Perceptions of electoral fraud in the 2006 presidential election in Mexico

The Mexican party system can be considered fairly institutionalized compared to the rest of the party systems in the region (Mainwaring Reference Mainwaring2018). Although Mexico has only been considered a democracy since 2000, its political parties have existed for decades. The authoritarian successor party (PRI), the center-right PAN, and the center-left PRD opposition parties were key actors during Mexico’s democratic transition—negotiating electoral reforms and channeling social demands—from 1988 to 2000. The 2006 presidential election constituted the first major political crisis in Mexico’s young democracy. In that election, the PAN candidate, Felipe Calderón, narrowly defeated Andrés Manuel López Obrador, supported by the PRD and a coalition of leftist parties, in a highly contested race (Klesner Reference Klesner2007). Calderón’s margin of victory was only about 0.58%. Although the IFE, international observers, and most parties in Mexico—except for López Obrador’s PRD—confirmed that the electoral process had been free and fair, López Obrador denounced the election as fraudulent and contested the results (Ugues Reference Ugues2010). The Mexican Electoral Tribunal ordered that the IFE undertake a partial ballot recount but did not find any significant or systematic bias against López Obrador. Because there was never a full recount of the votes as demanded by López Obrador’s team—and some administrative controversies emerged during the campaign (Schedler Reference Schedler2007)—his campaign delegitimized the election. ^{Footnote 1} While it is unclear exactly how any fraud supposedly took place, since López Obrador and his campaign team changed their accusations multiple times (e.g., from “cybernetic” fraud to allegations that there were thousands of hidden votes or that votes were incorrectly counted; see Aparicio, Reference Aparicio2009, for a review), the accusations triggered a significant postelection crisis.

The accusations about electoral fraud became one of the most polarizing issues in post-2006 Mexico politics. ^{Footnote 2} In fact, as many political observers argue, election denialism became a founding myth of MORENA and the lopezobradorista movement, eventually playing a central role in Lopez Obrador’s rhetoric to mobilize his voters (Bruhn Reference Bruhn and Mudde2012; Castro Cornejo, Beltrán, and Ley Reference Castro Cornejo and Ley2020; see also Espino Reference Espino2021). While the country experienced a long history of systematic electoral fraud during the PRI’s hegemonic era (1929–2000), the IFE was a product of the 1996 electoral reform in which the PRI and opposition parties—the center-right PAN and the center-left PRD—created an independent electoral institution to organize national elections. Since then, Mexican elections are widely recognized as free and fair (Eisenstadt Reference Eisenstadt2007). Until 2006, the IFE was widely trusted by both winners and losers of the elections (Moreno Reference Moreno2022). It was not until the 2006 postelection crisis that a partisan bias emerged. After that election, on average, PAN and PRI partisans displayed high levels of trust while AMLO voters (PRD partisans between 2006–2015 and MORENA partisans beginning in 2015) reported low levels of trust. This partisan gap began to fade after 2018, when AMLO finally won the presidency on his third attempt. By 2021, the gap between losers and winners’ trust of the IFE shrank considerably: between 60 percent and 70 percent of the Mexican public now trust the National Electoral Institute (Instituto Nacional Electoral, the successor of the IFE).

As Bruhn (Reference Bruhn and Mudde2012) argues, the 2006 presidential election exposed that Mexico remained an unconsolidated democracy. López Obrador denounced the results as fraudulent without proof of irregularities, organized massive protests, refused to accept the outcome of the election, and declared himself the “legitimate President of Mexico.” As Bruhn (Reference Bruhn and Mudde2012) argues, the rules of democratic competition were not universally accepted, adhered to, and defended by all actors. Since then, López Obrador has consistently claimed that a political mafia in Mexico “stole” the presidency from him and “kidnapped” the country, particularly during his 2012 and 2018 presidential campaigns. While other studies have focused on the perceptions of a corrupt elite in Mexico and their implications on voters’ electoral behavior (Castro Cornejo, Ley, and Beltrán Reference Castro Cornejo and Ley2020), the following sections explore why some voters believe that there was fraud in 2006 even when that belief contradicts the best available evidence and even though more than fifteen years have passed since the election.

This perception, in fact, remains a strong grievance for a significant proportion of Mexican voters. According to an original survey conducted during the 2018 presidential campaign, 24 percent of the electorate believe that it is “very true” and 24 percent of voters consider it “somewhat true” that election fraud took place (question wording: “How true is it that there was fraud in the 2006 presidential election when Felipe Calderón won against López Obrador? Absolutely true, somewhat true, somewhat false, absolutely false”). In response to an open-ended question about how the election was stolen in the 2006 presidential election, 18 percent of voters reported that the votes were incorrectly counted or that the electoral counting system was manipulated, 15 percent reported that ballots were manipulated, another 15 percent that ballots were stolen, 14 percent cited vote buying, and 5 percent declared that the IFE was “bribed” (29 percent did not know). These open-ended answers mirror the accusations that López Obrador’s team raised against the IFE back in 2006 (Aparicio Reference Aparicio2009; Ugues Reference Ugues2010). Like other political events that have been studied in the literature—such as birtherism (Jardina and Traugott Reference Jardina and Traugott2019) or the existence of weapons of mass destruction in Iraq (Nyhan and Reifler Reference Nyhan and Reifler2010)—this perception of a fraudulent election has proven resilient and remained a polarized topic of public debate, even after the 2018 presidential election, when López Obrador finally won the presidency.

Why do some voters believe in events that contradict the best available evidence?

Past research on electoral integrity has studied the general public’s attitudes toward electoral fraud by focusing on institutional and individual variables. For example, administrative performance (Bowler et al. Reference Bowler, Brunell, Donovan and Gronke2015) and the independence of the electoral management body, as well as contextual factors such as countries’ levels of democracy and corruption (Birch Reference Birch2010), explain variations in perceptions of electoral integrity. Moreover, a level playing field for all contestants increases citizens’ confidence and supports the idea that institutions follow the rules (Birch Reference Birch2008). A second body of work considers perceptions of electoral fraud as a product of individual-level variables such as race (Wilson and King-Meadows Reference Wilson and King-Meadows2016), sociodemographic characteristics (Norris Reference Norris1999), political sophistication (Karp, Nai, and Norris Reference Karp, Nai and Norris2018), citizens’ direct experience with elections (Atkeson and Saunders Reference Atkeson and Saunders2007), their interaction with poll workers (Cantú and García Ponce Reference Cantú and García-Ponce2015), and underlying political predispositions Edelson et al. Reference Edelson, Alduncin, Krewson, Sieja and Uscinski2017) such as partisanship (Kernell and Mullinix Reference Kernell and Mullinix2019)—which plays a prominent role in the literature. For instance, depending on the elections’ results, partisanship adjusts voters’ confidence in the electoral procedures. Partisan identification has an amplifying effect for winners and losers (Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug2005), whereby an electoral defeat diminishes the legitimacy of the electoral process among party supporters (e.g., on Mexico, see Cantú and Ponce Reference Cantú and García-Ponce2015; Ugues and Medina Vidal Reference Ugues and Vidal2015; Monsiváis Reference Monsiváis-Carrillo2023).

The literature on the winner-loser gap, or the “sore loser” hypothesis, argues that voters who support a losing party or candidate differ from those who support a winning party or candidate: electoral losers hold more negative opinions about the elections’ integrity (Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug2005; Hernandez Huerta and Cantú Reference Hernández-Huerta and Cantú2022) and think there has been manipulation of the results (Sances and Stewart Reference Sances and Stewart III2015). However, while we know that the winner-loser gap shapes postelection attitudes in both long-standing democracies and young democracies (Carreras and İrepoğlu Reference Carreras and İrepoğlu2013; Maldonado and Seligson Reference Maldonado, Seligson, Norris, Frank and Martinez i Coma2014; Cantú and Ponce, Reference Cantú and García-Ponce2015; Ugues Reference Ugues2010), comparative studies on election fraud, as Kernell and Mullinix (Reference Kernell and Mullinix2019) argue, rarely integrate theories of preference formation. It is important to analyze how winners and losers systematically differ in their postelection attitudes. It is also necessary to understand the factors that shape those perceptions, particularly in a context—young democracies—where it has not been systematically studied. While there is evidence that motivated reasoning (Kernell and Mullinix Reference Kernell and Mullinix2019) shapes voters’ interpretations of election integrity and vote counting in long-standing democracies like the United States, we have less evidence of similar behaviors in Latin America and Mexico in particular.

Motivated reasoning theory argues that citizens are often determined to defend their prior beliefs and seek out and accept information consistent with them (Kunda Reference Kunda1990). In politics, voters are motivated to interpret information through a partisan lens (Bartels Reference Bartels2002; Zaller Reference Zaller1992), which makes them reject information that goes against their long-standing beliefs (Taber and Lodge Reference Taber and Lodge2006). While it is possible for voters to update their beliefs in a Bayesian way and carefully revise their opinions as they consume more information (Gerber and Green Reference Gerber and Green1999), this research argues that voters are particularly likely to see political events through a partisan lens in polarized settings after highly contested elections (Bartels Reference Bartels2002). This means that voters do update their beliefs, but the process is biased by partisan motivations: voters protect their in-group by choosing interpretations that rationalize their partisan beliefs, failing to consider that their partisan team might have lost the elections. Moreover, partisan-motivated reasoning affects voters’ interpretations of factual information (Gaines et al. Reference Gaines, Kuklinski, Quirk, Peyton and Verkuilen2007), altering voters’ perceptions of election integrity and, consequently, making them more prone to view the elections as fraudulent. This behavior has no instrumental benefits; rather, it is based on expressive benefits rooted in partisanship.

The literature on opinion formation has analyzed different episodes of political controversy viewed through a partisan lens. For example, Gaines and colleagues (Reference Gaines, Kuklinski, Quirk, Peyton and Verkuilen2007) study the Iraq War and show that both Democrats and Republicans updated their factual beliefs as conditions changed but interpreted the same facts quite differently. The literature has also analyzed perceptions that have persisted among the American public throughout the years, such as those about climate change (Hart and Nisbet Reference Hart and Nisbet2012) and birtherism (Hartman and Newmark Reference Hartman and Newmark2012) among others. For instance, survey researchers, beginning in 2008 and as recently as 2017, have provided evidence that a significant proportion of the American public believes that former president Barack Obama was born outside the United States and was raised as a Muslim (Jardina and Traugott Reference Jardina and Traugott2019). As Jardina and Traugott (Reference Jardina and Traugott2019) highlight, birtherism perceptions have persisted despite news media outlets’ efforts to correct the record and even after the release of two versions of Obama’s birth certificate. Like other pieces of misinformation (e.g., on healthcare reform, see Berinsky Reference Berinsky2015; on weapons of mass destruction, see Nyhan and Reifler Reference Nyhan and Reifler2010), such perceptions remain in the public debate for years and persist despite a lack of evidence.

Regardless of the specific issue, these studies show that facts or events do not speak for themselves; people must still interpret them (Gaines et al. Reference Gaines, Kuklinski, Quirk, Peyton and Verkuilen2007; Castro Cornejo Reference Castro Cornejo2023a), leading voters to interpret those facts and their implications differently. When directional motivations are more salient (e.g., accusations of election fraud in a polarized party system like Mexico), respondents are likely to process information in a manner that is consistent with their partisan preference rather than maximizing accuracy (Flynn, Nyhan and Reifler Reference Flynn, Nyhan and Reifler2017; Taber and Lodge Reference Taber and Lodge2006). Consistent with motivated reasoning theory, these voters do not judge the integrity of the election based on available evidence but choose partisan interpretations to explain the outcome of the election. Thus, the main hypothesis of this study, on partisan motivated reasoning, is the following:

To understand why voters believe in misperceptions, this research argues that some individual-level factors—affective polarization, information levels, and conspiratorial thinking—can exacerbate or diminish voters’ tendency to engage in motivated reasoning, for example, making them more likely to believe in electoral fraud even if it contradicts the best available evidence. If partisanship is important, affective polarization would be even more consequential given the in-groups (MORENA) and out-groups (the PAN and the PRI) that have recently formed in Mexico (e.g., lopezobradorismo or MORENA versus the “PRIAN” as López Obrador commonly refers to both parties) to explain political events (Castro Cornejo Reference Castro Cornejo2023b). Affective polarization is the tendency to view copartisans positively and opposing partisans negatively (Druckman and Levendusky Reference Druckman and Levendusky2019; Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012), which makes ordinary partisans dislike and distrust one another and treat each other with prejudice (Iyengar and Westwood Reference Iyengar and Westwood2015). This means that polarization is not primarily rooted in ideology but identity (Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012; Mason Reference Mason2018). Given polarized partisans’ tendency to develop in-group and out-group identity (Tajfel Reference Tajfel1981), they resort to motivated reasoning to view their party as acting with good intentions ( Edelson et al. Reference Edelson, Alduncin, Krewson, Sieja and Uscinski2017; Lodge and Taber Reference Lodge and Taber2013) and others as acting immorally and engaging in fraudulent behavior (Miller, Saunders, and Farhart Reference Miller, Saunders and Farhart2016). In other words, the more polarized partisans’ affective responses toward opposing parties are, the more influence these negative feelings are likely to have on their perceptions, which can promote the endorsement of inaccurate beliefs, causing them to ignore or misinterpret relevant evidence.

Affective polarization increasingly influences voters’ behavior and attitude formation in old and new democracies. For instance, the 2016 US presidential election represents an example of strong affective polarization rooted on identity (Mutz Reference Mutz2018). Donald Trump activated latent grievances in the electorate, particularly based on sexism (Valentino, Wayne, and Oceno Reference Valentino2018), racial resentment (Sides, Tesler, and Vavreck Reference Sides, Tesler and Vavreck2019), and white identity (Jardina Reference Jardina and Traugott2019). In fact, animus toward the Democratic-leaning social groups strongly predicted support for Trump’s candidacy (Mason, Wronski, and Kane Reference Mason, Wronski and Kane2021). In Brazil, Jair Bolsonaro’s victory was rooted in strong sentiment against the Workers’ Party (i.e., anti-petismo; Samuels and Zucco Reference Samuels and Zucco2018) as well as disenchantment with political institutions amid corruption scandals involving major parties. Bolsonaro was able to mobilize anti–Workers’ Party voters (Amaral Reference Amaral2020) by relying on a radical conservative rhetoric (Renno Reference Renno2020). In Mexico, Andrés Manuel López Obrador’s historic victory in 2018 was an outcome of a strong rejection of the country’s major parties, the PAN and the PRI, which were increasingly perceived as the same political alternative (Castro Cornejo Reference Castro Cornejo2023b).

Conspiratorial thinking can lead voters to engage in motivated reasoning, as some voters not only reject relevant evidence but also believe that pieces of evidence are concealed. Conspiratorial thinking refers to the belief that powerful people’s secretive and malevolent actions explain certain events (Vermeule and Sunstein Reference Vermeule and Sunstein2009). Voters who engage in conspiratorial thinking are more likely to believe unsubstantiated accusations that small groups are working in secret for their own benefit and against the common good (Uscinski and Parent Reference Uscinski and Parent2014). These voters are particularly resistant to information that challenges these misleading narratives (Berinsky Reference Berinsky2015; Nyhan and Reifler Reference Nyhan and Reifler2010; Uscinski and Parent Reference Uscinski and Parent2014). As Edelson et al. (Reference Edelson, Alduncin, Krewson, Sieja and Uscinski2017) argue, perceptions of election fraud are a good example of a belief that can be exacerbated by conspiratorial thinking. In the end, a conspiracy theory is not necessarily “wrong.” It is still a theory and requires evidence to either support or reject. However, conspiracy theories are troubling because, more often than not, they interpret the lack of confirming evidence as support. Since evidence of electoral fraud is normally hidden and fraud is typically accomplished by the machinations of powerful groups, the lack of evidence demonstrates the fraud—it shows just how widespread and concealed it really is (Edelson et al. Reference Edelson, Alduncin, Krewson, Sieja and Uscinski2017). This means that voters with strong predispositions to conspiratorial thinking are more likely to view elections with skepticism, expect fraud, and develop explanations for the results they are unhappy with.

This is particularly important when it comes to the 2006 presidential election in Mexico. The accusation of fraud in that election, in fact, fits the definition of a conspiracy theory. For example, after Lopez Obrador challenged the election results, the Mexican Electoral Tribunal ordered a partial recount of the votes (in precincts more likely to report mistakes) instead of a complete recount across the entire country—as López Obrador’s team demanded. While the partial recount did not find any systematic bias against López Obrador—in fact, Calderon’s margin of victory slightly increased—this decision prompted some accusations from López Obrador’s team that the tribunal was hiding the evidence of the fraud. In this scenario, consistent with the literature on conspiracy theories (Enders et al. Reference Enders, Uscinski, Seelig, Klofstad, Wuchty, Funchion and Stoler2021), losers argued that the corrupt elites and electoral institutions worked together in secret to violate electoral procedures for their own benefit.

Levels of information are also expected to exacerbate motivated reasoning. From a normative perspective, we expect that voters with higher levels of information will be less likely to believe in rumors without solid evidence (Bolsen, Druckman, and Lomax Cook Reference Bolsen, Druckman and Lomax Cook2014). For instance, some studies have found that trust in elections depends on the degree to which voters have political knowledge (Norris, Garnett, and Gromping Reference Norris, Garnett and Gromping2020). Nevertheless, studies on misinformation challenge the idea of higher levels of knowledge as a panacea. In fact, they have found that higher levels of information exacerbate motivated reasoning (Miller, Saunders, and Farhart Reference Miller, Saunders and Farhart2016), because highly informed voters tend to be more partisan and are highly motivated to protect their beliefs, which makes them more likely to engage in motivated reasoning (Lodge and Taber Reference Lodge and Taber2013; Miller, Saunders, and Farhart Reference Miller, Saunders and Farhart2016). As a result, this study proposes an additional set of hypotheses:

Finally, the last hypothesis studies the persistence of perceptions, which is particularly relevant when studying election denialism. If motivated reasoning affects voters’ updating of their beliefs, voters are likely to believe in events that contradict the best available evidence and resist efforts to reduce voters’ belief inaccuracy, such as fact-checking, even if this event happened a long time ago. While some literature has found that voters’ can update their beliefs as they engage with new information (Wood and Porter, Reference Wood and Porter2018), most studies suggest that fact-checking fails to make respondents’ beliefs more accurate (Garrett, Nisbet, and Lynch Reference Garrett, Nisbet and Lynch2013; Nyhan, Reifler, and Ubel Reference Nyhan, Reifler and Ubel2013) and, in some cases, even strengthens misperceptions (“backfire effect”; Nyhan and Reifler Reference Nyhan and Reifler2010). For example, Nyhan and Reifler’s corrections on the belief that Iraq had weapons of mass destruction failed to reduce misperceptions and even strengthened such views among congenial ideological groups. Based on this discussion, the last hypothesis is the following:

The next section applies motivated-reasoning theory to the results of the 2006 presidential election in Mexico, a political event that continues to activate important grievances among voters and political elites to this day and one that played a key role in López Obrador’s rhetorical strategy during the 2018 election.

Perceptions of fraud in 2006 and the “corrupt elite” during the 2018 presidential election

As previously mentioned, the literature on public opinion highlights that attitudes are not formed in a vacuum: they reflect a marriage between political predispositions and political elites’ communication (Zaller Reference Zaller1992; Druckman Reference Druckman2014). In Mexico’s case, López Obrador’s rhetoric has consistently divided Mexican society between the “people” and a corrupt elite that stole the presidency from him and impoverished Mexico with neoliberal policies and rampant corruption (Bruhn Reference Bruhn and Mudde2012). This accusation began during the postelection crisis in 2006 and remained part of his political rhetoric for the next electoral cycles. In 2015, relations between López Obrador and his party deteriorated markedly after the PRD decided to join the Pact for Mexico (Pacto por México)—with the PAN and the PRI—which sought to create an agreement between political forces to approve various structural reforms in Congress. López Obrador denounced the PRD for betraying its members by joining the same “political mafia” as the “PRIAN,” the term he uses colloquially to conflate the PRI and the PAN. López Obrador resigned from the PRD and cofounded a new party—the National Regeneration Movement, or MORENA. Accusations of electoral fraud in 2006 served as a foundational myth of the new party. In other words, López Obrador has been communicating a clear and consistent message for many years: a corrupt elite stole the presidency from him in 2006.

In addition to clear elite cues, partisanship is also likely to bias voters’ perceptions. Recent literature highlights that partisanship is fairly strong among the Mexican electorate (Castro Cornejo Reference Castro Cornejo2019)—about two-thirds of the electorate self-identify with a political party since the country’s transition to democracy, according to the Mexican Election Study (Beltrán, Ley, and Castro Cornejo Reference Beltrán, Ley and Castro Cornejo2020). The emergence of a new major party, however, transformed the partisan attachments of an important part of the electorate. While partisan loyalties were fairly stable between 2000 and 2015, when priístas formed the largest partisan group, and panistas and perredistas followed in their wake, in 2018, morenistas became the largest partisan group in the country. According to the 2018 Mexican Election Study, 22 percent of the electorate self-identified with MORENA, 17 percent with the PAN, 14 percent with the PRI, 4 percent with the PRD (López Obrador’s former party)—which lost a significant part of its base after MORENA’s appearance in 2015—and 6 percent with smaller parties. In addition, 33 percent self-identified as “independent”—a proportion of voters that has remained fairly consistent since previous elections. In other words, the decay of the Mexican party system has not led to the demise of partisanship—as was the case in other Latin American countries such as Venezuela and Perú (Morgan Reference Morgan2011; Seawright Reference Seawright2012)—but to the development of new partisan loyalties, particularly to MORENA. While identification with MORENA as a party is still evolving (Aruguete et al. Reference Aruguete, Calvo, Cantú, Ley, Scartascini and Ventura2021), recent studies have found that MORENA partisanship moderates voters’ attitudes and behavior, such as their retrospective evaluations of the economy, security, corruption, belief in a corrupt elite, and turnout (Castro Cornejo, Ley, and Beltrán Reference Castro Cornejo and Ley2020). This partisan bias effect is independent of voters’ opinion of López Obrador (Castro Cornejo Reference Castro Cornejo2023a).

While the Mexican electorate has experienced some shifts in party identification, the same data suggests that affective polarization has increased in Mexico. Following survey strategies in American politics (Lelkes and Westwood Reference Lelkes and Westwood2017), Figure 1 reports the feeling thermometer that rates respondents’ opinion towards major parties in Mexico. Affective polarization—the difference between the in-group’s average rating and the out-group’s average rating (on a 0–10 scale)—has increased. Between 2000 and 2015, the average rating of in-groups and out-groups were quite stable. However, after 2015, there is a very important increase in out-party animus. The average rating of the opposing parties was 4.1 in 2000 and had decreased to 2.4 by 2018, almost 2 points on the 0–10 scale (Figure 1). This context makes partisans who self-identify with López Obrador’s MORENA more likely to engage in motivated reasoning, particularly on a highly polarized issue such as the outcome of the 2006 presidential election.

Figure 1. Affective Polarization in Mexico (2000–2018).

Note: Average ratings on the feeling thermometer (0 = very bad, and 10 = very good) among voters who self-identify with a political party.

Source: Mexican Election Study 2000–2018 (Beltrán, Ley, and Castro Cornejo Reference Beltrán, Ley and Castro Cornejo2020).

Observational data

This section relies on an original telephone survey and the 2021 Mexican Election Study. The telephone survey is nationally representative and was conducted by the polling firm BGC Beltrán y Asocs between April 3 and 11, 2018, the first month of the 2018 presidential campaign in Mexico (N = 1,000). The 2021 Mexican Election Study is a nationally representative survey and part of the Comparative Study of Electoral Systems (CSES). It was conducted as a face-to-face survey two weeks after the midterm election (N = 1,800).

To identify the proportion of voters who believe that the 2006 presidential election was fraudulent, both studies rely on the following survey question: “How true is it that there was fraud in the 2006 presidential election when Felipe Calderón won against López Obrador? Absolutely true, somewhat true, somewhat false, absolutely false.” To facilitate interpretation, the variable is recoded from 0 to 1, and the average answer was 0.59 in the telephone survey (13 percent didn’t know or didn’t answer) and 0.63 in the 2021 Mexican Election Study (20 percent didn’t know or didn’t answer). It is important to highlight that in the 2021 survey, this question was included for only half the sample. Table A2 in the appendix (see supplementary material) reports descriptive statistics of each variable included in the models.

To measure partisanship, the study relies on the following question: “Regardless of the party you normally vote for, do you consider yourself panista, priísta, perredista, a supportor of MORENA, or some other party?” To estimate affective polarization, both studies follow survey strategies in American politics (Druckman and Levendusky 2019) and rely on the feeling thermometers: “On a 0–10 scale, where 0 means having a very bad opinion and 10 means having a very good opinion, what is your opinion of the PRI/PAN/MORENA?” From these questions, a new variable estimates the difference between respondents’ favorability of MORENA (in-group attitudes) and the mean favorability of the PAN and the PRI (out-group attitudes). ^{Footnote 3}

$${\rm{AFFECTIVE}}\;{\rm{POLARIZATION}} = {\rm{MORENA\!{\hbox '}s}}\;{\rm{FT}} - ({\rm{PAN{\hbox '}s}}\;{\rm{FT}} + {\rm{PRI{\hbox '}s}}\;{\rm{FT}}){\rm{/2}}.$$

This new variable ranges from –10 to 10. To facilitate interpretation the variable has been recoded from 0 to 1. Both surveys measure voters’ levels of political information by looking at their ability to correctly answer questions about factual information. It uses an additive index based on three general knowledge questions about the Mexican political system: “What are the chambers of Mexico’s Congress? What is the term length of a member of Congress? What is the name of the current governor of your state?” This question was also recoded from 0 to 1 to facilitate interpretation.

As previously discussed, conspiratorial thinking increases the chances that a voter will believe in election fraud. To measure respondents’ susceptibility to conspiracy theories, the telephone survey (2018) relies on an additional additive index based on the following two questions: “In your opinion, how true is it that the Mexican government knew that the 2017 earthquakes were going to happen and refused to warn the citizens? How true is it that worldwide secret conspiracy groups control events such as world wars? Absolutely true, somewhat true, somewhat false, absolutely false.” In terms of the 2021 Mexican Election Study (Beltrán, Castro Cornejo, and Ley Reference Beltrán, Castro Cornejo and Ley2023), the analysis relies on the following question: “As far as you know, is it true or false that the coronavirus exists and it has caused the thousands of deaths that have been registered in Mexico during the present pandemic?” These questions are not related to respondents’ political behavior in order to avoid close conceptual or empirical connection to partisanship and voting behavior or to the dependent variable of this study. ^{Footnote 4}

The upcoming models also include control variables to make sure that third variables are not driving the results. The models control for respondents’ opinion about Andrés Manuel López Obrador to ensure that such attitudes are not driving their belief in electoral fraud in 2006. While we can expect voters’ opinion of López Obrador to be correlated with their belief in the alleged fraud, including the feeling thermometers makes it possible to identify whether partisanship and affective polarization have an independent effect on this study’s dependent variable: “On a 0–10 scale, where 0 means having a very bad opinion and 10 means having a very good opinion, what is your opinion of Andrés Manuel López Obrador.”

The following models also control for voters’ distrust of Mexican elections. As previously discussed, given the history of electoral fraud during the PRI hegemonic regime before the country’s democratization, it is expected that an important proportion of voters are skeptical about elections—even though elections are recognized to have been free and fair when organized by the IFE. For these reasons, the following questions seek to measure such beliefs. The telephone survey (2018) relies on the question “Talking about the presidential elections in Mexico since the year 2000, on a scale from 0 to 10, where 0 means that electoral fraud is not committed and 10 that electoral fraud is committed in Mexico, where would you place the presidential elections in Mexico?” In turn, the Mexican Election Study (2021) relies on the following statement: “In some countries, people think their elections are fair. In other countries, people think that their elections are not fair. Using the following 1–5 scale, where 1 means that the elections were not fair and 5 means that the elections were fair, where would you place the last elections?” These two variables were also recoded to range between 0 and 1 to ease interpretation.

The models also control for left-right ideology (“In politics, people sometimes talk of left and right. Using a scale from 0 to 10 where 0 means the left and 10 means the right, where would you place yourself?”) to ensure that ideology is not driving the results instead of partisanship. Finally, the models also control for voters’ sociodemographic characteristics, such as age, gender, and levels of education.

Table 1 displays the results of the ordinary-least-squares models. ^{Footnote 5} Their dependent variable is the respondents’ probability to engage in election denialism. In both the 2018 survey and the 2021 Mexican Election Study, consistent with Hypothesis 1, election denialism varies across partisan groups (Table 1). MORENA partisans are the most likely to believe in electoral fraud in 2006 (p < 0.01, models 1 and 3). To better understand the substantive impact of partisanship, Figure 2 presents the probabilities of each partisan group. For example, in 2018, while PAN partisans had a 0.37 chance of believing that there was election fraud, MORENA partisans had a probability of 0.82, a 45-percentage-point difference. This comparison is relevant because the partisan groups were the two leading parties during the 2006 presidential election and report the largest differences. The few respondents who self-identified in 2018 with the PRD—López Obrador’s former party (2 percent of the sample)—report a probability of 0.66, which suggests that they partially believe the allegations of fraud even though they do not currently identify with López Obrador’s party. In fact, for the 2018 presidential election, the PRD nominated the PAN presidential candidate as part of the PAN-PRD electoral coalition. Finally, PRI partisans report a probability of 0.47 and independents of 0.60. Interestingly, independent voters’ perceptions are closer to those from MORENA than those from the PAN and PRI. This result is probably an outcome of the strong support that López Obrador enjoyed among independents in 2018. Unlike during the previous two presidential elections in which López Obrador lost the independent vote, in 2018 he finally won a strong majority of independents (Aparicio and Castro Cornejo Reference Aparicio and Cornejo2020). In 2021, the differences are not as large, but the general results are consistent with 2018: MORENA supporters are the partisan group most likely to believe in election fraud. Independents, once again, have an intermediate position between MORENA partisans and out-partisans. In this survey, since less than 1 percent of the sample identified as PRD partisans, the models do not report results for this partisan group.

Table 1. OLS models: Probability of believing that there was electoral fraud in 2006

Note: Standard error in parentheses. Values range from 0 = “not true at all” to 1 = “very true.”

***p < 0.01. **p < 0.05. *p < 0.1.

Figure 2. OLS models on the probability of believing that there was electoral fraud in 2006.

Note: values range from 0 = “not true at all” to 1 = “very true.”

Models 2 and 4 estimate the probability of believing in the 2006 fraud across levels of affective polarization, conspiratorial thinking, and political information in order to understand the conditions under which voters are more likely to engage in election denialism—while controlling for opinions of Andrés Manuel López Obrador and voters’ distrust of elections in Mexico (the results also hold when distrust is excluded from the models; Table A6 in the appendix; see supplementary material). Affective polarization is a strong predictor of believing in events that contradict the best available evidence. Moreover, even though the operationalization of conspiratorial thinking differs in both studies, voters with higher levels of conspiratorial thinking are more likely to engage in election denialism (p < 0.01). Political information is loosely associated with such beliefs (p < 0.10) but only in one of the surveys, and the sign is the opposite direction of the expected one. This is, in fact, consistent with the recent literature on misinformation, which finds that highly informed voters tend to have strong political predispositions and are highly motivated to protect their beliefs. It was expected that respondents’ opinion of Andrés Manuel López Obrador would have an important association with election denialism, but it is statistically significant only in 2018, not in 2021. Distrust in elections, in turn, is statistically significant in both years; in other words, voters who report a general distrust in Mexican elections are more likely to believe in the 2006 election fraud.

Figures 3 and 4 show that only two variables have a consistently strong substantive impact on the belief that the presidential election was rigged: affective polarization and voters’ levels of conspiratorial thinking. For example, in 2018 the difference between the lowest value of affective polarization—that is, voters who have a very positive opinion of the PRI and the PAN and a very negative opinion of MORENA—and the highest value—voters who have a very positive opinion of MORENA and a very bad opinion of the PRI and the PAN—is 42 percentage points (from 0.34 to 0.76). The difference in 2021 is about 30 percentage points. In the case of the voters’ levels of conspiratorial thinking, in 2018 the difference between the highest and the lowest values comes down to 18 percentage points (from 0.50 to 0.68); in 2021 the difference is about 12 percentage points. In other words, affective polarization is the variable that best explains why some voters believe in allegations that are inconsistent with the best available evidence (Hypothesis 2a)—followed by voters’ conspiratorial thinking (Hypothesis 2b)—even when controlling for their opinion of López Obrador, ideology, and socioeconomic variables. Distrust in elections, while statistically significant in both years, does not report a strong substantive impact in 2021.

Figure 3. Probability of believing that there was electoral fraud in 2006 (2018 Survey).

Figure 4. Probability of believing that there was electoral fraud in 2006 (2021 Mexican Election Study).

Experimental data

A survey experiment was included in a second nationally representative telephone survey. The survey was conducted between June 26 and June 29, 2018, the week before the presidential election held on July 1. It was conducted by the polling firm BGC Beltrán y Asocs. It included a sample of 782 respondents randomly divided into two groups. ^{Footnote 6} These groups are balanced across observed covariates (Table A7; see Supplementary material). The survey experiment followed an indirect strategy to fact-check the belief that there was fraud in the 2006 presidential election. The survey randomly assigned a question asking whether respondents were aware of a statement made by the president of the INE (the successor of the IFE). Respondents who were part of the control group were not asked that question. It reads as follows: “Did you know that the president of the National Election Institute, Lorenzo Córdova, recently declared that in Mexico there is no possibility of electoral fraud during presidential elections and that there was no evidence that votes were improperly counted in the 2006 presidential election?”

After reading the vignette, respondents were asked the following question, which constitutes the dependent variable of this section: “On a 0–10 scale, where 0 means that there was no fraud in the 2006 presidential election and 10 that there definitely was electoral fraud, how true is it that there was electoral fraud in the 2006 presidential election when Felipe Calderón defeated López Obrador?” (Table A2 reports descriptive statistics; see supplementary material). Random assignment to the treatment condition guarantees that, on average, respondents exposed to treatment and control groups will be identical in both observable and unobservable characteristics. Any systematic difference between the two groups’ answers to the dependent variable question provides an estimate of how fact-checking the accusation that there was an electoral fraud affects voters’ attitudes. It is worth mentioning that while a potential treatment highlighting that a MORENA politician declaring that there was no fraud in that election would have been stronger (or even López Obrador declaring there was no fraud), it would have lacked external validity since most politicians that belong to López Obrador’s party endorse the misperception that there was fraud. To enhance external validity, INE was chosen as a source of information since it in charge of organizing elections in Mexico.

Table 2 displays the results of the OLS models, in which the dependent variable is the respondent’s belief that electoral fraud took place in the 2006 presidential election. Consistent with Hypothesis 1, the likelihood of engaging in election denialism varies across partisan groups. MORENA partisans are the most likely to believe that there was election fraud compared with PAN, PRI, and PRD partisans. Moreover, consistent with Hypothesis 3, a motivated-reasoning effect took place: none of partisan groups updated beliefs, and voters resisted the information about the 2006 presidential election (Figure 5). Only among PRD partisans was there some belief updating—about 3.6 points on the 0–10 scale—but it is not statistically significant under conventional thresholds (p > 0.10), given the few observations that are part of this partisan group (less than 2 percent of the sample self-identified as a PRD partisan). ^{Footnote 7}

Table 2. OLS models: Probability of believing that there was electoral fraud in 2006

Note: Standard errors in parentheses, DV = measured on a 0–10 scale; Party ID (Reference Category = MORENA)

***p < 0.01. **p < 0.05. *p < 0.1.

Figure 5. Probability of believing that there was electoral fraud in 2006.

Note: Values range from 0 = “not true at all” to 10 = “very true.”

Consistent with studies in advanced industrial democracies, the survey experiment highlights that fact-checking is sometimes unable to shift voter perceptions. Particularly for highly polarizing issues like the belief in election fraud in the 2006 election, a partisan bias is activated to resist such efforts. As previously discussed, López Obrador voters—PRD partisans after the 2006 presidential election and MORENA partisans after 2015—have reported low levels of trust in the IFE after the 2006 postelection crisis (Figure A1 in the appendix), making them less likely to be responsive to information produced by such source. Therefore, it was a difficult case of belief updating. The results among PAN and PRI voters are particularly interesting. Even though they report high levels of trust in the IFE, they also did not update their beliefs. Respondents identifying with MORENA, PAN, and PRI resisted any kind of belief updating on this topic, one of the most polarizing issues in Mexico.