Fiddling while Democracy Burns: Partisan Reactions to Weak Democracy in Latin America

Abstract

Democracy is weakened when citizens and elites do not criticize actions or actors that undermine its principles. Yet this study documents a widespread pattern of partisan rationalization in how elites and the public evaluate democratic performance in Latin America. Survey data show that those whose party controls the presidency consistently express positive evaluations of the current state of democratic competition and institutions even when democracy in their country is weak. This pattern emerges in both mass survey data and among elected elites. These data have a worrying implication: if only the political opposition is willing to publicly acknowledge and sound the alarm when democracy is under attack, public pressure to protect democracy is likely to be dramatically reduced.

In many countries, democracy is under attack. Global indicators track marked declines in the level of democracy, with one recent report concluding that one-third of the world’s population lives in countries where democracy is in decline (Lührmann et al. 2018). In the majority of these cases, the threat to democracy is not from outsiders looking to overthrow it via a coup but instead comes from elected insiders who weaken democracy from within by restricting electoral competition and curtailing the rights of their opponents (Lührmann and Lindberg 2019). In many cases, this also includes leaders taking steps to limit media freedom and otherwise curtail criticism of their rule (Kellam and Stein 2016). Between 70% and 80% of democratic breakdowns since 1990 have resulted from an elected executive consolidating power (Lührmann and Lindberg 2019; Svolik 2019). These breakdowns are not limited to a single region, but have occurred in the Americas (e.g., Venezuela, Nicaragua), Europe (e.g., Russia, Belarus), Africa (e.g., Benin, Tunisia, Tanzania), and Asia (e.g., India, the Philippines). These concerns are also not limited to developing countries; in recent years observers in the United States have expressed worries about whether elected leaders will respect the limits on their office and not politicize state agencies against their political opponents (e.g., Levitsky and Ziblatt 2018; Kaufman and Haggard 2019; Haggard and Kaufman 2021).

In light of these weaknesses, the question becomes whether political elites and the mass public will act in defense of democracy when it is under threat. In an ideal world elites and citizens in countries where the incumbent government is undermining democracy would 1) recognize that government actions have compromised democracy in important ways, 2) express public disapproval of those actions, and 3) vote against and otherwise hold accountable actors who continue to act undemocratically (Lührmann 2021). Yet this ideal process often does not occur. For example, self-interested actors are often willing to accept undemocratic reforms perpetuated by candidates that they otherwise agree with to achieve their preferred policy goals or to ensure that their electoral enemies are defeated (e.g., Ahlquist et al. 2018; Singer 2018; Svolik 2019; Fossati, Muhtadi, and Warburton 2021). This tolerance for undemocratic action creates opportunities for aspiring dictators to restrict political competition and weaken democracy.

In this paper, however, I ask whether citizens and elites will even acknowledge and publicly admit to the breakdown of democracy while it is happening, a necessary precursor to acting to stop that slide. Specifically, I look at whether citizen and elite evaluations of democratic performance in surveys are lower when democracy is weak than when it is strong. I argue that partisan rationalization often leads elites and the mass public who are winning under the current institutional arrangements to express satisfaction with democracy and confidence in election management in countries even when democracy’s objective level is low. As a result, these groups have incentives to ignore and not acknowledge the deterioration of democratic norms and institutions, creating space for incumbent actors to further restrict democratic competition.

I explore these questions using survey data at the elite and mass level from Latin America, a region where democracy’s strength varies significantly both between countries and within them over time. The data show that among both the elites and masses there is a connection between the level of democracy and evaluations of democratic institutions among those who don’t belong to the president’s party. But these groups are the outliers; evaluations of democratic performance by elites who belong to the president’s party or by citizens who voted for the incumbent in the last election have no connection to the level of democracy in their country. In weak democracies, opinions become polarized on the basic questions of whether elections are trustworthy or on whether the country is a satisfactory democracy because winners do not acknowledge that democracy is weak.

These results highlight the difficulty facing many struggling democracies. The unwillingness of electoral winners to criticize democratic abuses by “their side” may create opportunities for aspiring autocrats. In their recent review, Lührmann and Lindberg (2019) argue that “the present reverse wave—starting after 1993—mainly affects democracies, unlike prior waves. What is especially worrying about this trend is that historically, very few autocratization episodes starting in democracies have been stopped short of turning countries into autocracies.” While the implication of these survey data cannot be directly observed, the empirical results analyzed here provide a potential mechanism for why autocratization crises are so rarely arrested as they are happening. When election winners and government supporters in flawed democracies will not publicly acknowledge that democracy is deteriorating, this silence may deepen the crisis of democracy. When only one side sees democracy as being under attack, then attempts at democratic reform may become a partisan issue that the ruling party can dismiss as an attack on their power by self-interested and disloyal actors instead of a general societal project worth pursuing. Democracy needs champions when it is under attack, but there seems to be little evidence, based on their responses to pollsters asking them about the level of democracy, that elites and voters on the winning side are willing to be that champion.

How Do Citizens Evaluate the State of Democracy in their Country?

A large literature tracks how people describe the state of democracy in their country. Some evaluations of democratic performance have little to do with democracy itself. Satisfaction with democracy, for example, tends to be lower among the educated and other groups with high expectations about what democracy should provide (e.g., Norris 1999; Booth and Seligson 2009). Satisfaction with democracy’s performance is also often buoyed when the economy is strong (Wagner, Schneider, and Halla 2009; Cordero and Simón 2015; Magalhaes 2016; Quaranta and Martini 2016). Yet most studies find that specific support for democracy reflects the quality of democratic governance: public evaluations of the democratic system being higher when institutions are of high quality and perceived as fair (Wagner, Schneider, and Halla 2009; Magalhaes 2016; Christmann 2018) and lower when corruption is common (Evans and Whitfield 1995; Rose, Mishler, and Haerpfer 1998; Bratton and Mattes 2001; Weitz-Shapiro 2008; Linde 2012).

This correlation between the quality of a country’s governance and citizen satisfaction means that dissatisfaction is a bellwether for the kinds of political dissatisfaction that might lead to political change. Sometimes that change can be negative, as falling satisfaction with democracy in regions where democratic norms are already weak can create opportunities for aspiring dictators (Booth and Seligson 2009; Lührmann 2021). Yet criticisms of how democracy is being practiced can also lead citizens and elites to advocate for reforms that would deal with democracy’s perceived shortcomings (e.g., Karp 1995; Dalton 2004; Bowler and Donovan 2007). Elite acknowledgment of undemocratic actions can also trigger a public reaction that further constrains autocratic leaders (Christenson and Kriner 2020). This suggests that both political elites and the public can potentially act as a corrective of democratic decline if they are unhappy about that change.

Yet while citizen evaluations of democratic performance reflect how their country is being governed, they also seem to be shaped by partisan considerations. Citizens evaluate democratic performance through the lens of personal representation: voters who feel represented are more likely to support democracy (Aarts and Thomassen 2008). As a result, citizens who voted for losing political parties are less satisfied with democratic institutions and more ambivalent about the overall desirability of democracy (e.g., Anderson et al. 2005; Anderson and Guillory 1997; Anderson and Tverdova 2003; Listhaug, Aardal, and Ellis 2009; Booth and Seligson 2009; Dahlberg and Linde 2016). Explanations of the gap in system support between election winners and losers have traditionally focused on how defeat shapes the voters who supported a losing party. Electoral losers, it is argued, suffer both the emotional blow of having been defeated and the real policy consequences of living under a government whose views and interests are different from their own.

It is an open question if elites behave in the same way because most evaluations of specific support focus on mass data while the study of elite attitudes usually focuses on their commitment to the democratic norms of competition and tolerance (e.g., Miller, Hesli, and Reisinger 1997; Stevens, Bishin, and Barr 2006). Moreover, differences between how elites and the mass public experience elections may minimize the difference between winners and losers at the elite level. Losing political candidates are more likely to have negative evaluations of the electoral arrangements that resulted in their defeat (Bowler, Donnovan, and Karp 2002, 2006) but elected legislators from non-governing political parties are not really “losers.” They do not control the head of government, but opposition legislators enjoy the perks of office, including some access to resources they can distribute to allies and supporters, some policy influence, and the opportunity to actively build their case against the government through legislative and public debates. Some authoritarian leaders use these benefits of elected office to co-opt moderate opposition elements, incentivizing them to invest in the status quo (Levitsky and Way 2002; Lust-Okar 2005; Gandhi and Przeworski 2007). Thus, the effect of losing control of the government may be mitigated by the perks of personally winning in a way that leads to co-opted public elected officials having a smaller response to losing an election than voters do even when they believe that the election is corrupt. On the other hand, non-governing elites also feel personally the lack of policy influence from being excluded from ruling. Indeed we show later that there is a winner-loser gap in evaluations of democratic performance even among elected officials.

The fear this literature often raises is that political losers might sour on democracy and support its overthrow. As a result, democracy promoters think about steps and reforms that can maintain “the loser’s consent” by empowering them and ensuring them that they will be competitive in the future (Anderson et al. 2005). As a result, the possibility that election winners could be a problem for democracy was largely overlooked until recently. Yet there is evidence that winners often are as large of a problem for democracy as losers are. The impact of corruption scandals on government support is smaller for voters who previously supported the winning party; incumbent co-partisans continue to support their preferred party even after corruption allegations (Anderson and Tverdova 2003). Voters for losing parties in Africa resist elite manipulation of democratic institutions while winners are often “submissive subjects, granting unconditional support to their current leaders” (Moehler 2009, 345) and also support the weakening of national courts who might otherwise check the executive (Bartels and Kramon 2020). In a similar vein, citizens who voted for the winning president in Latin America are more likely to support limitations on free speech while also supporting expansions in presidential power (Singer 2018; Cohen et al. forthcoming). Finally, recent experiments suggest that voters presented with candidates who espouse anti-democratic reforms are likely to support those candidates if they belong to the same party as them and advocate for policies they agree with (Ahlquist et al. 2018; Svolik 2019; Graham and Svolik 2020; Fossati, Muhtadi, and Warburton 2021).

These recent studies point to a dynamic whereby election winners who benefit from the status quo have incentives to resist democratic reforms and to support undemocratic candidates who favor their side. For those with ties to the existing regime, steps that enhance the power of the executive or which may favor them in elections are not evaluated in the abstract but are seen as distributive choices, not as steps that weaken democracy but as ones that enhance the representation of their group. Authoritarians benefit when “their supporters would rather tolerate their authoritarian tendencies than back politicians whose platform these supporters abhor” (Svolik 2019, 23).

While these recent studies explicitly make undemocratic actions salient for respondents evaluating a hypothetical candidate, a fundamental question for how this process plays out in the real world is whether party members making this tradeoff to support a favorable leader undermining democracy perceive that their democracy is being weakened or will admit that the leader is weakening democracy? The existing literature showing that bad institutions and governance lead to lower satisfaction with democracy and its institutions implicitly assumes that all citizens are responding to these threats. Yet a large literature suggests that people often are ignorant of how the world is evolving and, more troublingly, systematically perceive and describe the world in ways that diverge from reality if it benefits their political side. This may lead them to either be ignorant of or to not acknowledge violations of democratic norms by their side as violations.

Examples of motivated reasoning abound. Partisans diverge in their responses to survey questions on topics as varied as whether Iraq had weapons of mass destruction (Jacobson 2010), whether climate change is real (Dunlap, McCright, and Yarosh 2016), how large inauguration crowds were (Schaffner and Luks 2018), and how well the economy is doing (e.g., Kramer 1983; Duch, Palmer, and Anderson 2000; Evans and Anderson 2006). ¹ These differences in belief about the world often persist even after respondents are presented with information designed to correct their misperceptions (Nyhan and Reifler 2010). Individuals are also more likely to know more about events that put their party in a favorable light than those that make their party look bad (Jerit and Barabas 2012). While these gaps have been most commonly documented and studied in contexts where partisanship is strong, they also exist in countries where partisanship is comparatively weak (Carlson 2016; Lee and Singer forthcoming; Fossati, Muhtadi, and Warburton 2021).

While a large literature has documented these misperceptions, there is not a consensus about whether these differences represent sincere differences in belief or are an artifact of respondents not wanting to state publicly to a pollster what they know. Evidence that respondents do not fully believe their answers comes from studies that provide respondents with inducements to be accurate, e.g., paying them if their answers to objective questions about the unemployment rate of the number of war casualties (Bullock et al. 2015; Prior, Sood, and Khanna 2015; Hill 2017). This implies that many respondents are engaging in some sort of cheerleader effort instead of expressing their true opinion, perhaps because they feel social pressure to give answers that bolster their side (Connors 2020) or they want to express solidarity with their party (Yair and Huber 2020). Other studies, in contrast, have observed that sizable differences in opinion remain even when researchers give respondents incentives to give accurate answers (Peterson and Iyengar 2021; Allcott et al. 2020). This suggests that while part of observed partisan polarization about the state of the world may reflect partisan cheerleading, part of it also reflects how political identities shape how people receive and process information.

Several mechanisms explain why political identity changes how people perceive the world. Partisans may wish to avoid the cognitive dissonance that would occur if they were to admit that they were supporting the incumbent despite his or her policy failures and so they are predisposed to see the world through biased lenses. Or they may be drawing on information sources that promote stories favorable to their sides and ignore those that put their party in a negative light. Studies show that individuals tend to seek out information that corresponds to their preconceptions (Taber and Lodge 2006) even when incentivized to do otherwise (Peterson and Iyengar 2021). Individuals also are less critical of information that reinforces prior beliefs (Ditto et al. 1998) and argue against facts that go against their preconceptions (Gaines et al. 2007).

I argue that these dynamics are likely to generate partisan rationalizations concerning the state of democracy among winners in countries where democracy is under attack. Partisan media and candidates have few incentives to publicize authoritarian actions by the government and people may choose to only consume media that cheerleads their preferred party and downplays their misdeeds. Then even if regime supporters are exposed to that information they might not respond to information they do receive about a weakening democracy. Instead, they may discount information that would put the regime in a bad light as untrue “fake news” that overstates the nature of the problem, diminishes its impact by arguing that outcomes in previous regimes were worse, or simply ignore the problematic outcomes and continue to claim that all is well to justify their continued support for the regime. Or regime members and supporters may be aware of the undemocratic actions but choose to not acknowledge them publicly to avoid describing their party in a bad light, to express solidarity with their party, or because they fear social stigma if they do so. Yet even if the public and elites do not believe their positive statements about the level of democracy to pollsters, by not publicly challenging the regime they are not acting as a check on its actions. Private discontent with a would-be dictator that is never publicly expressed upon will not stop autocratization. Any of these mechanisms would generate a scenario whereby public evaluations of democratic performance diverge from reality among those who support the ruling party.

Of course, partisan rationalization may not be the exclusive domain of the ruling party. Some opposition partisans may deny that institutions are working well. Election losers, for example, may be prone to see election fraud even in clean elections, especially if party leaders claim that such fraud occurred (Beaulieu 2014; Cantu and García-Ponce 2015; Stewart Ansolabehere, and Persily 2016). Partisan media on the losing side may amplify these claims. Yet it is also possible that losers in weak democracies will have personal experiences with democratic violations that are more tangible than the claimed violations losers are presented with in strong democracies and thus losers’ opinions of democracy will diverge between strong and weak democracies despite the potential for partisan-based rejection of even a legitimate election. The degree to which the government or the opposition is more likely to engage in partisan rationalization is an open empirical question that I will explore.

The implication of this argument is that previous work arguing that citizens take governance quality into account when assessing democratic performance may be too optimistic. They may respond to generalized corruption that does not benefit them, but tolerate democratic norm violations that benefit their political side. If partisan rationalizing is occurring, then the ability of poor democratic performance to generate a reform movement may be blunted by partisans who not only reject the reform as a threat to their power but who deny that there is even a problem.

Three previous studies have examined these questions and reached contradictory conclusions. Dahlberg and Linde (2016) find that the gap between election winners and election losers in a cross-section of European countries is smaller in countries where democracy is stronger and that this gap exists because the evaluations of election winners are not affected by the level of democracy. This suggests partisan rationalization is occurring. Fortin-Rittberger, Harfst, and Dingler (2017), in contrast, look at countries included in the CSES and find that there is no gap in how winners and losers perceive democracy in countries where election quality is low, which implies that winners can admit that election fraud happened. Finally, Nadeau, Daoust, and Dassonneville (2021) also look at CSES data and fine that winners become less satisfied with their victory in very democratic countries and losers become more satisfied with it, suggesting that losers recognize improving democratic quality and winners engage in partisan rationalization. The divergent results between these three studies, each of which is based on a limited number of country-years, suggest that further research is needed to establish the empirical regularity.

The Context: Democratic Fluctuations in Latin America

To explore these questions, we focus on how democracy is evaluated in Latin America. Latin America is an ideal place to study these questions because the strength of democracy varies across the region and within countries over time. On the one hand, there has been an expansion of democracy across Latin America in the third wave; as Levitsky recently noted, “The last three decades have been the most democratic in Latin American history. Never before has so much of the region been so democratic, for so long” (2018, 102). Yet some countries have more perfectly approached the democratic ideals than others have, and there have been serious setbacks with democracy, including Venezuela and Honduras in the 2000s, and Nicaragua and, to a lesser extent, Ecuador, Brazil, Bolivia, and El Salvador in the last decade. These wide variations in democratic performance across and within countries are apparent in the Variety of Democracy’s (Coppedge et al. 2019) annual ratings of electoral democracy, where experts rate election cleanliness, media environment openness, and the protection of civil liberties. These data document significant perceived declines in democratic performance in many countries (figure 1). While coups still occasionally happen, most of these declines have occurred in contexts where the problem has been incumbent parties abusing the powers of their offices to restrict electoral competition, to compromise the ability of the opposition to organize or mobilize, or to commit electoral fraud.

Figure 1 Trends in electoral democracy in Latin America

These differences in democratic strength make Latin America the perfect setting for exploring whether partisan motivations lead individuals to ignore democracy’s weakening. Expressed satisfaction with democracy and evaluations of democratic performance should ideally be lower in the countries where democracy is weaker, but if individuals are engaging in partisan rationalization, then their reflections of democratic performance and trust in its institutions will systematically diverge from these tendencies. The two empirical analyses that follow thus explore how elite and mass evaluations of democracies inside Latin American countries match up with democratic trends.

Data: Parliamentary Elites

I begin by looking at how political elites evaluate democracy in their countries. I focus on elites because of their key role in the actual passage of institutional reforms and their ability to mobilize the public to protect or to attack democratic institutions. Elites also have more experience with the institutions that regulate and manage electoral competition and civil liberties than the average citizens do and thus are well positioned to directly evaluate the performance of electoral democracy. Data come from Parliamentary Elites in Latin America (PELA) surveys conducted by the University of Salamanca since 1997. These surveys of members of legislative lower chambers are representative of the party composition of each country and are either conducted face-to-face or via telephone. ² The specific evaluations of democratic performance vary by survey wave and so the number of country-years differs across analyses (online appendix 2 lists questions by country-year), but this project has been conducted in a variety of democratic contexts that allow me to check if political elites are willing to acknowledge when democracy is weak.

The PELA surveys contain three main questions about how legislators view the current state of democracy. The first is whether they are satisfied with the state of democracy in their country:

In general, would you say that you are very satisfied, satisfied, unsatisfied, or very unsatisfied with the way that democracy functions in your country?

This question has been criticized by some for capturing both overall regime performance and specific performance by the president (Canache, Mondak, and Seligson 2001), but this measure captures how the respondent views the specific aspects of democratic performance “at a low level of abstraction” (Anderson 1998, 583) and it is the most commonly used indicator of specific support (e.g., Bratton, Mattes, and Gyimah-Boadi 2005; Norris 1999). High values represent more positive views of democracy.

The second and third questions on democratic performance focus specifically on the conduct of elections. Members of the legislature were asked to evaluate the national election agency and also about the quality of the most recent elections in their country (in which that legislator was elected):

Thinking about the following people, groups, and institutions, I would like to know how much confidence does each deserve with regards to its actions in public life—the electoral agency? No confidence, a little confidence, some confidence, a lot of confidence.

How much confidence do you have in the electoral process in which you were elected? Use the scale that goes from 1 “minimal confidence” to 5 “maximum confidence.

Confidence in the electoral process and the agencies that run it help legitimize democracy and these kinds of questions are a widely used indicator of satisfaction with democratic processes (e.g., Norris 1999; Anderson et al. 2005; Birch 2008; Rosas 2010). The second question may result in legislators attempting to justify their own election as fair even in the context of biased elections, but the hope is that legislators who were elected in the last elections can recognize when electoral processes fall short and admit that the electoral agency and elections themselves are under-deserving of confidence. Both variables are coded so that high values express more positive views of electoral practices.

Our main interest is in how these considerations reflect the level of democracy in the country and also partisan considerations. To measure democratic performance, I use the Variety of Democracy’s (Coppedge et al. 2019) electoral democracy index, which combines data from expert assessments of election cleanliness and openness, freedom of association, and freedom of speech to classify the degree to which democracy in a country meets the procedural minimum. This variable has a theoretical range of 0–1 with high values representing a more complete democracy and ranges from 0.307 (Nicaragua in 2017) to 0.933 (Uruguay in 2010) in this sample. In general, the hope is that in countries where democracy is weak and being undermined legislators will be less satisfied with democracy and admit to being less confident in the electoral administration and election processes. The fear, in contrast, is that either winners will ignore poor democracy or losers will deny that elections are clean. I thus test if any relationship between democratic performance and evaluations of democracy is contingent upon legislator partisanship, using a dummy variable that distinguishes those who belong to the president’s party from those who belong to another party and interacting it with the electoral democracy index to see if the two groups respond to the level of democracy in divergent ways. ³

The survey does not contain many demographic control variables, but I control for the legislator’s gender, age, and education level. Full details of how these variables are coded are in online appendix 1 and their estimated coefficients are in onlinme appendix 3. ⁴ I also control for the state of the economy in the twelve months before the survey, taking the weighted average of the GDP growth in the year of the survey and the previous year. ⁵ Because the model includes variables estimated at the country-year level and that then likely cluster by country as well (as democracy scores often tend to be stable), I estimate the model as a three-level multi-level model, nesting individual members of parliament (MPs) inside of legislature years and then nesting the legislature inside of countries. For ease of presentation and interpretation, I estimate the models as hierarchical linear models; in online appendix 4 I present the results of hierarchical ordered logits as the substantive results on the polarization of opinion are generally consistent. ⁶

Results: Parliamentary Elites

The results in table 1 confirm that even among elites, partisan interests are correlated with evaluations of democracy in the average country. In particular, the results in models 1, 3, and 5 show that evaluations of democracy, the election agency, and elections tend to be significantly lower for those members of the legislature that are not presidential co-partisans. The implication of these pooled models is that members of the losing party have a drop in democratic support even after personally winning election.

Table 1 Legislative elites’ evaluations of the state of democracy in their country

Notes: Multilevel linear models, standard errors in parentheses. Includes demographic controls; Refer to online appendix 3. * p<0.05, ** p<0.01, *** p<0.001 (two-tailed)

Yet the interactive models (models 2, 4, and 6) suggest that for elites who don’t belong to the president’s party, their evaluation of the system is shaped by the level of democracy in their country in a way that differs from that of presidential co-partisans. The coefficient for the level of democracy variable is the estimated effect of democracy when the opposition partisanship variable equals zero or, in other words, when the MP belongs to the president’s party. In all three interactive models, the non-significance of those variables implies that there is no association between the level of democracy in a country and how members of the president’s party evaluate democracy. Thus, there is no evidence that presidential co-partisans in weak democracies are more likely to admit dissatisfaction with their democracy or have lower trust in the electoral process when democracy is poor than when it is good. Members of non-ruling parties, in contrast, significantly differ in their evaluations of democracy when it is poor compared to when it is weak. The significant and negative coefficient for the dummy variable distinguishing non-presidential partisans from presidential ones implies that when the level of democracy is low, these legislators are more critical of the current state of democracy than are those from the ruling party. The positive and significant interaction terms between non-presidential partisanship and the level of democracy means that the association between opposition partisans’ views of democracy and the level of democracy is significantly stronger than is the same relationship for presidential co-partisans. As a result, for all three dependent variables non-presidential partisans’ evaluations of democratic conditions worsen when democracy does. ⁷

The difference in how presidential and non-presidential co-partisans evaluate their countries’ democracies is illustrated by predicted outcomes for presidential co-partisans and other party members in figure 2, which graphs the predicted outcome for an otherwise average respondent. For each of the measures of democratic evaluations, there is no significant gap between how government partisans and other partisans view the state of democracy in the most democratic countries. This suggests that those who don’t belong to the president’s party can generally recognize when democracy is clean and elections are fair even when they lost. ⁸ Yet as democracy becomes weaker, the predicted evaluation of democracy by presidential co-partisans does not change even as members of other parties become less satisfied with democracy or trusting in elections. Thus, while non-government elites become increasingly willing under more democratic conditions to say that they are satisfied with democracy and election administration even though their preferred presidential candidate did not win, winning elites will not admit democracy has problems when it is weak.

Figure 2 Predicted evaluations of democratic performance by party and level of democracy, parliamentary elites

The results in table 1 and figure 2 reflect all cases from which we have data. I performed two robustness tests to ensure that these results are not being driven by outliers. First, I ran the models, dropping one country at a time. The same polarized response emerges in weak democracies. I also ran the models with country-specific fixed effects that isolate country-specific factors and cause the models to only focus on changes within countries in the level of democracy. The fixed-effects results confirms that changes in democratic quality within countries over time do not affect how winners evaluate democracy while changes in election quality within countries affect how losing parties’ members evaluate democracy.

Data: Latin America Mass Public

The previous data suggested that the ruling party’s politicians’ evaluations of democracy correspond to their partisan interests. To test whether the public behaves in the same way, I analyze the biennial AmericasBarometer surveys conducted by Vanderbilt’s Latin American Public Opinion Project. Roughly 1,500 respondents are selected via a stratified probabilistic sample and are interviewed face-to-face using either paper surveys or, increasingly, electronic devices that protect against interviewer error and fraud. ⁹ While the survey is conducted in more than twenty-five countries in the Americas and Caribbean, I restrict the attention to the eighteen presidential countries in Latin America. I use data from the 2006, 2008, 2010, 2012, 2014, 2016/17, and 2018/19 waves, although the availability of specific questions varies across surveys. The survey’s wide geographic and temporal scope allows for variation in democratic contexts. The largest advantage of the survey, however, is its extensive battery of questions on how democracy is perceived to be functioning.

Just as with the elite data, I begin with the standard question on whether respondents are satisfied with democracy:

In general, would you say that you are very satisfied, satisfied, dissatisfied, or very dissatisfied with the way democracy works in (country)? Very satisfied, Satisfied, Dissatisfied, Very dissatisfied.

High values represent satisfaction with democracy.

The second variable asks respondents an even more basic question: how democratic is their country. Specifically, respondents were asked:

In your opinion, is [country] very democratic, somewhat democratic, not very democratic, or not at all democratic? Very democratic, Somewhat democratic, Not very democratic, Not at all democratic.

Again, high values represent a belief that the country is democratic.

Finally, there are two questions about the degree to which respondents trusted electoral processes in their country. Respondents were asked:

To what extent do you trust elections in this country?

To what extent do you trust the Supreme Electoral Tribunal?

Both questions were answered on a 7-point scale where 1 is “not at all” and 7 is “a lot.”

Just as with the elite data I look at whether these evaluations reflect the level of democracy in a country or are shaped by partisan considerations. I measure the level of democracy using the V-Dem electoral of democracy indicator; in this sample, it varies from 0.23 (Nicaragua in 2018/19) to 0.93 (Uruguay in 2010 and 2012). I then interact this variable with a variable that measures whether the respondent voted for the incumbent president, for a losing presidential candidate, or abstained/cast a null vote in the last election. I focus on previous voter choices because levels of partisanship in Latin America tend to be lower and more volatile than in more established democracies (Lupu 2015). ¹⁰ In online appendix 7, I model support for the president as a function of self-identification with the president’s party, and those results mirror those presented later.

While the focus is on electoral considerations and the level of democracy in the country, I control for other variables that previous work shows correlate with how citizens view the democratic system (refer to online appendix 1 for the full question wording details). In doing so I focus on exogenous variables that should not be caused by partisanship. ¹¹ Given that specific support for democracy often reflects government performance, I control for the growth in GDP in the twelve months before the survey, using the weighted growth measure described earlier. I include a series of demographic controls, including gender, age, education level, the level of household wealth, living in a rural area, and ethnicity-details about question wording are in the appendix. Because these variables are all included as controls, I do not dwell on the results and present them in online appendix 3. Their associations with evaluations of democracy are consistent with previous analyses of specific support for democracy in the region (e.g., Booth and Seligson 2009).

I nest respondents inside of country-years and inside of countries because the level of democracy is constant for each survey wave and does not change much in many of these countries over time. I model the data with a linear specification for ease of presentation, with ordered logit specifications available in onliane appendix 4.

Results: Latin America Mass Publics

The pooled analyses in table 2’s models 1, 3, 5, and 7 show that there is a significant gap on average between voters who voted for the winning candidate and voters who voted for the losing candidate for all four measures of perceived democratic performance. Election losers tend to be more pessimistic about the state of democracy than are election winners. Yet just as with the elite data, the interaction terms in table 2’s models 2, 4, 6, and 8 demonstrate that the connection between the level of democracy and citizen evaluations of the state of democracy varies significantly by partisan groupings. The baseline partisan category is having voted for the president, which means that the baseline coefficients for the level of democracy in the interaction models capture the association between this variable and evaluations of the state of democracy among those who voted for the ruling party. There is no significant difference in how presidential voters evaluate the current state of democratic performance in countries where democracy is weak than in countries where it is strong. In contrast, the positive and significant interaction terms between support for a losing candidate and the level of democracy in all four models imply that the opinions of losing voters tend to be more strongly connected to the level of democracy in a country than are the opinions of winning voters. As a result, in all four models, the level of democracy is positively and statistically significantly associated with how losers evaluate the level of democracy in their country (refer to online appendix 15). When a country’s democracy is strong, election losers have higher levels of satisfaction with democracy and with electoral processes than when democracy is weak while those who won the election or those who abstained do not differ in their evaluation of democratic performance across democratic contexts. ¹²

Table 2 Mass publics’ evaluations of democratic quality in their country

Notes: Multilevel linear models, standard errors in parentheses, * p<0.05, ** p<0.01, *** p<0.001 (two-tailed), includes demographic controls. Refer to online appendix 3.

The effect of these differences for an otherwise average respondent is graphed in figure 3. The marginal effect of democracy on satisfaction with democracy is smaller for the mass public than it was for the elites documented in figure 2, which is not surprising given that over 83% of respondents chose one of the two middle categories, and the marginal effect of democratic performance on whether either winners or losers describe their country as a democracy is also not very large. Yet even in those cases, these data show that the mass public behaves in a similar way as do parliamentary elites, as election losers can acknowledge the strength of democratic institutions when democracy is strong and elections are clean and do not differ in their evaluations of democracy from election winners in the most democratic countries but election winners are unwilling or unable to admit dissatisfaction or distrust of democratic procedures that are biased in their favor in weak democracies. Thus in the weakest democracies, election winners have significantly more positive evaluations of democratic and institutional performance than do election winners.

Figure 3 Predicted evaluations of democratic performance by party and level of democracy, Latin American public

Just as with the elite data, I ran several tests to confirm that the results in table 2 are not driven by outliers. The substantive results remain the same even if drop each country one at a time or if we control for country-specific fixed effects that focus on changes over time. Deteriorations in the level of democracy result in increased polarization in how the public evaluates democracy.

Robustness Checks

These differences between winners and losers in the elite and mass data are robust to various modeling choices. While several of those tests are described above, further robustness tests are in the online appendix. They do not change if we control for left-right self-placement (online appendix 5), if the level of democracy is measured using Polity scores instead of the V-dem polyarchy measure (online appendix 12) or if we restrict the democracy measure to the V-dem measure of electoral integrity (online appendix 13). The results also do not change if we add controls for per capita GDP (online appendix 9) or for the level of ideological polarization in the country (online appendix 10). ¹³ Elite and mass surveys find no evidence that average presidential party supporters will negatively describe democratic performance in countries where democracy is weak and these results do not change across model specifications.

Nicaragua: An Illustrative Case

Nicaragua illustrates well how the opinion of presidential party supporters and the mass public increasingly diverge as democracy weakens. ¹⁴ While Nicaragua had begun to show signs of democratic decline in the decade before Daniel Ortega’s 2006 election (Martí i Puig and Serra 2020), since he has taken power he and his allies in the FSLN party have taken steps that have substantially weakened democracy. They have banned some opposition parties, replaced opposition party leaders, limited election observers, ensured that opposition parties were denied campaign funds, weakened the freedom of the press, took control of major television stations, politicized the judiciary, increased presidential decree powers, and been accused of electoral fraud (see Thaler 2017; Martí i Puig and Serra 2020). These changes led the OAS to declare that “the lack of freedoms in the country has become desperate, creating a critical situation with regards to civil and political liberties” (OAS 2019). The V-dem expert surveys concur, with the country going from a score of 0.627 in 2006 before Ortega took office, to 0.43 in 2012, and 0.21 in 2018. The weakening of democracy and increased corruption have resulted in several waves of mass protests, many of which have met with violent pushback by the state.

Ideally, FSLN members and supporters would also have criticized these actions as being anti-democratic. Yet regime insiders insist that Nicaragua is a democracy. After his most recent reelection in 2016, for example, Ortega declared that Nicaragua has “consolidated its democracy.” ¹⁵ A recent publication in the party newspaper argues that elections are free and fair and that “Nicaragua is constructing a democracy that the North America Empire [the United States] never wanted us to have.” ¹⁶ Instead of rejecting these changes, the FSLN has increasingly consolidated around Ortega and his family (Martí i Puig and Serra 2020; Buben and Kouba 2020), and the FSLN-controlled congress has recently passed laws weakening media freedom, empowering the president, and allowing Ortega’s wife to serve as vice president despite a constitutional ban on such intra-familial ties. Multiple key members of the regime have labeled the recent protests as a conspiracy coordinated by the United States whose participants are “terrorists” and “devils” and argued for their suppression. ¹⁷

Then the survey data show that both elite and mass opinion has polarized on partisan lines about the level of democracy in the country. ¹⁸ The PELA surveys show that non-FSLN elites are increasingly negative about the level of democracy in their country: among legislators who do not belong to the FSLN the level of satisfaction with democracy has fallen from 1.41 (on the 0–3 scale) in 2007 to 0.53 in 2017 (refer to table 3). The average level of trust in the election agency also fell in that period among non-presidential partisans. A similar decline in confidence in electoral institutions and belief that their country is democratic occurred at the mass level among those who voted against Ortega, although democratic satisfaction did not change much among this group (in contrast to the analysis earlier). These drops in specific support among non-FSLN supporters are consistent with democratic reality in the country.

Table 3 Elite and mass evaluations of democracy and democratic institutions in Nicaragua, 2006–2018

Yet the FSLN members and supporters do not describe the same trends in their survey responses. Instead, elite and mass surveys show that satisfaction with democracy, trust in the election management body, and beliefs that the country is democratic have not fallen but instead have increased among FSLN supporters. Despite the weakening of democracy that international organizations and experts have observed, supporters of the ruling party continue to describe Nicaragua’s democracy in positive terms. There is thus little likelihood that the bulk of the FSLN will check further power grabs because they believe democracy is performing well.

Conclusion

Democracy is increasingly under threat in many countries. Leaders threaten to undermine checks and balances, attack the press, and in some places manipulate election processes in their favor. When democracy is under threat, its defenders will often call upon the public to stand up for democratic institutions and to mobilize for democracy-enhancing reforms. Yet many previous studies show that voters support undemocratic actions by candidates and leaders that they believe provide benefits to their base. The data presented here suggest supporters of the ruling party will often not even acknowledge poor institutional performance or the low level of democracy in contexts where democracy has waned. Reported satisfaction with institutional arrangements and performance among political winners at both the elite and mass levels is not significantly lower in contexts where the ruling party is consolidating its power. Political losers, in contrast, generally have confidence that elections are trustworthy, and believe democracy is satisfactory in countries where democracy is strong despite their loss but are painfully aware when democracy weakens and subsequently are the ones who display low levels of specific support in weak democracies.

The analysis presented here has focused on Latin America. Yet Latin America is not the only region where regime insiders are weakening democracy. One needs to only look to the recent history of the United States, where President Trump repeatedly attacked the media, threatened to use the Department of Justice to investigate political rivals, and attempted to undermine the independence of electoral administration while votes were being counted to be reminded that executive actions to undermine democratic accountability are not unique to developing countries. Larger deviations from democratic norms have occurred in recent years in many East European countries as well as in India and other places. While these data do not directly examine those places, they suggest that we need to worry about whether regime supporters in those contexts will even recognize that leaders are undermining democracy or see their democracy as being under threat. Frequent Bright-Line Watch surveys have shown stark partisan divides on whether various hardline tactics are appropriate (Carey et al. 2019) and recent events in the United States where the public and elites have polarized on whether and how to investigate the January 6, 2021 Capital riot show that partisan rationalizations about whether actions are threats to democracy can emerge in even the most established of democracies.

While the present study identifies a partisan polarization in responses to democracy, it leaves several important questions open for further study. One of the most important is the mechanism that underlies these changes. On the one hand, the positive evaluations by presidential supporters may only be cheap talk to a pollster and not represent their actual opinions. Yet the fact that they will publicly signal support for the regime despite undemocratic actions suggests that they are willing to tolerate those kinds of actions as long as they keep getting their way. On the other hand, if election winners are ignoring the declining quality of democracy, we don’t know if this is because they are badly informed or are willfully ignorant. Further research should attempt to isolate these mechanisms.

The work in this paper also suggests several avenues of further empirical work. While this paper has focused on how these dynamics differ on average across countries, one might expect that these dynamics will be particularly pernicious in certain circumstances. Coalition politics, for example, may shift these dynamics, as presidential partners may be tempted to endorse steps that empower their “team” while also needing to potentially worry about the president becoming more powerful at the expense of his or her allies (refer to online appendix 11). Further research should explore these differences. Then while this research has focused on large breakdowns of democracy in developing countries, recent events in the United States and elsewhere show that even citizens living in established democracies might be tempted to ignore the consequences of norm-breaking leaders who violate the norms of democratic tolerance and restraint in pursuit of partisan goals. The question becomes if a deeper mass commitment to democracy might make some individuals more likely to criticize members of their party who violate democratic norms. Yet recent experience in the United States suggests that many people will ignore violations of democratic norms by their side in even established democracies. Thus while we might expect that similar partisan rationalization will occur in established democracies, further work should explore if these tendencies are more muted.

Yet while this paper has focused on describing certain empirical regularities, these results have troubling implications, albeit ones that we cannot directly observe with these data. They contribute to the growing literature showing that the winner-loser gap is not just driven by losers abandoning democracy and whose consent must be maintained but also may reflect the undemocratic opinions and public responses of political winners (e.g., Moehler 2009; Singer 2018; Bartels and Kramon 2020; Cohen et al. forthcoming). When democracy is being toppled from within, winners who won’t accept that these actions violate democratic norms become barriers to democracy reemerging. As a result, while political losers can sometimes be a threat to democracy, they may sometimes also be its last and best defenders. ¹⁹ If the losers lost the election illegitimately due to restrictions on their civil liberties or due to electoral fraud, then the losers should be strongly dissatisfied with democracy and loud in criticizing these failings. Protests by election losers against unfair elections have been a key element in the overthrow of electoral authoritarian regimes (e.g., McFaul 2005; Tucker 2007) and we want an actively monitoring public to be a check against democratic malfeasance. ²⁰ In systems where democratic practices are falling short, a gap between the winners and losers may be evidence that the winners may be the ones who are ignoring the decline in democracy that biases the system in favor of their party.

This polarization may create problematic dynamics. If only the political losers recognize when democratic quality is waning, then the quality of democracy becomes a “partisan fact” that incumbent party supporters can dismiss as “sour grapes” from disenchanted actors, making those claims easier to dismiss. Discussions of pro-democratic reforms may then become politicized as winners see losers as acting not to strengthen democracy (which they don’t see as being under threat) but to tip the scales in their favor. Then as election losers see winners abandon support for democratic norms and fail to acknowledge or criticize norm violations, these losers may begin to question the importance of continuing to play by democratic norms themselves. The result may be an escalation of political hardball as each side perceives that the other will not engage in moderation (Levitsky and Ziblatt 2018). If winners cannot be led to separate their self-interest from their evaluations of government actions and to acknowledge when democracy is failing, arresting that fall may become difficult. Unfortunately, the data in this study suggests that both elites and the mass public struggle to avoid the temptation to justify and ignore anti-democratic behavior by their side as it happens.