2007). From the Whole Citizen Body? The Sociology of Election and Lot in the
Athenian Democracy. Hesperia: The Journal of the American School of Classical Studies
at Athens, 76(2), 323-345. http://www.jstor.org/stable/25068022
Udani, A., Kimball, D. C., & Fogarty, B. (2018). How local media coverage of voter fraud
influences partisan perceptions in the United States. State Politics & Policy Quarterly,
18(2), 193-210.
Walter, N., Cohen, J., Holbert, R. L., & Morag, Y. (2020). Fact-checking: A meta-analysis of
what works and for whom. Political Communication, 37(3), 350-375.
Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag.
https://ggplot2.tidyverse.org
22
Supplementary Material
Appendix A: Information Letter and Consent Form
You are invited to participate in a research project. Before accepting to take part, please read the
following information which outlines the conditions related to your participation. Once you are
done reading the document, press “yes” to accept the terms and carry on with the research and
“no” if you do not wish to proceed. Clicking “yes” will be considered as your informed consent
to participate in the research.
This project seeks to better understand citizens’ beliefs and attitudes about UK elections. Your
participation will be completely anonymous, and the researchers will not be able to identify
participants. The questionnaire should take roughly 5 minutes to complete.
There are no foreseeable risks associated with participating in this research. There are no
foreseeable advantages to taking part in this research. There is, however, the gratification of
helping advance research.
Participants are not to provide their names or any information that could help identify them. The
only personal information to be collected is demographics such as age and gender. It will
therefore be impossible for anyone to identify participants.
Although the researchers will take all the necessary measures to protect the data, it is impossible
to guarantee an absolute confidentiality because they will be collected via a software that has
servers in the United Kingdom. Furthermore, the data will be made available publicly via the
Open Science Framework (OSF). Participants will not be identifiable.
You will be rewarded £0.50 for participating in this research. Your participation in this research
is entirely voluntary and you are allowed to end your participation at any time. If you decide to
end your participation, your data may not be used and will be destroyed.
Pressing the “Yes” button (below) will be interpreted as providing consent for participation in
this research. It will also be interpreted as indicating that you: realize that you are not required to
participate if you so choose, are free to withdraw from the study at any point during the
completion of the survey, and freely consent to participate in this research.
Yes No
For any questions regarding this study, please communicate with Jean-Nicolas Bordeleau at the following email address: jean-
nicolas.bordeleau@umontreal.ca
If you have any questions regarding your rights or the responsibilities of researchers regarding your participation in this project,
you can contact the Arts and Humanities Research Ethics Committee by email at cerah@umontreal.ca or visit the website:
http://recherche.umontreal.ca/participants. Any complaints regarding your participation in this research project can be addressed
to the University of Montreal Ombudsperson by calling 1-514-343-2100 or by email at ombudsman@umontreal.ca.
This project was approved by the Comité d’éthique de la recherche en arts et humanités at the University of Montreal. Project no
CERAH-2021-137-D
23
Appendix B: Sociodemographic Questions
Filter question
6
: Are you eligible to vote in UK elections? ☐ Yes ☐ No
Please answer the following sociodemographic questions.
1. Please, can you tell me your age: ______________
2. Please indicate your gender: ☐ Male ☐ Female ☐ Other ☐ Prefer not to say
3. What is the highest educational qualification you have?
o No qualification
o Below GCSE
o A-level
o Undergraduate
o Postgrad
4. Did you have the right to vote in the UK General Election of 12 December 2019?
7
☐ Yes, I had the right to vote ☐ No, I did not have the right to vote
5. We have found that a lot of people didn’t manage to vote in that election. How about you,
did you manage to vote in 2019 UK General Election?
☐ Yes, I voted ☐ No, I did not vote ☐ I prefer not to say
6. In politics people sometimes talk of left and right. Where would you place yourself on the
following scale? (0 = Left; 10 = Right)
7. If there were a UK general election held tomorrow, which party would you vote for?
o Conservative Party
o Labour Party
o Scottish National Party
o Liberal Democrats
o Democratic Unionist Party
o Sinn Féin
o Plaid Cymru
o Green Party
o Alliance Party
o UKIP
o Social Democratic & Labour Party
o Reform UK
o Other Party
o None
o I have no idea
o Prefer not to say
8. How interested are you in politics? (0 = Not at all; 10 = A great deal)
6
Participants who answer “No” will not take part in the rest of the questionnaire (participation terminated). This
question will be answered before the Information & Consent Form.
7
If participants answer “No” to question 4, they will proceed directly to question 6 (skip question 5).
24
Appendix C: Questionnaire
Please answer the following questions about your political beliefs.
1. How likely are you to vote in the next UK general election? (0 = Very unlikely; 10 = Very
likely)
2. All in all, how well do you think elections work in the UK? (0 = Not well at all; 10 =
Very well)
3. To what extent do you trust UK elections? (0 = Not at all; 10 = A lot)
4. How well do you think that the results of UK elections represent the people’s will? (0 =
Not at all; 10 = A lot)
5. It is important for us to make sure you are correctly reading the questions, please click on
‘6’ on the following scale. (0 = Not at all; 10 = A lot)
Please tell us if you think the following statements are true or false. If you don’t know, just say
so and we will skip to the next one.
1. Polling stations close at 10:00pm on election day. (Answer: True)
2. Only taxpayers are allowed to vote in a general election. (Answer: False)
3. The UK uses proportional representation for general elections. (Answer: False)
4. There are 650 elected members in the House of Commons. (Answer: True)
25
Appendix D: Codebook
Variable
Details
ResponseID
Unique respondent identification number
Group
Categorical variable which identifies which experimental group the participant is
in:
0. Control Group
1. Fraud Allegation Treatment
2. Fact-Check Treatment
FraudAllegation
Dummy variable for the Fraud Allegation Treatment Group
0. Control Group or Fact-check Treatment
1. Fraud Allegation Treatment
FactCheck
Dummy variable for the Fraud Allegation Treatment Group
0. Control Group or Fraud Allegation Treatment
1. Fact-Check Treatment
RightVote2019
Did you have the right to vote in the UK General Election of 12 December
2019?
0. No
1. Yes
Vote2019
If answer to RightVote2019 is ‘yes’:
We have found that a lot of people didn’t manage to vote in that election. How
about you, did you manage to vote in the 2019 UK General Election?
0. No
1. Yes
9. I prefer not to say
Ideology
In politics people sometimes talk of left and right. Where would you place
yourself on the following scale? (0 = Left; 10 = Right)
Party
If a general election were held today, which political party would you vote for?
1. Conservative Party
2. Labour Party
3. Scottish National Party
4. Liberal Democrats
5. Democratic Unionist Party
6. Sinn Féin
7. Plaid Cymru
8. Green Party
9. Alliance Party
10. United Kingdom Independence Party (UKIP)
11. Social Democratic & Labour Party
12. Reform UK
13. Other Party
14. None
98. I have no idea
99. Prefer not to say
Polinterest
How interested are you in politics? (0 = Not at all; 10 = A great deal)
Fraudmeasure
In your opinion, how much election fraud is there in UK elections? (0 = None at
all; 10 = A great deal)
VoteLikely
How likely are you to vote in the next UK general election? (0 = Very unlikely;
10 = Very likely)
Confidence1
All in all, how well do you think elections work in the UK? (0 = Not well at all;
10 = Very well)
26
Confidence2
To what extent do you trust UK elections? (0 = Not at all; 10 = A lot)
Confidence3
How well do you think that the results of UK elections represent the people’s
will? (0 = Not well at all; 10 = Very well)
ConfidenceIndex
Index made up of the mean answer to the three confidence items (Confidence1;
Confidence2; and Confidence3)
AttentionCheck
It is important for us to make sure you are correctly reading the questions, please
click on the ‘6’ on the following scale. (1-10 scale)
PolKnow1
Polling stations close at 10:00pm on election day. (Correct answer: True)
0. False
1. True
PolKnow2
Only taxpayers are allowed to vote in a general election. (Correct answer: False)
0. False
1. True
PolKnow3
The UK uses proportional representation for general elections. (Correct answer:
False)
0. False
1. True
PolKnow4
There are 650 elected members in the House of Commons. (Correct answer:
True)
0. False
1. True
PolKnowIndex
Index score from 1 to 4 indicating the number of correct answers to the four
Political Knowledge questions (PolKnow1; PolKnow2; PolKnow3; and
PolKnow4).
Age
Self-reported age.
[18-99]
Gender
Please indicate your gender:
0. Female
1. Male
2. Other
9. Prefer not to say
Education
What is the highest educational qualification you have?
0. No qualification
1. Below GCSE
2. GCSE
3. A-Levels
4. Undergraduate
5. Postgraduate
Dataset:
• Data was collected using a Prolific survey panel between 6-8 May 2022 (N = 3003).
27
Appendix E: Balance Tests
Table 5. OLS regression models for balance tests
Control vs
Fraud Treatment
Control vs
Fact-Check
Fraud Treatment
vs Fact-Check
(Intercept)
0.567***
1.134***
1.496***
(0.062)
(0.127)
(0.062)
Age
0.000
0.000
0.000
(0.001)
(0.002)
(0.001)
Gender
-0.006
0.033
0.022
(0.023)
(0.046)
(0.023)
Education
-0.016
-0.026
0.004
(0.012)
(0.023)
(0.011)
Ideology
-0.006
-0.020
-0.004
(0.006)
(0.011)
(0.006)
Political Interest
0.006
0.005
-0.003
(0.005)
(0.009)
(0.005)
N
1965
1967
1968
R2
0.002
0.002
0.001
F
0.824
0.970
0.290
* p < 0.05, ** p < 0.01, *** p < 0.001
28
Appendix F: Supplementary Analyses
Table 6. Regression coefficients for confidence in elections
Model 1
Model 2
Model 3
Model 4
(Intercept)
7.237***
5.797***
5.329***
4.452***
(0.106)
(0.360)
(0.222)
(0.375)
Fraud Perceptions
-0.439***
-0.402***
-0.423***
-0.399***
(0.025)
(0.026)
(0.025)
(0.025)
Age
0.031***
0.021***
(0.005)
(0.005)
Education
0.016
0.041
(0.063)
(0.062)
Gender
0.014
0.016
(0.014)
(0.013)
Ideology
0.267***
0.243***
(0.028)
(0.029)
Political Interest
0.126***
0.109***
(0.023)
(0.024)
N
996
995
994
993
R2
0.235
0.264
0.318
0.330
F
304.542
88.859
154.111
80.971
* p < 0.05, ** p < 0.01, *** p < 0.001
Note. Coefficients for simple linear regressions with confidence in elections as the
independent variable. Standard error in parentheses. Models computed with cross-
sectional data from non-treated participants (control group). Table generated using
the <modelsummary> package by Arel-Bundock (2022).
29
Figure 3. Regression plot for model of confidence in elections
Note. Regression line for Model 4 from Table 7. Perceptions of fraud measured on 11-point scale from 0
= no fraud at all to 10 = a lot of fraud. Confidence in elections measured on 11-point scale from 0 = I
will not vote to 10 = I will vote for sure. Simple OLS linear regression equation line. Graph created using
the <ggplot2> R package by Wickham (2016).
