https://orcid.org/0000-0003-2401-2914
Abstract
In this paper, we investigate how trust in traditional and social media correlate with misperceptions of electoral integrity. Relying on insights from political communication research on exposure to misinformation and selective exposure mechanisms, as well as insights on the different roles of traditional and social media in different regime types, we argue that misperceptions of election integrity are likely driven in large part by the interplay between the trust people have in different media sources and the context (i.e., the level of press freedom) in which the elections take place. Using data from a survey conducted in 25 countries across the world, we find that trust in information from traditional media decreases misperceptions, while trust in information from social media increases misperceptions. However, both these effects are smaller when press freedom is restricted. In countries with low levels of press freedom, trust in social media is even associated with lower levels of misperceptions.
Election integrity varies widely around the world, from elections marred by blatant manipulation and fraud to elections that are largely free and fair (Garnett et al. 2023). However, citizens’ perceptions of election integrity do not always match objective measures of election quality, as citizens may still distrust high-quality elections or, conversely, trust elections of low quality. In this paper, we seek to explain what drives such misperceptions of election integrity.
Accurate citizen perceptions of election integrity are essential for democracy. An informed citizenry would ideally notice if elections are rigged, subsequently mobilizing and protesting to hold their government accountable and demand clean elections. Conversely, citizens would ideally also notice when elections are free and fair. Fair and democratic elections legitimize the elected government, foster political trust and satisfaction with democracy, and generate acceptance of election outcomes by opposition parties and citizens (Moehler 2009; Moehler and Lindberg 2009; Norris, Frank, and Martinez i Coma 2014; Norris 2014, 2019; Mauk 2020).
Misperceptions of election integrity may lead to either unduly positive or unduly negative evaluations of election integrity by citizens. Rigged elections perceived as clean by citizens may help legitimize and stabilize electoral authoritarian rule (Donno 2013; Schedler 2013; Simpser 2013). Conversely, if elections are free and fair but citizens nevertheless perceive them to be rigged, legitimate elections may end up triggering unwarranted civil conflict and destabilizing democracy (cf. recent examples in the United States and Brazil).
The existing body of research on perceptions of electoral integrity suggests that traditional media plays an important role in the accuracy of perceptions of electoral integrity (Coffé 2017). Furthermore, previous empirical work underlines that the influence of the media is an important factor in the perception of electoral quality, especially given the recent decline in the quality of election media coverage (e.g., increased misinformation, digital propaganda, and cyberattacks) and the “erosion of public confidence” in the media (Norris and Grömping 2019, p. 10). While some research has been conducted, there is still an incomplete understanding of the relationship between trust in traditional and social media and misperceptions of electoral integrity. This is especially true across different countries and in the current high-choice media environment, in which there has been a sharp rise in social media usage. Indeed, Coffé’s study (2017), while seminal, only included general internet use and found no effect of internet use on electoral misperceptions. It remains to be seen whether such findings also hold in the current, more social media-focused environment. In fact, what follows will demonstrate they do not.
In this paper, we analyze the sources of misperceptions of electoral integrity, drawing on insights from political communication research on exposure to misinformation and selective exposure mechanisms, as well as those on the different roles of traditional and social media in different regime types. We argue that misperceptions of election integrity are likely to be largely driven by the interplay between (1) the trust people have in different media sources available to them; and (2) the context (i.e., the level of press freedom). The extent to which citizens trust media sources will likely determine how the information obtained through these sources affects their (mis)perceptions of election integrity. We also consider a wide variety of information sources, including social networks such as friends and family, politicians, and social and traditional media. In general, we expect that some of these sources, most notably social media, are more likely to spread misinformation about elections, and therefore are more likely to increase misperceptions of election integrity. Previous work demonstrated that health misinformation is more prevalent on social media (Suarez-Lledo and Alvarez-Galvez 2021). Moreover, during the 2016 election in the United States, 5 percent of social media posts on Twitter (tweets) containing political links were from fake news sources. It should be noted, however, that this was highly concentrated, with very few people being exposed to the tweets that contained misinformation (Grinberg et al. 2019). Conversely, traditional media should decrease misperceptions of election integrity. However, we expect these differences to be (partly) context dependent In contexts where media freedom is low and the traditional media is likely under significant control by the incumbent governments, social media frequently serves as an alternative source of correct information. Therefore, in low-media-freedom contexts, we expect trust in traditional media may lower misperceptions to a lesser degree. Conversely, in contexts with less press freedom, trust in social media may yield fewer misperceptions. For instance, Reuter and Szakonyi (2015) found that using Twitter and Facebook increased people’s (correct) perceptions of election fraud during the 2011 Russian parliamentary elections, as these were the platforms that were used to spread information about electoral fraud by the opposition.
We tested our hypotheses using novel survey data collected in February 2023 from 25 countries across the world with different levels of electoral integrity, which we combined with EIP’s Perceptions of Electoral Integrity expert data (Garnett et al. 2022). Results from multilevel analyses confirmed our initial expectations: overall, trust in traditional media reduced misperceptions while trust in social media increased misperceptions. These effects depend on the level of media freedom: in countries with low press freedom, the traditional media effect was significantly smaller, and for social media, the effect is even reversed.
The paper is outlined as follows. The next section provides a brief review of existing research on citizens’ perceptions of election integrity, outlining what we know and don’t know yet about (mis)perceptions of election integrity. We subsequently outline our theoretical expectations for how trust in different media sources affects misperceptions of election integrity in contexts of varying media freedom and present the hypotheses to be tested in this paper. The third section describes the data and methods used, and the fourth section presents our results. We conclude with a reflection on the implications of our findings for a better understanding of misperceptions about election integrity.
Explaining (Mis)Perceptions of Electoral Integrity
What causes misperceptions of electoral integrity? The burgeoning research literature on citizens’ perceptions of electoral integrity indicates that for most citizens in most elections, perceptions of electoral integrity are accurate (Norris 2013; Mochtak et al. 2021). However, misperceptions do occur, and the goal of this paper is to acquire a better understanding of why, when, and for whom misunderstandings about election integrity arise. Existing research on citizens’ perceptions of election integrity has focused on two broad explanatory approaches to understanding citizens’ perceptions: citizens’ access to information about the electoral process and citizens’ partisan perceptions of the electoral process.
In describing how people obtain information about the quality of elections, theories from psychology, communication, and political science point to several sources of information: direct personal experience, media (traditional and social), and social networks. Direct experience is an important source of information for citizens who follow the election campaign and vote, as they are able to directly observe any irregularities, such as long waiting lines at polling stations, ballot shortages, or, more seriously, unbalanced media coverage, voter intimidation, and ballot-box stuffing. Indeed, direct experience with problems in election administration has been shown to affect citizens’ perceptions of election integrity in contexts as varied as the United States, South Korea, Southeast Europe, and sub-Saharan Africa (Kerr 2014; Bowler et al. 2015; Cho and Kim 2016; Mochtak et al. 2021). In addition to direct experiences, citizens can also become informed about election integrity through the experiences of others in their social networks. Therefore, when a citizen’s family, friends, or neighbors experience irregularities at the polling booths, those experiences may affect the citizen’s perceptions of election integrity even though they did not experience the problems themselves. Information from political elites can also affect citizens’ perceptions of election integrity. Clayton et al. (2021) found that citizens who were exposed to Donald Trump’s election-rigging claims demonstrated significantly eroded trust and confidence in elections. More generally, the fact that elite accusations of electoral fraud are often successful in mobilizing electoral protests around the world suggests that cues by politicians about election integrity are important sources of information for citizens. Finally, the media are also essential in informing citizens about irregularities in the electoral process. Indeed, the accuracy of perceptions of election integrity is higher among citizens who are more frequent users of traditional media (Coffé 2017). Clearly, citizens obtain information about the quality of elections from a variety of different sources. However, the impact of these sources on (mis)perceptions of election integrity is likely to differ depending on the degree of accuracy of the sources and how they enable the spread of election misinformation (Norris et al. 2020). In an environment that is highly dependent on media and digital information, traditional and social media in particular play key roles in shaping people’s perceptions. We therefore focus on the trust people have in the information stemming from those two sources.
Trust in Traditional and Social Media and Perceptions of Electoral Integrity
Important sources for electoral information include traditional media (e.g., television, national and local newspapers, and radio) and online social media (cf. Coffé 2017). Elections are often exhaustively covered by the media. In addition, politicians and political parties frequently use digital media, such as organic and paid political content, to directly reach the electorate during election periods (Kruikemeier 2014; Stier et al. 2018). In general, in countries with higher levels of press freedom, scholars are in general optimistic about the role of the media in informing citizens about elections. When more citizens read or watch political information (e.g., news), particularly offline, citizens become informed, knowledgeable (De Vreese and Boomgaarden 2006), and aware of the most important current social and political issues (Beckers et al. 2021).
Traditional media, defined as mass media outlets where journalists function as gatekeepers, is often found to have a positive impact on knowledge about politics and current affairs (Beckers et al. 2021).1 For instance, Soroka and colleagues (2013) found that exposure to public news broadcasts leads to knowledge about hard news facts. Wei and Lo (2008) found that increased exposure to traditional media (i.e., television, newspapers, and online news) improved attention, elaboration, and knowledge about elections. Finally, Norris et al. (2020) found that in the United States, exposure to news and journalistic stories about political affairs—as opposed to simply “tuning in” (p. 118)—had a positive impact on perceptions of electoral integrity. The most important argument for this relationship is that in democracies, the traditional media has high journalistic standards. Journalists consider it essential to be autonomous, accurate, and reliable in their reporting (e.g., through verification and fact-checking), compared to social media, where no journalistic gatekeeping takes place (for an overview, see Gil de Zúñiga and Hinsley 2013). When people have more trust in traditional media, they are also more likely to consume it (Kalogeropoulos et al. 2019).
Some people have a less positive view of traditional media, arguing that television, for instance, is mainly used for entertainment purposes and is not likely to lead to an informed electorate (e.g., media malaise theories; for an overview, see Newton 1999). Others have argued that a more partisan and polarized media system, which makes politics sensational, hostile, and less trustworthy, increases division and exposure to misinformation (Tucker et al. 2018), which could lead to more misperceptions about elections and election integrity. However, empirical evidence for these media malaise theories is contested at best (Van Aelst 2017). Moreover, when examining the available empirical evidence, Coffé (2017) found that “the frequency of the use of traditional media relates significantly more positively to citizens’ capacity to make accurate judgments about the electoral process in societies with open media environments compared with societies with restricted media environments” (p. 292). Based on this finding, as well as previous work in political and communication science on the positive association between traditional media usage and political knowledge, we expect the following: people who put more trust in information from traditional media are more likely to be exposed to more accurate and reliable information about elections. As a result, we expect them to have fewer misperceptions about electoral integrity. It is important to note that we cannot completely rule out the possibility of effects running in reverse causal order, as could also be the case for social media. However, previous research has examined whether media trust can both be an antecedent variable steering news selection and play a conditional role in the relationship between news usage and perceptions of societal problems (Shehata and Strömbäck 2022). Using longitudinal panel data, they demonstrated that citizens’ use of public service and alternative news affected their perceptions of societal problems. They also found that media trust influences news selection and (partly) conditions media effects. In addition, a survey experiment found that exposure to fact-checks about Trump’s false claims about the integrity of the US electoral system increased participants’ confidence in the integrity of the electoral system (Bailard et al. 2022). While these results do not offer any formal causal interference testing, they support our expectations regarding the correlation between trust in traditional media and lower levels of misperception.
H1: The more trust people have in traditional media, the fewer misperceptions they have about electoral integrity.
For social media, defined as “online platforms that allow a user to send, share, and consume information” (Jacobs and Spierings 2016, p. 3), we expect to see the opposite effect. It is often asserted that social media contains more inaccurate information because there is no journalistic mechanism that verifies or checks the information. Also, direct communication via social media fuels this spread of (mis)information. Populist politicians and parties regularly use the label “fake news” (Egelhofer and Lecheler 2019) to undermine journalistic evidence and objective knowledge (McIntyre 2018; Norris and Grömping 2019). Such politicians also use social media to “name and shame” journalists and question their ethical standards (Jacobs et al. 2020, p. 615). Previous work also shows that, for example, in the United States, one out of four Americans visited a fake news website during the 2016 presidential election (Guess et al. 2018). Grinberg et al. (2019), which focused on Twitter data for the 2016 election, found that fake news accounted for approximately 6 percent of all news consumption. However, both studies found that this was heavily concentrated among more conservative voters.
We expect that when individuals trust social media and thus consume more social media information, they run a higher risk of being exposed to incomplete, deceptive, and inaccurate information about the electoral process, which could lead to more misperceptions about election integrity. Hence, when social media propagates unfiltered, repeated exposure to distortions of reality, this leads to “deep-seated misinformed beliefs” that may have severe and detrimental consequences for democracy (Norris and Grömping 2019, p. 10).
This mechanism might be reinforced online because citizens who are already doubtful about, for instance, an election campaign are more likely to be exposed to misinformation about the election. Reinforcing existing beliefs and biases and excluding information that is not in line with what someone already believes can create an environment where (mis)perceptions about the election are formed and perpetuated, or even strengthened through exposure to like-minded individuals. While empirical assessments regarding online echo chambers (e.g., like-minded people) and filter bubbles (e.g., algorithmic decision-making) are more complicated than can be explained here (Zuiderveen Borgesius et al. 2016), the psychological mechanisms of bias reinforcement should not be neglected in examining how misperceptions form and grow.
While exposure to attitudinal confirming information might reinforce misperceptions, the opposite can also be true, even in social media environments. One of the longstanding debates in the field of political communication is whether and how new technologies affect democracy. An important segment of the literature argues that new technologies offer opportunities for marginalized voices typically excluded from traditional media (Zhuravskaya et al. 2020). Using new technologies, such as social media, these marginalized people can bypass traditional media to forge direct connections with their audience. Social media also increases access to new, alternative, and more diverse sources, for example, those that challenge the status-quo messaging of state-owned media in countries with less press freedom. It has been argued that social media can be a “soft weapon” that is particularly important during political unrest, protest, and political conflicts (Stockemer 2018, p. 44). Therefore, one might also expect that in some instances, social media contributes to fewer misperceptions about elections, as it offers more alternative (and perhaps more accurate) perspectives. Overall, however, the empirical evidence reviewed in the section above suggests that more trust in social media leads to more consumption of social media, which may expose users to more inaccurate election information, leading to more misperceptions about electoral integrity. Consequently, we formulate the following hypothesis:
H2: The more trust people have in social media, the higher their misperceptions about electoral integrity.
Regardless of which media sources people trust, we expect citizens in countries with limited media freedom to have more difficulty finding accurate information about election integrity. In contexts of limited media freedom, both traditional media and a significant portion of social media are likely to be either under government control or subject to self-censorship. Indeed, previous work found that internet exposure did not have a positive impact on citizens’ perceptions of electoral integrity in this context (Coffé 2017). Coffé (2017) argued that at the time, the internet did “not seem to provide an alternative source” (p. 293) in countries with less press freedom. Kerr and Lührmann (2017) also found that when media freedom is limited, citizens are more likely to express confidence in elections, whereas public trust in elections is lower when media freedom is higher. This suggests that in contexts of low media freedom, we should expect to find that a comparatively higher proportion of citizens is unduly positive about election integrity. Conversely, in contexts of high media freedom, we should expect accurate information about electoral conduct to be more readily available, as media is not under government control. However, precisely because the media is free in these contexts, it might also provide more fertile ground for spreading misinformation about elections. In contexts of high media freedom, we would therefore expect to find higher proportions of citizens who are unduly negative about election integrity.
H3. The lower the level of media freedom in a country, the higher the portion of citizens who are unduly positive about election integrity.
Cross-Level Interaction: The Interplay between Trust in Media and Media Freedom
The power of the media to affect citizens’ perceptions of elections is neither inherently democratic nor inherently autocratic (Tucker et al. 2017, p. 48). In a context of extensive media restrictions, social media can, for instance, be a place to reveal and share messages about attempts at election fraud, whereas, in a context of media freedom, it can be a place to share unsubstantiated doubts about free and fair elections. At the same time, when it comes to the effects of such messages on citizens, trust in media may matter more than actual exposure to information. In a low media freedom context, governments and state-controlled media offer one-sided, centralized, and propagandist accounts aimed at putting a positive spin on the regime and the elections, even if international observers detect unfairness in both (Coffé 2017). State-controlled traditional media and governments might also try to cast doubt about the veracity of the information on social media (Tucker et al. 2017). In such a context, one can expect citizens who still trust social media to have more accurate perceptions of election integrity than those who trust the official state-controlled media. Furthermore, according to Coffé (2017), citizens who have abundant trust in traditional media might be less aware of criticisms of the election and thus hold more misperceptions about electoral integrity.
Yet in contexts of high media freedom, one can expect the reverse effect. Here, traditional media and governments tend to react to misinformation spread via social media by fact-checking claims and introducing regulations to reduce the spread of misinformation (i.e., the watchdog role of the media; see Coffé 2017). Citizens who trust traditional media more than social media are also expected to have fewer misperceptions about electoral integrity. Thus, in a context of high media freedom, one can expect citizens who trust social media to have less accurate perceptions of election integrity than those who trust traditional media, due to possible exposure to misinformation. Coffé (2017) tested this assumption and found that in countries with a higher level of press freedom, the frequency of traditional media usage had a positive effect on citizens’ ability to make accurate judgments about the electoral process. She found the opposite effect in countries with low press freedom, as in such contexts the one-sided information had a negative effect on citizens’ ability to make a correct assessment of electoral integrity. Hence, we hypothesize that:
H4a. The higher the level of media freedom, the more traditional media reduces misperceptions.
H4b. The higher the level of media freedom, the more social media increases misperceptions.
Method
Sample
We used data from a survey conducted in 25 countries. The countries were selected to capture a wide variety of electoral settings with varying degrees of democratic freedom. A complete overview of countries is provided in Supplementary Material figure 1, and Supplementary Material table 1. Data were collected January 25–31, 2023, in the United Kingdom and February 8–21, 2023, in all other countries. The survey focused on a range of political issues, including electoral integrity. The survey was conducted by the research company Kantar, which used its respondent pool in each country, or that of a partner. Kantar’s respondent pools are composed through self-registration and use a range of recruitment and other quality checks. From the larger pool a selection of respondents is invited, stratified based on age and gender. When needed, additional respondents are invited until in each country, a minimum of 550 respondents completed the survey. Upon completion, respondents received an incentive from Kantar. The response rate of the invited respondents differed across countries, ranging from 13.5 percent (United States) to 51.7 percent (Japan), with an overall response rate of 21.5 percent. Of the people who started the survey, 87.7 percent completed the full questionnaire. For our analyses, we relied on a total of 14,723 observations. The full questionnaire can be found in Supplementary Material section B.
Electoral integrity and media freedom in 25 countries. See Supplementary Material table 1, for list of country abbreviations.
Multilevel models predicting misperceptions on electoral integrity.
| Model 1: Random intercept model | Model 2: Random slope model | |||||
|---|---|---|---|---|---|---|
| Absolute misperceptions | B | SE | p | B | SE | p |
| Individual level | ||||||
| Age | −0.01 | 0.00 | 0.001 | −0.01 | 0.00 | 0.000 |
| Female | 0.33 | 0.08 | 0.000 | 0.31 | 0.08 | 0.000 |
| Education | −0.04 | 0.03 | 0.104 | −0.04 | 0.03 | 0.197 |
| Political interest | −0.13 | 0.03 | 0.000 | −0.12 | 0.03 | 0.000 |
| Left-right | 0.13 | 0.02 | 0.000 | 0.12 | 0.02 | 0.000 |
| Left-right extremity | 0.12 | 0.02 | 0.000 | 0.12 | 0.02 | 0.000 |
| Trust in national media | −0.70 | 0.04 | 0.000 | 0.61 | 0.24 | 0.011 |
| Trust in local media | −0.52 | 0.04 | 0.000 | 0.32 | 0.26 | 0.218 |
| Trust in social media | 0.74 | 0.03 | 0.000 | −2.25 | 0.21 | 0.000 |
| Trust in politicians | −0.33 | 0.03 | 0.000 | −0.07 | 0.20 | 0.747 |
| Trust in acquaintances | −0.07 | 0.03 | 0.029 | 0.55 | 0.20 | 0.006 |
| Mainstream | 0.03 | 0.03 | 0.396 | 0.03 | 0.03 | 0.368 |
| Country level | ||||||
| Media freedom | 0.14 | 0.02 | 0.000 | 0.16 | 0.03 | 0.000 |
| Cross-level interactions | ||||||
| Trust in national media*media freedom | −0.02 | 0.00 | 0.000 | |||
| Trust in local media*media freedom | −0.01 | 0.00 | 0.002 | |||
| Trust in social media*media freedom | 0.04 | 0.00 | 0.000 | |||
| Trust in politicians*media freedom | −0.00 | 0.00 | 0.242 | |||
| Trust in acquaintances*media freedom | −0.01 | 0.00 | 0.002 | |||
| Constant | 2.05 | 1.66 | 0.217 | 0.49 | 1.73 | 0.779 |
| Log likelihood | −43872.80 | −43741.62 | ||||
| Model 1: Random intercept model | Model 2: Random slope model | |||||
|---|---|---|---|---|---|---|
| Absolute misperceptions | B | SE | p | B | SE | p |
| Individual level | ||||||
| Age | −0.01 | 0.00 | 0.001 | −0.01 | 0.00 | 0.000 |
| Female | 0.33 | 0.08 | 0.000 | 0.31 | 0.08 | 0.000 |
| Education | −0.04 | 0.03 | 0.104 | −0.04 | 0.03 | 0.197 |
| Political interest | −0.13 | 0.03 | 0.000 | −0.12 | 0.03 | 0.000 |
| Left-right | 0.13 | 0.02 | 0.000 | 0.12 | 0.02 | 0.000 |
| Left-right extremity | 0.12 | 0.02 | 0.000 | 0.12 | 0.02 | 0.000 |
| Trust in national media | −0.70 | 0.04 | 0.000 | 0.61 | 0.24 | 0.011 |
| Trust in local media | −0.52 | 0.04 | 0.000 | 0.32 | 0.26 | 0.218 |
| Trust in social media | 0.74 | 0.03 | 0.000 | −2.25 | 0.21 | 0.000 |
| Trust in politicians | −0.33 | 0.03 | 0.000 | −0.07 | 0.20 | 0.747 |
| Trust in acquaintances | −0.07 | 0.03 | 0.029 | 0.55 | 0.20 | 0.006 |
| Mainstream | 0.03 | 0.03 | 0.396 | 0.03 | 0.03 | 0.368 |
| Country level | ||||||
| Media freedom | 0.14 | 0.02 | 0.000 | 0.16 | 0.03 | 0.000 |
| Cross-level interactions | ||||||
| Trust in national media*media freedom | −0.02 | 0.00 | 0.000 | |||
| Trust in local media*media freedom | −0.01 | 0.00 | 0.002 | |||
| Trust in social media*media freedom | 0.04 | 0.00 | 0.000 | |||
| Trust in politicians*media freedom | −0.00 | 0.00 | 0.242 | |||
| Trust in acquaintances*media freedom | −0.01 | 0.00 | 0.002 | |||
| Constant | 2.05 | 1.66 | 0.217 | 0.49 | 1.73 | 0.779 |
| Log likelihood | −43872.80 | −43741.62 | ||||
Note: N = 14,723.
Multilevel models predicting misperceptions on electoral integrity.
| Model 1: Random intercept model | Model 2: Random slope model | |||||
|---|---|---|---|---|---|---|
| Absolute misperceptions | B | SE | p | B | SE | p |
| Individual level | ||||||
| Age | −0.01 | 0.00 | 0.001 | −0.01 | 0.00 | 0.000 |
| Female | 0.33 | 0.08 | 0.000 | 0.31 | 0.08 | 0.000 |
| Education | −0.04 | 0.03 | 0.104 | −0.04 | 0.03 | 0.197 |
| Political interest | −0.13 | 0.03 | 0.000 | −0.12 | 0.03 | 0.000 |
| Left-right | 0.13 | 0.02 | 0.000 | 0.12 | 0.02 | 0.000 |
| Left-right extremity | 0.12 | 0.02 | 0.000 | 0.12 | 0.02 | 0.000 |
| Trust in national media | −0.70 | 0.04 | 0.000 | 0.61 | 0.24 | 0.011 |
| Trust in local media | −0.52 | 0.04 | 0.000 | 0.32 | 0.26 | 0.218 |
| Trust in social media | 0.74 | 0.03 | 0.000 | −2.25 | 0.21 | 0.000 |
| Trust in politicians | −0.33 | 0.03 | 0.000 | −0.07 | 0.20 | 0.747 |
| Trust in acquaintances | −0.07 | 0.03 | 0.029 | 0.55 | 0.20 | 0.006 |
| Mainstream | 0.03 | 0.03 | 0.396 | 0.03 | 0.03 | 0.368 |
| Country level | ||||||
| Media freedom | 0.14 | 0.02 | 0.000 | 0.16 | 0.03 | 0.000 |
| Cross-level interactions | ||||||
| Trust in national media*media freedom | −0.02 | 0.00 | 0.000 | |||
| Trust in local media*media freedom | −0.01 | 0.00 | 0.002 | |||
| Trust in social media*media freedom | 0.04 | 0.00 | 0.000 | |||
| Trust in politicians*media freedom | −0.00 | 0.00 | 0.242 | |||
| Trust in acquaintances*media freedom | −0.01 | 0.00 | 0.002 | |||
| Constant | 2.05 | 1.66 | 0.217 | 0.49 | 1.73 | 0.779 |
| Log likelihood | −43872.80 | −43741.62 | ||||
| Model 1: Random intercept model | Model 2: Random slope model | |||||
|---|---|---|---|---|---|---|
| Absolute misperceptions | B | SE | p | B | SE | p |
| Individual level | ||||||
| Age | −0.01 | 0.00 | 0.001 | −0.01 | 0.00 | 0.000 |
| Female | 0.33 | 0.08 | 0.000 | 0.31 | 0.08 | 0.000 |
| Education | −0.04 | 0.03 | 0.104 | −0.04 | 0.03 | 0.197 |
| Political interest | −0.13 | 0.03 | 0.000 | −0.12 | 0.03 | 0.000 |
| Left-right | 0.13 | 0.02 | 0.000 | 0.12 | 0.02 | 0.000 |
| Left-right extremity | 0.12 | 0.02 | 0.000 | 0.12 | 0.02 | 0.000 |
| Trust in national media | −0.70 | 0.04 | 0.000 | 0.61 | 0.24 | 0.011 |
| Trust in local media | −0.52 | 0.04 | 0.000 | 0.32 | 0.26 | 0.218 |
| Trust in social media | 0.74 | 0.03 | 0.000 | −2.25 | 0.21 | 0.000 |
| Trust in politicians | −0.33 | 0.03 | 0.000 | −0.07 | 0.20 | 0.747 |
| Trust in acquaintances | −0.07 | 0.03 | 0.029 | 0.55 | 0.20 | 0.006 |
| Mainstream | 0.03 | 0.03 | 0.396 | 0.03 | 0.03 | 0.368 |
| Country level | ||||||
| Media freedom | 0.14 | 0.02 | 0.000 | 0.16 | 0.03 | 0.000 |
| Cross-level interactions | ||||||
| Trust in national media*media freedom | −0.02 | 0.00 | 0.000 | |||
| Trust in local media*media freedom | −0.01 | 0.00 | 0.002 | |||
| Trust in social media*media freedom | 0.04 | 0.00 | 0.000 | |||
| Trust in politicians*media freedom | −0.00 | 0.00 | 0.242 | |||
| Trust in acquaintances*media freedom | −0.01 | 0.00 | 0.002 | |||
| Constant | 2.05 | 1.66 | 0.217 | 0.49 | 1.73 | 0.779 |
| Log likelihood | −43872.80 | −43741.62 | ||||
Note: N = 14,723.
Measurement
The dependent variable was a citizen’s degree of misperception about electoral integrity. It was calculated based on two variables. First, respondents’ perceptions of electoral integrity were captured through 10 questions taken from the Electoral Integrity Project (and also used in the World Values Survey). Participants were asked: In your view, how often do the following things occur during election campaigns in your country? They were then asked to describe how often various scenarios occur, with answers ranging from 1 (never) to 5 (very often). They were asked to provide ratings for the following: (1) votes are counted fairly, (2) opposition candidates are prevented from running (reversed), (3) TV news favors the governing party (reversed), (4) voters are bribed (reversed), (5) journalists provide fair coverage of elections, (6) election officials are fair, (7) rich people buy elections (reversed), (8) voters are threatened with violence at the polls (reversed), (9) voters are offered a genuine choice in the elections, and (10) women have equal opportunities to run for office. Questions were randomized and presented to the respondents on a single screen (as a grid). The items formed a reliable scale (Cronbach’s alpha = 0.77). Scores were summed and yielded a scale from 10 (lowest perceived electoral integrity) to 50 (highest perceived electoral integrity) (M = 32.63, SD = 6.61; see scores per country in figure 1). The second variable we used to calculate misperceptions of electoral integrity was expert assessments of the same items, which were taken from the expert survey conducted by the Electoral Integrity Project (see Garnett et al. 2022).
We used the difference between the assessments of citizens and experts to assess misperceptions, with positive scores indicating unduly positive citizens and negative scores indicating unduly negative citizens. Average country-level scores ranged from -11.51 (France) to 0.94 (Kenya), with an overall mean of -7.05 (SD = 6.85), indicating that overall, citizens were less positive about the electoral integrity in their country than the experts. In the first set of analyses, we used absolute numbers to assess the degree of misperceptions. In the second set of analyses testing Hypothesis 3, we differentiated between respondents that scored higher than the experts (optimists) and respondents that scored lower than the experts (pessimists).
The key independent variable was the degree of trust respondents have in different types of sources. Respondents were asked: To what extent do you trust information that comes from the following sources? Sources listed were (1) national media, (2) local media, (3) social media, (4) politicians, and (5) friends, family, and colleagues. Respondents rated their trust for each source from 1 (not at all) to 7 (completely). Average scores were 4.06 for national media (SD = 1.66), 4.10 for local media (SD = 1.54), 3.49 for social media (SD = 1.64), 3.20 for politicians (SD = 1.65), and 4.77 for family, friends, and colleagues (SD = 1.41). Correlations between the various trust measures ranged from 0.27 (politicians and friends, family, and colleagues) to 0.74 (national and local media). Additionally, a principal component analysis demonstrated that all trust variables load on a single factor (first factor eigenvalue 2.91, with factor loadings ranging from 0.33 to 0.49). This indicates on the one hand that people are in general more or less trusting and thus general trust levels are captured by the variables in the model, but on the other hand shows that there is considerable variation across sources. Correlation between national and local media and social media is modest (0.45 and 0.50, respectively) and is less strong in countries with lower levels of media freedom (interaction term between social media and media freedom in a regression predicting national media trust is negative and significant: b = −0.009, p < 0.001, and also when predicting local media: b = −0.008, p < 0.001). This corroborates the idea that in contexts with low media freedom, social media serves as a viable alternative to traditional media.
As a country-level moderator, we considered the most recent World Press Freedom Index as reported by Reporters Without Borders (2022) as a measure of media freedom. The index is based on an elaborate analysis of political context, legal framework, economic context, sociocultural context, and safety, yielding a standard country-level score between 0 and 100. The average score for the countries we considered was 70.6, with scores ranging from 47 (Mexico) to 90 (Denmark). We differentiated between countries with relatively low media freedom (scores below 60, including Mexico, Indonesia, Israel, Chile, and Brazil) and countries with relatively high media freedom.
We included the following individual-level control variables in our models: Age as measured by the question What is your age? (M = 44.7, SD = 42.6); Female, using information from the question What is your gender? (52.4 percent female); and Education, measured by the question What is the highest level of education you completed? with a five-point answer category ranging from “less than high school” (1) to “graduate degree” (5) (M = 3.11; SD = 1.20). Political interest was captured by the question To what extent are you interested in politics, with answer categories ranging from “not at all” (1) to “very much” (7) (M = 4.51, SD = 1.78). Left-right positioning was based on self-placement on an 11-point left-right scale, relying on the question In politics, people sometimes talk about the left and right. Where would you place yourself on a scale from 0 to 10, where 0 means left and 10 means right? (M = 5.31, SD = 2.50). Extremity left-right positioning was based on the self-placement on the left-right scale and was calculated as the absolute deviation from the midpoint of the scale (M = 1.82, SD = 1.75). Additionally, we considered the degree to which respondents considered themselves political outsiders, which affected their assessment of integrity. We constructed a variable mainstream based on the average agreement on the following statements: My interests are in line with most people in the country and My political views are shared by most people in the country, with ratings ranging from “fully disagree” (1) to “fully agree” (7) (Cronbach’s alpha = 0.70, M = 4.28, SD = 1.33).
Analysis Approach
We used multilevel models with respondents nested in countries to test our first two hypotheses. More specifically, we use multilevel regressions with a random intercept to test the main effects of trust in different sources of information. To test Hypothesis 3, we conducted an analysis at the country level and looked at the correlation between media freedom and the proportion of optimists—those people who scored higher than the expert assessment. To test our final hypothesis, we conducted an additional multilevel analysis where we allowed the slope of the trust variables to vary across countries and looked at the interaction effects of media freedom. We do not weigh our cases in the analysis.
Results
In all countries, except Kenya, overall perceptions of electoral integrity are lower among our respondents than among experts. This negativity bias is particularly strong in Canada and France, where citizens scored more than 11 points lower on average on a scale ranging from 10 to 50. Overall, misperceptions about electoral integrity seem to be particularly present in countries where electoral integrity and levels of press freedom are high.
We first examine individual differences. The intraclass correlation (ICC) from an empty multilevel model (0.11) demonstrates that most variance in the dependent variable (absolute misperceptions) is present at the individual level. Table 1 presents the random intercept model, where we focused particularly on the effects of the trust variables. Hypothesis 1 predicted that the more trust people have in traditional media, the lower their misperceptions of electoral integrity. Our analysis confirmed this hypothesis: both trust in national media (B = −0.70, SE = 0.04) and trust in local media (B = −0.52, SE = 0.04) had a significant and negative impact on misperceptions about electoral integrity. A one-point increase in self-reported trust in national media decreases misperceptions by 0.70; for local media, this decrease is 0.52. In addition, Hypothesis 2 (which states that trust in social media increases misperceptions) is confirmed (B = 0.75, SE = 0.03). From table 1, it can also be seen that the other trust variables exert an influence, with trust in politicians and friends, family, and colleagues both decreasing misperceptions.
The analyses also revealed that higher levels of media freedom yield higher levels of misperceptions. Furthermore, we found that younger people, women, and less educated people display higher levels of misperceptions, though the differences are relatively small. Political interest decreases misperceptions. The more right-wing and the more extreme individuals are, the higher levels of misperceptions they hold. Self-identification as a political outsider does not impact misperceptions.
Hypothesis 3 predicted a negative association between the levels of media freedom in a country and the proportion of unduly optimistic citizens. Overall, the number of optimistic citizens (those that assess electoral integrity higher than experts) was low: 20.8 percent of our respondents. This number ranged from 7.4 percent in France to 63.3 percent in Kenya. We found a moderate negative correlation between media freedom and the presence of optimistic citizens (r = −0.37, p < 0.10). Thus, in countries where experts consider media freedom low, a significant number of citizens make more positive assessments than experts.2
Finally, we tested whether the effects of trust in different information sources depended on media freedom. Here, we used a random slope model and focused on cross-level interactions. Again, our hypotheses were confirmed: the higher the levels of media freedom, the smaller the negative effects of national and local media (4a), and the larger the positive effect of social media (4b).
Figures 2 and 3 show the effects of national and social media use in countries with differing levels of media freedom and provide more insight into the interactions. Figure 2 demonstrates that while slopes differ substantially across countries with different levels of media freedom, trust in national media contributes to lower misperceptions across the board. Figure 3 demonstrates that the effect of trust in social media can actually reverse: in countries with high levels of media freedom, it contributes to high levels of misperceptions, but in countries with lower levels of media freedom, it actually decreases misperceptions.
The varying impact of trust in national media on misperceptions. Linear prediction based on fixed portion of random slope model; 95 percent CI are displayed.
The varying impact of trust in social media on misperceptions. Linear prediction based on fixed portion of random slope model; 95 percent CI are displayed.
Conclusion
This paper set out to understand varying levels of misperceptions on electoral integrity. By employing a unique dataset based on a survey conducted in 25 countries, our results are in line with previous research but also provide important additional insights. Indeed, various types of media are important sources of information that shape electoral misperceptions (Coffé 2017). The trust people hold in information from different media sources is a key explanatory variable for misperceptions—this trust is likely to correlate with both actual use as well as the degree to which citizens accept (mis)information from these sources, which might indicate that people differentially assess the quality and importance of information from different sources. While previous research has suggested that internet usage has no effect on the accuracy of electoral integrity perceptions (Coffé 2017), we demonstrate that trusting information from social media yields higher misperceptions. We contend that the prevalence of misinformation on social media might well account for this finding. However, we also find that effects are largely context dependent—the impact of trust in traditional media decreases significantly when press freedom in a country is lower and the effect of trust in social media even reverses in countries with lower levels of media freedom, decreasing misperceptions of election integrity. This is in line with our expectations that in contexts of low media freedom, government control of media means that traditional media can actually lead citizens to believe their elections are reasonably free and fair, when in fact they are not. We expected that in such settings, social media might offer an opportunity to spread accurate information about election irregularities, and our findings suggest this is indeed the case. Our sample did not include countries with exceptionally low levels of press freedom; the reversal of effects would likely be even more prevalent in those contexts.
Another noteworthy finding is that the share of unduly negative citizens is highest in contexts of high media freedom. This is a cause for serious concern, as it suggests that high levels of media freedom, while essential for democracy, also enable the spread of misinformation about elections. Citizens’ exposure to such misinformation can result in widely believed yet incorrect allegations of electoral fraud, which can lead to unrest and violence.
While our study provides relevant insights to understand the functioning of democracies, it also invites further research. For example, we did not measure exposure to misinformation directly, but rather used a set of proxy variables: trust in local and national media and trust in social media. As such, we were able to provide correlational evidence that is suggestive of such an effect but were not able to dig deeper into the impact of misinformation itself. To do so, future research should explore the actual presence of misinformation in different sources and across contexts. For instance, we now assume that traditional media may also be responsible for propagating misinformation about elections, even in countries with high press freedom. While research indicates that misinformation is prevalent on social media (e.g., health misinformation; Suarez-Lledo and Alvarez-Galvez 2021), we cannot rule out that misinformation is also present in traditional media. For instance, right-leaning media outlets in the United States were more likely to disseminate inaccurate information regarding the origins and treatment of COVID-19. Consequently, people who consumed more right-leaning news were more likely to hold misinformed views (Motta et al. 2020). Nevertheless, we still assume that due to the journalistic gatekeeping principles and verification norms, the prevalence of misinformation is much lower in traditional media than in social media. For instance, Canadian journalists use various verification techniques for information, such as source triangulation and the analysis of primary data sources or official documents, and are also often more critically aware of their own limitations and biases (Shapiro et al. 2013). These journalistic strategies are more likely to prevent (but not totally mitigate) misinformation in traditional media.
In addition, we did not include the winner-loser gap in perceptions of electoral quality in this study. This strand of research shows that supporters of the winning side perceive elections to be of higher quality than supporters of the losing side. Hence, supporters on the losing side believe elections are less credible (see, e.g., Bush and Prather 2017; Whitt et al. 2023). While such an effect might be less straightforward in multiparty systems (as government coalitions are formed after the election and the winner-loser gap might be less clear-cut than in two-party systems), this variable should be considered in future work on the perceptions of electoral integrity, as it may be an important additional predictor.
Furthermore, two methodological points are worth mentioning. First, our country samples of respondents are representative of characteristics such as age and gender, but might not be representative of other background variables. Within the scope of this research, it was not possible to also reach representativeness on, for example, political preferences or educational background. As a result, we must treat our findings with some caution. However, given the robustness of our results, we are confident that they are generalizable, but replication with higher-quality samples should confirm this. Second, at the country level, we selected countries with considerable variation in press freedom but did not include any countries with authoritarian regimes where the media is completely restricted. Obtaining reliable data from those countries is a complex enterprise, but assessing the generalizability of our findings is a worthwhile effort.
Despite its limitations, our study provides important insights into who holds misperceptions about electoral integrity as well as the role played by media freedom. The study underlines that simplistic qualifications of information sources as “good” or “bad” for democracy are unwarranted. It is the interplay between those sources and the larger (information) environment that determines whether people obtain accurate election knowledge or believe falsehoods.
Footnotes
Alternate terms for traditional media are “old media” and “legacy media.” There are problems with both these terms. The term “old media” does not convey the fact that almost all modern “old” media makes use of “new” technologies in producing and distributing its content (Langer and Gruber 2021). The term “legacy media” suggests that these media predated the internet. However, some television programs, radio stations, and newspapers were founded after the advent of the internet, and therefore cannot be accurately termed “legacy.” For these reasons, we opted to use the term “traditional media” in this study.
One might argue that this finding is a statistical artifact because low levels of media freedom coincide with lower levels of electoral integrity and floor effects limit individuals to reporting negative misperceptions. Indeed, levels of electoral integrity and media freedom are correlated to a considerable degree (r = 0.76 on the country level). However, our sample does not include countries that score at the very low end of the electoral integrity scale—the lowest-scoring countries are near or well above the absolute midpoint of the scale. This is also reflected in the distribution of “optimists” and “pessimists.” Even in the countries with the lowest scores, we see that there are more pessimists than optimists, with the exception of Kenya, where 63 percent of the respondents are unduly positive.
Supplementary Material
Supplementary Material may be found in the online version of this article: https://doi.org/10.1093/poq/nfae021.
Funding
Data collection was sponsored by funding from the DATADRIVEN. The project DATADRIVEN is financially supported by the NORFACE Joint Research Programme on Democratic Governance in a Turbulent Age and co-funded by Economic and Social Research Council (ESRC), Austrian Science Fund (FWF), the Dutch Research Council (NWO), and the European Commission through Horizon 2020 [grant agreement No. 822166 to S. K.].
Data Availability
Replication data and documentation are available at https://doi.org/10.17605/OSF.IO/942UY.
