Abstract
Introduction
Disinformation remains a resonant topic of public debate, not only in the EU but also globally. Many discussions have focused on the role of disinformation in elections and its impact on voting behaviour (Schünemann 2022). Since free and fair elections represent the very fundament of the democratic political system, it is no wonder that so many people are concerned that the intentional spread of falsehoods might discredit them. Furthermore, the thought that the beliefs and political choices of fellow citizens are not genuine but motivated by misguided false stories is harmful to social cohesion and can undermine open deliberation about political matters. Therefore, the impact of disinformation on elections has to be a matter of rigorous research interest since a proper understanding of this phenomenon will enable us to better understand how the political process functions in the twenty-first century. Equally, looking at disinformation within the context of elections is beneficial since such a specific focus allows us to gain better insights into the role of this phenomenon in contemporary societies.
With these thoughts in mind, the authors decided to join forces to investigate the role of disinformation in the Czech parliamentary elections in 2021 and its interaction with voting behaviour. The research team comprised two entities: the think tank the Prague Security Studies Institute, with its notable expertise in the analysis of disinformation in the online information space, with a special emphasis on elections; and the STEM Institute for Empirical Research, which has been analysing the opinions of Czech citizens since 1992. The multidisciplinary setting of the project, combining skills in media monitoring and interpretation of public opinion, and including insights from political science, sociology and psychology, has allowed us to uncover the role of falsehoods during the elections and gain new knowledge about how people interact with them. With our research we did not aim to address only the context of Czechia in autumn 2021 but to suggest a general model which other researchers in different settings—such as the 2024 European elections—could utilise.
The main purpose of our model was to measure not solely exposure to or belief in disinformation but the resilience of respondents to falsehoods. In our understanding, the key indicator of resilience is the ability to orient oneself in the public debate—not only to recognise false statements but also to identify the factually correct criticisms of running parties. To make this measurement even more precise, we also included fabricated placebo statements about Czech political parties. These statements followed a similar logic to the misinformation that we noticed in the information space in the run-up to the elections. The reason for using placebo statements was to measure the ability of respondents to distinguish between a political bias (e.g. belief in any statement made against a party they dislike or disbelief of all news in general) and the influence of disinformation campaigns. While the combination of factually correct statements, misinformation and placebo headlines had already been used in research (see Allcott and Gentzkow 2017), our model, emphasising the ability to distinguish between them is, to our knowledge, a methodological innovation. We aimed to use this model not only to evaluate the impact of falsehoods prior to the 2021 elections, but also to better understand the audiences that are receptive to disinformation. For this reason we reimagined the position of disinformation in causal relations and did not consider belief in falsehoods to be the single cause of a change in electoral behaviour but perceived it as part of a self-confirming cycle, in which it reinforces pre-existing stereotypes, opinions and emotions. In particular, our aim was to understand these interactions in the context of the broader socio-economic dynamics that are currently shaping the opinions and behaviour of Czech society.
Before elaborating further on our argumentation, a terminological remark is in order. Since we were not closely investigating the intent of the spreaders of the falsehoods appearing in the course of the election campaign period, we do not use the term
The article is structured as follows. First, we explain our methodology and demonstrate how our model is able to measure resilience to misinformation while overcoming bias related to the selection of the test statements. Second, we show that resilience to misinformation has important sociological underpinnings, which allows us to make an argument for the need for a holistic approach to tackling this threat. Finally, we return to the topic of voting behaviour and demonstrate that—in the right circumstances—everyone can be prone to believe in misinformation. Based on this knowledge, we formulate some general policy-oriented recommendations, which are presented in the final section of the article.
How to design a useful model?
The obvious research question about misinformation in the context of elections concerns its impact on the final results. While this might be the logical question, it is highly difficult to answer, since the academic community even disagrees about the impact of electoral campaigns on election results. Some scholars have observed that political preference tends to be stable over time and highlight the importance of socio-economic factors in electoral decision-making. Even those who are more inclined to believe that political campaigns play a role in influencing voters’ behaviour argue that they face significant constraints since they not only have to reach out to voters but also change their existing opinions (Fisher 2018). The same consideration can be applied to the impact of misinformation on voting behaviour. Therefore, the research questions should be reframed thus: in which circumstances might misinformation impact voters and which voters are likely to be impacted?
The question of exposure to misinformation is also not as banal as it seems. Sociological research shows that people are not good at remembering if they have noticed a specific news item, or if they have, where they noticed it (Cardenal et al. 2022). Exposure also has to be put into the context of the broader dynamics of the information space (see, e.g. Allen et al. 2020). A good illustration is the case of the so-called Pandora Papers, which we also touched on in our research. Shortly before the 2021 elections, journalists released the results of their investigative work into the offshore companies used by rich individuals to lower their taxes. Former prime minister and leader of majority party Yes (ANO), Andrej Babiš, was allegedly involved in such a scheme (Goodley et al. 2021). Unsurprisingly, this news was circulated widely in the Czech media space and became one of the important topics of political discussion. Yet our polling shows that 31% of our respondents were not aware of this news item. This finding forces us to reconsider the impact of misinformation, since there are groups of people who are able to ignore the news completely, including any circulating misinformation. As shown in the chart below, misinformation related to the elections only resonated with a small proportion of the population—only 14% of respondents believed two of the test statements, while just 11% believed three or four of them. This shows that the Czech public is, in general, less prone to believe misinformation related to elections than it is to believe in general conspiracy theories, which can affect the beliefs of about 40% of the population. Another notable finding is the existence of a sizeable group—28% of respondents—that did not believe any of the factually correct test statements. Our approach was based on the assumption that it is not sufficient to draw conclusions from the total number of people believing tested misinformation since there are significant groups of the population which believe either nothing or everything (meaning misinformation as well as factually correct statements).
Number of headlines people believed.
Therefore, we decided to define resilience to misinformation as the ability to orient oneself in the public debate, which we measured by the ability to distinguish between the factually correct and factually incorrect statements related to the running parties. Being able to make such distinctions not only allows one to identify misinformation but may also help to control personal bias, since it demonstrates the ability to recognise the difference between the true and false criticisms of individual candidates.
When compiling the list of test statements, we utilised real-time media monitoring conducted by the Prague Security Studies Institute within the framework of the project ‘Czech Elections in the Era of Disinformation’ (Syrovátka and Šefčíková 2021). In total we tested 12 statements of 3 different kinds (real news, misinformation and placebo) for each of the 4 main political parties or coalitions running in the 2021 elections. The sociological polling was conducted in two waves—at the beginning of September and in the week after the elections (middle of October)—on a representative sample of the population. Based on their differing abilities to distinguish the reliability of the test statements, the respondents were divided into four clusters whose differing opinions and values were analysed. The most relevant cluster for this text, comprising the most vulnerable part of the population, is represented by the number 1 (= lowest ability to distinguish between different test statements) in the subsequent figures.
Societal factors as the key indicator of vulnerability
The definition of the vulnerable group of respondents allowed us to reach a deeper understanding of its characteristics. Perhaps surprisingly, the economic situation, level of education, age and gender of these respondents did not seem to play a significant role in their ability to distinguish between factually correct and false statements. The more relevant indicator was the perception of the respondents with regard to their personal situation and their general assessment of Czech society, which differed from other clusters. People vulnerable to misinformation felt personally unappreciated in their communities, and they were very sceptical in their evaluation of the future—54% of these respondents looked to the future with no or only very little hope. This negative perception of the situation was not limited only to the personal level but was also manifested in their assessment of the overall situation in Czechia. For instance, the quality of democracy was rated negatively by 44% of the respondents from the vulnerable group—they gave it just two marks out of five.
Perception of the future.
Generally, the vulnerable respondents felt unhappy and frustrated, which made them receptive to narratives hostile to the existing political system. These people were significantly more likely to believe in general conspiracy theories that had been circulating in the information space for a long time—81% were also prone to endorse them. This percentage is significantly higher than in the other clusters of respondents. This finding is consistent with the monological belief system thesis, which suggests that people who start to believe in one conspiracy theory are significantly more likely to endorse more of them over time (Goertzel 1994).
Trust in general conspiracy theories.
It is important to point out that people vulnerable to misinformation, despite their negative perception of the system in which they live, were not completely disengaged from its processes. The percentage of respondents in the vulnerable group who voted in the 2021 elections did not differ significantly from the percentages in the other clusters. The difference was that they were more likely to vote for far-right or populist parties that promised to revise the current political system (similar to the effect demonstrated by Cantarella et al. 2023). This group seems to be formed of staunch supporters of these parties since 53% of respondents knew how they would vote several months before the elections. Here we found an interesting paradox—a notable proportion of the respondents vulnerable to misinformation were actually fairly immune to falsehoods appearing in the run-up to the elections since they were not likely to be swayed in their voting behaviours. For this group, the more important factor is long-term interaction with general conspiracy theories that reinforce their negative perception of the situation in society—which, to remind readers, might be understandable due to the living conditions of the given individual—and makes them open to the rhetoric of anti-system parties. The exact interaction of variables in this vicious circle should be of further interest to researchers. A single survey such as ours is not able to determine causality and therefore it would be warranted to use different methods, such as a natural experiment or a longitudinal study (see Eady et al. 2023), to establish this.
Intent to vote.
Opinion first, misinformation second
At the beginning of the article, we reframed the main question regarding the impact of disinformation on elections and tried to define the circumstances in which voters’ decisions could be influenced by misinformation. As we demonstrated in the previous section, in the case of people unable to distinguish between factually correct and false statements, this question requires taking a long-term view and disentangling the interactions between unsatisfying personal situations, dissatisfaction with the political system and the influence of conspiracy theories. However, it was not only respondents from the vulnerable group who were prone to believe falsehoods that appeared in the run-up to the elections. In fact, a significantly higher number of respondents—including those who did not endorse general conspiracy theories—were open to believing misinformation in specific circumstances. To better understand their reasoning, we find it useful to perceive the belief in misinformation as a consequence of pre-existing positions.
Regardless of political preference, people who felt disappointed by a specific party were more likely to believe misinformation about this party. This did not make them change their vote—since they would not have supported this party anyway—but it did reinforce their already formulated position. In explaining this tendency, we refer to the concept of motivated reasoning, which suggests that individuals select the information they believe pragmatically, considering its relevance for achieving the desired goal regardless of its factual correctness (Bolsen et al. 2014). And since, in the case of the 2021 elections, the main goal for disappointed voters was to reinforce or maybe rationalise their negative position towards a specific party, they were keen to believe any negative statement about it—even if it was misinformation or a placebo statement.
A case in point is that belief in both the misinformation and the placebo headlines about the SPOLU coalition 1 did not differ significantly among disappointed voters. The effect of voter disappointment on belief in misinformation is also visible in relation to the PirSTAN coalition 2 —35% of people who decided not to vote for the coalition also believed the false test statement, a significantly higher proportion than believed it among loyal voters. We are convinced that while misinformation can play a role in creating bias against a particular party, it is only one of the relevant factors in the long-term process. Therefore, we hypothesise that the misinformation that appeared directly before the elections did not play a significant role in affecting voting behaviours since it was very likely to resonate most among those who would not have voted for the targeted party anyway. Again, the long-term interaction between party preferences, feelings of disappointment and belief in misinformation should be the topic of further, more precise academic scrutiny.
Trust in headlines related to individual parties: the difference between loyal and disappointed voters.
Conclusions
Our research describing the impact of misinformation on the electoral behaviour of Czech voters in the context of the 2021 parliamentary elections could also be useful in other contexts. We believe that our model, based on the ability to distinguish between factually correct, false and placebo statements, should inspire other research since it overcomes several limitations related to survey methods and the self-reporting of beliefs and behaviours. It might also have more practical uses since it suggests an easy way to operationalise the concept of resilience to misinformation (for various approaches to this topic, see Hassain 2022). Therefore, our model could serve as a tool to assess the impact of projects aiming to increase resilience, such as media-literacy workshops or online educational games. While the main limitation remains the selection of appropriate statements for testing, we believe that this limitation could be overcome by interdisciplinary cooperation.
Using the case study of the 2021 elections, we also aimed to contribute to the burgeoning debate about determinants of belief in misinformation. Our approach suggests that the endorsement of falsehoods should be treated as a consequence of pre-existing positions and perceives this decision to believe them as part of the process of reinforcing already-made decisions—for instance in voting behaviour.
It is important to distinguish between two different groups affected by misinformation. First, there are those who are generally distrustful of the political system, who are stressed and frustrated, and who are open to conspiracy theories and the anti-systemic voices of far-right and populist parties. To increase the resilience of this group to misinformation, a holistic approach is needed that addresses the societal conditions which have alienated these people from the democratic system. Second, on a more general level, everyone can become vulnerable to misinformation if it fits with his or her biases—for instance by fuelling resentment against a particular political party. Since recognising and filtering one’s personal biases represents a significant challenge, this ability should be addressed and developed alongside existing educational activities that aim to tackle misinformation. Perfecting such a skill could be helpful in decreasing the polarisation of European societies and cultivating public discussion, which is always one of the most important tools in the fight against the spread of manipulative falsehoods.