Misinformation, political preferences, and cognitive traits: a look at the Brazilian electorate

To contribute to the expanding literature on misinformation in contexts beyond developed countries, this article seeks answers to questions such as: Who are the individuals more susceptible to misinformation? What determines such a vulnerability? This research employs a sample of Brazilian voters surveyed between May and June 2019, and it concludes that partisan preferences alone do not explain susceptibility to misinformation. The impact of partisanship on proneness to misinformation is moderated by analytical ability, need for cognition, and political knowledge. People with high levels of these attributes tend to be more capable of evaluating information through the lens of their political beliefs. Furthermore, the findings suggest that individuals with the highest trust in professional journalism may be less prone to misinformation.


Introduction 2
Information disorders are a threat to democracy because they can undermine its foundations.If under normal conditions, the average person cannot evaluate political issues in line with their real interests (Converse, 1964;Gilens, 2001), imagine how distorted these judgments might be when wrong beliefs are predominant.For instance, how bad will people's evaluations of the public policies or institutions typical of the democratic state be if they make them based on misconceptions?People do not think or process information evenly (Chaiken, 1980;Kahneman, 2013).Therefore, in circumstances of viral propagation of inaccurate information, which individuals would be more likely to believe falsehoods?What would determine such a vulnerability?
There is an increasing number of studies on this topic in political science, especially considering the case of the United States.Findings have consistently shown that 1 Universidade de Brasília (UnB), Instituto de Ciência Política (IPOL).Brasília (DF), Brasil.E-mail: <caroliveira.work@gmail.com>. 2 First of all, I would like to express my gratitude to my wife, Alessandra, and our children, Maitê and Noah, for their unwavering support in every circumstance.I am also immensely thankful to Professor Mathieu Turgeon for his exceptional guidance and invaluable advice throughout my doctoral journey.Additionally, I extend my heartfelt appreciation to my friend, Alex Hudson, for his meticulous review and insightful comments on this study.Moreover, I want to acknowledge Meta for generously funding this research through a Facebook Foundational Integrity Research grant, with Principal Investigator Mathieu Turgeon from the University of Western Ontario.Lastly, I would like to extend my sincere thanks to the anonymous reviewers for their precious contributions that greatly enhanced the quality of this article.partisanship plays a significant role in shaping people's susceptibility to misinformation (Kuklinski et al., 2000;Berinsky, 2017;Min, 2021;Motta, 2021).Nevertheless, a point to be raised is what would the picture be like in a circumstance in which political parties are of little relevance, and the party system is not strongly institutionalized -as is the case in Brazil (Carreirão;Kinzo, 2004;Samuels, 2008;Carreirão, 2014;Samuels;Zucco, 2018)?
Much of the partisan effect on the proneness of misinformation is associated with the level of polarization (Zollo et al., 2015), and social media can exacerbate such antagonism (Tucker et al., 2018).However, in ideological terms, there is no empirical confirmation of this kind of polarization in Brazil (Borges;Vidigal, 2018).On the contrary, the evidence suggests an affective polarization (Fuks;Marques, 2023), with Brazilians showing sympathy or aversion toward the Workers' Party (PT -Partido dos Trabalhadores).This is commonly referred to as petismo versus antipetismo3 (Samuels;Zucco, 2018).
At least in the last three decades, the political disputes in Brazil have been all about the PT, supporting or rejecting it.This refusal of the PT can also be taken as what is called negative partisanship in the American case (Abramowitz;Webster, 2018).It is worth mentioning that the level of animosity toward a specific political party likewise impacts attitudes and behaviors (Caruana;McGregor;Stephenson, 2015;Haime;Cantú, 2022).
Specific worldviews can exert a powerful influence on cognition and behavior (Koltko-Rivera, 2004).Still more, endorsing misinformation has a lot to do with previous motivations, especially political ones (Lewandowsky et al., 2012).Thus, from these situational and dispositional environments, this research primarily tests whether partisan preferences (petismo and antipetismo) determine the susceptibility to misinformation among the Brazilian electorate.
In this framework, one must consider that some cognitive traits reinforce the personal ability to link new data to their political tendencies (Carl, 2015).An example is people with high analytical capacity (Kahan et al., 2017).Another might be the need for cognition because such an attribute is associated with deep thinking (Cacioppo;Petty, 1982).Yet, more strongly linked to political reasoning is the degree of political knowledge (Kam, 2005).People knowledgeable politically are greatly inclined to analyze new information under the lens of their political likings (Taber;Cann;Kucsova, 2009;Miller;Saunders;Farhart, 2016).Do these variables moderate the impact of partisan preferences on vulnerability to misinformation?Still in the individual's perspective, the level of trust in information sources plays a relevant role in information processing (Nyhan, 2020).With a constant and overwhelming influx of political content on the internet, distinguishing between truth and falsehood requires significant effort.The origin and credibility of information sources heavily influence how people evaluate them.In this environment of limitless information sources, traditional journalism is generally perceived as more trustworthy than other sources (Daniller et al., 2017).Consequently, are the individuals who place a high degree of trust in professional journalism less likely to fall prey to misleading information?This article addresses these issues using a national non-probabilistic sample of the Brazilian electorate surveyed between May and June 2019 (Turgeon et al., 2019).More than eight thousand adults over the age of 18 were recruited, being people from all social strata who had the habit of accessing social media at least once a month -and agreed to participate in the research.
Statistical analyses of these data indicate that partisan preferences (petismo and antipetismo) in isolation do not explain someone's tendency to be misinformed.However, the effect of partisan preferences on the susceptibility to inaccurate information seems to be moderated by analytical capacity (measured by the Cognitive Reflection Test -CRT), need for cognition (NFC), and the amount of political knowledge (PK).If someone is more analytical, high in NFC, and politically knowledgeable, they will be more prone to evaluate information based on political bias.It suggests that for individuals with such a profile, the political usefulness of content may matter more than its precision.Besides, results indicate that those who rely more on professional journalism can be less inclined to be misinformed.

What is misinformation?
Misinformation can occur due to failures in processing information -when someone processes factual content mistakenly (Marsh et al., 2016) -or because of endorsement of information disorders 4 (rumors, fake news, disinformation, mal-information).A misinformed person firmly holds wrong beliefs, acts politically based on them, and resists facts when they contradict their beliefs (Kuklinski et al., 2000).In summary, misinformation arises when individuals accept inaccurate information, leading to the adoption of delusive convictions that can strongly impact their attitudes and behaviors.
In a political setting, misinformation undermines democracy as it impedes citizens from being adequately informed (Pereira;Harris;Bavel, 2018).Still worse, misinformed individuals are unaware of the mistakes made by political leaders (Hochschild;Einstein, 2015).It is a personal lapse that is highly likely to spread among the masses and proliferate in environments with a high level of political polarization (Zollo et al., 2015).According to Zollo and co-authors, the more misinformed people are, the more fertile the atmosphere for radical antagonism to run rampant.
An aggravating factor is that misinformation has a continuous effect, regardless of debunking attempts (Lewandowsky;Van Der Linden, 2021).Once set in memory, its implications for individual inferences are almost unquenchable, even if there is a provision of truthful information at some point afterward (Johnson;Seifert, 1994;Lewandowsky, 2020).Something almost subversive is that misinformed people can make decisions contrary to their own interests and are more politically active (Hochschild;Einstein, 2015).
What makes misinformation a much more complex problem is that misinformed people rarely are willing to accept correction (Nyhan;Reifler, 2010).For many, admitting correction is conflated with compromising their identity (Clarke, 2002) and threatens their feeling of belonging (Jost, 2017).Often, accepting correction of an erroneous belief will mean breaking long-term social bonds (Hochschild;Einstein, 2015).There are always specific motivations inducing ratification or refusal of newly received messages (Kunda, 1990).Generally, it is not the truth that matters, but how pieces of information confirm pre-established views (Nickerson, 1998).
In this dynamic, political preferences play a substantial role (Van Bavel;Pereira, 2018).This makes sense because partisanship has to do with affective settings (Greene, 1999); it is a psychological link with some political party (Dalton;Weldon, 2007).
As a consequence, if inauthentic information is relevant to political disputes, a partisan person can use it as weaponized or incendiary rhetoric, even if they know the information is false (Hochschild;Einstein, 2015;Lewandowsky, 2020).Individuals favor their own groups in the face of rival causes (Greene, 1999).Partisanship can have damaging outcomes that are challenging to reverse, even with copious amounts of factual information to eventually discredit it (Bartels, 2002).Political preferences reduce the cost of information (Jones;Hudson, 1998), working as a shortcut to facilitate interpretation in a way consistent with prior beliefs.
But the extent to which political parties are relevant can vary considerably from one country to another.In some countries, they are vital; in others, they are irrelevant.In general, parties matter more in developed nations, where party systems are vigorously institutionalized (Mainwaring;Torcal, 2006).There is also a denial orientation, even in countries where identification with a party has been guiding political life.According to Abramowitz and Webster (2018), it is negative partisanship, or, in the American scenario, "individuals line up against one party instead of affiliating with the other" (p.119).This party anti-orientation also can influence citizens' political attitudes (idem).
Negative partisanship has been an important variable that politically orients voters in Latin America (Brazil included), and it is a personal characteristic that distinguishes them from nonpartisans (Haime;Cantú, 2022).Brazil, especially, has been a case where political parties have not been prominent for too many people (Carreirão;Kinzo, 2004;Samuels, 2008;Carreirão, 2014).To be sure, the Brazilian party system is not stable, and is not a long-term institution (Fleischer, 2007).Conversely, there is a claim that since the beginning of re-democratization in the 1980s, the political disputes in Brazil have been centered on the approval or rejection of the Workers' Party (Samuels;Zucco, 2018).Such polarization cannot be classified as ideological, but as an affective5 antagonism (Fuks;Marques, 2023) -at least until 2018, when Jair Bolsonaro ascended as a conservative leader (Borges;Casalecchi;Rennó, 2020).Despite this movement around President Bolsonaro, Borges, Casalecchi and Rennó also point to anti-Workers' Party attitudes, known in Brazil as antipetismo, as a foundational component of this new rightwing orientation in Brazil.
Petismo versus antipetismo can be a sui generis example of this kind of political dispute because antipetismo is not a political party but a heterogeneous and emotional phenomenon, in which the common tie is the rejection of anything linked to the PT.And, as discussed, endorsing information disorders also has much to do with affective connections.This is a relevant variable to explain one's susceptibility to misinformation.Therefore, bearing in mind the general importance of the PT for Brazilian political contests and this potential to ignite more emotive disputes, this article tests the following hypothesis: (H1) support or rejection of the PT explains susceptibility to political misinformation.Depending on how some political issue is attached to the PT, more or less likely to be misinformed a partisan person will be.For instance, it is expected that an antipetista6 is more likely to be misinformed about the Bolsa Família Program (BFP) -an asset of petismo.It could be that the feeling of disapproval toward the PT induces a wrong belief about this policy.By contrast, a petista7 is less likely to be misinformed about the BFP.
Of course, people are not homogenous, and many may have better cognitive skills to distinguish false content from factual or empirical truth.There are, for example, people with a greater "need for cognition" (NFC), or "chronic thinkers" (Cacioppo;Petty, 1982); such a trait surely shapes the way individuals process information.Someone high in NFC has the intrinsic demand to think more, feels happy with challenges to their mind and prefers to face complex situations with reasoning (Cohen;Stotland;Wolfe, 1955;Suedfeld;Tetlock, 2001).According to Cohen and co-authors (idem), when there is not enough data to fill the need for in-depth scrutiny, frustration emerges, possibly blocking the effects of NFC.
Partisan individuals' judgments, including their support of or opposition to public policies, can be affected by NFC (Bullock, 2011).Bullock's findings show that a Democrat with a high NFC (compared to one with a low) favors liberal policies more than conservative ones.The reverse is true of Republicans.In terms of public security policies, high NFC is negatively associated with punishment among less authoritarian people, however positively among very authoritarian ones (Tam;Leung;Chiu, 2008).Thus, it is reasonable to think that people who are high in NFC may be the ablest to evaluate information through the lens of their political preferences.From this perspective, I outline hypothesis (H1a): the need for cognition will moderate the effect of partisan preferences regarding the vulnerability to misinformation.For instance, if a person is a petista and high in NFC, she/he might be less likely to be misinformed about the BFP because she/he is more capable to see that policy as an asset of the PT.The opposite can happen for an antipetista who is also high in NFC, exactly because of their expected ability to better link the BFP to the PT government.
Another personal trait frequently studied in information processing is analytical capacity (measured by the Cognitive Reflection Test -CRT).It is an attribute related to decision-making and impulsiveness in the evaluation of information (Thomson;Oppenheimer, 2016).Individuals with low CRT scores interpret new messages based mainly on intuition, while those with high scores can perceive information nuances (Frederick, 2005).This cognitive peculiarity, though, reinforces ideological links (Carl, 2015;Kahan et al., 2017).Considering this, I propose the following hypothesis (H1b): analytical capacity similarly will moderate the effect of partisan preferences regarding the vulnerability to misinformation.
Evidence suggests that political sophistication greatly matters for shaping political opinion (Kam, 2005).Thinking exclusively in terms of misinformation, this variable tends to impact susceptibility to it (Nyhan;Reifler, 2010).Consequently, I can presume an additional hypothesis (H1c): political knowledge will moderate the effect of partisan preferences regarding the vulnerability to misinformation.Politically knowledgeable people also manage to have a greater capacity to evaluate information in harmony with their political predilections (Taber;Cann;Kucsova, 2009).
Accepting or not a piece of information likewise depends on trusting the source.
This can be the strength or weakness of the message and affects how much someone will pay attention to any content (Halpern et al., 2019).For this reason, I include trust in professional journalism as a control variable.The idea is to test a research question of whether trust in journalism can prevent individual susceptibility to misinformation.
Journalism has been losing credibility, but compared to other sources of information, it tends to be more trustworthy (Daniller et al., 2017).

Data
This article uses an original database (Turgeon et al., 2019) with data about sensitive issues in Brazilian politics.I chose themes that have been highly relevant in the political debates in the last two decades: the direct conditional cash transfer program known as Bolsa Família (BFP), the voting machine system (VM), and the tax burden (TX).
These topics were chosen due to the ardent controversies underlying them, mainly the BFP and the VM.
Despite being a public policy already accepted and defended even by opponents of the PT, before its consolidation the BFP was a constant target of lies and misinformation.
An example is the rumor that circulated on social media saying that poor women with kids intentionally were seeking to be impregnated in order to receive more money from the BFP.
The controversy over electronic voting machines does not have the same long history.The issue gained notoriety when then candidate for president Aécio Neves, who faced with defeat in the 2014 presidential race against Dilma Rousseff, challenged the validity of the electoral process and its outcome.Such conspiracy theories about the electoral process gained even greater force in recent years as former president Jair Bolsonaro has similarly cast doubts on Brazilian elections and its technologies.
In the 2018 election, a rumor circulated on social media that the electronic voting machines used at the polls had been manufactured in Venezuela and were rigged to produce an unfavorable result for Bolsonaro.The Superior Electoral Court (TSE, in Portuguese) issued a note refuting this false information (Available at: <https://bit.ly/3imnooX> 8).Therefore, I believe that such political/emotional connotations and the broad reach of these topics (BFP and voting machines) are good subjects for such analyses.The issue of tax burden is a more neutral matter that I do not expect to be of great consequence to the present work.
Data collection was carried out between May and June 2019.The sample was nationally representative and non-probabilistic, with more than eight thousand Brazilians over the age of 18 recruited as respondents.People from all social strata who used social media at least once a month and agreed to participate in the research were selected.
Participants answered questions about their media consumption, political knowledge, political attitudes, party identification, and demographics.Moreover, there were batteries of questions about the CRT and that aimed to measure NFC.

Methodology
To test the hypotheses, I used OLS regression models 9 .The dependent variable is misinformation.Specifically, part of the participants (n = 2346) received a correct piece of information regarding the BFP: "Families that received Bolsa Família benefits had fewer children than average in Brazil between 2003 and 2013" 10 .A second subset (n = 2267) read factual information about voting machines: "Various countries other than Brazil use electronic voting systems" 11 .Finally, the third subset (n = 2279) read the following factual text about the tax burden: "Compared to the Organization for Economic Cooperation and Development (OECD) member countries, Brazil is the country that charges the least in 8 Accessed on: July 21, 2019. 9The descriptive statistics for the variables in the model are in taxes" 12 .
After reading the factual information, they answered the question: "in your opinion, how likely is this to be true?"The response options were "not likely at all", "not very likely", "somewhat likely", and "very likely".Those who answered "not likely at all" were considered more prone to be misinformed, and those who said "very likely" were considered less inclined.Finally, to facilitate analyses, I added up these answer options and recoded them as a continuous variable, from 0 (informed) to 1 (misinformed).
In other studies about misinformation, especially those in which correction strategies are tested, researchers present false information (often made up for the study) and check the acceptance of correction (Nyhan;Reifler, 2010).My goal is quite different: to evaluate people's knowledge of actual sensitive political issues.That is why I provided only exact information and measured how much people believed in its authenticity.It is a tactic like that used by Kuklinski et al. (2000).
Undoubtedly, numerous individuals might not be misinformed but, instead, be unfamiliar or unaware of the facts.However, due to the controversies around the BFP and voting machines, especially the former, false information about these subjects has circulated widely.Furthermore, many question or flat-out reject correct information because of party ties or for any number of dispositional and situational reasons (Lewandowsky et al., 2012).These conditions, as discussed, can lead to misinformation.
Consequently, I believe that, for the purposes of this work, the strategy adopted here to measure propensity to misinformation is adequate.
I mainly estimate the effect of partisan preferences (petismo and antipetismo) on the chances of a Brazilian voter being misinformed.Nevertheless, I presume that the effect of partisanship on the likelihood of someone being misinformed will vary due to analytical capacity (measured by the CRT), the need for cognition (NFC), and the degree of political knowledge (PK).For this reason, I include interactions between these variables and those partisan inclinations in the models.
Given the specificity of Brazilian politics, I added three variables: antipartisans (people who dislike parties); partisans (those who identify or affiliate with a party); and nonpartisans (individuals who do not have a preferred party).I do not have many expectations regarding these variables.
It is essential to remember that my sample focuses on people who frequently access social media.Thus, among them, it is possible that those who trust more in professional journalism are less likely to be misinformed.To account for this, I also added trust in professional journalism as a control variable.
To measure partisan preferences, I operationalized some dimensions of partisanship and created a variable with three categories: antipetismo, petismo, and other.The respondents were asked to express their feelings about some political parties 13 on a 10-point scale of support to those parties, where 0 means "I don't like it at all", 5 "I neither like nor dislike it", and 10 "I like it a lot".Likewise, they indicated their preferred party and the party they would never vote for.
Then, to classify petismo, I selected the participants who indicated that the PT was their preferred party, rated it higher than 5, and systematically and consistently ranked it above the others.For antipetismo, those respondents who reported that they would never vote for the PT, gave the party a score less than 5, consistently ranked the party below the others, and are not antipartisan.The remaining respondents were classified as other.
Concerning the other variables associated with partisan preferences, I considered as antipartisan those respondents who stated: "I don't have a preferred political party" and scored all the parties in the ten-point scale equal to 0 (I don't like it at all).On the other hand, I coded as partisan those who ranked all the parties above 5 and said that there is no party for which they would never vote.Nonpartisans are those who indicated NOT having a party of preference and who ranked all the parties above 5.
The variable analytical capacity is an adaptation to the Portuguese of five questions used in the CRT 14 .To operationalize it for the analyses, I added up all the answers and created a continuous range from 0 (intuitive) to 1 (analytical).
To measure the NFC, I have adapted to Portuguese five questions 15 from the battery created by Cacioppo and Petty (1982) to assess how much of a "chronic thinker" a respondent is.For example, if someone likes or does not like to solve complex problems; about the level of pleasure got from thinking; preference for daily or long-term projects; and if someone likes or dislikes having responsibility for handling situations that require lots of thinking.To get a unique measurement, I also added up all the answers and created a continuous range from 0 (low NFC) to 1 (high NFC).
For political knowledge (PK), I presented five questions about general political subjects, and participants who answered properly were taken as politically knowledgeable.
For example, those who said the Supremo Tribunal Federal (Brazilian Supreme Court) is responsible for ensuring the constitutionality of a law.As with the other cases, I created from all the answers a continuous measure varying from 0 (little PK) to 1 (high PK). 13 The parties who launched candidates to President in the 2018 electoral races include: Partido dos Trabalhadores (PT), Movimento Democrático Brasileiro (MDB), Partido Social Liberal (PSL), Partido da Social Democracia Brasileira (PSDB), Partido Socialismo e Liberdade (PSOL), and Partido Democrático Trabalhista (PDT), in addition to Democratas (DEM) because it had the presidencies of the Chamber of Deputies (national legislature) and Senate. 14(1) If you are running in a race and you pass the person in second place, what place will you be in?(2) A farmer had 15 sheep and all except for 8 died.How many sheep are still alive?(3) Carlos' mother has three children.The first is named April.The second is named May.What is the third named?(4) How many cubic meters of dirt are in a hole that is 3 meters deep, 3 meters long, and 3 meters wide?(5) A bat and ball cost $1.10 in total.The bat costs $1 more than the ball.How much does the ball cost? 15The questions are available in detail in the Appendix.
Regarding the level of trust in journalism, the question was how much the respondents trust the news about politics produced by professional journalists.The answer is a four-item Likert scale that I similarly recoded to form a continuous scale of 0 (never trust) to 1 (trust).

Results
Firstly, the percentages of respondents considered more inclined to misinformation are in Figure 1 (note that the closer to 1 on the scale, the higher the probability of misinformation).
Paying attention to the extreme values in the scale, around 21% of participants tend to be more misinformed about the BFP (n = 2346, M = 0.6, SD = 0.2).Regarding the voting machine system, almost 15% of participants are more likely to be misinformed (n = 2267, M = 0.4, SD = 0.3).Finally, concerning the tax burden, the proportion of people prone to misinformation is much higher: close to 60% (n = 2279, M = 0.8, SD = 0.3).In general, the mean values show a steeper inclination to misinformation about the level of taxes paid by Brazilians and the BFP, respectively.In the case of the voting machines, the distribution shows a minor variance.

Figure 1 Percentage of Misinformation among the Respondents
Source: Self-elaboration.Dataset prepared for this project (Turgeon et al., 2019)

Misinformation TX
The BFP is a public policy that is already a key part of Brazilian politics and a constant target of political vitriol, and the subject of heated debates and clashes.The controversy over electronic voting machines, however, does not have a history as extensive as the BFP.As mentioned, it emerged when the candidate for president in 2014, Aécio Neves, cast doubt on the election results.More recently, the rumors gained intensity when former president Jair Bolsonaro started to similarly speculate or conjecture that election results were unreliable and, allegedly, even rigged.
Concerning taxes, less rigor might be necessary to interpret the high percentage of propensity to misinformation.Taxation is labyrinthine subject in Brazil, and even more complicated when compared to another country.It is difficult for most citizens to know the actual tax burden in their own country let alone assess or compare taxation in Brazil and other countries.Furthermore -except for those identified as the richest or wealthiestfew people are interested in this issue (Alm;Mcclelland;Schulze, 1992).The vast majority are more likely to be entirely uninformed and not simply misinformed.
Obviously, it is possible to criticize the measure of misinformation used here.
However, when considering the trend pointed out by these results, the possibility that so many people are likely to be misinformed about substantial matters of Brazilian politics is a very concerning sign.So, it is necessary to advance in understanding the peculiarities of such a predisposition to misinformation.
In this sense, in terms of proclivity to misinformation, the expectation is that petistas will be less inclined to be misinformed about the BFP compared to antipetistas (as the program is one of the most notable and celebrated achievements of the PT's governance).The same result is expected for the voting machine coefficients (the PT has won many elections with this system and never doubted its reliability).Likewise, I believe that the effect of these partisan preferences regarding the chances of someone being or not being misinformed will be reinforced by the NFC, CRT, and PK.In isolation, I expect that those people who trust professional journalism the most will be less susceptible to misinformation than those who do not trust it.
Table 1 shows the results for the OLS regressions.The model for the BFP suggests that respondents who identify as petista are more likely to be misinformed compared to their antipetista peers.This is completely against what would be expected in terms of partisan preferences.As the BFP is an asset to the PT, if identification with petismo or antipetismo (without considering other factors) had any effect on the susceptibility to misinformation regarding that policy, the inclination should be in the opposite direction.Of course, sympathy with the PT does not mean that someone will be informed about policies linked to petismo, and there is a risk that my measure for petismo had not captured well how petista a person is.On the other side, considering the heterogeneous character of antipetismo, many antipetistas can be informed about the BFP.It is not possible to comment on the other models because of the lack of statistical significance.Source: Self-elaboration.Dataset prepared for this project (Turgeon et al., 2019).
Individuals who hold an antipartisan stance, expressing a lack of affinity for any political party, are more prone to misinformation concerning the BFP compared to those who are not antipartisans.Considering Fuks, Ribeiro, and Borba's (2020) theorization, antipartisanship in Brazil has intensities, and part of the phenomenon has to do with the level of tolerance of the PT.It is possible that for those who are most intolerant, some political subjects intrinsically associated with petismo (or the PT's platform) might fall on deaf ears, which may be the case of the BFP.It is counterintuitive or contradictory to expect antipartisans who are highly intolerant of the PT to admit the accuracy or truthfulness of a piece of information considered an asset to petismo (future research would do well to contemplate this).Regarding the models for VM and TX, there is no statistical significance.
The coefficients for partisan respondents are statistically significant only in the model for VM and show that they tend to be less inclined to misinformation.As the great majority of the political parties do not cast doubt on the electoral system, this result is not a surprise.So, if someone is a partisan, it is more likely that they will not be misinformed on this matter because they might have greater trust in and access to state institutions.
Non-partisans are more likely than their partisan peers to be misinformed about the voting machine system and taxation.Maybe not having a preferred political party makes someone less knowledgeable on such themes.Normally, parties are an important shortcut for processing political issues.It is possible that the electoral system and taxes are not relevant subjects for someone who does not identify with any party.Concerning the BFP, the coefficient is not statistically significant.
Keeping the rest constant, results from the VM model indicate that individuals with the highest CRT scores are more likely to be misinformed.Analytical individuals are best equipped to evaluate political issues through the lens of their preexisting beliefs, which tends to impact how they interpret political content.As for the other regressions, I cannot make any claims, as the effects were not statistically significant.
In relation to the need for cognition, note that the results indicate that people with more NFC are less likely to be misinformed about the BFP.As there is a lot of information available about the BFP, this element may fill the chronic thinkers' need to think more; thus, they may be significantly more likely to be informed about this issue.In the other models, no statistical significance was observed.
Finally, more political knowledge tends to increase the perspective of misinformation related to VM and TX.As discussed, PK is a strong variable for political thinking.This individual attribute therefore influences information processing.Perhaps accuracy is not the most important feature when politically knowledgeable people are evaluating a piece of information.It is possible that they consider how useful the content is for their political disputes.For the BFP model, the tendency is the same but lacks statistical significance.
Conversely, for all the models and keeping all other things constant, relying more on professional journalism implies a lower chance of misinformation.One can expect it really happens because, as mentioned, when compared to other sources of information, traditional journalism is considered more trustworthy.
Concerning the interactions, in the regression referring to BFP, keeping all the other things constant, as the CRT, NFC, and PK scores increase, respectively, the expectation of misinformation for participants identified with petismo decreases compared to those with antipetismo who are also high in those attributes.None of the other models achieves statistical significance.
In the case of the BFP, then, the coefficients suggest that the effect of these partisan preferences on susceptibility to misinformation is moderated by analytical abilities, the need for cognition, and political sophistication.These individual characteristics worked as if they were bringing things back to normality.After all, on average, it is more expected for antipetistas to espouse misinformation and misguided beliefs about the BFP than for a petista, because it is a public policy significantly associated with the PT.On its own, the effect of these partisan preferences was not in this direction.
Nevertheless, to improve the interpretation of interactions and increase the certainty of the results, I calculated conditional slopes16 for the BFP and VM models 17 .In doing so, I tested the effect of partisan preferences (petismo and antipetismo) on the propensity to misinformation by setting CRT, NFC, and PK at the mean and at -1 and 1 standard deviation from the mean.
When CRT is 1 standard deviation below the mean (0.33), its impact on the effect of partisan preferences regarding misinformation about the BFP is not statistically significant.Contrarily, when setting CRT on the mean, the petistas are less likely to be misinformed than their antipetista peers (Est.= -0.03,S.E = 0.01, t val.= -3.4,p = 0.00).
Towards the variable PK, this tendency repeats.The effect is statistically significant only when the PK is set on the mean (0.49) and 1 SD above it, respectively (Est.= -0.03,S.E = 0.01, t val.= -3.4,p = 0.00; Est.= -0.05,S.E = 0.01, t val.= -3.9,p = 0.00).In other words, petistas with high scores of PK tend to be less prone to misinformation about the BF program than antipetistas with similar profile.
Consequently, there is evidence to support the hypotheses that the effect of partisan preferences (petismo and antipetismo) on the susceptibility to misinformation about the BFP is moderated by analytical capacity (CRT), NFC, and PK.It is naturally expected that the Workers' Party's (PT) partisans will be less likely to be misinformed about the BFP than antipetistas.However, this is only the case when petistas with the highest scores of CRT, NFC, and PK are compared with antipetistas who have the same profile.Now, I calculate the conditional slopes for the interactions of the VM model.When the CRT scores are set at 1 SD below the mean (0.32), the petistas are less likely to be misinformed than the antipetistas (Est.= -0.05,S.E = 0.01, t val.= -3.4,p = 0.00).If they are set on the mean, the tendency is similar (Est.= -0.04,S.E = 0.01, t val.= -3.3,p = 0.00).However, when setting the CRT score at 1 SD above the mean, there is no statistical significance (Est.= -0.02,S.E = -0.02,t val.= -1.3,p = 0.36).This suggests that it is not possible to say that this effect is not distinguishable from zero, although the inclination is in the expected direction.Thus, I cannot accept with certainty that high analytical capacity shapes the effect of petismo and antipetismo on the likelihood an individual is misinformed or holds misguided beliefs about the voting machine system and unfounded accusations of fraud.
Petistas high in NFC are less inclined to be misinformed than antipetistas also high in NFC.
For political knowledge, when the slope is set at -1 SD, petistas are less likely to be misinformed about the voting machine system than antipetistas (Est.= -0.03,S.E = 0.02, t val.= -2.01,p = 0.04).The same trend occurs in the mean (Est.= -0.03,S.E = 0.01, t val.= -2.45,p = 0.01); however, despite developing in the expected direction, it is not statistically significant at 1 SD above (Est.= -0.03,S.E = 0.02, t val.= -1.58,p = 0.18).In conclusion, the evidence suggests that petistas who have average political knowledge are less likely to be misinformed about the VM than antipetistas with the same profile.At 1 standard deviation above the mean, the estimate is not statistically significant, which impedes us to have certainty of the effect.

Discussion
The results reported in this article reveal a considerable propensity for misinformation among Brazilian voters concerning the BFP, voting machine system, and tax burden.However, differently from hypothesized, partisan preferences in isolation do not explain such a vulnerability.Purely being identified with the petismo or antipetismo does not imply a higher or lower probability of someone being misinformed on those politically sensitive subjects.In contrast, the findings suggest that the effect of partisan preference (petismo and antipetismo) on the susceptibility to misinformation is moderated by the degree of analytical capacity, the need for cognition, and political sophistication.
In general, for the great majority of people, being misinformed has nothing exclusively to do with liking or disliking the PT.In other words, among people who have identify or are sympathetic to the PT or those who reject it, only those who are similarly highly analytical, chronic thinkers, or politically knowledgeable seem to evaluate information as expected to someone who has those partisan preferences.
These findings reinforce the claim that the historical political polarization around the PT is not ideological for the mass public but might be an emotional antagonism.If ideological presuppositions structured information processing in a systematic way, one should expect that someone identified as petista would not to be misinformed about the BFP and the voting machines.The opposite would be expected for antipetistas.But this does not happen in isolation.
Theories about vulnerability to information disorders predict that the rhetoric implicit in various political disputes dictates information usefulness.The most important consideration is not the accuracy of a piece of information, but how "good" it might be in the daily political contests.It is expected that political preferences matter to the political content processing.If it worked for the Brazilian political setting, it would not be strategic for an antipetista to endorse and for a petista to reject the authenticity of precise information concerning the BFP.In this hypothesis, there would be a contradiction in terms of political pragmatism.
In a more specific approach, there is no evidence that only partisan preferences (petismo) or negative partisanship (antipetismo) affect the susceptibility to misinformation among Brazilians.It does not work as it does in the United States, where political rivalries between the Democrats and Republicans usually impact the tendency to be misinformed (Clemm von Hohenberg, 2023).New research on this should test if the effect of partisan preferences is different for someone who merely likes a party and for those who are affiliated and engaged in partisan activities.
Hence, a more encompassing contribution of this article lies in its presentation of evidence, which proposes that, even within an environment of affective polarization, merely aligning with a specific side of this antagonism is insufficient to account for Brazilians' susceptibility to misinformation.Only partisans displaying elevated scores in crucial cognitive traits, coupled with political sophistication, demonstrate a propensity to interpret information through a partisan perspective.This framework should be taken into consideration in future studies.
Another contribution is related to possible strategies to face the problem.For all the regression models, keeping everything else constant, the coefficients suggested that people who have greater trust for professional journalism are less likely to be misinformed.
It is an encouraging picture as a mechanism to curb misinformation.It is promising once it points to a filter when processing information.In a moment where many people access information on the Internet, maybe some strategy to face misconceptions is to highlight that the newsgathering process was carried out in a professional way.Of course, depending on the level of polarization, many might reject the source credibility, even if it is an established and trustworthy news company.
Despite the theoretical and empirical plausibility of these results, there are some limitations to this study.For instance, the themes addressed (especially the BFP and voting machines) are closely related to PT's terms in the Presidency of the Republic.Moreover, they are lively subjects in recent political clashes involving the PT as a central focus.Such a setting may have impeded verifying the Brazilian electorate's susceptibility to political misinformation.
Another weakness, as discussed earlier, is the measure of misinformation itself.
There are, in truth, many false stories concerning the Bolsa Família and the voting machines circulating within Brazil.Many people can receive those fabrications as if they were correct information and become misinformed.However, it could also be that there is no misinformation, but only ignorance about the topics.Also, as postulated, politically motivated reasoning interferes in the processing of information.As a way to correct this deficiency, to get some filter, it could be possible to design an experiment capable of measuring in advance the respondents' attitudes regarding the political objects under analysis.
Table A of the Appendix.