Abstract
Many Latin American countries democratized between 1975 and 2000, and research has confirmed that contemporary Latin Americans hold democratic political attitudes. Using AmericasBarometer surveys of 18 countries from 2008, we examine the commitment of Latin Americans to three democratic attitudes – preference for democracy over other forms of government, support for general participation rights, and tolerance for participation by system critics. We also explore the impact of personal resources, crime and corruption, evaluation of system performance, social capital, and the sociopolitical context on democratic attitudes. A preference for democracy and support for citizens’ participation rights are strong, but tolerance is lower than the other attitudes. Evidence is found for acculturation – that Latin Americans acquire democratic attitudes by living in democratic regimes and through education.
Introduction
Latin America between 1975 and 2000 experienced an unprecedented wave of democratization (Doorenspleet 2005; Huntington 1991; Smith 2005) 1 . Most of these newly democratic nations have so far conserved their democratic rules of the game, but not all. Observers have noted an erosion of democracy in certain countries in the region (Freedom House 2010; Marshall and Jaggers 2010), and others have questioned both the quality of democracy (Smith 2005) and the commitment of Latin Americans to basic democratic values (Camp 2001; Lagos Cruz-Coke 2001, 2008; Smith 2005). Lagos Cruz-Coke, for example, reported that from 1996 to 2000 in eighteen Latin American countries the average level of support for democracy was about 60 percent and that it remained unconsolidated and “in serious danger” (Lagos Cruz-Coke 2001). Such assessments raise concerns for anyone who supports democracy in the region, especially after its lengthy track record of authoritarian rule. If citizens’ attitudes matter at all in keeping democracy, it is worthwhile to assess how committed Latin Americans are to democratic values, and what factors might strengthen or undermine them.
An earlier draft of this paper was presented at the Midwest Association of Latin American Studies (MALAS), Dallas, Texas, 20 November, 2009. We thank three anonymous reviewers of this journal for suggestions on an earlier draft of the paper.
We employ survey data collected in 2008 to analyze Latin Americans’ commitment to three important democratic attitudes – expressed preference for democracy over alternative systems of government, support for basic political participation rights, and tolerance of participation by regime critics. We find evidence that levels of expressed preference for democracy are higher than previously cited findings suggest, but that other attitudes related to democracy had somewhat lower support. Because factors that might alter these democratic attitudes are critically important to their survival, we also examine what might shape the stability or consolidation of such democratic attitudes among Latin Americans.
Theory
Theorists have offered many explanations for citizens’ commitment to democratic attitudes, and varying interpretations of how important such attitudes are to the development and persistence of democracy itself. Many focus on the characteristics of individuals, often sharing the assumption that aggregated individual attitudes strongly shape regime characteristics. Schumpeter, Adorno et al., and Lipset have argued that mass publics are inherently illiberal and a threat to democratic systems, that poverty contributes to holding authoritarian norms, and that poor countries with many poor people are therefore less likely to be democratic (Adorno et al. 1950; Lipset 1959a, 1959b; Schumpeter 1943). Vanhanen and Rueschemeyer et al. extended this argument to the resources for political engagement available to individuals or to classes of individuals. Key resources include literacy, various types of wealth, and organization (Rueschemeyer, Stephens, and Stephens 1992; Vanhanen 1997). Smith, Rose et al., and Booth and Seligson have argued that, once democratic rules of the game are in place, citizens’ experiences under the democratic regime and their evaluations of it may strengthen or weaken support for democracy (Booth and Seligson 2009; Rose, Mishler, and Haerpfer 1998; Smith 2005). Experiencing official corruption, for example, might undermine a citizen's support for democracy by eroding confidence in political actors, institutions, or the system of government in general (Cruz 2008; Seligson 2002, 2006; Seligson et al. 2000). Putnam, Coleman and others contend that social capital, especially participation in civil society organizations and one's level of interpersonal trust, enhance democratic attitudes (Coleman 1976; Putnam 2000, 1993).
Other theorists who assume that the sociopolitical context exercises a strong influence over individual attitudes have focused on how system-level economic and political factors may affect democratic values. Among the contextual factors that may strengthen support for democracy are the quality of the democratic rules of the game (Muller and Seligson 1994), national economic performance (Jackman and Miller 2004; Lipset 1959b), and spatial effects such as geographic proximity to other democracies (Brinks and Coppedge 2006; Huntington 1991; O'Loughlin et al. 2004). 2
In previous research we found an effect of geographical proximity to the United States on democratic attitudes in Latin America (Salinas and Booth 2009a, 2009b). We had anticipated a simple linear effect (higher democratic attitudes in nations nearest to the United States). Instead, we found a U-shaped relationship between proximity to the United States and higher democratic attitudes. Citizens of several Andean nations had relatively lower democratic attitudes and showed slightly less support for basic participation rights and political tolerance than citizens of countries that lay both closer to and further from the United States.
Thus arguments vary on whether the main direction of causality for democracy goes from citizens’ attitudes to system characteristics or vice versa. Almond and Verba (Almond and Verba 1963), Lipset (Lipset 1959b), and Putnam (Putnam 2000, 1993) generally assume that the main direction of causality is for citizen attitudes to shape system rules. Muller and Seligson and Jackman and Miller, in contrast, contend that political culture is adaptive and mainly reacts to system rules and characteristics imposed by elites. Inglehart argues for both paths of influence (Inglehart 1988). Indeed, it seems likely to us that both paths are possible, and that the influence is recursive.
We propose in Figure 1 a model of factors likely to influence citizen democratic attitudes in Latin America. First, and most obviously, several micro-social factors, such as personal experiences, socioeconomic status, and political behavior, have been used to explain attitudes supportive of democracy (Booth and Richard 1998b, 1996; Booth and Seligson 2009). We begin with resources. Education associates positively with democratic attitudes in many contexts, so we expect to observe a positive influence on democratic attitudes (the direction of influence is indicated in Figure 1 by a positive sign). Personal wealth facilitates the acquisition of information. Moreover, Adorno et al. and Lipset reported in the mid-twentieth century that working classes tend to have authoritarian attitudes (Adorno et al. 1950; Lipset 1959a). Subsequent research has partly discredited this finding (Altemeyer 1981; Dekker and Ester 1987), and recent empirical findings from developing countries depict the poor as being just as democracy-supporting as their wealthier fellow citizens (Krishna 2008a). (We indicate the disagreement on this influence by signing wealth both positively and with a question mark in Figure 1).
Expected Influences on Democratic Attitudes
Personal experiences may shape democratic attitudes. One line of research argues that those who experience bad things under democracy may lose their democratic orientations. To the extent that citizens hold government responsible for honoring the social contract by keeping them safe, the experience of political corruption or being victimized by crime under democracy should erode citizens’ loyalty to democratic practice. Those victimized by corrupt public officials or criminals, it is widely assumed, could lose faith in democracy and its benefits. The Latin American experience has proven complex, however. Recent comparative surveys reported that being victimized by crime and the perceptions of personal insecurity have little influence on citizens’ democratic attitudes. Being a victim of corruption also has had little effect on democratic norms. In contrast, the perception of widespread corruption lowers democratic attitudes (Cruz 2008; Seligson 2001). For our model we hypothesize that crime and corruption victimization and the perception of both insecurity and widespread corruption may erode democratic attitudes (hence a negative sign in Figure 1).
Evaluations of the political system and its performance may also affect citizens’ democratic attitudes (Booth and Seligson 2009; Easton 1975, 1965; Lipset 1981; Norris 1999). Those whose experiences, knowledge and attitudes give them low confidence in key aspects of the political system (leaders, institutions) may demonstrate diminished support for democracy. Indeed, Smith (2005) has argued that the performance of Latin American democracies in general is so poor that it may erode both satisfaction with democracy and democratic values. Ironically, however, empirical analysis demonstrates that evaluations of institutional performance in Latin America have little influence on democratic attitudes. In contrast, citizen evaluations of economic performance and of political actors’ performance contribute to stronger democratic attitudes (Booth and Seligson 2009). Seligson and Booth in another study have linked the simultaneous presence of three evaluation factors – high low economic performance evaluations, low institutional support and weak regime principles – to political breakdown and to the Honduran coup of 2009, and to high risks of institutional instability (Seligson and Booth 2010, 2009). Thus we hypothesize that positive evaluations of political actors, of economic performance, and of democracy itself will contribute to stronger democratic attitudes.
Social capital is another set of possible influences on democratic attitudes. Coleman, Putnam and Inglehart argue that participation in organizations (civil society) contributes to interpersonal trust of the bridging variety that helps citizens, thus indirectly contributing to other attitudes supportive of democracy (Coleman 1990; Inglehart 1990; Putnam 2000, 1995). By allowing citizens to collaborate in democratic contexts, participation in organizations may also directly increase democratic attitudes (Booth and Richard 1998a, 1998b). Thus we anticipate that both civil society engagement and interpersonal trust will contribute to higher levels of democratic attitudes.
The final category is the sociopolitical context, which many scholars contend shapes political attitudes and culture (Cutright 1963; Inglehart 1988; Jackman and Miller 2004, 1996a, 1996b; Lipset 1959b; Muller 1997; Muller and Seligson 1994). As noted, many scholars have argued that higher levels of economic productivity contribute to system-level democracy, whether by reducing the share of authoritarian citizens, by increasing levels of education (almost axiomatically found to enhance democratic attitudes), or by promoting more widespread “participant” orientations and political efficacy. Without identifying particular paths of influence (we will control for many of these factors in our model), we expect Latin America's more democratic countries to have citizens with higher levels of commitment to democracy.
Another contextual influence is geographic – specifically, proximity to other democratic states. As noted above, democratization is a somewhat “sticky” process. Clusters of democratic regimes tend to form and to exist in proximity with each other (O'Loughlin et al. 2004). Reasons for this may be that elites in one country imitate models of governance in a neighboring state to reduce possible dissonance with neighbors, and that attitudes of elites and mass publics diffuse across borders from democracies to less-democratic states. In a previous study cited above, we tested for the possible influence of proximity to the Americas’ oldest democracy, the United States, and found that democratic attitudes were higher among residents of the countries closest to and farthest from the United States. This suggested to us that something was missing from our first geographical proximity measure (distance from the United States). Here we refined the argument by assuming that exposure to (sharing a border with) any country that has been a democracy for a long time would likely enhance the democratic attitudes of citizens in the less-democratic nation. Latin America has several countries that could exert such influence – Costa Rica, Venezuela, Uruguay, and the United States, for example.
As final theoretical consideration we ask: How might citizen attitudes matter for the practical processes of adopting and keeping democracy? As noted above, one approach argues that mass culture matters for the formation of democracy. Unfortunately, we lack evidence in the form of good opinion surveys from Latin America extending back into the 1970s or earlier that might allow us to evaluate democratic attitudes during the formation of democratic regimes. Thus in this analysis we must focus only on how attitudes might contribute to keeping democratic rules of the game once in place.
We believe it reasonable to assume that citizens’ attitudes and preferences may exercise some influence on key political elites. While there is much evidence that elites disregard citizens’ preferences, at a minimum mass attitudes indicate popular preferences for elite behavior. For example, mass attitudes strongly supportive of democratic principles and rules of the game would signal to elites that to deviate from such principles could incur political costs in the form of lost future votes, demands placed on government, and protest. 3
An example of that occurred in Honduras in 2009 when a coup d'etat by the military, supreme court and legislative leadership resulted in many weeks of mass protest, some of it violent. Having deviated from the constitution to depose President Zelaya, the architects of the coup reverted to the constitutional order following the regularly scheduled election a few months after the coup (Pérez, Booth, and Seligson 2010; Seligson and Booth 2009).
Drawing from this model and discussion, we specifically hypothesize that democratic attitudes are the product of individual resources, personal experiences, evaluations of the political system, social capital, and the sociopolitical context.
The Latin American Cases and the Data Set
We employ data from the 2008 round of the LAPOP AmericasBarometer surveys in 18 countries in Latin America. 4 LAPOP interviewed approximately 1,500 respondents per nation to provide a nationally representative sample of citizens of voting age. The surveys were conducted face-to-face by interviewers trained and supervised by the AmericasBarometer team. 5 In each national sample employed, the data have been weighted to adjust the sample size to 1,500 per country, giving an error margin of approximately 2.5 percent at the 95 percent confidence interval.
Mexico, Guatemala, El Salvador, Honduras, Nicaragua, Costa Rica, Panama, Colombia, Ecuador, Bolivia, Peru, Paraguay, Chile, Uruguay, Brazil, Venezuela, Argentina, and the Dominican Republic.
The data were collected using a multi-stage probabilistic sample design based on regional and urban/rural area stratification within each country, and employing age and gender quotas at the household level to assure balance among population groups. For details on technical information these surveys and questionnaires with all items included, see www.AmericasBarometer.org.
The countries in the study are 18 Latin American nations, 17 sharing Spanish colonial heritage, and Brazil being a former Portuguese colony. In populations, their sizes ranged from Uruguay (smallest at 3.3 million) to Brazil with (very large with 190.1 million). All have developing economies. Per capita gross domestic product (GDP) in 2007 ranged from less than 880 USD in Nicaragua to 9,353 USD in Argentina (ECLAC 2009). Although all 18 nations had formally democratic governments at the time of our surveys (Smith 2005), several of them were emerging or young democracies still struggling with institutionalization and surprisingly high levels of political and social violence. Mexico's democracy was the youngest, dating from 1997. Costa Rican democracy was the oldest in continuous operation, having been established in 1950.
Variables in the Analysis
Democratic Attitudes
The range of attitudes related to democracy is broad, 6 but for clarity and precision we focus on three attitudes governed by the definition of democracy itself. Democracy is a system of government based on citizen participation in rule (Aristotle 1962; Cohen 1973; Dahl 1998; Pateman 1970). By definition, citizens of democracies should be able to elect their leaders, take part in politics in other ways, criticize their governments, and mobilize to support challengers to incumbents. Citizens of a democracy should, therefore, hold values that support or reinforce these basic elements of democratic rule. Irrespective of the direction of causality assumed, such attitudes would either (a) strengthen democracy by shaping elite and mass behavior to value and embrace democratic rules and citizens’ rights, or (b) be strengthened by the operation of democratic rules of the game.
Scholars have analyzed legitimacy (Booth and Seligson 2009), justification of military and executive coups (Booth, Wade, and Walker 2010; Seligson and Booth 2010), and combinations of attitudes (Seligson and Booth 2010).
To be specific, we believe that citizens should first prefer democracy over alternative systems of rule. (This item is very similar to the measure employed by Lagos Cruz-Coke 2001.) We measure this variable with an index constructed from four separate items indicating an explicit or stated preference for democracy (see Appendix for questions and details of construction). The resulting index is scaled from 0 (does not prefer democracy on any of the items) to 100 (prefers democracy on all four items).
Mass publics should embrace the rights of citizens to participate in politics, and they should be tolerant of the participation rights of regime critics. We measure support for general participation rights with an index composed of (dis)agreement with three general political participation rights (the right of citizens to take part in an organization, a political campaign, and a legal protest). We measure tolerance of regime critics as an index composed of (dis)agreement with four participation rights to critics of the regime (the right to vote, demonstrate, seek public office, and give a speech on television). Both these indices are also constructed ranging from zero (lowest level of support on all items) to 100 (highest level of support on all). (See Appendix for details).
How strongly did Latin Americans hold these democratic norms in 2008? Figure 2 presents the distribution of these measures for our 18 countries. There we see that the expressed preference for democracy gray bars) is, with a single exception of Paraguay, the most strongly embraced democratic norm. Preference for democracy ranges from a low of 64 points out of 100 in Honduras to 88 points in Venezuela. Because scale points are functionally equivalent to percentages, the regional mean of 76 points is considerably higher than the 60 percent support found by Lagos Cruz-Coke (2001) a decade earlier. This finding no longer justifies, in our opinion, Lagos’ characterization of Latin American democratic attitudes as unconsolidated and “in danger.”
Democratic Attitudes among Latin Americans, 2008
But what of the other two democratic attitudes, which are conceptualized differently from expressed support? Citizens’ agreement with general political participation rights (white bars, Figure 2) is also rather high, with an average of 70 out of 100 for the 18 countries. Hondurans report the lowest national average (57 points above the scale midpoint) and Paraguayans the highest (83), followed by Argentines (81). Again, this level of adherence to democratic norms does not in our opinion pose a generalized danger to Latin American democracies.
The democratic attitude for which support is weakest is tolerance of participation by regime critics (the black bars in Figure 2), with a regional average of 52 out of 100. The lowest tolerance means are in Bolivia and Guatemala (approximately 44 out of 100). Five other countries also have national means below the scale midpoint of 50 – Honduras, Peru, Chile, Ecuador and Colombia. Argentines report the highest tolerance for regime critics (68), followed by Paraguayans at 64. We believe that tolerance of regime critics is important for keeping political space free for contestation and in providing an open marketplace for political ideas. To the extent that we are correct, the low levels of political tolerance in several countries arouse cause for concern. A populace unsure about whether it would allow critics of the government to engage in politics does not fully embrace a right essential for well-functioning democracy.
Independent Variables
We begin with individual resources: Education has been widely observed in diverse settings to correlate with higher democratic attitudes. We employ a simple measure of the number of years of education completed. Personal wealth and income may (or may not) affect democratic attitudes, as the literature review pointed. We employ an index of the wealth of the respondent's household to measure the economic resources available to the respondent. 7 Our analysis will allow us to separate the effects of overall output at the system level from individual wealth.
We employ wealth instead of family income because the surveys revealed an unacceptably high rate of respondent refusal to answer the income question, raising a serious problem of missing data. Some 12.5 percent of cases in our sample would have dropped out had we used family income as the economic status measure. Our measure of wealth of the household of residence is based upon the possession of certain artifacts indicative of a range of household economic status ranging from indoor plumbing to the ownership of automobiles, and computers. The bivariate correlation between the wealth measure and income is .499, indicating that wealth clearly captures some part of family income.
In the realm of personal experience, several recent studies posit a link between experiencing certain social pathologies and democratic attitudes. The argument is that crime or corruption victims could lose their faith in democracy. We employ two measures of such experiences – crime victimization and corruption victimization: each is an index constructed from multiple items (see Appendix for details). Perhaps even more important than actual experience as a crime or corruption victim for shaping attitudes about democracy is one's perception of these problems. A perception of being insecure in one's own neighborhood and a belief that corruption is rampant are also likely to undermine democratic attitudes (see Appendix for an explanation of each item).
We employ three measures of evaluation of the political system. The first is evaluation of the incumbent government's performance. This is an index composed of items asking for appraisal of the government's performance in five policy areas. The second measures evaluation of economic performance, an index combined from two sociotropic items assessing the respondent's view of the current condition of the economy and projecting the national condition of the economy a year in the future. The third evaluation measure is a single item that asks for the respondent's satisfaction with “how democracy functions” in his/ her country (see Appendix).
We measure social capital with two items. We construct an index of civil society activism from the respondents’ levels of involvement in four different types of organizations – church-related, school-related, communal improvement groups, or business/professional/producers’ association. 8 A second social capital variable is a single-item measure of interpersonal trust (see Appendix).
Chronbach's alpha for this measure is .405, indicating a weakly cohesive scale. This problem arises because of the difference in levels of participation in each type of organization, which ranges from 54.5 percent for church-related groups down to only 12.7 percent participating in business/ professional/ producer groups.
We employ four measures to explore the impact of the system-level environment on democratic attitudes. In order to test whether the absolute level of systemic economic activity in a country will affect citizens’ attitudes and behaviors, we employ a measure of GDP per capita in 2007, the year before our surveys. A second system-level context variable seeks to assess the short-term impact of the regime's characteristics on citizens’ attitudes. To measure this we utilize the level of democracy in 2007 as measured by the Polity IV score for the year prior to the surveys. A second democracy variable attempts to assess the impact on citizens’ democratic attitudes of having more or less experience with system-level democracy over time. We assume that longer exposure to democratic rules of the game will lead to higher democratic attitudes. As a surrogate for this acculturation process, we measure the total number of years of democratic rule since 1950 in each country. This allows us to probe for the degree to which acculturation to democratic rules of the game may have occurred in recent decades. We selected 1950 as the base year because it is the year of the foundation of Latin America's oldest continuously democratic regime, that of Costa Rica (Booth 1998). 9
Chile, Uruguay, and Argentina, of course, experienced democracy before Costa Rica's democratic regime began, but in each one democracy was violently interrupted for a number of years after 1950.
A fourth context measure seeks to assess the influence that geographic proximity to a strongly democratic regime might have upon the attitudes of citizens of a neighboring nation. We assume that mass publics living in a country next to a strong democracy could absorb democratic attitudes from across the border via interpersonal contacts and via the media. We construct a variable that scores a country one point for sharing a border with another nation that has been democratic for at least 45 years since 1950.
We also include three control variables known to affect levels of democratic attitudes. We employ an ordinal measure for the size of the community of residence of each respondent (e.g. rural, small town, small city, large city, capital city). We include two socio-demographic variables that reflect individual resources: sex and age. These measure aspects of the individual respondent's position within society that might shape democratic attitudes.
Finally, we recognize that certain factors in the macro-social context may interact with individual characteristics. The possibilities here are extensive, but we confine ourselves to four that to us appear theoretically justified. First, given the observed (albeit disputed) linkage between both overall national economic output and individual wealth and democratic attitudes, we will interact GDP per capita in 2007 with individual wealth, which we expect to positively associate with democratic attitudes. Because absorbing democratic norms from observing a proximate nation's democratic system is likely to be a facilitated by the respondent's ability to gather information, we relate proximity to a strong democracy with the respondent's education, which we expect to have a positive influence on democratic attitudes. An individual's opportunity for acculturation to democratic attitudes likely depends on the amount of exposure to democratic rule. We therefore interact years of democracy since 1950 with age, which we anticipate will positively influence democratic attitudes. Finally, because we believe that the socialization effect of system-level democracy may be weakened by an individual's perception of corruption, we relate democracy level in 2007 with perceived corruption and expect a negative effect on democratic attitudes.
We recognize that including context variables in an OLS regression model has certain weaknesses. 10 The use of interaction terms will provide a robustness check on the modeled relationships without interactions and thus minimize this problem. We will also conduct national fixed-effects analyses as robustness checks for the impact of our individual-level variables on democratic attitudes.
The repetition of a single value for all members of a national subsample tends to artificially reduce standard errors and thus to exaggerate the apparent influence of contextual variables.
Analysis
Expressed Preference for Democracy
Table 1 outlines factors predicted by our model to contribute to Latin Americans’ preference for democracy over its alternatives. The three models estimated present t-ratios for each respective independent variable. 11 Model 1 in Table 1 simply presents an OLS analysis of preference for democracy regressed on each of the independent variables, individual and aggregate. We present two ways of assessing the possible overweighting of contextual variables by OLS regression. Table 1's Model 2 excludes the context variables as such, but includes each of them interacted with an individual-level variable as explained above. This eliminates the standard error problem by mediating country value by its individual-level interaction partner. Table 1's Model 3 simply deletes both the context variables and interaction variables and substitutes national-level fixed effect dummies (Costa Rica is the excluded case). Because of fixed national effects (shorthand for national history, culture, and other particularities), Model 3 allows evaluation of whether the microlevel findings in Models 1 and 2 are distorted by the context variables, and by comparison provides some perspective on how well the other two models are specified.
Prefers Democracy over Alternatives
Note: Significant t-scores (absolute value greater than 2.0) are highlighted in boldface.
Source: AmericasBarometer by LAPOP, 2008 surveys.
T-ratios indicate the direction of association (by the sign), the strength of the independent contribution of the independent variable to the dependent variable (the absolute value of the t-ratio itself), and the significance of the relationship (significant if above the absolute value of 2.0).
Age shows strong and significant positive association with preference for democracy in Latin America as predicted. To the contrary, wealth shows no association (Models 1 and 3) and a negative association with preferring democracy (Model 2). This calls into question for Latin America as a whole whether the working class of any particular country prefers democracy less than that country's middle and upper classes.
The experience of victimization by crime has no impact on preference for democracy (all models in Table 1), contrary to expectations, but the perception of insecurity negatively associates with it as hypothesized. When we examine the experience of victimization by corruption we also find no effect (all models). The perception of widespread corruption, however, associates positively with preference for democracy rather than negatively as predicted (all models). Thus the expectation that perceived corruption could lead Latin Americans to lose faith in the democratic system of government proves incorrect. Indeed, these findings show that perceived corruption may strengthen the preference for democracy.
We hypothesized that evaluations of the performance of both the current government and the economic system, as well as satisfaction with democracy in general, would contribute to a stronger preference for democracy over alternatives. Evidence supports the hypothesis for evaluation of the incumbent government's performance and for satisfaction with democracy (all 3 models in Table 1). Thus preference for democracy is, to some extent, contingent upon evaluation of political system performance. Economic performance, in contrast, associates slightly positively with higher democracy only in Model 3 (fixed effects).
Social capital variables turn in a mixed performance in Table 1. Civil society activism has a very weak positive association with democracy preference for Model 1 only. In contrast, interpersonal trust behaves as predicted, elevating preference for democracy on all three models.
Turning to the context variables in Table 1, we see that GDP per capita in 2007 strongly associates with higher preference for democracy (Model 1). Years of democracy since 1950 also increases democracy preference as hypothesized. The two models that fail to conform to prediction are those for proximity to a longstanding democracy (no relationship), and the level of democracy in 2007 (significantly negative relationship). The latter is on its face confusing, but further analysis suggests why this apparent anomaly occurs. First, the level of democracy in Latin America has risen sharply and the systems have become more similar since 1980, when the mean Polity IV score (range from −10 to +10) for our 18 countries was −1.06 (standard deviation = 6.82). By 2007 the mean Polity IV score was 8.11 (s.d. = 1.10). Thus, with all the other factors in the analysis held constant, the small amount of variation of the 2007 Polity IV score has a negative effect. We suspected that having a higher democracy level for several years instead of just one year might capture some acculturation to preference for democracy. We reran Model 1 substituting Polity IV 2000 for Polity IV 2007 (not shown due to space constraints), resulting in a t-score for the 2000 democracy measure of 8.43, strong and signed in the predicted direction; other relationships in the model remained essentially unchanged. This confirmed our suspicion that the negative t-score for Polity IV 2007 was likely an artifact of low variation or a lack of time for system democracy to have an impact, other factors held constant.
Model 2 in Table 1, which considers the interacted system-individual variables, reveals no effect for proximity to a longstanding democracy interacted with education. Thus, considering the absence of any simple effect of proximity (Model 1), we conclude that our hypothesis about cross-border diffusion of attitudes does not hold up for preference for democracy, at least as measured. In contrast, the years of democracy since 1950 – age interaction term has a significant moderate positive effect on preference for democracy, as predicted. Looking further, in Model 2 age loses the effect observed in Models 1 and 3, indicating that the acculturation explanation (age plus time under democracy) has more power than age alone, as does the increased model strength (R-squared).
We observe a strong effect from GDP per capita 2007-personal wealth interaction term, while personal wealth changes from having no impact to a significant negative impact. The zero-order correlation between GDP per capita 2007-personal wealth interaction term and personal wealth alone is .72; however, it indicates excessive collinearity between these variables for OLS regression. We thus hesitate to trust the model's apportionment of explained variance (or the signs) for wealth and the GDP-wealth interaction term.
Turning to the control variables in Table 1, other factors held constant, women do not significantly differ from men in their preference for democracy. Older citizens have a stronger preference for democracy than the young, whether modeled directly (Models 1 and 3) or indirectly (interacted with years of democracy since 1950 – Model 2). Clearly being older and in a country with greater democratic experience contributes to a preference for democratic rule. Should this finding arouse concern about the democracy preference of the 25.4 percent of the respondents in our 18 countries who are between 16 and 25 years of age? In a word, no. The preference for democracy in this younger cohort averages 75 out of 100, well on the positive end of the scale. Size of community of residence is negatively related to preference for democracy in two of three models. This finding undermines an old but widely held belief that rural-dwelling Latin Americans harbored authoritarian attitudes. In fact, big-city residents, it appears, prefer democracy less.
Finally in Table 1, we note that Model 3, which places national fixed effects in the stead of context variables and interaction terms, gains several percentage points in explained variation (R-squared). We have thus evidently somewhat underspecified our model. National history and culture, summed up as “countryness” by the fixed effects, explain more variation than the simpler models. As a robustness check for individual-level findings, Table 1's Model 3 t-scores generally confirm other findings. They follow the patterns observed in Model 1 for 9 of 11 individual-level variables. Of the remaining two, civil society activism loses significance (very weak in Model 1), and evaluation of the national economic performance gains a positive association with preference for democracy.
Several system-level variables confirmed our hypotheses, correlating strongly and positively with the preference for elected rulers over authoritarian rulers (Table 1, Model 1). These were GDP per capita and age of current democratic regime. So, while personal wealth had no discernible effect on preference for elected rulers, national wealth contributed measurably and positively to it. One explanation for this could be that wealthier nations socialize their citizens to support democracy more than poorer ones do.
Support for Basic Political Participation Rights
Table 2 identifies variables contributing to citizen support for basic participation rights. It contains 3 models like those in Table 1. We note that education associates strongly and positively for all 3 models, confirming our hypothesis. Personal wealth is not related (Models 1 and 3); given the collinearity issue identified in the previous analysis, we will withhold comment on this part of Model 2. As observed for preference for democracy, neither crime nor corruption victimization lowers support for basic participation rights. Perceiving oneself to be unsafe does, however, contribute to weaker support for participation rights. For two of three models, perceiving widespread corruption increases support for participation rights (Models 1 and 3).
Support for Basic Political Participation Rights
Note: Significant t-scores (absolute value greater than 2.0) are highlighted in boldface.
Source: AmericasBarometer by LAPOP, 2008 surveys.
Evaluations of system performance produce contradictory evidence in Table 2. First, satisfaction with how democracy functions is significantly but weakly positively associated with support for participation rights. In sharp contrast, higher citizen evaluations of the government's performance have a significant negative effect on support for participation rights for Models 1 and 2, and no effect for Model 3 – all three findings run counter to expectations. Virtually identical results also appear for citizen evaluation of national economic performance. One way to interpret this pair of findings is as good news for those who value democratic attitudes. That is, support for democratic participation rights in Latin America appears not to be contingent on citizens’ evaluation of quality of government or economic performance. This suggests that support for participation rights may be generalized and unmoved by evaluations of performance. Support for participation rights is also broadly distributed irrespective of personal wealth, size of community of residence, and crime and corruption victimization. Latin American governments thus appear to enjoy a generous margin for error in their performance before citizens abandon support for participation rights.
The social capital variables perform strongly as hypothesized for all models. All three models in Table 2 show citizens who are more active in organizations to support participation rights (which include the right to take part in community organizations). Similarly, for all three models, more trusting respondents express more support for participation rights. This supports a basic premise of the social capital literature – that engaged and trusting citizens will hold attitudes supportive of democracy.
Three of four context variables in Table 2 (Model 1) perform as predicted. Higher GDP per capita in 2007, more years of democracy since 1950, and a higher Polity IV democracy score in 2007 predict higher levels of support for basic participation rights. Living in a nation proximate to a longstanding democracy has an unexpected negative effect. We interrelate proximity to a strong democracy with education (Model 2) on the premise that having more information about a neighboring democracy could enhance support for participation rights. This produces a negative t-score for the interaction term, suggesting that the effect is not an artifact.
The control variables in Table 2 reveal women to be less supportive of participation rights than men (all 3 models). Older people are less supportive of participation rights in Models 1 and 2. These results are inconsistent with those for preferring democracy, where women were no different from men and older citizens more democracy-preferring. Community size has no effect.
Turning to Table 2's Model 2, we see that years of democracy related to age performs as hypothesized, giving support to the idea that acculturation into democratic values develops the longer one lives within a democratic system. The interaction of democracy level in 2007 with perceived corruption, which we anticipated would negatively affect support for participation rights, performs contrary to expectations. We also note that the interaction of GDP per capita with wealth has an apparent strong positive effect on support for participation rights. 12 Including the related terms in the models added very little to explained variation, indicating little progress toward explaining support for basic participation rights.
We recall, however, that there is a collinearity problem with this relationship and so withhold judgment about the relationship.
As a robustness check, the national fixed effects Model 3 in Table 2 confirms the patterns observed in Models 1 and 2 for 8 of 11 individual-level independent variables. The large gain in explained variation (R-squared increasing almost threefold) for Model 3 over the other two indicates that the other models for support for basic participation rights remain inadequately specified. Factors that we are not capturing are shaping this model.
Tolerance of Regime-critics’ Rights
Table 3 presents the same three multiple regression models for tolerance of regime critics’ rights to participate in politics. Consistent with the other democratic attitudes, education exerts a strong positive influence and performs as hypothesized. The resource of wealth correlates positively with tolerance of critics’ rights in Models 1 and 3. This fits Lipset and Adorno's hypothesis about working class authoritarianism and diverges from the findings for preference for democracy and support for general participation rights.
Tolerance for Participation Rights of Regime Critics
Note: Significant t-scores (absolute value greater than 2.0) are highlighted in boldface.
Source: AmericasBarometer by LAPOP, 2008 surveys.
Table 3 reveals different patterns among the variable pertaining to the perceptions of crime and corruption levels than what we have seen before. Crime victims are slightly more tolerant (counter to hypothesis). Those who perceive themselves to be insecure are, as predicted, slightly less tolerant (Models 2 and 3). Corruption victims are also slightly less tolerant of rights for regime critics (all three models), also as predicted. Perception of widespread corruption associates with higher levels of political tolerance.
Perhaps the most noteworthy finding for tolerance of regime critics’ rights is that those who negatively evaluate government performance and economic performance and who express dissatisfaction with democracy all express higher levels of this important democratic attitude. On reflection, we believe this actually makes considerable sense. We suspect that citizens who are unhappy with system performance may sympathize with tolerance for dissenters because they themselves may wish to criticize the system.
Civil society engagement has no effect on tolerance (all models), but interpersonal trust strongly associates with it (all models). Clearly citizens who trust others also are more politically tolerant. This makes sense on its face in that a generalized feeling of more trust of others would, we believe, also make it easier to allow a fellow citizen to strive to challenge or criticize the political system.
Turning to the context variables, we observe a strong positive effect for GDP per capita in 2007 and a positive effect for years of democracy since 1950, as predicted. Proximity to a longstanding democracy performs significantly negatively and counter to expectations. With all other factors controlled, the level of democracy (Polity IV score) in 2007 had a small but significant negative effect on tolerance (counter to hypothesis). Suspecting that learning tolerance might require longer exposure to higher levels of democracy, we reran the models substituting Polity IV levels for earlier dates (1980, 1990, and 2000 – not shown due to space constraints). None of them exercised a significant positive influence on tolerance for regime critics.
Women are less tolerant than men (all models), and older Latin Americans somewhat less tolerant than younger citizens (Models 2 and 3). Residents of larger communities are consistently more tolerant of regime critics than smaller community and rural residents.
The fixed effects model (3) for political tolerance largely confirms the findings provided by the other two models in Table 3 that include context variables and interacted variables. The gain in explained variance for the fixed effects model over Models 1 and 2 is smaller than for preference for democracy and for support for general participation rights. The tolerance model thus appears to be comparatively better specified than the others.
Turning finally to the interacted variables in Table 3, we observe that they produce no increase in explained variance (the R-squared for Models 1 and 2 are the same). For tolerance, proximity to democracy-education produces a result counter to expectations, and years of democracy-respondent age performs as expected, again underscoring that exposure to democracy over time strengthens democratic attitudes. The democracy in 2007 – perceived corruption interaction term performs weakly, but negatively as hypothesized.
Our final analysis delves more deeply into the effect of both personal wealth and system-level economic development on democratic attitudes. Recall that in the regression analyses individual wealth exerts little or no influence on attitudes, other factors controlled, including system-level wealth. GDP per capita in 2007, in contrast, whether simply or joined with individual wealth, appeared to exert a strong positive effect, but collinearity issues left that finding in doubt. Our deeper analysis here tests differences of means on our three democratic attitudes by countries sorted into three cohorts of national development, and by personal wealth divided into three cohorts (surrogates for the working, middle and upper classes). We employed both graphing and statistical analysis of variance to test for significance of differences (not shown due to space constraints).
We find that GDP per capita and personal wealth each have a significant impact on all three democratic attitudes (absent controls for other variables). However, Latin Americans average well above the scale midpoints on preference for democracy and support for basic participation rights. Moreover, the differences between poor and rich within any national development level are very small (0 to 5 points out of 100 – that is, the slopes of democratic attitudes across ascending class levels are quite shallow). In sum, isolated by level of national economic development (i.e. comparing citizens who live in the poorer countries with each other and so on) we find a small but substantively unimportant difference in democratic attitudes by economic class.
For tolerance, the overall mean for all 18 countries is just above the scale midpoint. Scores on democratic attitudes among the poorest countries, and the intermediate GDP per capita countries, range from 46 to 52 (poorest countries) to 47 to 57 (middle GDP per capita countries). For the wealthiest cohort of countries, tolerance for regime critics’ participation rights ranges from 52 to 56. All differences of means are significant. The slopes across personal wealth levels for tolerance are somewhat greater than for preferring democracy and supporting basic participation rights. In sum, the biggest effects of both economic development levels and personal wealth are found for political tolerance. Moreover, the tolerance variable is substantially lower (centered from just below to a few points above the scale midpoint) than preference for democracy and support for participation rights.
Thus we cannot rule out a national economic development effect on democratic attitudes, but the effect of individual wealth within countries is modest even without introducing controls for education and other factors. When such controls are applied (regression analyses), the effects of social class on democratic attitudes are clearly negligible. So we have a complex finding: Within countries, the working classes hold attitudes slightly but not meaningfully less democratic than the middle and upper classes. Nevertheless, between countries, higher GDP per capita leads to democratic attitudes.
Discussion and Conclusions
Table 4 summarizes the findings of the multiple analyses. Overall, our model of influences on democratic norms performs only moderately well, with just over half of the independent variables exerting significant influence in the expected direction.
Summary of Findings
Notes: Dark gray cells: hypotheses strongly confirmed, white cells: inconsistent or inconclusive or contradictory findings light gray cells: consistent and significantly counter to hypothesis signs within the cells indicate the actual direction of association found (if significant).
Source: Own compilation.
Individual resources have mixed effects. Education powerfully predicts stronger democratic attitudes, but social class does not.
The experiences of crime and corruption victimization do not lower democratic attitudes. Perceived insecurity, in contrast, lowers all three democratic attitudes while perceived corruption has the unexpected opposite effect of correlating with higher democratic attitudes.
Evaluations of system performance have inconsistent effects or impacts contrary to what we expected. The strongest pattern is that satisfaction with democracy correlates positively with two attitudes. We conclude that the relationship among system evaluation and democratic values is more complex than theorists have suggested. We are, however, comforted that support for democracy is not strongly contingent on system performance in these nations (almost all democracies), and that to some extent citizens who are unhappy with system performance remain committed to democracy. This suggests that acculturation to democratic values may have advanced considerably in Latin America, rendering these attitudes stronger and less volatile than pessimists believe. That is, a reservoir of legitimacy may exist for Latin America's democracies. Indeed, Booth and Richard (Booth and Richard 2011) report that Latin Americans retained democratic attitudes despite the sharp negative economic downturn of 2008-2009.
Social capital variables reveal a mixed record in Table 4. The results strongly affirm that interpersonal trust contributes to holding democratic attitudes. In contrast, civil society engagement predicts only higher support for participation rights. This finding makes sense given that the attitudeapproval of organizing to resolve a community problem constitutes one component of the rights measure.
Contextual factors produce mixed effects. Higher overall national economic productivity has a positive influence on all three democratic attitudes. Years of democracy since 1950 also has a consistently positive effect, underscoring that acculturation to democratic attitudes occurs because of system-level rules of the political game. Living in a country adjacent to a longstanding democracy exerts none of the positive effects expected, and indeed proves negative for both support for participation rights and tolerance of regime critics. We suspect that the latter may occur because non-democracies adjacent to democracies may engage in repression that socializes their citizens against democratic attitudes (Booth and Richard 1996). 13 The democracy level in 2007 proves a weak predictor. On reflection, employing democracy levels measured only a few months before the 2008 surveys may simply not allow a sufficient lapse for any socialization to take place. Further investigation into the lag between national democracy levels and citizen attitudes seems warranted.
Of course this apparent anomaly may also occur because we have poorly measured the likelihood of democratic diffusion. This finding warrants further research.
Interrelated terms meet with mixed success. Proximity to a strong democracy coupled with education has the opposite effect expected, again suggesting an unanticipated anti-democratic socialization effect. Years of democracy paired with age has a strong positive effect on all three democratic attitudes, again underscoring the likelihood that living in a democracy has pro-democratic socialization effects. Finally, and as predicted, the democracy level variable joined with perceived corruption produced an expected negative effect on preference for democracy and political tolerance.
Other factors held constant, education, interpersonal trust, GDP per capita, democratic experience since 1950, and “years of democracy” with “age” behave as predicted by elevating all three democratic attitudes. The age, education, and longevity of democracy variables factors are elements of a democratic socialization process. We suspect that schools contribute strongly to democratic socialization in democracies by imparting lessons emphasizing the benefits of democracy. Education may also allow citizens to perceive how much better human rights outcomes are in democracies than non-democracies, which we believe would also tend to strengthen democratic attitudes.
Our finding that higher interpersonal trust correlates with holding higher democratic attitudes appears to vindicate Putnam's arguments about social capital (1993, 2000). However, the direction of causality remains unclear in this relationship. Muller and Seligson, Tarrow and Jackman and Miller have argued and provided empirical evidence of a system-to-attitudes causality – that is, that democratic regimes encourage higher trust levels (Jackman and Miller 2004; Muller and Seligson 1994; Tarrow 1996).
Within countries, we failed to confirm Lipset's hypothesis that the poor are more authoritarian than the middle and upper classes. We believe our findings should lay to rest the idea that Latin American working classes are generally authoritarian. At most, there are minor gradations in each democratic attitude based on social class. We have seen that preference for democracy and support for basic participation rights places Latin Americans strongly in the pro-democracy end of the scales. This replicates for a larger sample of countries what Booth and Seligson reported previously for a smaller set of Latin American countries (Booth and Seligson 2008) and was also found in Africa and India (Krishna 2008b).
System-level development effects on democratic attitudes are stronger than social class effects, but again we emphasize that the overall bias of Latin Americans is favorable toward democracy. The commitment to tolerance is the weakest of the three attitudes, and the one most affected by system-level development. On balance, however, we conclude that one must regard with skepticism the received wisdom about working class authoritarianism.
Findings on crime and corruption victims provide good news for the prospects of democratic stability, especially given the crime waves afflicting northern Latin America. Even more remarkably, the perception of widespread corruption strengthens all three democratic attitudes. For observers worried that crime and corruption in Latin America might be undermining support for democracy in the region, these findings offer some comfort. We cannot say that such experiences will never have negative effects, but as recently as 2008 and 2010, democratic attitudes seemed to be resilient despite these problems (Seligson and Smith 2010).
Turning to the context-level variables, we find that having more years under democracy consistently predicts stronger democratic attitudes. The longer the citizens of a country live under democratic rules of the game, the stronger democratic attitudes they absorb. Insofar as mass attitudes constrain elite behavior, this finding suggests that democracy is self-reinforcing. Moreover, the fact that the working classes of Latin America have strong democratic attitudes irrespective of national wealth is good news. Democratic attitudes and system-level democracy itself may be much more amenable to low national economic development levels and to low personal wealth than has been widely believed for decades. This means that the lagging wealth of many countries in the region and the widespread poverty of many citizens do not undermine popular support for democracy.
Footnotes
Variables Used in the Analysis
| Variable | Description | Mean | Std. Dev. |
|---|---|---|---|
| Expressed Preference for Democracy |
Index constructed from the mean of 4 items: (1) “There are people who say that we need a strong leader that does not have to be elected by the vote. Others say that although things may not work, electoral democracy, that is, the popular vote, is always the best. What do you think?”; coded 0 = supports unelected strongman ruler…100 = supports elected ruler add other items; (2) “Democracy may have its problems, but it is better than any other form of government”; coded 0 for very much disagrees … 100 for very much agrees; (3) “Which of the following do you agree with most? It doesn't matter to someone like me whether there's a democratic or nondemocratic government (coded 0), democracy is preferable to any other form of government (coded 100), in some circumstances an authoritarian government might be preferable to a democratic one” (coded 50); and (4) “Do you believe that our country needs a government with a strong hand (coded 0), or do you believe that problems are best resolved with the participation of all?” (coded 100). Chronbach's alpha = .509. |
76.34 | 24.32 |
| Support for Basic Democratic Participation Rights | Index of respondents’ mean support for basic democratic participation rights (taking part in election campaigns, organizations, and demonstrations); 0 = does not support any basic participation rights…100 = fully supports all three basic participation rights. Chronbach's alpha = .750. | 70.07 | 24.36 |
| Tolerance for Regime Critics’ Participation Rights | Index of respondents’ mean support for participation rights for regime critics (voting, demonstrating peacefully, seeking public office, and giving a speech on television); 0 = does not support any participation rights for critics … 100 = fully supports all participation rights for critics. Chronbach alpha = .871. | 52.37 | 28.19 |
| Crime Victimization | Index based on the percent of respondents who reported having been a victim of a various types of crime within previous year (range 0 to 100 percent). | 17.85 | 38.30 |
| Perception of Insecurity | “Speaking of the neighborhood you live in and thinking about the possibility of becoming a victim of a robbery or assault, do you feel very safe (coded 0), somewhat safe (coded 33), somewhat unsafe (coded 67) or very unsafe (coded 100)?” | 42.89 | 30.97 |
| Corruption Victimization | Index constructed from the number of different ways in which one was asked for a bribe or illegal payment in previous year (range 0 to 5). | .27 | .68 |
| Perceived Level of Corruption | “Taking into account your experience or what you have heard, corruption of public officials in this country is: not at all widespread (coded 0), a little widespread (coded 33), somewhat widespread (coded 67), or very widespread (coded 100)?” | 73.88 | 28.01 |
| Evaluation of National Economic Performance | An index composed of the mean of two items: (1) “Now speaking of the economy, how would you describe the country's economic situation?” Coded very bad = 0, bad = 25, neither good nor bad = 50, good = 75, and very good = 100; (2) “Do you believe the current economic situation of the country is better (coded 100), the same (coded 50), or worse (coded 0) than it was twelve months ago?” Chronbach's alpha = .584. | 38.25 | 25.99 |
| Evaluation of Incumbent Government's Performance | An index composed of the mean of five items asking how well the government (1) combats poverty, (2) promotes and protects democratic principles, (3) fights corruption, (4) protects citizen security, and (5) fights unemployment. Cronbach's alpha = .923. | 42.64 | 27.42 |
| Satisfaction with Democracy | “In general, how satisfied would you say you are with the way democracy functions in (respondent's countr)?” Very unsatisfied = 0; unsatisfied = 33; satisfied = 67; very satisfied = 100. | 50.39 | 24.31 |
| Civil Society Activism | Index of frequency of participation in four types of organizations (church-related, school-related, community improvement, and business or professional groups); coded 0 = no involvement in any group … 100 = participation once per week in all four. Chronbach's alpha = .405. | 18.02 | 15.43 |
| Interpersonal Trust | “Now, speaking of the people from here, would you say that they are generally very trustworthy, somewhat trustworthy, not very trustworthy, or untrustworthy?” 0 = untrustworthy, 33.3 = not very trustworthy, 66.7 = somewhat trustworthy, 100 = very trustworthy | 58.18 | 30.04 |
| GDP Per Capita in 2007 | Gross domestic product per capita in 2007, expressed in 2000 constant dollars (ECLAC 2009) | 3911.56 | 2441.35 |
| Years of Democracy since 1950 | The total number of years of democratic rule (a Polity IV score of +6 or higher) since 1950 (Marshall and Jaggers 2010) | 26.39 | 12.90 |
| Proximity to a Longstanding Democracy | Sharing a border with a country that has had at least 45 years of democracy (Polity IV +6 or more) since 1950; yes = 1, no = 0. The longstanding democracies bordering at least one of our countries are the United States, Costa Rica, Colombia, and Venezuela | .44 | .50 |
| Democracy Level in 2007 | Polity IV score in 2007 (Marshall and Jaggers 2010) | 8.11 | 1.10 |
| Size of Community of Residence |
Size of the respondent's community of residence; 0 = rural; 1 = small town, 2 = small city, 3 = large city, 4 = capital city |
1.97 | 1.54 |
| Sex | Respondent's sex; male = 1, female = 2 | 1.52 | .50 |
| Age | Respondent's age in years | 38.98 | 15.97 |
| Education | Number years of formal education completed | 8.90 | 4.572 |
| Wealth | Number of artifacts indicative of a range of economic levels ranging from indoor plumbing to the ownership of refrigerators, television sets, microwave ovens, motorcycles, automobiles, and computers | 4.10 | 1.98 |
Source: AmericasBarometer 2008, unless otherwise specified above.