key: cord-0068264-ixqtbwui authors: Segovia, Joselin; Pontarollo, Nicola; Orellana, Mercy title: Discontent with democracy in Latin America date: 2021-09-30 journal: nan DOI: 10.1093/cjres/rsab020 sha: 07e1d055f25212909fe7da3e44d9d88b9073c89e doc_id: 68264 cord_uid: ixqtbwui Satisfaction with democracy registered its lowest global score in 2019, with Latin America being the worst-performing region and the most seized by social protest. This paper briefly surveys and assesses the main narratives that attempt to explain the causes of this discontent in the region. The results show that discontent has different roots, which are mostly explained from an individual point of view rather than from a contextual one. Inequality of opportunity and institutional weaknesses play the main role. Considering the COVID-19 social and economic crisis, our results shed light on the key elements to address whether discontent is to be contained. Latin America is a heterogeneous region in terms of ethnicity, religion, climate and geography, yet some important traits are common, namely poverty, uneven development and income inequality (Kingstone, 2018) . During the "Golden Era" spanning roughly between 2003 and 2013, rising prices of commodities (Gruss, 2014) and high international liquidity prompted economic growth and an unprecedented bonanza in most Latin American countries (Sánchez-Ancochea, 2021), allowing for substantial social progress. Unemployment and poverty decreased significantly, while real wages and access to social rights and to all levels of education increased in most countries (CEPAL, 2019; OIT, 2017; Sánchez-Ancochea, 2021; UNESCO, 2014) . By the end of this cycle, however, the sustainability of this progress became a major concern. A combination of economic slowdown, corruption scandals, adjustment programs and stagnated income set the scene for social conflict and complex governance (Lustig, 2020; Morlino et al., 2016) . In 2019, street protests engulfed the region, turning it into the annus horribilis (Matera and Despradel, 2020) and affecting countries both with relatively thriving economies and institutional strength, such as Chile and Colombia, and those with difficulties, such as Venezuela. While it is not easy to identify the causes of this discontent, The Economist Intelligence Unit (EIU, 2020) describes it as the result of overall dissatisfaction with the political status and politicians' decisions, boosting citizens' willingness to protest and express disagreement with the status quo. Therefore, it comes as no surprise that the region's dissatisfaction with how democracy works is the highest in the world (Foa et al., 2020) , leading experts to talk about a democratic crisis (Azpuru, 2019; Shifter, 2020; Vlaicu, 2019; Zovatto, 2020) . Two main narratives attempt to shed light on the potential drivers of this discontent in Latin America. The first identifies political and institutional factors as the main drivers (Azpuru, 2019; EIU, 2020; Shifter, 2020; Vlaicu, 2019; Zovatto, 2020) and the second, inequalities (Ferreira and Schoch, 2020; Lustig, 2020) . Testing these conjectures has not yet been tackled however. In this paper, we aim to identify whether inequality or institutional quality is more important for dissatisfaction with democracy. To this end, we carry out a comprehensive analysis of discontent, approached through dissatisfaction with democracy, by integrating the main features of these narratives, as well as other characteristics of the Latin American context. In light of the projected economic and social crisis resulting from the COVID-19 pandemic, identifying the factors affecting discontent is imperative. Our results can inform policymakers on how to take concrete actions to restore trust in and satisfaction with representatives, which is crucial in times requiring a high level of compliance with government measures. Furthermore, our results identify ways to strengthen the quality of democracy and consolidate it in the region. This paper is structured as follows. First, we give an overview of discontent with democracy in Latin America and survey the main drivers that have been descriptively linked to discontent, as well as others that should be part of any analysis of this type. Then we present the data, the variables and the estimation strategy. The subsequent section describes the results, and the final sections offer a discussion and some concluding remarks. The Democracy Index compiled by the EIU shows that global democracy is in its worst state since data of this kind has been available (EIU, 2020) , and, according to Foa et al. (2020) , citizen dissatisfaction with democracy is at its highest levels, indicating that a "global democratic recession" is occurring. While there is variation across and within regions, Latin America is the worst-performing in both democracy measures. Figure 1 is based on data from the Latinobarómetro, and more precisely on the question "In general, would you say you are very satisfied, fairly satisfied, not very satisfied or not satisfied at all with the way democracy works in your country?" . Based on the sum of the share of "not very satisfied" and "not at all satisfied" responses in each country, dissatisfaction with democracy has been increasing over the past decade. From 2015 to 2018, it experienced an increase of 15 percentage points and reached its highest point in the decade, with 74% of citizens reporting being unsatisfied in 2018. A comparison with other transition economies like those in Eastern Europe, where between 2015 and 2018 dissatisfaction with democracy decreased from 60% to 51%, demonstrates how important the situation is. 1 Figure 2 reports the share of respondents not satisfied with democracy in 2018 by quintiles (data are from the latest Latinobarómetro, 2018) . With the exception of Uruguay, Costa Rica and Chile, in all other countries, the percentage of dissatisfied citizens is above 61.8%. In three out of the four most populated countries of the region, Mexico, Peru and Brazil, more than 80% of respondents are not satisfied with democracy. As pointed out by Moisés (2011) , in most Latin American countries the democratic tradition is fragile and discontinuous, and long-lasting and excessive political distrust may in some cases lead to a partial rejection of the role of political parties and the parliament (Park and Shin, 2006) . Dissatisfaction or discontent with democracy is, however, a complex and multifaceted phenomenon with important social, economic and political features. In Latin America, the "Golden Era" generated a stronger and more empowered citizenship, which generated a "revolution of expectations" (Morlino et al., 2016) . As a result, people are demanding better economic and political management by their representatives and are less tolerant of corruption (Kingstone, 2018) . In this sense, recent subsidy cuts, electoral fraud, economic stagnation and austerity measures have frustrated citizens and may have led them into protest on the streets (Lustig, 2020; Matera and Despradel, 2020) . According to the EIU (2020), however, while economic distress is almost a necessary feature for social uprising, it is not sufficient in itself and unrest depends on other structural features such as income inequality, governance capability, levels of social provision, ethnic tensions and trust in institutions. Several hypotheses have been developed by scholars and political analysts to explain the causes of social unrest in Latin America. For instance, Lustig (2020) and Ferreira and Schoch (2020) focus on discussing social discontent as a result of inequality. They hold that if approached through the Gini coefficient, income inequality could not have triggered social protests since data shows that, despite a slowdown, it is declining in the region. However, Lustig (2020) also raises the question of the suitability of this indicator for approaching discontent, showing that the data available to calculate it are deficient and that correcting this might lead to different patterns from those observed in the publicly available data. Hence, the author infers that income inequality might explain social discontent and urges for an improvement in the availability of data sources to better capture and analyze it. Ferreira and Schoch (2020) argue that instead of income inequality, inequality of opportunity (IOp) might be the trigger for uprisings. In this respect, data show that IOp is overall higher in Latin America with respect to other regions (Brunori et al., 2013; Ferreira and Schoch, 2020) , which means that opportunities are restricted to only a portion of the population, depending on background features such as birth and residence location, family income and education . In fact, the empirical evidence finds that in Latin America childrens education opportunities are strongly correlated with that of their parents' (Neidhöfer, 2019) . In this line of research, inequality beyond traditional measures may be correlated with social and political discontent. Other authors, including Azpuru (2019); Shifter (2020); Vlaicu (2019) and Zovatto (2020), have described uprisings as the result of a lower tolerance for corruption, a lack of trust in institutions, low approval ratings for governments and anger towards the economic and political elites. Corruption is indeed an ingrained feature of Latin American politics, and in 2018 more than 50% of all citizens felt that corruption was increasing (Pring and Vrushi, 2019) . Furthermore, in 2018, 79% reported the belief that governments work only for the benefit of a few powerful groups, and trust in institutions was remarkably low especially for legislatures and political parties, who were trusted by only 22% and 13% of the population, respectively (Latinobarómetro, 2018) . The role of public safety and media in social unrest in Latin America has also been highlighted. First, public safety has become one of the main concerns of citizens, and with a regional homicide rate four times that of the global average, 43% of the population report being constantly afraid of being a victim of crime (Chinchilla et al., 2018) . This may reflect the inability-or even unwillingness-of states to protect citizens, and especially the poor, undermining trust in political parties and other democratic institutions. Second, the media has come to play an important role. On the one hand, the mainstream media has been increasingly scrutinised for being an ally of the economic and political elites, as well as of incumbent governments (Magallanes-Blanco and Treré, 2019), but on the other hand, this has caused the rise of social media as a catalyst of social protests (Valenzuela et al., 2016) . These conditions set the scene for what experts have termed "defective" or "flawed" democracies, characterised by an underdeveloped political culture and problems with governance and media freedom, leading citizens to "lose faith" in how democracy works. This is corroborated by the EIU (2020) according to which, by 2019, most countries across the region fell within this category and only three of them fit under the category of "full democracy" , namely Uruguay, Costa Rica and Chile. Considering the complex and multifaceted nature of social discontent, at the individual level we include three categories of factors potentially associated with dissatisfaction with democracy: (i) personal characteristics, (ii) inequality-related variables and (iii) institutional quality variables. At the contextual level we have (i) contextual economic characteristics, (ii) inequality and (iii) institutional quality measures. Data at the individual level are taken from the latest Latinobarómetro (2018), a public opinion survey conducted annually in 18 Latin American countries plus Spain, representing more than 650 million inhabitants. The 2018 wave consists of 20,204 interviews collected between 15 June and 2 August 2 2018, with representative samples of the populations of each of the 19 countries. A multistage stratified probability sampling design is applied (usually four stages): the first three stages correspond to a random selection of the sampling units (city, area sector or block and household), while the last stage corresponds to a non-probabilistic selection of the individuals to be surveyed, with the use of control quotas (generally according to region, area and gender characteristics). The information used to select the sampling units comes from the latest population census available in each country. The survey is 100% representative at the country level (the only level of disaggregation available) in all countries, with the exception of Honduras (99%) and Panama (98%). The sampling error ranges between ±2.8% and 3.5%. The number of observations in each country is between 1000 and 1200. We exclude Chile, for which we detected a problem in the variable city size, and Venezuela, for which various contextual indicators are unavailable and which has been classified by the International Institute for Democracy and Electoral Assistance (IDEA) (2019) as a non-democracy since 2017 and as authoritarian regime by the EIU (2020). According to the latter, Venezuela (together with Cuba) is the only country that scores zero in the field electoral process and pluralism. 2 Our sample of 16 countries results in 17,804 observations. However, due to the exclusion of observations for which we have "don't know/no answer" for at least one independent variable or for the dependent variable, the final sample we used consists of 13,166 observations, ranging from 65% of respondents of Guatemala to 84% of the Dominican Republic. In the Appendix, we report a full description of the variables included in the model (Supplementary Tables A1 and A2 ) and descriptive statistics (Supplementary Table A3 ). Among personal characteristics, we include sex, age, civil status, ethnicity, education, size of the city of residence, employment status and socio-economic status. The latter provides a proxy of income level through two subjective variables, given that the Latinobarómetro does not have an objective measure of it. The first variable is self-perceived socioeconomic status (low, middle and high). However, as selfranking variables usually suffer from respondents over-selecting middle values, our second measure is the interviewer's assessment of the interviewee's socioeconomic status (low, middle and high), based on the quality of their dwelling, their furniture and the interviewee's general appearance. Among these variables, the literature in Europe has often labelled age, education and income as the "holy trinity" of social discontent (Dijkstra et al., 2020) . Hence, apart from being basic relevant information, the inclusion will allow for comparison between Latin America and Europe, even though our measure of discontent differs from that literature, where it is generally approached through the populist vote. The second category, inequality-related variables at the individual level, aims to test the hypothesis of Ferreira and Schoch (2020) which sees inequality of opportunity as playing a key role in discontent. Literature on IOp states that inequality can be broken down into a portion that can be considered fair (related to effort/ merit) and another that can be considered unfair (due to opportunities/circumstances) (Roemer and Trannoy, 2016) . To approach this, we introduce three variables: the perception of fairness of inequality, a measure of intergenerational mobility and the family economic outlook. The perception of fairness of inequality is a key measure in the IOp literature. According to Checchi et al. (2016) , when inequality is judged unfair, that is, the result of opportunities rather than effort, it will be largely objected to by citizens. According to our dataset, a staggering 80% of citizens across the region consider that the income distribution is unfair. This leads us to hypothesise that the perception of the unfairness of the income distribution would increase dissatisfaction with how democracy works. Furthermore, according to Dahl and DeLeire (2008) , a key aspect in the study of IOp is intergenerational mobility, and education is widely used to measure this (Neidhöfer, 2019) . Thus, we introduce a second variable that consists of the individuals' education compared to their parents, that is, whether they have more or fewer education years than their parents, obtained by combining a variable about the individual's own education and one about their parents' education (S10 and S11 in Supplementary Table A1 , respectively). In light of the literature, having a higher (lower) education level than one's parents suggests upward (downward) mobility in a society, which would in turn be associated with less (more) IOp. Following Ferreira and Schoch (2020), we expect less (more) IOp to cause less (more) dissatisfaction with democracy. 3 A third variable used to study inequality is family economic outlook, which is addressed by the individual's expectations regarding the family's economic situation in the coming twelve months. Its inclusion finds justification in the fact that nearly two-thirds of Latin Americans can be classified as living in poverty or vulnerability and, as observed by Scheidegger and Staerklé (2011) , an individual's experience of financial difficulties not only produces perceptions of illegitimate and wider, inequality but also contributes to political resentment and electoral upsets. Therefore, the economic outlook may influence the perception of fairness and hence can also be seen as a potential confounding factor. The third category, that is, institutional quality variables at the individual level, is aimed at testing the narrative developed by Azpuru (2019); Shifter (2020); Vlaicu (2019) and Zovatto (2020) . We include variables that measure trust in media, government support, public safety, corruption perception and a variable on the perception that governments are a coalition of a few powerful groups working for their own benefit. The inclusion of this set of variables follows directly from the discussion in the previous section. We take advantage of direct questions the Latinobarómetro asks regarding these matters. Following Zmerli and Castillo (2015) , in this category, we also incorporate a composite indicator on institutional trust. Our indicator is a combination of five variables denoting the level of trust in the parliament, government, political parties, elections and courts. The five (ordinal) variables are combined by means of a principal component analysis (PCA) based on a polychoric correlation (Kolenikov and Angeles, 2009 ). Hence, our trust indicator corresponds to the first component, accounting for 56% of the total variance. With respect to contextual variables, the first aspect to be stressed is that, unfortunately, we had to use indicators at the national and not the subnational level. There are two reasons for this. The first is the fact that the Latinobarómetro is not regionally representative, and the second is the lack of contextual data at the subnational level. While we are aware that this does not allow for within-country variation, we will interpret the coefficient of contextual variables with care. In fact, with variation only from 16 countries, it could be the case that an important correlation between individual and national variables exists but is not observed due to insufficient statistical power. In the first category of contextual variables, labelled contextual economic characteristics, we have the factors likely to affect individuals' lives as they determine favorable or unfavorable economic conditions and hence affect their dissatisfaction with democracy. Among these factors, we include the three-year average rate of inflation representing the post-boom conditions from 1 January 2016, to 31 December 2018 (The Heritage Foundation, 2020), and the GDP per capita and unemployment rate in 2019. In addition, trade openness in 2018 is included as a proxy for globalisation (World Bank, 2020) , and government expenditures during 2016-2018 as a share of GDP to represent the government's role in the economy (The Heritage Foundation, 2020). The second category, that is, inequality variables, is aimed at testing the hypothesis of Lustig (2020) . These include traditional inequality measures provided by the World Bank, such as the Gini coefficient and the share of income held by the top 10% and 20% of the population and by the bottom 10%. The last set of contextual variables includes institutional quality measures aimed at testing the narrative by Azpuru (2019); Shifter (2020); Vlaicu (2019) and Zovatto (2020) beyond the individual's perspective. We incorporate two sets of data taken from the World Bank (Kaufmann et al., 2011) and Polity (Marshall and Gurr, 2020) , respectively. From the first source, we select six indicators: voice and accountability, political stability and absence of violence, government effectiveness, regulatory quality, rule of law and control of corruption. All vary between a governance performance of −2.5 (weak) to 2.5 (strong). The institutional quality measures from the World Bank are based on 30 different data sources, including international organisations, government organisations, NGOs, universities and commercial consultancies, in addition to objective and perceptions-based measures. Furthermore, the different data sources involve a variety of different respondents ranging from foreign experts and country analysts to individuals and national firms, mitigating the risk of response and perceptional bias. Polity indicators, on the other hand, account for the quality of democracy. We include institutionalised democracy scores from 0 (no democracy) to 10 (full democracy), institutionalised autocracy scores from 0 (no autocracy) to 10 (autocracy) and the combined Polity score, obtained by subtracting the autocracy score from the institutionalised democracy score. Since our dataset includes observations at the individual level nested at the country level, as is customary in the literature (see Steenbergen and Jones, 2002 , and for recent applications, Loveless and Binelli, 2020; Wiesehomeier and Doyle 2013) we employ a multilevel ordered logit model to account for country-level variance. The model is as follows: (1) where i denotes the individual observation (13,166) and j the country (16). Dissatisfaction i,j captures dissatisfaction with democracy. We consider the 4 categories (c = 4) ordered as follows: Not at all satisfied > Not very satisfied > Rather satisfied > Very satisfied. X is a vector including the three cut points and individual level characteristics, and β is a vector of associated coefficients; Z and δ are vectors of national contextual variables-with those referring to inequality and institutions included one by one to avoid multicollinearity issues-and associated coefficients, respectively; γ j is the random intercept representing level 2 (nation-specific) residuals. ε i,j are level 1 (individual-specific) residuals assumed to be mutually independent and normally distributed with zero mean and variance equal to σ ε 2 . Level 2 residuals are assumed to be uncorrelated with ε i,j , mutually independent and normally distributed with zero mean and variance equal to σ γ 2 . ε i,j residuals represent the unexplained variability of the individual discontent after considering measurable individual characteristics and the national clustering, whereas γ j residuals represent unexplained heterogeneity at the national level. Indeed, individuals are "clustered" in the respective countries, and in our multilevel framework γ j is an extra parameter that controls for this clustering. Instead of one general random effect that captures how each observation deviates from the predicted fixed effects, we have multiple random effects that capture how observations deviate within a cluster and how each cluster deviates from the overall group. In our context, having only a random intercept γ j indicates variability in the level of discontent by country but a constant relationship between the explanatory variables and discontent. 4 Finally, the nature of the data we use, unfortunately, leaves open the issue of endogeneity. The dependent variable and the right-hand-side institutional proxies are likely to be endogenous since they are derived from individual's subjective perspective and since the strong correlations can be due to omitted variables or to simultaneity/reverse causality. To this extent, given that the Latinobarómetro is constructed as a single cross-sectional survey for each wave and that it collects data mainly on perception, it should be stressed that the regression results do not identify drivers per se, but rather they are cross-sectional correlations. This may be linked to the fact that various perceptions are all driven by the same underlying (unobserved) variable or to the fact that that this correlation is indicative of poor institutions motivating dissatisfaction with democracy. Nevertheless, we comment on our results following the second and less restrictive perspective, although we cannot rule out the first, more conservative interpretation. Results are reported in Table 1 and are robust to different specifications. When models (1) and (3) are estimated considering countries as fixed effects, the results are further confirmed. The results also hold when including Chile and excluding the variable City Size, as well as when incorporating Venezuela and/or removing Nicaragua. For reasons of space, all of these estimates are available upon request. A likelihood ratio test rejects the null hypothesis that σγ 2 is equal to zero, confirming the choice of the random intercept model. With regard to personal characteristics, results show that citizens who self-identify as belonging to Afro-descendant populations report higher discontent than the reference group of mestizos. This relationship is expected in the Latin American context, since this ethnic group is among the most disadvantaged (CEPAL, 2017). Results for age and for secondary and higher education levels are not statistically significant, (3) Zmerli and Cartillo (2015) . Only primary education is found to have a decreasing effect on discontent, marking a difference with the literature for Europe where low levels of education have the opposite effect (Abreu and Öner, 2020; Dijkstra et al., 2020; Gordon, 2018; Schoene, 2019) . This, however, is not completely unexpected. In Latin America, increasing dissatisfaction levels are a result of higher levels of education because better educated citizens "raise the bar" . Thus, we might expect lower levels of education to have the opposite effect (Foa et al., 2020; Norris, 2011) . Furthermore, citizens who identify as having a high (low) socioeconomic status are less (more) discontent than those reporting a middle socioeconomic status. This result is robust to the alternative measure of socioeconomic status based on the interviewer's assessment, allowing us to exclude problems of over-selection of middle values by respondents self-ranking their socioeconomic status. The outcome might reflect an association between discontent and the great vulnerabilityrepresented by the lack of access to services and opportunities-that low socioeconomic status groups usually face. This finding is in line with the mainstream literature that uses income per capita as a measure of socioeconomic status to study discontent, mainly in the European context (see, for example, Dijkstra et al., 2020; Gordon, 2018; Schoene, 2019) . Unemployment and city size are not significantly different from zero. Regarding the former, its lack of association with discontent contrasts Algan et al. (2017) ; Dijkstra et al. (2020) and Nicoli and Reinl 2020) , who find that unemployment is one of the main factors behind discontent in Europe. In the Latin American context this might be explained by the high rate of informal employment, which sits around 60% (OECD, 2020a). City size does not play a role in dissatisfaction with democracy, and this may be due to the fact that in the region, it is not the place that matters but other factors. However, when studying Latin America from a territorial perspective, it must be considered that despite poverty incidence historically being higher in rural areas, in the past decade it has increased relatively more in urban areas, and particularly in the suburbs. In this regard, the Latinobarómetro has some limitations, especially because of its lack of representativeness at the subnational level. Hence, assessing these characteristics of the region may require a different approach that considers differences in discontent within urban territories. Together, these results suggest that dissatisfaction with democracy could be more associated with economic features rather than demographics, supporting allegations about the role that post-boom stagnation and economic vulnerability play in the region's social unrest (Lustig, 2020; Morlino et al., 2016) . Regarding the set of inequality-related variables at the individual level, as expected, the perception of fairness of the income distribution shows a positive and statistically significant effect on discontent with democracy. This result supports the claim by Ferreira and Schoch (2020) regarding how anger about unfair inequality, namely IOp, elevates discontent. Related to this, both upward and downward mobility also increase discontent. It is worth mentioning, however, that while those experiencing downward mobility correspond just to 6% of our sample, those with an education higher than their parents comprise around 23%, a non-negligible share of the population. While higher discontent in those who experience downward mobility is expected given the theoretical background provided, the result for upward mobility might seem counterintuitive. A possible explanation is a fact that more education usually implies greater economic opportunities. This has been limited in the region due to the decline of education premiums and the high demand for low-skill labor (see Messina and Silva, 2018) . Furthermore, processes such as upward mobility can be linked to citizens demanding more from their governments, demands the latter have failed to meet (Kingstone, 2018) . These circumstances indicate that upward mobility per se might not be sufficient to decrease social discontent in the region since the economic, as well as the political and institutional, structure might generate unfulfilled expectations. The final variable in this category, that is, family economic outlook, is found to be positively and statistically correlated with discontent, confirming the findings of Loveless and Binelli (2020) for Italy. However, while these authors restrict the analysis to highly skilled jobless youth, our result holds for a representative sample of the Latin American population, offering a warning that the association between negative economic expectations and dissatisfaction with democracy may affect the likelihood of sustainable democratic governance. With respect to institutional quality variables at the individual level, we find that citizens' perceptions of increased corruption are associated with greater discontent with democracy. This effect can be explained from two perspectives. The first relates to the inherent anger corruption produces, which has resulted in a lower tolerance (Kingstone, 2018) . In recent years, numerous scandals have been uncovered in the region, leading to high-ranking representatives being formally charged and committed to trial. This prompted public manifestations of discontent in a number of countries such as Ecuador, Brazil, Bolivia, Guatemala and Chile. The second perspective through which to analyze corruption is related to some of its potential consequences. Corruption negatively affects the functioning of markets and governments, hindering the performance of both public and private institutions (Kingstone, 2018) . Our data show remarkably low levels of trust in public institutions, and the results suggest that increasing it is related to a decrease in dissatisfaction with democracy. Regarding government support, our results show that those who are opposed to the government are more dissatisfied with democracy compared to those who support it. If democracy was perceived as an "absolute value" , we would expect a lack of significance from this variable, meaning that it would be judged independently from support for the ruling parties. Note: *p < 0.10; **p < 0.05; ***p < 0.01. Standard errors are in parentheses. Results are based on multilevel ordered logit regression including individual and contextual variables. Note: *p < 0.10; **p < 0.05; ***p < 0.01. Standard errors are in parentheses. Results are based on multilevel ordered logit regression including individual and contextual variables. As this is not the case, it might be a sign that in Latin America there is a partisan position that leads to a sort of identification between personal ideology and democratic values, probably due to issues like clientelism, personalism and the lack of autonomy of the bureaucratic system (Kline et al., 2018) . Another variable related to government and institutions is the perception of how these are configured and how they operate. Table 1 shows that citizens who hold the view that countries are governed by a few powerful groups for their own benefit report less satisfaction with democracy. Kline et al. (2018) point out that, historically, across the region, the first political parties were founded by powerful groups, representing the economic and social elites. Throughout the years, these groups have changed or lost influence, in part due to anti-establishment claims. From this, however, new political elites have emerged, maintaining the link between private interests and the political class (Shifter, 2020) . We also observe that citizens who have been victims of assault report greater discontent, reflecting the importance of strengthening public safety institutions and law enforcement. Citizen safety is in fact one of the biggest challenges in the region, despite years of economic growth and socioeconomic improvement. For example, Latin American cities remain the most violent in the world (UN, 2017) . While the phenomenon is multidimensional, one of the major underlying causes is the lack of state capacityfrom police and prosecutors to the prison system-to address it (UNDP, 2013) . Trust in the media also has a negative relationship with discontent. Based on the work by Magallanes-Blanco and Treré (2019) , this may be read as a result of citizen distrust in mainstream media and in the emergence of new alternative channels that have gained a key role in providing space for counter-hegemonic discourses and information that is otherwise unavailable. Social media, in particular, has become a major channel through which social movements build bridges, organise their tactics and strategies and challenge mainstream media (Treré and Magallanes-Blanco, 2015; Valenzuela et al., 2016) . Moreover, in the 2019 wave of social protests across Latin America, social media allowed individuals to channel different forms of participation into mobilisation by creating spaces for easy and immediate organisation and action (Mascarell and Rodríguez, 2019; Shifter, 2020) . At the contextual level, contextual economic characteristics are not statistically significant. However, it must be stressed that the correlation is based on variation from 16 countries and thus the statistical power is low. In particular, if statistically different from zero, people in countries with higher government expenditure and trade openness would show more discontent than people in countries with lower levels of these. The contrary would be true for inflation, GDP per capita and unemployment. In Tables 2 and 3 , we report variables regarding inequality and institutional quality measures, respectively, that have been included individually in the regression models. We avoid including these results alongside individual regressors as in Table 1 due to the lack of space. However, the tables are available upon request. As observed in Table 2 , the country-level inequality measures we tested were not found to be statistically significant, although their signs are as expected. In this respect, our results support the claim by Lustig (2020) about questioning the relevance of the Gini coefficient as well as other traditional measures of inequality used to study social discontent. In Table 3 , among institutional quality measures the only (highly) significant variable is control of corruption. This result corroborates what is found for the variables at the individual level. It is worth mentioning, however, that the fact that contextual variables have no effect on discontent is not new. For example, Wiesehomeier and Doyle (2013) found a non-significant correlation between variables at the national scale and voting for leftist parties in Latin America. Our empirical analysis shows that discontent has different roots, which can mostly be explained from an individual point of view rather than from a contextual one. Our results confirm some of the main narratives that have emerged to explain social discontent in Latin America. Namely, we find that inequality of opportunity and poor institutional quality are associated with discontent, while traditional variables such as the Gini index may not be the most appropriate to analyze this phenomenon. Furthermore, personal characteristicsapart from socioeconomic status-are mostly non-significant. The relevance of institutional factors shows that despite the corruption and institutional weakness at all levels that have historically characterised the region, citizens are increasingly aware and dissatisfied with this situation. It should be noted that the increase in discontent has taken place in the midst of an unprecedented persistence of democracy in Latin America, lasting from the early 1990s, which has created opportunities for popular mobilisation and social pressure for more accountability. 5 In this context, new counter-hegemonic media spaces have emerged, unveiling some of the largest corruption scandals to have plagued the region recently (Kingstone, 2018) . Moreover, government configuration and operation aggravate discontent, since even in countries where new political groups have formed to break with the traditional political elites, these have managed to enforce regulations that allow them to hold onto power (Kingstone, 2018; Shifter, 2020) , undermining legitimacy and trust in politicians. It comes as no surprise that, tired and exasperated by the advantages to a few (old and new) elites that shape institutions with the aim of holding on to power, 73% of citizens across the region consider protest to be the only instrument available to them to fight corruption (EIU, 2020) . In addition to this, one of the main grievances during the 2019 protests was the regressive tax regimes in the region. Experts claim that Latin American countries have historically failed to tax their economic elites (Kingstone, 2018) and that their fiscal regimes may not be as redistributive as other regions of the world (Holland, 2018; Lustig, 2017) . Citizens, therefore, called not only for less impunity for representatives involved in corruption scandals but also for fairer tax systems (Ferreira and Schoch, 2020; Lafuente, 2019) . Taken altogether, the 2019 wave of social protest is also seen as a "rebellion against the political and economic elites in Latin America" (Shifter, 2020) and poses serious challenges to the political class. This is reflected in our results showing that the perception of inequality matters. Despite income inequality decreasing, as shown by measures such as the Gini coefficient, other features of inequality are persistently experienced, that is, opportunity and social mobility inequality. In this sense, we confirm the narratives developed to explain social discontent in Latin America by Ferreira and Schoch (2020) and Lustig (2020) . At the contextual level, however, we found no significance for almost all variable-categories including inequality and institutional quality, which could be due to the low variability in our dataset due to only 16 countries being included. We observe that only the control of corruption might help explain discontent with democracy. With the diffusion of participatory institutions in Latin America, which can be a means to improve government efficiency and reduce corruption (Goldfrank, 2021) , corruption scandals have been brought to the fore, undermining citizens' trust in democracy. To this extent, efforts to ensure that public policies are designed around the core principles of transparency, participation, accountability and integrity-and that they are implemented fairly-are required (OECD, 2020b). Our results also mark important differences with the parallel literature on social discontent in Europe. Out of the three usual variables of age, education and income that explain discontent in that context, income (approached through socioeconomic status) shows the only robust and strong correlation with discontent with democracy. Furthermore, that literature has highlighted the geographic component of discontent (Dijkstra et al., 2020; Rodríguez-Pose, 2018) , which we approached by including city size in the analysis but which is not statistically relevant in our case, probably due to the characteristics of the data we use. In the situation in which the world currently finds itself due to the COVID-19 pandemic, the relevance of considering and analyzing the factors affecting discontent becomes even clearer. A transparent debate on the multiple facets of inequality (particularly on IOp) and the design of comprehensive policies to face it has become more pressing. In Latin America, one in three families lives in deficient housing conditions, among which the lack of basic services (water, sanitation, electricity) and overcrowding (Libertun de Duren et al., 2018) make it extremely difficult to cope with stay-at-home policies. Furthermore, 47% of the population lacks access to social security, and the risk of impoverishment due to health expenditures is a threat for many households in the region, increasing their vulnerability (OECD, 2020c) . From this perspective, institutional reforms are also becoming increasingly urgent. Citizens' trust, however, is historically low, which not only limits the enforcement of measures to contain the health crisis, as well as the economic crisis derived from it, but also threatens the consolidation of democratic governance in the region. As part of an agenda to restore confidence in institutions, the UNDP emphasises the importance of transparency and distributional fairness (Sapienza, 2020) . Transparency in public administration must not only be enforced but transmitted to the population. Unfortunately, this seems rather far from becoming the reality: during the first months of the pandemic, disinformation, limited access to public information, excessive government reactions and overuse of force and the involvement of state and municipal officials, as well as politicians, in corruption scandals characterised the region (Harris et al., 2020) . Furthermore, the estimated social and economic effects of the pandemic are highly concerning. By the end of 2020, the poverty rate is expected to increase from 30% to 37% and unemployment is expected to rise by 5 percentage points to 13.5%. Inequality is expected to increase by 1.1% and 7.8% in several countries (UN, 2020) . Unfortunately, the fiscal capacity of governments across the region is constrained, limiting their ability to financially assist both companies and individuals. All of these new features aggravate the existing social discontent and have already led citizens in many countries to defy lockdown restrictions to hold public protests, including in Ecuador, Bolivia, Argentina, Brazil, Colombia, Mexico and Panama (Lister and Pozzebon, 2020) . One thing is clear: COVID-19 has exacerbated the region's structural inequalities and institutional weaknesses (Arnson, 2020; UN, 2020) , proportionally increasing the urgency of addressing them. Our empirical analysis offers an overview of the factors associated with discontent in Latin America by combining a number of features previously addressed independently or not explicitly linked to the outcomes of the 2019 annus horribilis. The results support the main narratives surrounding the phenomenon in the region and reveal important challenges for Latin America's society and political class, while confirming its complexity. Furthermore, our results show that discontent may share some factors globally, while others might be context-specific. The medium-and long-term effects of the COVID-19 crisis on social discontent are yet to be seen. However, given the importance of the socioeconomic factors in the region, we can infer that the longer and deeper the effects of the pandemic, the harsher the economic and social crisis that will be faced, triggering the conditions for greater social discontent. In any case, government action is essential and addressing the drivers of discontent has become imperative. For international organisations such as the OECD (2020b) and the United Nations (UN, 2020), the response to the pandemic should include taking the occasion to reconsider and change the region's development model and, ergo, the region's social contract. In light of our results, we believe this would involve at least a few facets. The first is building trust and credibility in representatives and institutions so that policy implementation eases, advancing inclusion, efficiency and compliance. Corruption needs to be tackled so that international inflows of financial and technical aid, and any resources overall, are handled for the benefit of all citizens. Furthermore, governments should move towards building comprehensive welfare systems that consider the complexity of the inequalities faced across the region. Lastly, and to support the previous item, countries should engage in building and strengthening progressive tax systems while also decreasing tax evasion and avoidance. However, advancing towards these objectives seems like a tough challenge due to their complexity and the restricted capability of governments. Forecasts in fact predict that the effects of the pandemic will indisputably deepen inequalities and increase social division, conflict and discontent (Lister and Pozzebon, 2020) . The starting point of this paper was the 2019 wave of protests in Latin America, a period that led many scholars to claim that citizens "lost faith in democracy" , that there was a crisis of democracy, and that it had never before been so threatened (Azpuru, 2019; Vlaicu, 2019; Zovatto, 2020) . Results from this analysis, as well as the current conditions and prospects for the region, lead us to believe that 2019 might have only been a preview of what is yet to come if the factors associated with discontent are left unaddressed. international conditions (Levitsky and Way, 2010; Mainwaring and Pérez Liñán, 2014) . Disentangling the Brexit vote: the role of economic, social and cultural contexts in explaining the UK's EU referendum vote The European trust crisis and the rise of populism What Covid-19 is Revealing About Latin America's Politics. Wilson Center Latin American Democracy may be in Trouble. The Protests are a Symptom of Increasing Mistrust Handling correlations between covariates and random slopes in multilevel models Social mobility and the demand for redistribution: the POUM hypothesis Taxing higher incomes: what makes the high-income earners consent to more progressive taxation in Latin America? Getting Development Right. Structural Transformation, Inclusion, and Sustainability in the Post-Crisis Era Inequality of opportunity in Europe: is there a role for institutions? Inequality: Causes and Consequences Citizen Security in Latin America and the Caribbean: Challenges and Innovation in Management and Public Policies Over the Last 10 Years Situación de las personas afrodescendientes en América Latina y desafíos de políticas para la garantía de sus derechos Comisión Económica para América Latina y el Caribe (CEPAL). (2019) Social Panorama of Latin America and the Caribbean Social mobility and the demand for income redistribution in Latin America The Association between Children's Earnings and Fathers' Lifetime Earnings: Estimates Using Administrative Data. Discussion Paper No. 1342-08 The geography of EU discontent Economic Mobility and the Rise of the Latin America Middle Class Inequality and Social Unrest in Latin America: The Tocqueville Paradox Revisited The Global Satisfaction with Democracy Inclusion without power? Limits of participatory institutions The Inclusionary Turn in Latin American Democracies In what sense left behind by globalisation? Looking for a less reductionist geography of the populist surge in Europe After the Boom -Commodity Prices and Economic Growth in Latin America and the Caribbean, Working Paper 14-154 Un) happiness in transition Coronavirus Corruption Cases Spread across Latin America International Institute for Democracy and Electoral Assistance (IDEA). (2019) The Global State of Democracy 2019: Addressing the Ills, Reviving the Promise The Worldwide Governance Indicators: Methodology and Analytical Issues The Political Economy of Latin America: Reflections on Neoliberalism and Development After the Commodity Boom Latin American Politics and Development, 9th edn Socioeconomic status measurement with discrete proxy variables: Is principal component analysis a reliable answer? La desigualdad moviliza a América Latina Banco de Datos Competitive Authoritarianism: Hybrid Regimes after the Cold War Vivienda ¿Qué viene?: De pensar la unidad a construir la ciudad Protests across Latin America Reflect a Toxic Cocktail of Pandemic and Recession Economic expectations and satisfaction with democracy: evidence from Italy, Government and Opposition El impacto del sistema tributario y el gasto social en la distribución del ingreso y la pobreza en América Latina. Una aplicación del marco metodológico del proyecto Compromiso con la Equidad (CEQ), El trimestre económico Desigualdad y descontento social en América Latina Contemporary social movements and digital media resistance in Latin America Democracies and Dictatorships in Latin America: Emergence, Survival, and Fall Political Regime Characteristics and Transitions, 1800-2018 Dataset Users' Manual Opinión | América Latina: Un grito contra la desigualdad urbana Latin America and the Caribbean in the New Decade. How Did We Get Here? Centre for Strategic and International Studies Wage Inequality in Latin America: Understanding the Past to Prepare for the Future Political discontent in new democracies: The case of Brazil and Latin America The Quality of Democracies in Latin America Intergenerational mobility and the rise and fall of inequality: Lessons from Latin America A tale of two crises? A regional-level investigation of the joint effect of economic performance and migration on the voting for European disintegration Organisation for Economic Cooperation and Development (OECD). (2020a) COVID-19 in Latin America and the Caribbean: An Overview of Government Responses to the Crisis Organisation for Economic Cooperation and Development (OECD). (2020b) COVID-19 in Latin America and the Caribbean: Regional Socio-economic Implications and Policy Priorities Organisation for Economic Cooperation and Development (OECD). (2020c) Government at a Glance: Latin America and the Caribbean 2020 Panorama Laboral 2017 América Latina y el Caribe. Tema Especial: Evolución de los salarios en América Latina en el periodo 2005-2015 Popular support for democracy and its institutions in Korea: The dynamics and source of regime support and institutional trust Global Corruption Barometer Latin America and The Caribbean The revenge of the places that don't matter (and what to do about it) Equality of opportunity: theory and measurement The surprising reduction of inequality during a commodity boom: what do we learn from Latin America Taking the Pulse of the COVID-19 Response in Latin America and the Caribbean Political trust and distrust in Switzerland: a normative analysis European disintegration? Euroscepticism and Europe's rural/urban divide The rebellion against the elites in Latin America, The New York Times Fixed and random effects Modeling multi-level data structures The Economist Intelligence Unit (EIU). (2020) Democracy Index 2019 A Year of Democratic Setbacks and Popular Protest The Heritage Foundation. (2020) 2020 Index Data United Nations Development Programme (UNDP). (2013) Regional Human Development Report 2013-2014. Citizen Security with a Human Face: Evidence and Proposals for Latin America. United Nations Development Programme Latin America and the Caribbean Education for All 2015. Regional Review. Oficina Regional de Educación para América Latina y el Caribe United Nations (UN). (2020) The Impact of COVID-19 on Latin America and the Caribbean (Policy Brief) Social media in Latin America: deepening or bridging gaps in protest participation? What Can Restore Latin Americans' Faith in Democracy. Ideas Matter IDB's Research Department Blog Lost in transition, found in recession? Satisfaction with democracy in Central Europe before and after economic crises Discontent and the left turn in Latin America World Bank Open Data Income inequality, distributive fairness and political trust in Latin America The Rapidly Deteriorating Quality of Democracy in Latin America The authors would like to thank two anonymous reviewers for their constructive and helpful comments, which greatly enhanced the coherence of the arguments. Funding for this research was granted by the XVIII call for research projects by the Research Department of the University of Cuenca (DIUC), Cuenca, Ecuador. Supplementary data are available at Cambridge Journal of Regions, Economy and Society Journal online. 1 Transition economies are generally characterised by lower levels of happiness (Guriev and Zhuravskaya, 2009 ) and satisfaction with democracy (Vlachová, 2019) . 2 Another potentially problematic country that we nevertheless decided to keep in our analysis is Nicaragua, which is classified as a hybrid regime by IDEA (2019) and as authoritarian by the EIU (2020).3 This variable, according to Benabou and Ok (2001) , can be seen together with the previous one on fairness given that an expectation of upward mobility would reduce support for high levels of redistribution: the so-called prospect of upward mobility (POUM) hypothesis. Empirical evidence is not uniform with respect to the Latin American context, with, for example, Berens and Schiller (2017) confirming the POUM hypothesis while da Fonseca Silva and de Figueiredo (2013) do not. Therefore, although we do not aim to test the POUM, we consider the inclusion of both variables important as this allows us to reduce misinterpretations of one or another due to omitted variable bias.