Introduction

The second half of the 20th century in Latin American history was marked by long periods of military and civil dictatorships. The last three decades, on the other hand, have been the most democratic ones in the history of the con­tinent (^{Levitsky, 2018}). However, although democracy is almost the norm, several analysts posit that democracy is still a “troubled system of governance” in the region (^{International IDEA, 2017a}, p 40).

Citizens’ attitudes are crucial in preventing or driving democratic backsliding. Positive attitudes to democracy can bolster collective resistance to state mis­use and help to avoid democratic backsliding, while negative attitudes can induce democratic backsliding and even the fall of democracy in a country (^{International IDEA, 2017b}). This is why studying the factors that affect atti­tudes towards democracy is important.

This paper aims to analyse how country-level and individual-level character­istics are related to citizens’ satisfaction with democracy in Latin America. Specifically, we focus on the role of a set of political and electoral variables such as the age of democracy in the country, how enforced voting is, the electoral system used to choose the parliament, the system used to elect the president, and how federal the system of government is. As a measure of the level of satisfaction with democracy, we use a survey question from the Latinobarometer ²⁰¹⁵.

Using survey questions to measure the level of satisfaction with democracy around the globe has been very popular in recent decades. But what do such questions actually measure?⁴ We concur with Linde and Ekman (²⁰⁰³) who argue that satisfaction with democracy “is an indicator of support for the performance of the democratic regime” (p. 399). In other words, it measures people’s evaluations of the political regime (^{Klingemann, 1999}). It is not, how­ever, an indicator of the level of support for the principles of democracy (^{Linde & Ekman, 2003}), which is actually measured in survey questions regarding support for democracy or the political regime.

Previous studies point out the role of the current macroeconomic conditions as well as socio-demographic factors on satisfaction with democracy. Thus, we control for a large set of macroeconomic and socio-demographic variables. Of equal or even more importance, as shown in the literature, is whether a citizen’s ideology is represented in the government. Thus, we also control for whether the survey respondent supports a party that is represented in the government.

We find that people are more satisfied with democracy when members of the parliament are elected by a proportional electoral rule, while the rule used to elect the president is only partially associated with satisfaction with democracy. Furthermore, the level of satisfaction is lower in older democracies: i.e. the age of democracy is negatively related to satisfaction with democracy. Whether voting is compulsory and how strictly it is enforced, however, is not related to satisfaction with democracy. The same happens with the degree of federalism.

Since we interpret satisfaction with democracy as an indicator of support for the performance of the democratic regime, and because changes in electoral rules and institutions could be the result of citizens’ dissatisfaction with the political regime, an endogeneity problem is a possibility. We discuss this issue in Section II.

This paper makes two contributions to the literature. To the best of our knowl­edge, we are the first to consider an extensive set of institutional and electoral variables for Latin American countries in order to study their relationship with satisfaction with democracy. We are also the first to investigate the nexus between the degree of federalism, the electoral system used to choose the president, and the level of satisfaction with democracy.

The paper is organized as follows: The next section reviews the literature and presents our hypotheses. Section II presents the data and describes the meth­odology we use. Section III details our results, and the last section discusses our findings and contains the concluding remarks.

I. Literature review and hypotheses

In this section, we present the relevant literature regarding our variables of interests, i.e. variables related to a country’s political system, and some control variables at the country level. We also formulate seven hypotheses about the relationship between the characteristics of the political system and satisfaction with democracy.

While on the one extreme Costa Rica has been under democratic rule for more than six decades, democracy has been the norm in many Latin American coun­tries only for less than three decades. So, how does how old the democratic system is relate to the level of satisfaction with democracy in Latin America?

As Keefer (²⁰⁰⁷) posits, young democracies suffer from the “inability of polit­ical competitors to make broadly credible preelectoral promises to voters” (p. 804), resulting in higher levels of corruption and political violence and more social fragmentation. As a democracy gets older, politicians experience more democratic competition that eventually improves their political credibility.

There are several studies analysing the determinants of satisfaction with democracy in Europe, which also focus on the role that age of democracy plays. These studies include both Western European countries and also more recently democratized Eastern European countries. Armingeon and Guthmann (²⁰¹⁴) expect to observe stronger support for democracy in older democracies and confirm this empirically analysing 26 EU countries. Reher (²⁰¹⁵) also finds that how old the democracy is has a positive effect on satisfaction with democ­racy. She also points out that there is a strong correlation between the age of democracy and the quality of democracy in the 28 EU countries she analyses. Moreover, she notes that for the case of the relatively recently democratized countries in Europe, the transition was not only a change in the political sys­tem but in also the economic system.

On the other hand, Anderson and Tverdova (²⁰⁰³), using survey data for Australia, Canada, Japan, New Zealand, the U.S., five EU, and six (then) non- EU European countries, and Stockemer and Sundström (²⁰¹³) who analyse 23 EU countries, Israel, Ukraine, and Russia, do not find that the age of the democracy has any significant effect on satisfaction with the democracy. The difference between these authors’ country samples and those of Armingeon and Guth­mann (²⁰¹⁴) and Reher (²⁰¹⁵) may explain the differences in their findings.

Fernandez and Kuenzi (²⁰¹⁰) argue that the lower level of satisfaction with democracy in Latin America compared to that in Africa could be due to the fact that in Latin America people have had more time to become disillusioned with democracy. The transition from authoritarian regimes to democracy might initially lead to high levels of satisfaction with democracy due to increasing freedoms, more participatory decision-making mechanisms, and several other factors. As pointed out by Wagner, Schneider and Halla, (²⁰⁰⁹), people living in more democratic countries are more satisfied with democracy. The fact that most Latin American democracies are still far from ideal in terms of the qual­ity of democracy, reflected for instance in the Freedom Index, might have led to disillusionment over time in terms of how people perceive their democra­cies. Thus, we would expect satisfaction with democracy to decrease over time.

Hypothesis 1: In Latin America, satisfaction with democracy decreases with how old the democracy is.

In Latin America, voting in both presidential and parliamentary elections is compulsory in some countries but not in others. Berggren, Fugate, Preuhs and Still, (²⁰⁰⁴) argue that citizens may resent being forced to vote and the cost of voting may outweigh the potential benefits. Using the Eurobarometer sur­vey, they find that satisfaction with democracy is lower in countries where voting is compulsory.

The fact that voting is compulsory, however, should not play a major role, unless the law is enforced. How strictly voting is enforced, however, varies across the continent. Singh (²⁰¹⁸) posits that compulsory and enforced vot­ing may have two effects on satisfaction with democracy. On the one side, participation will be higher if voting is enforced and a higher participation rate would improve how citizens perceive the legitimacy of the democratic process and would enhance attitudes toward democracy. On the other hand, forcing people to participate in the democratic process may strengthen the negative feelings towards democracy, especially for those who have more anti-democratic attitudes.

In the light of the arguments by Berggren et al. (²⁰⁰⁴) and Singh (²⁰¹⁸), and combined with the fact that satisfaction with democracy has been declining in Latin America (^{Levitsky, 2018}), we would expect satisfaction with democ­racy to be lower the higher the level of voting enforcement.

Hypothesis 2: The stricter voting enforced is, the lower the satisfaction with democracy.

The composition of the parliament in a country clearly depends not only on for whom or for which party voters cast their votes but also on how those votes are transformed into seats: the electoral rule. Thus, one could expect that the electoral rule used to choose parliamentary members might affect the level of satisfaction with democracy. Berggren et al. (²⁰⁰⁴) argue that smaller districts establish a closer tie between voters and their representatives, and, thus, they conjecture that satisfaction with democracy would be lower in large-district proportional representation systems than in single-member district systems.

Their analysis confirms the argument. Several other studies, conversely, such as Guldbrandtsen and Skaaning (²⁰¹⁰), Listhaug, Aardal and Ellis, (²⁰⁰⁹), and Rich (²⁰¹⁵) do not find any significant effect of the electoral rule used to choose the legislature on satisfaction with democracy. Lijphart (²⁰¹²), however, concludes that citizens in countries with proportional representation systems have higher levels of satisfaction with their democracy. Singh (²⁰¹⁴) confirms the conclusion.

Aarts and Thomassen (²⁰⁰⁸) find that satisfaction with democracy depends positively on the perception of representativeness.⁵ One might argue that the more proportional an electoral system is, the more representative the parlia­ment is in terms of different political views, ethnicities, minorities, etc. This, in turn, implies that consensual systems are more inclusive in their power-sharing (^{Listhaug et al., 2009}). Thus, we expect that more extensive repre­sentation goes together with a higher level of satisfaction with democracy.

Hypothesis 3: Satisfaction with democracy is higher when a proportional elec­toral rule is used for parliamentary elections.

All the Latin American countries we consider in our analysis are presidential democracies. There is, however, variation in how presidents are elected. Some countries, such as Uruguay and Chile, use a two-round majority system (also called absolute majority rule), others such as Argentina and Costa Rica use a qualified plurality rule, whereas others such as Panama and Paraguay use a plurality or first-past-the-post (FPTP) rule. Theoretically, in a country using the plurality rule for presidential elections where, say five candidates com­pete, the president might be elected with as little as twenty percent of the votes. On the other hand, if the rule used is an absolute majority system, the elected president will have received the absolute majority of the votes, at least in the second round. In other words, it is more likely that the elected president has been supported by an absolute majority, or, at least, it is much less likely that s/he is a “bad choice” for a vast majority. Thus, we would expect an electoral rule for choosing the president that prevents the election of a candidate who has little support to be associated with a higher level of satisfaction with democracy.

Hypothesis 4: Satisfaction with democracy is higher when an absolute majority system or a qualified plurality rule is used to choose the president rather than a plurality (FPTP) rule.

In Latin America, there are a few countries such as Argentina and Brazil that have a federal system of government; the majority are unitary states. Norris (¹⁹⁹⁹) studies the effect of institutional characteristics on trust in institutions and she argues that federal states are more flexible and therefore would be able to integrate more diverse interests and thus create more trust in its insti­tutions.⁶ If a higher degree of federalism implies more integration of diverse interests, we would, in these countries, expect to observe higher levels of sat­isfaction with democracy than in unitary states.

Hypothesis 5: The higher the degree of federalism in a country, the higher the satisfaction with democracy.

As stated before, we interpret satisfaction with democracy as a measure of how people evaluate the performance of the regime. A democratic regime’s performance greatly depends on the performance of the governing party or parties. The evaluation of such a performance, in most cases, would be of a greatly subjective nature. Therefore, in their seminal work, Anderson and Guillory (¹⁹⁹⁷) argued that people’s satisfaction with democracy would be influenced by whether they are among the winners or losers of the latest elections, that is, whether they voted for the party (or a party) in government in the last election. They found this to be true. Following their work, several other authors, including Anderson, Blais, Bowler, Donovan and Listhaug (²⁰⁰⁵), Anderson and Singer (²⁰⁰⁸), Anderson and Tverdova (²⁰⁰³), Banducci and Karp (²⁰⁰³), Curini, Jou and Memoli, (²⁰¹²), Farrer and Zingher (²⁰¹⁸), Leiter and Clark (²⁰¹⁵), Listhaug et al. (²⁰⁰⁹), Lühiste (²⁰¹⁴), Schäfer (²⁰¹²), Singh (²⁰¹⁴), and Rich (²⁰¹⁵) all obtained the same result. Therefore, we would expect to observe the same phenomena in Latin American countries. However, since we do not have information in our database about the party the respondent has voted for in the last election, we use as a proxy a variable created by the Latinobarometer organization: if the party the respondent would vote for is ruling the country. Anderson et al. (²⁰⁰⁵) discuss the advantages and disad­vantages of using any of those two types of survey questions. Using data from a Eurobarometer survey, they conclude that there is little difference between classifying survey respondents as winners or losers on the basis of the past vote or future vote. Therefore, we expect that respondents who would vote for a party that is ruling the country at the moment the survey was conducted, would be more satisfied with the way democracy works.

Hypothesis 6: People show higher satisfaction with democracy if the party they would vote for is in government.

One serious problem that institutions and probably many of the societies as a whole have been suffering in Latin America is corruption. Corruption has seri­ous negative impacts on societies for several reasons. On the one hand, cor­ruption negatively impacts the economy (^{Rose-Ackerman, 1999}; ^{Lambsdorff, 1999}), and on the other, it erodes the perception of fairness and equality that citizens believe institutions must adhere to (^{Chang & Chu, 2006}). This ero­sion, then, may lead to distrust in institutions and hence a lower satisfaction with the system. Empirically, the impact corruption has on satisfaction with democracy has already been established. Wagner et al. (²⁰⁰⁹), using the World Bank Control of Corruption Indicator, and Stockemer and Sundström (²⁰¹³) using the same indicator as well as individual perceptions of corruption find that a higher corruption perception implies lower satisfaction with democ­racy. By using the Corruption Perception Index undertaken by Transparency International as a measure of corruption perception, Anderson and Tverdova (²⁰⁰³) obtained the same finding.

As they are more likely to benefit from corruption, Anderson and Tverdova (²⁰⁰³) argue that citizens who are closer to the current government have a more benign view of corruption. In their empirical analysis, they find that cor­ruption has a less negative effect on the evaluation of the system of voters for a party in the government. Thus, we expect to observe the same behaviour.

Hypothesis 7: While satisfaction with democracy is lower in countries with a higher corruption perception level, the negative effect of corruption on satis­faction with democracy is lower for voters of the parties in government.

Besides the nexus between the political system and satisfaction with democ­racy, several authors have also studied the role that macroeconomic variables play on people’s level of satisfaction. For instance, regarding the effect of the unemployment rate on satisfaction with democracy, while Friedrichsen and Zahn (²⁰¹⁴), Leiter and Clark (²⁰¹⁵), and Wagner et al. (²⁰⁰⁹) found that a higher unemployment rate implies lower satisfaction with democracy, Anderson and Singer (²⁰⁰8) and Armingeon and Guthmann (²⁰¹⁴) did not find any significant relationship between those two variables. While Wagner et al. (²⁰⁰⁹) found a significant negative relationship between the infla­tion rate and satisfaction with democracy, Friedrichsen and Zahn (²⁰¹⁴) did not verify that finding. In terms of the effect of growth, while Anderson and Tverdova (²⁰⁰³) and Friedrichsen and Zahn (²⁰¹⁴) found a positive relation­ship between the annual GDP growth rate and satisfaction with democracy, Anderson and Singer (²⁰⁰⁸) and Stockemer and Sundström (²⁰¹³) found no significant relationship.

Considering all the above-mentioned findings, we include the inflation rate, the unemployment rate, and the annual GDP growth rate in our analysis as control variables.

While some studies found the unemployment rate had a negative effect of on satisfaction with democracy, the effect should be stronger if it personally affects the individual. In fact, Anderson and Singer (²⁰⁰⁸), Friedrichsen and Zahn (²⁰¹⁴), and Schäfer (²⁰¹²) found that being unemployed has a signifi­cant negative impact on satisfaction with democracy. For that reason, we also include whether or not a respondent is unemployed as one of our socio-demographic variables.

II. Database and methodology

Our empirical research is conducted using data from 18 Latin American coun­tries (Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, and Venezuela). The data source is the Latinobarom­eter Survey (²⁰¹⁵), which provides information about individuals’ opinions, attitudes, behaviours, values, and socio-demographic characteristics (includ­ing religious affiliation and practices). The countries’ sample sizes is either one-thousand (Costa Rica, Dominican Republic, El Salvador, Guatemala, Honduras, Nicaragua, and Panama), 1200 (Argentina, Bolivia, Chile, Colombia, Ecuador, Mexico, Paraguay, Peru, Uruguay, and Venezuela), or 1250 (Brazil). We use the weights provided by the Latinobarometer, which change from country to country. Details for the methodology used are presented in a methodology report written by Latinobarometer.

Our dependent variable is based on the following question: “In general, would you say that you are very satisfied, quite satisfied, not very satisfied or not at all satisfied with the working of the democracy in (country)?” The possi­ble answers range from 1 (Not at all Satisfied) to 4 (Very Satisfied). We have inverted the original answers to make it easier to interpret the results. Thus, a larger value of the variable implies more satisfaction with the workings of democracy. We call this variable Satisfaction with Democracy (SWD).

We have run multilevel ordered logit models with three sets of independent variables: country-level political variables, other country-level variables, and personal socio-demographic control variables. Tables A1 and A2 in the appen­dix show a summary description of the variables used in our analysis. Table A1 shows the variables we take from the Latinobarometer (including the depen­dent variable) and Table A2 shows the rest of the variables.

Our main independent variables are the ones reflecting the political system in our set of countries. They include a variable reflecting how old the demo-cracy is (Age democracy), a variable showing the level of enforcement of vot­ing (Voting enforced),⁷ a dummy variable that takes value 1 if the election of the parliament follows proportional representation (Elections-Proportional representation), a dummy variable that takes value 1 if the election of the president follows an absolute majority rule (Elections for president – AMR), a dummy variable that takes value 1 if the election of the president follows a qualified plurality rule (Elections for president – QPL) –meaning that the reference category is plurality (FPTP) rule, where the candidate who receives more votes wins the elections–, and a variable that shows the political sys­tem’s degree of federalism (Degree of federalism).⁸

The second set of variables consists of country-level variables: Gini index, infla­tion rate, unemployment rate, openness to trade index, size of the shadow economy, GDP growth, corruption level, freedom index, and human develop­ment index.

The Gini coefficient is included because Anderson and Singer (²⁰⁰⁸), Berry and Tello Rodriguez (²⁰¹⁰), and Wu and Chu (²⁰⁰⁷) explore the effect of income inequality measured by the Gini coefficient on SWD and find that higher inco-me inequality leads to lower SWD. Regarding the size of the shadow economy, Wagner et al. (²⁰⁰⁹) find that the smaller the weight of the shadow econo-my the higher SWD.

The data for these country-level variables, including the variables in the first group, come from different sources, which are indicated in Table A2 in the appendix.

The variables in the third group are from the same source as our dependent variable, i.e. the Latinobarometer. This group includes: the sex of the respon­dent (Male, takes value 1 if the respondent is a man), the age (Age), the level of education –we include four dummy variables representing educational attainments from incomplete secondary or technical education to complete higher education–, two variables reflecting the respondent’s religion (Catho­lic and Other Christian religion), two dummies about the level of religiosity (Religious person and Very religious person), two variables reflecting marital status (Married⁹ and Single), and a dummy variable that takes value 1 if the respondent is unemployed (Unemployed).

The sex of the respondent is included as a control variable because Leiter and Clark (²⁰¹⁵) and Schäfer (²⁰¹²) found that males are more satisfied with democracy. We control for age because Anderson and Singer (²⁰⁰⁸), Friedrichsen and Zahn (²⁰¹⁴), and Leiter and Clark (²⁰¹⁵) found a negative relationship between age and SWD. The dummy variables reflecting the edu­cation level of the respondents have been included because Leiter and Clark (²⁰¹⁵) and Schäfer (²⁰¹²) found that more educated subjects are more satis­fied with democracy, while Berry and Tello Rodriguez (²⁰¹⁰) found the opposite. The religion and religiosity of the respondents are control variables because Schäfer (²⁰¹²) found that those with stronger religious beliefs are more sat­isfied with democracy.

In the second model, we also include a variable that shows if the subject sup­ports one of the parties in the government or not (Winner),¹⁰ and an interac­tion variable between Winner and the level of corruption in the country.

We study how the characteristics of the political system are related to the level of SWD. There is, however, the possibility of an endogeneity problem between the variables reflecting the electoral system, the degree of enforcement of vot­ing, the degree of federalism, and our dependent variable (SWD). That is, the electoral rules or the rules concerning voting might have been changed as a reaction to low levels of satisfaction with the previous system.

Analysing why countries that have compulsory voting rules adopted such rules, Singh (²⁰¹⁸) observes that they have been adopted for a variety of reasons. In most Latin American countries, compulsory voting was adopted at the same time as universal (or universal male) suffrage. He argues that this suggests such rules were not adopted as a reaction to a specific need, but instead as part of wider changes towards more democratic governments. On the other hand, Singh (²⁰¹⁸) argues that compulsory voting in Latin America is also tied to authoritarian forms of government: i.e. such rules were implemented by or kept by authoritarian regimes. Furthermore, Helmke and Meguid (²⁰¹⁰) observe that implementing compulsory voting is the result of strategic con­siderations of those in power as such rules would benefit them. Singh (²⁰¹⁸) concludes that all the examples across the globe cast doubt on the argument that countries with dissatisfied electorates implement compulsory voting rules.

Regarding electoral rules, the literature analysing changes in electoral rules empirically and theoretically argues that the rules would be changed when such changes would help to keep those in power in government or to reduce their electoral losses (see for instance ^{Benoit, 2004}; ^{Boix, 1999}; ^{Colomer, 2005}; ^{Ergun, 2010}; ^{Matakos & Xefteris, 2015}, and ^{Remmer, 2008}, for Latin America).

Selee (²⁰⁰⁴), observing that many Latin American countries went through a decentralization process during the 1980s and 1990s, states that the reason why those countries transitioned was to make sure that the central state would not have such power as was the case during the authoritarian regimes.

In light of the above arguments, we consider the potential endogeneity prob­lem not to be a very likely problem although we acknowledge that we cannot completely rule out this possibility. Hence, in what follows, we refrain from establishing strong causal relationships regarding these variables.

III. Results

This section presents a brief descriptive analysis of the data followed by an econometric analysis. Figure 1 shows the average SWD for all the 18 coun­tries in our sample. The average level of SWD in the 18 countries considered is 2.31. The minimum level is observed in Mexico (1.86) and the highest level in Uruguay (2.89); i.e. Mexicans are the less satisfied and Uruguayans are the most satisfied with the way their democracy works.

Source: Latinobarometer (²⁰¹⁵) based on the question: “In general, would you say that you are very satisfied, quite satisfied, not very satisfied, or not at all satisfied with the working of the democracy in (country)?” The answers range from 1 (Not at all Satisfied) to 4 (Very Satisfied).

Figure 1. Average satisfaction with democracy per country 

The following presents the results of two multilevel ordered logit models. The variables in Model I are organized into three groups. The first two groups include variables at the country level: the political variables group and the other country variables group. The third group consists of personal socio-demographic variables. Since this last group of variables is not crucial in our analysis but it comprises control variables, the results are not included in the tables, but we briefly comment on the variable reflecting the respondents’ job status. Model II includes the same variables as Model I plus a dummy variable that shows if the party the respondent supports is ruling the country (Winner) and an interaction between this variable and the country’s level of corrup­tion. Both models are shown because including the variable Winner implies reducing the sample size by half. For that reason, in what follows, we inter­pret the results of Model I except when we discuss the effect of supporting a party that forms part of the government. The estimates are presented in Table 1 and the marginal effects in Table 2.

Table 1. Multilevel regressions 

* significant at 10%, ** significant at 5%, *** significant at 1%Control variables included: sex, age, education, religion, religiosity, marital status, job status.AMR: Absolute majority rule, QPL: Qualified plurality rule.

Table 2. Marginal effects of the ordered logit models 

** significant at 10%, ** significant at 5%, *** significant at 1% Control variables included: sex, age, education, religion, religiosity, marital status, job status. SWD: Satisfaction with democracy, AMR: Absolute majority rule, QPL: Qualified plurality rule.

We mainly focus our attention on the country-level political variables, namely: how old the democracy is, how strictly enforced voting is, the electoral sys­tems used to choose the parliament and the president, and how federal the country is.

Regarding how old the democracy is, i.e. how many years have passed since the country returned to democracy, we find that it has a negative relation­ship with satisfaction with democracy: the more years that have passed since the end of a non-democratic regime, the less satisfied with democracy the respondents are. The estimate of the marginal effects shows a positive varia­tion of 0.2 percentage points (pp) of the probability of answering “Not at all Satisfied with Democracy” (NSD) if Age of democracy increases by one year. Conversely, there is a negative variation of 0.1 pp in the likelihood of a “Very Satisfied with the Democracy” (VSD) result. This supports Hypothesis 1. In Latin America, most of the countries have been democratic for a quite long time now, and it seems that the initial enthusiasm has been washed away with time. Our result is in line with Fernandez and Kuenzi’s (²⁰¹⁰).

The second hypothesis refers to the level of voting enforcement in the coun­try: the stricter that voting is enforced, the lower the SWD would be. Table 1 shows that there is no correlation with SWD, i.e. we cannot find support for Hypothesis 2.

The third political variable considered is the rule used to elect the parliament. Table 1 shows that having a proportional representation system has a posi­tive significant relationship with SWD, which supports Hypothesis 3: SWD is higher in countries which use a proportional electoral rule for parliamentary elections. The most likely explanation lies in the fact that a more proportional system represents a larger range of political views, ethnicities, minorities, etc. A proportional representation system increases the likelihood of responding VSD by 10.1 pp and decreases answering NSD by 25.6 pp.

Regarding the system used to elect the president, the dummy variable Elections for president - AMR does not have a significant relationship with SWD, while the variable Elections for president – QPL does: There is a positive relationship between having a Qualified Plurality Rule to elect the president and SWD in a country. Having QPL increments the likelihood of responding VSD by 6.5 pp and decreases answering NSD by 8.9 pp. Hypothesis 4 is partially supported.

The last political variable analysed is how federal the political system is. We find no correlation with SWD, which implies that no evidence to sustain Hypothesis 5 is found.

These six variables are our main focus of interest. What have we found? We found that the more time elapsed since the last non-democratic regime, the less satisfied the citizens are with democracy; the use of a proportional rep­resentation rule to elect the parliament correlates with a higher SWD; having a QPL to elect the president is positively related to SWD; and when there is an AMR to elect the president, the level of voting enforcement and the degree of federalism in a country are not related to SWD.

After running the ordered logit multilevel model, we ran a Wald test in order to test whether the coefficients of the variables related to the political sys­tem are simultaneously equal to zero. Based on the p-value (0.0000), we are able to reject the null hypothesis, which indicates that the coefficients of the variables related to the political system are not simultaneously equal to zero. This means that including these variables leads to a statistically significant improvement in the fit of the model.

In Model II, the variable Winner is included; it takes value 1 if the respondent would vote for a party that is in the current government. Its positive and sig­nificant effect implies that when the party the respondent supports is ruling (alone or with other parties) s/he is more satisfied with democracy. This result supports Hypothesis 6: people show higher satisfaction with democracy if the party they would vote for is part of the government. Being a winner increases the probability of VSD by 9.8 pp and decreases NSD by 13.1 pp.

Following Anderson and Tverdova (²⁰⁰³), we also include an interaction between the variable Winner and the level of corruption in the country (Corruption level is one of the variables included in the group “Other coun­try variables”) as an independent variable in model II. We find support for the first part of Hypothesis 7: corruption is negatively related to SWD; however, the interaction variable has a statistically insignificant effect, meaning that there is no support for the second part of Hypothesis 7.

The rest of the variables are mainly control variables: some at the country level and others at the personal level. They are included because they have been found to be related to SWD in previous research papers.

Regarding the country-level control variables, only some have a significant effect on SWD. We find that the Gini index and the Corruption level are nega­tively related to SWD. Conversely, Growth rate, Inflation rate, and Human devel­opment index are positively related to SWD. Surprisingly, Size shadow economy has a positive effect while Freedom index has a negative effect. As a smaller shadow economy might be reflecting higher-quality institutions (^{Wagner et al., 2009}), one would expect to observe a negative relationship between SWD and the size of the shadow economy. This is precisely what Wagner et al. (²⁰⁰⁹) found in the case of European countries. Maciel and de Souza (²⁰¹⁸), on the other hand, do not find any significant relationship between these two variables for 27 EU countries. The reason why we find a negative relationship requires further analysis, which is beyond the scope of this paper. Regarding the Freedom Index, even though it would be expected to find a positive rela­tionship with SWD, as some authors found (for instance, ^{Ariely, 2013}), many other researchers found insignificant results (^{Anderson & Tverdova, 2003}, ^{Listhaug et. al. 2009}, ^{Guldbrandtsen & Skaaning, 2010}). The negative relation­ship is rather unexpected, which shows that this issue warrants more investi­gation. Since the Freedom Index is a multi-dimensional variable, it would be interesting to study the relationship between each of its components and SWD.

Regarding the personal control variables: most do not have a significant effect on SWD. Being a male and being Catholic have a positive effect on SWD. On the other hand, being unemployed has a negative effect.

We also run two Logit models with the same set of variables where the depen­dent variable (Satisfied with democracy) takes value 1 if SWD is equal to 3 (Quite satisfied) or 4 (Very satisfied) and value 0 otherwise. The results are almost the same with only some slight changes in the statistical significance of just two control variables. The estimates can be found in the appendix (Table A3).

IV. Discussion and concluding remarks

Our objective in this paper was to analyse the nexus between the rules and characteristics of the political system with satisfaction with democracy in Latin America. We found that some of the characteristics of the political sys­tem that we hypothesized to be potentially influential on satisfaction with democracy, such as the electoral rule used to choose the legislature or how old the democracy is, are indeed related to satisfaction with democracy. One important exception, however, is the degree of federalism.

We conjectured that people’s satisfaction with democracy would be higher in countries which have a more federal system because a more federal system would allow better integration of diverse interests. The fact that we did not find a significant relationship between the degree of federalism and satis­faction with democracy could potentially be explained by two arguments. As pointed out by Norris (¹⁹⁹⁹), it is possible that federalism is a poor measure of decentralization. Additionally, accountability of the ruling elites would be higher in more unitary states (^{Norris, 1999}), which could potentially offset the positive link between federalism and satisfaction with democracy.

Regarding the electoral rule used to choose the president, as expected, we found that satisfaction with democracy is higher in countries that use a quali­fied plurality rule rather than a plurality (first-past-the-post) rule. Surpris­ingly, we did not find any statistically significant difference in the level of satisfaction with democracy in countries which use an absolute majority rule and with those using a plurality rule. This is an interesting issue which cannot be explained by our current research. More research on the topic could shed light on why this is observed.

We found a negative relationship between the age of democracy and satis­faction with democracy, that is, citizens in countries with a longer uninter­rupted period of democratic rule are less satisfied with the way democracy works. A fading satisfaction with democracy over time might have worrisome implications such as an erosion in support for democracy. This was pointed out by Sarsfield and Echegaray (²⁰⁰⁶), who show that satisfaction with democ­racy affects support for democracy. In other words, the higher the dissatisfac­tion with democracy, the more likely support for democracy will be eroded.

Finally, what do our findings suggest about what could be done to increase satisfaction with democracy and stop the decline of support for democracy in the region? On the political side, a more representative legislature, as indi­cated by the finding regarding the electoral rule used to choose the legisla­ture is a potential factor that could increase satisfaction with democracy. On the economic side, our results suggest that a stronger fight against corruption might improve the level of satisfaction with democracy. Moreover, observing structural discontent across Latin America, Levitsky (²⁰¹⁸) argues that high degrees of social inequality may be a driving force. His observation combined with our findings suggests that policies aimed at reducing income inequal­ity in the region could boost citizens’ satisfaction with the political regime.