The sharing of disinformation in cross-national comparison: analyzing patterns of resilience

ABSTRACT Although the problem of disinformation is on the rise across the globe, previous research has found that countries differ in the extent of widespread disinformation. In this study, we examine the willingness to disseminate disinformation across six countries (Belgium, France, Germany, Switzerland, the U.K. and the U.S.). We use a model by Humprecht et al. (2020) to study to what degree various systemic-structural factors influence individual behavior and contribute to resilience to disinformation. We draw on uniformly collected primary survey data and use regression analyses to examine which factors may explain citizens’ decisions to not further propagate disinformation. The results of our cross-national study show that resilience factors are country-specific and are highly dependent on the respective political and information environments. While in some countries extreme ideology weakens resilience, in others low education can have such an effect. Cross-national resilience factors include heavy social media use, the use of alternative media, and populist party support. We discuss what kind of tailored measures in combating online disinformation are needed to improve social resilience across different countries.


Introduction
Recent events, such as the outbreak of the novel coronavirus (SARS-CoV-2) and election campaigns, have demonstrated that disinformation spreads quickly on social media with potentially severe effects (Gruzd & Mai, 2020;. False information can be shared without the intent to cause harm (misinformation) or with the intent to cause harm (disinformation) (Wardle & Derakhshan, 2017). Disinformation involves purposeful manipulation, decontextualization, or fabrication of untrue information to achieve a predefined political goal (e.g., Bennett & Livingston, 2018;Freelon & Wells, 2020). It is used, for example, by political actors to spread conspiracies and alternative realities, fuel mistrust in the mainstream media and the established political order, and mobilize support for alternative (radical) agendas (Marwick & Lewis, 2017). In particular, the combination of the information setting of social media and a high degree of uncertainty can create a discursive opportunity structure for agents of disinformation who want to reinforce distrust and hostility towards outside groups (Guo & Vargo, 2020).
Although social media platforms are heavily used in many countries, not all countries are equally affected by the spread of disinformation (Fletcher et al., 2018;Newman et al., 2020). Humprecht et al. (2020) have argued that some countries have proven stable, adaptive, and resilient in times of social and technological transformation while other countries, such as the United States, appeared more vulnerable. Those authors argue that while there is presumably a great amount of disinformation in many countries, structural factors may cause individuals to disseminate it further-or refrain from doing so. However, there have been few comparative studies of whether and why individuals' willingness to disseminate disinformation differs across countries.
To close this gap, we aim for a rigorous test of the analytical framework proposed by Humprecht et al. (2020) by operationalizing disinformation and resilience and measuring these concepts using cross-country surveys. By doing so, we systematically analyze the influential relationships between theoretically founded explanatory factors, individual behavior, and the extent of resilience in different countries. We find that factors that promote resilience to disinformation are country specific. In other words, the political, media, and economic environments play a major role in how citizens react to disinformation, and policy solutions to tackle the problem must be tailored to the particular social environment.
Understanding disinformation, its dissemination, and demonstrating resilience Recent conceptualizations have distinguished between misinformation and disinformation (Wardle & Derakhshan, 2017): while misinformation can be understood as any kind of inaccurate information that is false without necessarily being intentionally misleading, disinformation is false or inaccurate information that is intended to deceive (Wardle & Derakhshan, 2017). Although the distinction between intentionally false information (disinformation) and false information in general (misinformation) is both conceptually and socially relevant (Ha et al., 2021), it is difficult for users to decipher the motivations and goals of disinformation agents based on the mere content of their messages. However, because disinformation is considered more problematic due to its strategic manipulative potential, we focus our attention on this type of false information in this article.
The intentions behind the spreading of disinformation may vary but can be broadly described as ranging from demobilization (e.g., attacking or delegitimizing political actors) to mobilization (e.g., cultivating support for certain perspectives) (Bennett & Livingston, 2018). Disinformation often aims at disrupting the established political order by arousing distrust and cynicism; it is meant to incite hostility and cultivate (partisan) divisions between social groups (Mourão & Robertson, 2019). Disinformation is similar to propaganda, which aims to control public opinion (Lasswell, 1927), for instance by attributing blame out of manipulative intent (Humprecht, 2019).
Social media facilitates the spread of disinformation (Allcott et al., 2019;Cinelli et al., 2020;Tandoc et al., 2020;Zhao et al., 2020). One reason for this situation is that social media platforms are consumption environments, in which users primarily seek entertainment and encounter information somewhat randomly (Boczkowski et al., 2018). In this situation, affective reactions, such as commenting and sharing, are more likely than in a situation that is primarily focused on information consumption (Metzger et al., 2021). This behavior leads to a spiral of dissemination: liking, sharing, and commenting lead to a higher visibility of the content for other users due to the algorithms of social networks. Moreover, frequent dissemination signals popularity which might increase the chances that (dis-)information is spread further. This chain of reactions is reinforced when opinion leaders with a particularly large number of followers, such as politicians or celebrities, pick up and disseminate the information (Pennycook et al., 2021). Finally, as Van Bavel et al. (2021) demonstrate, disseminating disinformation is not necessarily associated with belief in disinformation. Rather, citizens spread disinformation because it confirms their personal opinions.
Against this backdrop, the question arises of how to build resilience that leads to a lower prevalence of disinformation. Expert recommendations, e.g., by the COVID-19 Vaccine Commination Handbook (Lewandowsky et al., 2021), indicate that citizens and journalists should not repeat or disseminate disinformation, even if they want to refute it, because they otherwise increase their reach and possible consequences. In this sense, we operationalize resilience as the unwillingness to share, like, and comment on disinformation. This behavior of not propagating potentially harmful information can be interpreted on a collective level as 'the capacity of groups of people bound together in … communities or nations to sustain and advance their well-being in the face of challenges'which corresponds to the understanding of resilience put forward by Hall and Lamont (2013, p. 2). Humprecht et al. (2020) developed a model based on this definition with framework conditions that promote resilience and are suited for international comparison. The authors argue that resilience factors can be attributed to the political, the media, and the economic environment. In political terms, the authors argue that the extent of societal polarization and the amount of support for populism in a society is crucial (the lower these are, the more resilient the society is). In terms of media, they argue that trust in news media, the variety in individual's main news consumption sources and exposure to public media are crucial in a society (the greater these are, the more resilient a society is). Finally, they consider dependence on social media for political information as crucial (the more dependence there is, the less resilient the society is). While the heuristic value of the framework is high, Humprecht et al. (2020) admit that they measured both the dependent and independent variables unsatisfactorily: for resilience (DV), they had to rely on distortion-prone self-reporting, and for influencing factors (IVs), they had to rely on collated, incomplete secondary data. With the present study, we respond to the appeal of Humprecht et al. (2020, p. 509) 'to validate [their] framework with better data. ' We answer this call by measuring the dependent and independent variables at the individual data level and integrating them into a uniformly designed, cross-country comparative survey. We present participants with real-life replications of actually circulated deceptive news articles posted on social media, and ask the participants whether they would like, share, or comment on the post (thus participating in its publicity) or whether they would refrain from any reaction (thus demonstrating resilience). To increase the generalizability of our conclusions, we use three controversial topics, namely the COVID-19 pandemic, climate change, and immigration. Our selection of topics followed the suggestions by Bennett and Livingston (2018) and Marwick and Lewis (2017) that disinformation is often spread to (de-)mobilize people and blame certain actors, such as foreign governments, news media, or activists, in order to shake the social order. This measurement of resilience, which we consider more valid than a mere self-declaration, is intended to answer our first research question, namely how citizens from different countries vary in their willingness to disseminate disinformation (RQ1).

Factors influencing resilience against disinformation
The specific research interest of this study is to determine the explanatory power of factors that lead citizens to disregard and ignore disinformation (demonstrate resilience) or to actively disseminate it. Based on the groundwork laid by Humprecht et al. (2020), we assign these impact factors to three spheres (politics, media, social media platforms) and determine their influenceas well as that of other control variablesat the individual level. This approach allows us to analyze the relative weight of these influence blocks comparatively and to observe and explain the resilience situations of different countries side by side.

Political influence factors
Societal polarization and populist communication can reduce resilience to online disinformation. Research on the first factor has shown that disinformation is most effective when it is consistent with a person's previous beliefs (Nyhan & Reifler, 2010;Thorson, 2016), e.g., when a confirmation bias is activated (Knobloch-Westerwick et al., 2020). The effect of opinion-congruent disinformation is most relevant in the political sphere, where party supporters often have strong attitudes. If strong attitudes are accompanied by a high level of emotionality in political communication, this can further promote the deliberate dissemination of disinformation Shin & Thorson, 2017). Based on this reasoning, we hypothesize that extreme ideology leads to lower resilience against partisan disinformation and increases the likelihood of its dissemination (H1).
Furthermore, studies have frequently linked disinformation to populist communication because populists use communication styles that can be misleading (Boberg et al., 2020;Hameleers, 2020). For example, populist political actors often use ingroup/out-group narratives and blame certain groups, such as elites or immigrants, for current problems (Engesser et al., 2016). In their communication, populists combine a distrust of democratic institutions and elites and their claim to speak for and represent the people with the use of social media that allow them to engage directly with 'the people' (Engesser et al., 2016). This approach has been evident during the coronavirus pandemic, for example, in attacks on the WHO or national governments . Populists also try to convince the audience of their message by leaving out certain details. The line between true and false information is blurred, suggesting to the audience that there are 'alternative' truths . Moreover, populist leaders frequently insist that they are the arbiters of truth and that only they can be trusted (Reinemann et al., 2016). Supporters of populist politicians or parties are frequently exposed to this communication style and are more likely to internalize it. Against this background, we hypothesize that supporters of populist parties are less resilient and are therefore more likely to spread disinformation (e.g., by liking, sharing, or commenting) (H2).

Media influence factors
Trust in the news media also appears to be related to resilience against disinformation. For example, Zimmermann and Kohring (2020) showed that low trust in news media increases belief in disinformation. The reasons for this result may be related both to the producers of disinformation and to the conveyers. Producers of disinformation often claim that they present information that is deliberately hidden by established news media and politicians (Holt et al., 2019). News media, on the other hand, often do fact-checking as part of their daily reporting. If citizens do not trust the news media and its fact-checking, they are likely to be more receptive to alternative news sources that may spread disinformation (Stier et al., 2020). Moreover, distrust in news media is often associated with distrust in democratic intuitions-which are frequently criticized by producers of disinformation . Besides its relationship to democratic institutions on the macro-level, trust is also related to resilience on the meso level of intergroup conflicts, and individuals' generalized distrust towards power on the micro-level, i.e., individuals' conspiracy mentality (Miller et al., 2016). Thus, we assume that trust in news media leads to higher resilience against disinformation and reduces the willingness to spread it (H3a). In addition, low trust in the established media may lead citizens to use alternative media outlets that publish disinformation (Boberg et al., 2020). Therefore, we postulate that the regular use of alternative media outlets reduces resilience and increases the willingness to circulate disinformation (H3b).
In this vein, several authors have argued that conscientious media use can strengthen resilience to disinformation (Livingstone, 2018). For example, users of PSB were found to be more knowledgably regarding public affairs issues (Aalberg & Cushion, 2016) and are more likely to be exposed to high-quality information and the verification of potentially false claims (Cushion, 2012). Moreover, public service broadcasters in different countries demonstrate higher levels of news performance compared to private news outlets (Humprecht & Esser, 2018). One reason for this difference in performance is, that public media often have transparent ethical guidelines and professional norms, and they have sufficient resources to invest in high-quality journalistic production (Cushion, 2012). Therefore, PSB enjoys high levels of trust in many European societies . Based on this reasoning, Livingstone (2018) advocated for a strong role of PSB in increasing media literacy among citizens to fight the dissemination of mis-and disinformation. Thus, we postulate a positive relationship between the use of PSB and resilience against disinformation (H4).
In line with the idea of a conscientious media use, scholars have argued that a broad media diet can provide citizens with different perspectives, e.g., pro-and counter-attitudinal views on polarized topics, and thereby leading to depolarization and possibly limiting confirmation biases (Beam et al., 2018;Chadwick et al., 2018). Citizens who obtain their news from multiple established media sources are more likely to be confronted with a broader range of opinions compared to citizen who only use one news source (Iyengar et al., 2009). Moreover, selective exposure to attitudinal-congruent news content lowers the likelihood of acknowledging disinformation as such ). Finally, cross-cutting exposure has been found to limit dysfunctional information sharing (Rossini et al., 2020). Against this background, we hypothesize that a broader diet of main news contributes to the formation of resilience and increases the probability that disinformation will be disregarded (H5).

Social platform influence factors
Disinformation spreads rapidly on social media because algorithmic message curation can create homogeneous opinion environments (Bennett & Livingston, 2018;Marwick & Lewis, 2017). Within these insulated spaces, misleading information can accumulate in large quantities and be made accessible to users who, being repeatedly exposed to it, may form the impression that certain statements and viewpoints are undisputed (Chan et al., 2020;Elcheroth & Drury, 2020). Regular users of social media are more likely to be confronted with disinformation and may be less critical of it (Vraga & Tully, 2021). We therefore assume that the heavy use of social media is detrimental to resilience and increases the probability of the further dissemination of disinformation (H6a). Moreover, the way social media is used may also play a role. Users who rarely disseminate information may also be reluctant to spread disinformation-even if the message appeals to them. Therefore, we assume that rather passive social media usage habits decrease the willingness to disseminate disinformation (H6b).

Design
To test our hypotheses, we collaborated with an international survey company that interviewed representative population samples in six countries using equivalent methods (Belgium (Flanders) = 1,063; France=1,255; Germany=1,019; Switzerland (German-speaking regions) = 1,251; UK=1,380; US=1,038). We selected these countries because they systematically differ in terms of their resilience factors based on the study by Humprecht et al. (2020). Moreover, the chosen countries varied in the way they managed the pandemic (e.g., regarding infection rates, confirmed deaths, and test rates) according the COVID Performance Index (Lowy Institute, 2021). When investigating the relationships between our independent variables and our main dependent variable (resilience), we use country comparisons as a robustness check to determine which relationships hold in various situations. This helps us determine the conditional factors that are most advantageous for resilience.
The preregistered survey was conducted in April and May 2020, a period in which lockdowns were being imposed in most countries under study and disinformation related to the pandemic was being widely disseminated . The survey research company drew quota samples from their online access samples according to gender, age, and education. Respondents accessed the survey via a link that was distributed by the polling company. After indicating their informed consent, participants completed the pretreatment section of the survey, which included standard demographics, measures for news media trust, news consumption, and social media use. In a second step, each participant was presented with three news articles that included various false claims. To make the user experience as realistic as possible, the articles were presented as posts on social media. To keep the influence of the source cue constant, we used a fictional and therefore untainted name as the source for all presented articles ('news.com'). To increase ecological validity, false claims were taken from fact-checking websites and were manipulated in a way that matched our research interests (see Figure  A1, Appendix). The article on COVID-19 claimed that the virus is a bioweapon. The article on immigration claimed that the media conceals outbreaks of violence in refugee camps. The third article, on climate change, claimed that environmentalists leave rubbish in public parks after their protests.
After reading the posts, participants were sent to the posttreatment survey, which included measures for the dependent variable, manipulation checks, and voting intention. In a last step, respondents were debriefed. The average time to complete the survey was 18.6 min (SD = 7.09). After removing careless respondents identified by a quality failure question, their response times and response patterns, we obtained an adjusted sample of 7,006 respondents.

Measures
Nonreaction to disinformation. The willingness to react to disinformation on social media was measured with three different items. On a 7-point Likert scale (1 = very unlikely, 7 = very likely) participants indicated how they would usually react to the presented manipulated post: 'like the post', 'share the post' or 'leave a (positive or negative) comment'. Since the three types of reactions were highly correlated (Cronbach's α = .73), we combined them into an additive 'no reactions index' (M = 2.56, SD = .86) in which high values indicate the unlikelihood of a reaction. In line with expert recommendations on how to deal with disinformation (e.g., Lewandowsky et al., 2021), the index captures a certain form of resilience, namely to avoid disinformation and deny it publicity.
Populist party support. We asked about respondents' voting intention in the next national election and presented them with a list of all parties currently represented in their national parliament. Based on expert ratings by Rooduijn (2019) and Timbro (2019), we identified the populist parties relevant for this study (see Table I, Appendix). No populist party was identified for the United States. Respondents had to rate their voting intention for a populist party on a 7-point Likert scale (1 = 'very unlikely', 7 = 'very likely'; M = 2.58, SD = 2.19).
Trust in News Media. We used established scales  to measure on a seven-point scale (1 = 'strongly disagree', 7= 'strongly agree') whether respondents think (i) they can trust most news most of the time and (ii) whether they think they can trust most of the news they consume most of the time. The two items were combined into an index (Cronbach's α = .86; M = 4.40, SD = 1.4).
Alternative media use. To measure alternative media use, we asked participants which news outlets they had used in the last month to inform themselves about political and societal issues, such as migration, the environment or health. Participants were presented with a list of the 10 most used news outlets in their country (based on the Digital News Report, Newman et al., 2019), including public, established, and alternative news media (see Table A1 in the Appendix). Respondents were informed that it did not matter which device they used (laptop, smartphone, television, etc.) or whether they accessed the media brands online, offline or via social media. The use of alternative media was measured on a seven-point scale (1 = 'never', 7 = 'daily'; M = 1.47, SD = 1.1).
Broader main news use. Using our media usage measurement, we created a new variable and included all respondents who indicated they used at least two mainstream news media outlets at least monthly (M = .48, SD = .50) Social media news use. We asked participants how frequently they used different social media platforms for following the news, i.e., Twitter, Facebook, Instagram, and YouTube on a scale ranging from 1 to 5 (1 = never, 2 = less often, 3 = monthly, 4 = weekly, 5 = daily; M = 2.67, SD = 1.47) Social media activity. Participants were asked four questions regarding their social media use, namely, whether they respond to the personal posts of friends or family members with likes, whether they read news updates about political and societal issues, and whether they comment, like or share political and societal news (1 = 'never', 7 = 'very often'). These items were subsequently combined into an index (Cronbach's α = .76; M = 3.61, SD = 1.46).
Controls. We added the variables of gender, age, and education to our questionnaire in addition to our focal variables.

Results
Our main research goal is to determine which factors explain an individual's decision to pause and not propagate disinformation. The extent of resilience referred to here is probably not equally strong everywhere. Therefore, RQ1 asked how citizens from different countries vary in their willingness to disseminate disinformation. To answer this question, we ran analyses of variance (ANOVAs) to estimate the mean differences across the countries under study, broken down by disinformation topic (see Table 1). The results show significant country differences regarding reactions to the coronavirus article (F(5,7000) = 19.332, p = .000), the immigration article (F(5,6997) = 15.944, p = .000), and the climate change article (F(5,6998) = 13.564, p = .000). Swiss citizens punished these three cases of disinformation with the highest degree of indifference and disregard. In contrast, U.S. citizens showed the greatest inclination to participate in the spread of these false reports (Table 1). The relative outsider role of the U. S., which is evident in the comparison of the national user populations calculated here at the individual data level, confirms, in principle, a country typology for resilience, which Humprecht et al.
(2020) sketched out with a rougher, secondary statistical analysis at the macro level. Hypotheses 1-6 concern different predictors to explain the variance in the degree of resilience. An overview of the distribution of all explanatory variables per country can be found in the Appendix (Table AII). To compare their relative weights, the predictors were tested in multivariate ordinary least squares (OLS) regression with clusterrobust-standard errors for each country, and with the 'no reactions index' as the outcome. Independent variables were entered in three blocks: (1) controls and political predictors, (2) media predictors, and (3) social platform predictors. Table 2 presents the results of the various models for each of the six countries. We examined the relationships between different predictors and the intention to refrain from liking, sharing, and commenting on false information (dependent variable). Our first hypothesis H1 postulates a negative relationship between extreme ideology and resilience against disinformation. Our results show that extreme ideology significantly predicts the outcome in most countries (FR: β = -.07, p⩽ .05; DE: β = -.14, p⩽ .001; UK: β = -.10, p⩽ .001; US: β = -.11, p⩽ .01), with the exception of Belgium and Switzerland. For the United Kingdom and United States, the effect disappears in the second model when media variables are added. Based on these findings, we partly accept H1.
Our second hypothesis stated that support for populist parties is negatively related to disregarding disinformation (or positively related to propagating it). Our analysis indeed shows a negative relationship between populist voting intentions and resilience in all studied countries in which a populist party was identified (BE: β = -.20, p⩽ .001; FR: β = -.11, p⩽ .01; DE: β = -.17, p⩽ .001; CH: β = -.10, p⩽ .01; UK: β = -.29, p⩽ .001). In France and Switzerland, however, the effect is no longer significant in the second model. Based on these results, we can accept H2.
The literature behind H3a expects trust in news to have a positive influence on resilience, since established news media often correct false information. Our analysis did not show this relationship between trust and resilience in any country. We found non-effects for four countries and weak negative effects for Germany (β = -.08, p⩽ .05) and the United Kingdom (β = -.09, p⩽ .01). However, the significant relationship for Germany disappears in the third, most comprehensive model. Thus, we must partly reject H3a for the most part.
Next, we expected that the use of public service broadcasting (PSB) would strengthen resilience against disinformation (H4). However, our analysis shows only a weak relationship for France (β = .06, p⩽ .05), indicating that French users of the national PSB were less willing to disseminate disinformation. Based on this result, we can accept the hypothesis only for France, but reject it for all other countries.
Following common assumptions in the literature, we expected that a broader main news diet, i.e., the regular use of more than two established news media outlets, would contribute to higher resilience against disinformation. However, we only found a    923 Note: OLS = ordinary least squares. Cells include robust standard errors (clustered on respondents) and betas, *p < .05; **p < .01; and ***p < .001. significant effect for Germany and the U. S. in the respective third models (DE: β = .07, p⩽ .05; US: β = .11, p⩽ .01), indicating that diverse media use increases the resilience of frequent social media users in these countries. Moreover, contrary to our hypothesis, a negative relationship was found for the United Kingdom in the second model, but it disappeared in the third, most comprehensive model. Based on these findings, we only partly accept the hypothesis for Germany and the United States.
Finally, we formulated relationships between the use of social media (H6a), activity on social media (H6b) and willingness to disseminate disinformation. These hypotheses are based on findings showing that social media facilitate the spread of disinformation and that very active users are more likely to share information, including disinformation, with their network (Rossini et al., 2020). Our results support these assumptions, as they show that across all countries, both the use of and activity on social media are negatively related to disregarding disinformation. In other words, the more frequently respondents use social media and the more active they are when using it, the less likely they are to refrain from spreading disinformation (BE use : β = -.14, p⩽ .001; BE act : β = -.16, p⩽ .001; FR use : β = -.13, p⩽ .001; FR act : β = -.19, p⩽ .001; DE use : β = -.23, p⩽ .001; DE act : β = -.14, p⩽ .001; CH use : β = -.15, p⩽ .001; CH act : β = -.10, p⩽ .001; UK use : β = -.23, p⩽ .001; UK act : β = -.20, p⩽ .001; US use : β = -.23, p⩽ .001; US act : β = -.20, p⩽ .001). Thus, we accept both H6a and H6b.
The explained variance increased substantially for all countries through the stepwise inclusion of different sets of variables related to the political, media and social media platform spheres. Model fits slightly differed across countries (BE: R 2 = .24; FR: R 2 = .19, DE: R 2 = .30; CH: R 2 = .24; UK: R 2 = .37; US: R 2 = .33).
Next, we will discuss our results against the background of previous research on the dissemination of disinformation. In addition, we will elaborate on some unexpected results, and finally, we provide recommendations that can be derived from our study.

Discussion
The aim of this study was to examine whether various country-level conditions can act as a kind of firewall protecting a society against the mass circulation of disinformation. We operationalized all relevant independent and dependent variables at the individual level and measured the presumed influence relationships by means of a six-country survey. The overall objective of our study was to determine the explanatory power of factors that are related to three spheres of influence (politics, media, and social media platforms) to understand why citizens decide to ignore or disregard disinformation (= demonstrate resilience) or why they like, share or comment on it (= promote its publicity further). In analyses within the six countries under study, we were able to determine the relative weight of the various influencing factors; in analyses between the countries, we were able to show, by means of parallel comparison of significant variable relationships, which explanations of resilience can be generalized.
The results confirm several of our assumptions and show that resilience factors are suitable for predicting individual behavior in different countries. Based on these results, we argue that the spread of disinformation depends on the particular social, media, and political environment. Across all countries, populist party support, alternative media use, and the use of and activity on social media significantly lowered the intention to refrain from giving additional exposure to disinformation. These findings led us to accept H2, H3b, H6a, and H6b. These results are consistent with those of previous work that argues that social media and its use for political interests increase the problem of disinformation (Marwick & Lewis, 2017). These findings have important democratic implications, particularly regarding the question of how citizens can build resilience against disinformation. They show that in different countries different groups of society are vulnerable to disinformation. Especially during the current 'infodemic', it seems important to equip these vulnerable citizens with weapons to combat attempts at deception and manipulation, which can have serious consequences (Islam et al., 2020).
Furthermore, our study showed that factors related to specific political and media environments are of great importance in the dissemination of disinformation. In more consensus-oriented democracies, such as Switzerland, extreme ideology seems to be a less relevant driver of the spread of disinformation. In other countries, however, where the rise of populist ideology is tearing open social rifts, extreme ideology diminishes the resilience of citizens. This is particularly evident in the United States, where the willingness to spread disinformation is highest overall. This finding implies that combating disinformation should involve broader structural measures tailored to each country. In this context, our control variables are also relevant. They show, for example, that while younger individuals tend to be more resilient in most countries, older individuals are more resilient in the U.S. and U.K. Moreover, higher education seems to strengthen resilience only in Belgium, France and Switzerland-while in other countries this effect is most likely cancelled out by the effects of extreme ideology and populist support. Gender also seems to play a role, as across all countries, women are more resilient to disinformation than men. Men are generally slightly more willing to share, like or comment on news. These differences and similarities can provide important impulses for policy makers in the fight against disinformation and should also be included in transnational strategies, for example by the European Union.
Contrary to our initial assumption, we only found a limited relationships between resilience and PBS use, a diverse media diet, and trust in news media with resilience. These findings are possibly related to the fact that during our data collection period in the midst of a pandemic, citizens in all the studied countries had a greater need for information which resulted in higher levels of trust in established news media . However, it is also possible that the relationship between media-related factors and disinformation dissemination is more complex than presented in previous research. As Strömbäck et al. (2020) point out, more research is needed to understand the reciprocal relationship between media trust, media use, political polarization, misperceptions and increasing knowledge resistance. In-depth country studies are needed to examine these relationships at different levels and with regard to different issues and actors. For example, more subtle differences become apparent in our study when the three topics (coronavirus, immigration, and climate change) are considered separately. In the U.S., for example, a positive relationship between trust and resilience (β = .07, p⩽ .05) was observed for the topic of immigration. Therefore, we assume that media trust is of particular importance for salient long-term topics. In addition, future research should look more closely at the direction of the relationship between trust and disinformation. For example, Valenzuela et al. (2021) examined the long-term relationship between trust and disinformation and found that lower levels of trust in the media were related to higher levels of disinformation; however, they found no evidence that media trust protects users from disinformation. Similarly, Ognyanova et al. (2020) found a negative effect of misinformation on media trust arguing that cynical and tabloid-style coverage and attacks on the press can erode trust in the news media. Thus, we conclude that the causal flow between media trust and disinformation has not yet been sufficiently explored.
Naturally, this study has limitations. First, we used three specific examples of disinformation to test our hypotheses. However, we cannot generalize our results to other forms and topics of disinformation e.g., messages that are directly disseminated by political actors or friends and family. Second, we used a fictional and untainted news website ('news.com') as the source for our disinformation articles; this may have given the impression of being a reputable, trustworthy news outlet. A third limitation is related to our research design. Some argue that using digital trace data instead of survey data is a superior strategy to measure social media reactions, such as liking, sharing, or commenting (Jungherr, 2015). Supporters of this view consider digital trace data to be more objective as it does not rely on participants' memories and allows for measuring immediate reactions (Otto et al., 2021). However, recent studies have shown that taking a demand-side perspective, as we did in our study, can yield important results regarding the importance of message, sender, and recipient characteristics in explaining reactions to social media content (Blassnig & Wirz, 2019;Rossini et al., 2020). Fourth, we have focused on Western European countries and the U.S., and our results are therefore not transferable to countries with completely different political, media, and economic environments. Further research should therefore broaden the scope and focus to include countries, such as Asian countries, in order to examine how different communication cultures or forms of social media activity influence the willingness to disseminate disinformation. Fifth, it has been pointed out that the U.S. stands out from the countries under study because of its political system, its strong societal and political polarization and its high number of social media users, among others reasons (Benkler et al., 2018). Therefore, more countries that share these characteristics should be included in future comparative analyses in order to understand the underlying conditions that distinguish these countries from each other and from other countries in, e.g., Europe. Finally, our analysis revealed significant differences between Switzerland and other European countries regarding the willingness of citizens not to spread disinformation. This finding is surprising in light of previous research that sees Switzerland as being close to other European countries, e.g., with regard to its political information environment (Esser et al., 2012). To understand the underlying reasons for these differences, in-depth case studies are needed.
Comparing the intention to ignore and not propagate disinformation across topics and countries, this study reveals distinct patterns of behavior that indicate resilience against disinformation. In some countries, such as Switzerland, resilience to disinformation seems to be more pronounced than in other countries, most importantly in the U.S. Our study offers some initial solid insights into factors that can explain resilience at the individual level across different countries and will hopefully provide impetus to further develop policy initiatives such as those aimed at strengthening news literacy.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the Swiss National Science Foundation and the Research Foundation -Flanders (Grant No. 100017L_182253).