Support for Polyarchy in the Americas - Department of Political

Support for Polyarchy in the Americas

Ryan E. Carlin Department of Political Science

Georgia State University 38 Peachtree Center Ave., Ste. 1005

Atlanta, GA 30303-2514 Phone: (404) 413-6190 Fax: (404) 413-6156

[email protected]

Matthew M. Singer Department of Political Science

University of Connecticut 341 Mansfield Road, U-1024

Storrs, CT. 06269-1024 Phone: (860) 486-2440 Fax:(860) 486-3347

[email protected]

Keywords: Support for Democracy, Polyarchy, Modernization theory, Political Culture, Presidential Approval, Latin America

Forthcoming in Comparative Political Studies Volume 44 (Nov 2011) Acknowledgements: This research was funded by the Latin American Public Opinion Project’s (LAPOP) Small Grants Program for Research and Publishing. The authors kindly thank Damarys Canache, Julio Carrión, François Gélineau, Jonathan Hartlyn, José Álvaro Moisés, Karen Remmer, Rodolfo Sarsfield, Mitch Seligson, Lee Walker, and the anonymous reviewers for constructive critiques on earlier drafts. All remaining errors are our own.

Abstract: This study measures support for the basic rights, liberties, and practices associated with polyarchy in 12 Latin American democracies. Specifically, it identifies five profiles of support for polyarchy’s core values and norms – public contestation, inclusive participation, limits on executive authority, and institutional checks and balances. Although citizens who fit the polyarch profile accept all of polyarchy’s principles, those who fit one of the four mixed support profiles (power constrainer, power checker, power delegator, power restrainer) only accept some of them while rejecting other core democratic principles. Long-run factors emphasized by modernization and cultural theories (e.g. education, wealth, political engagement) are closely associated with the polyarch support profile. However, short-range performance factors (e.g. economic perceptions; crime, discrimination, and corruption victimization; voting for losing presidential candidates; presidential approval) may better explain why citizens fit one particular mixed profile over another and particularly explain willingness to delegate authority to the executive at the expense of other institutions.

Democracy in Latin America has reached a turning point: its success or failure appears to

rest more firmly in the hands of its own citizens, institutions, and elected leaders than military

commanders (Pérez-Liñán 2007, McCoy 2006).1 Despite this monumental feat, democracy’s fate

in the region remains uncertain, as recent evidence of a “rollback” of (Diamond 2008) or

“pushback” against (Puddington 2007) basic democratic rights and liberties attests. Such

backsliding begs the question—does liberal democracy enjoy widespread popular support in

Latin America? If not, why? Answering these questions will help fully appreciate the roles

everyday citizens and their beliefs play in the region’s democratization.

Guided by these questions, this study develops a theoretically-grounded measure of

support for polyarchy, or the rule by many (Dahl 1971). The analysis is based on an emerging

consensus that democratic support is a multidimensional concept and that ideal-type democrats

coexist with citizens holding mixed or ambivalent belief systems. It finds five shared profiles of

support for polyarchy across twelve Latin American democracies. And while some citizens fully

support polyarchy’s ideals, most display one of four profiles that only partially embrace them.

To determine what distinguishes polyarchs, citizens with an unblemished profile of

support for polyarchy, from citizens who reject one or more of polyarchy’s core tenets, the study

tests competing hypotheses from modernization theory, political psychology, ideology, and

short-term regime performance. Polyarchs are more educated, wealthier, more cognitively

engaged with politics, and less ideological than their less democratic counterparts. They also

tend to hold lower evaluations of the president. Factors that lead voters to reject polyarchy’s

principles ion contrast tend to be much more heterogeneous and multidimensional. Relatively

stable long-run factors like education and political interest are the dominant predictors of

attitudes toward protecting civil liberties and allowing inclusive participation for all groups in

1

society. Short run factors, in contrast, are more important in leading citizens to reject democratic

institutions or support delegative democracy.

The current investigation makes three main contributions to the study of democratization.

First, it probes beyond the ideal-typical polyarch profile to identify four heterogeneous profiles,

i.e. diminished sub-types of polyarchy support. Second, rather than assuming mixed belief

profiles represent non-attitudes, it examines what differentiates citizens with these profiles from

polyarchs empirically. Indeed, presidential approval, economic rationality, and ideology explain

mixed profiles as well as or better than political sophistication. Mixed support profiles, thus,

have more structure than previously thought. Moreover, this analysis suggests that rejection of

specific democratic principles can be linked to different short-term political considerations. In

contrast to much of the work on developed countries, but similar to work on Africa (Moehler

2009), voters who support the incumbent and are pleased with economic outcomes under the

current regime are more willing to delegate authority to the executive at the expense of checks

and balances and perhaps even civil liberties. Third, this study produces measures well-tailored

to testing the complex and contingent connections between mass orientations to polyarchy and

polyarchy as a regime outcome. Identifying the range of democratic belief systems, establishing

their sources, and tying them to political behavior are crucial to theory development (Booth and

Seligson 2009, Coppedge N.d, Norris 1999). In this sense, polyarchy support profiles are a sound

basis of “descriptive inference” upon which to build solid “causal inferences” (King, Keohane

and Verba 1994) about the relationship between democracy at the regime and mass levels.

Measuring Support for Democracy

Though democracy is an “essentially contested concept” (Gallie 1956), a procedural

minimum definition, based on Dahl’s (1971) conception of polyarchy, is gaining acceptance:

2

“fully contested elections with full suffrage and the absence of massive fraud, combined with

effective guarantees of civil liberties, including freedom of speech, assembly, and association”

(Collier and Levitsky 1997, 434). Absent one or more of these criteria, democracies can be

qualified “with adjectives” describing how they deviate from the procedural minimum (Collier

and Levitsky 1997) Like democracy measures themselves, valid measures of support for

democracy must flow from a clear a priori conceptualization of democracy. Yet most extant

measures of democratic support do not and, therefore, face two pitfalls.2

First, many survey measures directly inquire about support for “democracy” and, thus,

force respondents to answer from their own conception of “democracy.” Since citizens’

conceptions vary wildly, and may be vacuous or pejorative (Miller, Hesli and Reisinger 1997,

Ottemoeller 1998, Camp 2001, Moreno 2001, Seligson 2001, Bratton et al. 2005, Baviskar and

Malone 2004, Carrión 2008), comparisons and inferences based on such measures lack validity.

Moreover, many citizens express support for “democracy” in abstract terms but reject concrete

democratic freedoms, values, and norms (Prothro and Grigg 1960, McClosky 1964, UNDP

2004). Second, in addressing these issues, analysts must consider the possibility that support for

democracy, a multidimensional concept,3 might not converge on a single linear dimension.

While “democrats,” who fully support all conceptualized dimensions of democracy, occupy one

end of the continuum, and “non-democrats,” who reject all of democracy’s dimensions, occupy

the other end of the continuum, who occupies the middle? Consider a citizen who embraces

elections, waffles on government threatening the press, moderately favors banning extremist

parties, but champions free association. For this hypothetical citizen, the underlying dimensions

of support are not linear but bundled categorically. Scoring this citizen in the middle of a linear

single-dimension democracy scale—less democratic than “democrats” and more democratic than

3

“non-democrats”— obscures a great deal of potentially theoretically relevant information.

Thus an optimal measure of support for democracy would have several features. It would

begin with a concise conceptualization of democracy, and employ multiple indicators of support

for the dimensions of that conceptualization rather than asking point-blank about “democracy.”

Lastly, it would not restrict democratic values to a single continuum but rather explore the

systematic mixtures of orientations to democracy’s dimensions that may exist in the polity.

Measuring Support for Polyarchy

With these standards in mind, this study gauges support for a particular kind of

democracy, polyarchy. Dahl’s democratic ideal is “the continuing responsiveness of the

government to the preferences of its citizens, considered as political equals” (1971, 1). It requires

affording citizens “unimpaired opportunities” to formulate their preferences, signal them to each

other and to government individually and collectively, and to have their preferences taken into

equal consideration by the government (Dahl 1971, 2). To ensure these opportunities, Dahl posits

eight institutional guarantees—freedom of association, freedom of expression, right to vote,

expansive eligibility for public office, access to alternative information, free and fair elections,

and the dependence of government policy on citizens’ expressed preferences (1971, 3)—that

“might be fruitfully interpreted as constituting two somewhat different theoretical dimensions”

(4): public contestation and inclusive participation. Contestation means political competition

takes place in an atmosphere of free speech, organization, and participation. Inclusiveness

implies competition is, above all, open. Free and fair elections enhance both meaningful

participation and the effectiveness of contestation. While these two dimensions do not

necessarily covary, each makes the other more effective. Dahl terms a regime approximating this

ideal-type polyarchy instead of democracy to create a new typology to “serve as a basis for

4

estimating the degree to which various systems approach this theoretical limit” (1971, 2).

Dahl also theorizes that progress towards the exigent standard of responsive government

requires vigilance over the basic procedures and institutions that grant citizens effective

participation and control of the agenda. Recent research suggests a lack of checks on executive

power is democracy’s greatest threat (Kapstein and Converse 2008, Diamond 2008). While the

blame for unchecked executives surely falls on weak political and regulatory institutions and on

civil society, as least part of it could reflect mass values. In Latin America support for populist

(Seligson 2007) and delegative governance (Walker 2009) is widespread, even among citizens

purporting to prefer “democracy” (UNDP 2004, Carrión 2008). Executives who place “national”

priorities ahead of citizens’ rights or who emasculate (or close) the legislature and/or judiciary

nullify effective participation and compromise citizen control of the agenda. Citizens who accept

or are ambivalent to such political maneuvers augur against the democratic ideal of government

responsiveness conceived as “full procedural democracy with respect to an agenda and in

relation to its demos” (Dahl 1997, 44-45). Not surprisingly, Dahl included measures of

horizontal accountability via checks and balances in his measures of polyarchy (1971, 239-40).

Defining democracy as polyarchy allows this study to consider a citizen ideal-type

polyarch. A polyarch fully supports the political and civil freedoms, norms, and procedures that

bolster the twin dimension of public contestation and inclusive participation. Ideal-type

polyarchs also reject executive machinations to subvert, pervert, or reverse them or to otherwise

undermine government responsiveness. They value polyarchy’s principles in theory and in

practice. Yet just as many regimes have institutional shortfalls that deviate from polyarchy’s

standards without being fully authoritarian, many citizens’ belief systems will likewise deviate

from full support for polyarchy’s principles. How shall we characterize such citizens?

5

Taking an approach akin to Collier and Levitsky’s (1997) “democracies with adjectives,”

Schedler and Sarsfield identify “democrats with adjectives” based on orientations to multiple

dimensions of liberal democracy. Using national surveys in Mexico from 2003 (ibid 2007) and

2005 (ibid 2009), they find a liberal democratic belief system alongside five “diminished sub-

types” rejecting at least one—but not all—of these principles. The proportions of citizens that fit

these profiles are remarkably stable across the two studies: e.g. liberal democrats comprise just

13.6% of the sample in both analyses. Their results are fairly consistent with Dryzek and

Holmes’s (2002) findings on democratic discourse—a concept that overlaps with democratic

support—in 13 post-Communist societies. Only a handful of the discourses are either clearly

liberal (3) or clearly authoritarian (3) while multiple (8) discourses espouse a mixture of

democratic and authoritarian values.

It seems reasonable, then, to expect a range of sub-types of democratic support in Latin

America’s new democracies: polyarchs, non-polyarchs, and any number of mixed sub-types.

Whether this is indeed true, and the relative proportions of citizens who fit these democratic

belief profiles, are empirical questions this study seeks to answer.

Thus, the analysis measures support for polyarchy as a profile of multiple orientations to

polyarchy’s main dimensions—contestation and inclusiveness—and the institutions and norms

that foster responsive government and citizen control of the agenda. To support polyarchy is to

support inclusive political and civil freedoms as well as the institutional safeguards that make

them effective by checking power vertically (between government and the governed) and

horizontally (between government branches). The resulting multidimensional profiles of support

for polyarchy have major advantages over previous measures. Rather than asking citizens if they

prefer “democracy,” the profiles gauge support for a particular conceptualization of democracy.

6

Like regime-level research, polyarchy profiles can map mixed belief systems that diverge from

an ideal-type polyarch.

Data and Methods

Just as Dahl employed institutional metrics to compare governments’ progress towards

citizen responsiveness and limiting its own power (1971, Appendix A), this study seeks to gauge

the extent to which citizens support these ideals and institutions. Whereas Dahl compared

regimes on several measures of public contestation and inclusive participation, multidimensional

belief profiles capture citizens’ orientations to the basic institutions that undergird these twin

dimensions. And just as the conditions for responsive government can be threatened by

legitimately elected leaders, support for it rings hollow unless accompanied by support for the

vertical and horizontal checks and balances that ensure their maintenance.

Therefore the analysis employs four indicators tapping support for public contestation

and inclusive participation, and the institutions, procedures, and norms that make them effective.

The data come from face-to-face, nationally representative AmericasBarometer surveys

conducted by the Latin American Public Opinion Project (LAPOP) in 2006 and 2007.4

Unfortunately, the survey instruments in Ecuador, Bolivia, Paraguay, and Brazil lacked one or

more of the main question batteries. While these cases are regrettably excluded, the analysis

includes the following twelve countries: Chile, Colombia, Costa Rica, El Salvador, Guatemala,

Honduras, Mexico, Nicaragua, Panama, Peru, Uruguay, and Venezuela. The four indicators of

polyarchy support described below are summed into reliable scales5 and normalized with mean 0

and standard deviation 1 (see Appendix 1 for question wordings and coding rules).

• Public Contestation: An index combining orientations to (1) laws prohibiting (a) protests

and (b) the formation of social movements and (2) government censorship of (c)

7

television programs, (d) books in public libraries, and (e) critical media. Answers are

coded so that opposition to censorship and suppression of political organizing receive

high scores.

• Inclusive Participation: An index fashioned from questions regarding the tolerability of

citizens who speak poorly of the regime (a) to vote, (b) to conduct peaceful protests

expressing political ideas, (c) to run for public office, and (d) to voice political views on

television. Answers are coded so that more inclusive orientations are scored higher.

• Limits on Executive Authority: An index tapping beliefs about limits on executive power

via disagreement with the statements, “It is necessary for the progress of this country that

our presidents limit the voice and vote of opposition parties”, “Our presidents’ power

must be limited so as not to put our liberties at risk”, “The Congress hinders the work of

our presidents, and should be ignored”, and “Judges frequently hinder the work of our

presidents, and they should be ignored.” Each statement the respondent disagrees with

increases the value of the index, so high values represent preferences for restrictions on

executive authority.

• Institutions and Processes: Two questions about whether suspending the operation of (a)

the legislature or (b) the Supreme Court would ever be justifiable. Answers are re-coded

so that support for these institutions and processes are scored high.

The first two indexes focus on Dahl’s twin dimensions. While the first one captures

support for the political and civil freedoms necessary for public contestation, the second

concerns support for extending these rights, even to those who disapprove of the regime.6 The

third and fourth dimensions focus on checks and balances. Whereas the third index gauges

support for the president ignoring them, the fourth taps respect for the institutions charged with

8

exercising them. For Dahl, large-scale liberal democracy requires accountable executives,

representative institutions to aggregate preferences, and juridical institutions to interpret the

constitution and uphold the law. Absent these institutions, any political and civil freedoms

citizens enjoy will ring hollow and responsive government, citizen control of the agenda, and

effective participation will remain distant ideals.

The most common aggregation method in the democratic support literature is to collapse

the various measures into a single dimension of democratic support through a factor analysis

(Gibson et al. 1992, Moreno 2001, Mishler and Rose 2002, Inglehart and Welzel 2005, Norris

1999). This method assumes the data are structured on a single dimension with democratic

supporters at one extreme, citizens who reject democratic principles at the other, and comparable

individuals in the middle. This assumption is, nonetheless, frequently violated. For example,

Gibson et al. (1992) claim seven items tap a single latent factor named “support for democratic

rights, liberties, and institutions.” Democratic support is measured by respondents’ factor scores

on this single dimension. This first extracted factor has an eigenvalue of 3.19 and explains 46%

of the variance in the data. Just five of the seven variables load at 0.60 or above, with “political

tolerance” (0.25) and “rights consciousness” (0.41) loading much lower. Moreover, they discard

a second factor with an eigenvalue 0.99 – extremely close to the relatively arbitrary cutoff of 1.0

– that explains 14% of the variance. The ambiguity in this example suggests that, at the very

least, confirmatory factor analysis’s rigid linear assumptions must be tested empirically.

To test whether the linear assumptions hold for this data, the 15 indicators of polyarchy

support listed above undergo principle components analysis (see the web appendix for full

results). The results yield four principal components with eigenvalues greater than 1, with each

factor generally corresponding to one of the four indicators.7 Cronbach’s α scale reliability

9

coefficient for these four indicators is quite low 0.21 with missing data imputed8 – further

disconfirming the unidimensionality of these data and confirming the 4 separate dimensions we

use in our analysis. In other words, these indicators tap four separate dimensions and not a single

latent dimension and should not be combined into a single scale.

Given the theoretical proposition that citizens’ belief systems contain variegated

orientations to polyarchy’s core dimensions and the statistical evidence regarding the structure of

the data, cluster analysis appears superior to factor analysis for revealing the profiles of support

for polyarchy. Hierarchical agglomerative cluster analysis requires no a priori assumptions about

the nature or number of support profiles.9 Schedler and Sarsfield (2007) capture the distinction:

“While factor analysis allows us to discern how different variables hang together across cases,

cluster analysis reveals how cases hang together across different variables” (8, emphasis added).

It classifies cases that vary along multiple variables using numerical distances called similarity

measures. Members of the same profile (cluster) are most similar (numerically proximate) to

each other on the four orientations to polyarchy described above, but most dissimilar

(numerically distant) to members of other clusters on these same orientations. By relaxing the

linearity assumption, cluster analysis better detects mixed support profiles for citizens’ whose

orientations to polyarchy are only weakly “constrained” (Converse 1964).

Specifically, the analysis uses agglomerative hierarchical cluster analysis with squared

Euclidean distance as the similarity measure and Ward’s algorithm Ward’s method calculates the

sum of squared distances from each respondent to the mean of all variables and then minimizes

the sum of squares of any two hypothetical clusters that can be formed at each clustering step.

Cluster solutions are determined by Duda and Hart’s (1973) stopping rule.

The investigation proceeds in two steps. First, clusters are derived for the South

10

American countries (Chile, Colombia, Peru, Uruguay, Venezuela) and Mexico.10 Then, as an

out-of-sample check on the cluster solution,11 the analysis is replicated for six Central American

countries: Costa Rica, El Salvador, Guatemala, Honduras, Nicaragua, and Panama. Respondents

with missing values on the items making up the four variables in the analysis were excluded

from the cluster analyses. These respondents are subsequently imputed into clusters using

multiple imputations by chained equations (MICE) techniques (Royston 2005).12 Five

replications of each country panel were generated (cf. Rubin 1987), and the entries reported in

Tables 2 are averaged across them. The cluster distributions and, subsequently, significant

predictors of the profiles do not change if missing values are excluded.

Distribution of Profiles of Support for Polyarchy in the Americas

Cluster analyses of both the Central American and South American samples suggest five-

cluster solutions, i.e. five profiles of support for polyarchy. Across the two analyses the contents

of the profiles are nearly identical (see Appendix 2), a finding that enhances confidence in the

results’ robustness. Therefore results from the two samples are combined. To simplify the

presentation, variable values in the top third of the scale are considered “liberal”, in the middle

third “ambivalent”, and the bottom third “illiberal.”13

(Table 1 about here)

The profiles are named according to their content. The first is named “polyarch” because

citizens with this profile register support for each of the central tenets of polyarchy under study.

Compared to citizens from other profiles, polyarchs champion contestation rights on a wholly

inclusive basis. They are also fervently committed to the functioning of congress and the courts,

and opposed to executive power grabs that threaten citizen control of the agenda. Analysts find

similar belief systems in Argentina (Powers 2001), Mexico (Schedler and Sarsfield 2007, 2009),

11

Russia (Carnaghan 2007), Belarus, Romania, and Bulgaria (Dryzek and Holmes 2002).

The percentages reported at the bottom of Table 1 indicate that while polyarchs comprise

17.9% of the sample, most citizens do not neatly approximate this ideal-type. Yet none of the

profiles fits an ideal-type non-polyarch; all respondents accept at least one of polyarchy’s

principles. The other four support profiles, comprising over 80% of the sample, are either

ambivalent towards or reject at least one of polyarchy’s principles. Interestingly, most

respondents support contestation, are ambivalent to inclusive participation, and divided on limits

to executive authority. Highlighting how these sub-types of polyarchy support deviate from

ideal-type polyarchs, their names reflect distinct orientations to the use and restriction of power.

The discussion below divides the four profiles into sets of two, beginning with “power

constrainers” and “power checkers” and followed by “power delegators” and “power refrainers.”

Both power constrainers and power checkers advocate a system of checks and balances.

Their liberal status on the “Limits on Executive Authority” and “Institutions and Processes”

dimensions means they bristle at presidential power grabs and seek to limit executives from

encroaching upon congress and the judiciary. Like most Latin Americans with mixed support

profiles, power constrainers and power checkers have limited tolerance for the participation of

political dissidents. These two profiles diverge, however, on public contestation. Beyond

supporting formal checks on the executive, power constrainers wish to use public contestation as

an additional constraint on the executive. In the literature, power constrainers closely

approximate the “intolerant democrats” Schedler and Sarsfield (2007, 2009) identify in Mexico.

Power checkers, in contrast, are wholly ambivalent about public contestation. Although

power checkers do not relish polyarchy’s core political and civil freedoms, they hedge their bets

against presidents with unfettered power, as their liberal orientations to limits on executive

12

power and institutions and processes attest. Thus they mistrust both executives and citizens’

abilities to limit them, placing their faith instead in formal institutions. This profile hints at the

beliefs that may underlie Carrión’s (2008) “liberal authoritarianism” (33).

Power delegators and power refrainers comprise another set of mixed profiles. Unlike the

previous two profiles, these profiles are less willing to make executives completely beholden to

institutional checks. Power delegators prefer the executive obtain wide powers, yet disavow full

autocracy. They do, however, support formal checks to keep the president within certain bounds.

Specifically, they oppose closing congress and the courts, but favor endowing the president with

extensive power over the agenda. They also believe that public contestation should exist as a

check on the executive. Though not a perfect match, this profile shares some traits with belief

systems identified in recent analyses of support for other forms of delegative democracy (Walker

2009, Carrión 2008, Seligson 2007).

By contrast, power refrainers prefer that executives simply refrain from acquiring powers

and potentially subverting citizen control of the agenda. Yet they imagine scenarios that would

justify the executive suspending state institutions that check its power—the congress and

judiciary. Power refrainer is the only profile that tacitly supports such a virtual autogolpe. This

precarious support suggests a general distrust of democratic institutions of all stripes.

The foregoing analysis suggests three conclusions. First, the five support profiles do not

fit on a single continuum but rather vary vis-à-vis the areas they support. Second, most citizens

hold a mixed profile of support for polyarchy. Power refrainers and power constrainers are the

most common, each accounting for almost a quarter of the sample. Power checkers comprise

roughly one fifth of the sample, and power delegators about one seventh. Hence most Latin

Americans sampled neither fully accept nor fully dismiss polyarchy’s key principles. Finally, the

13

conviction polyarchs show towards the basic norms of participation stands in sharp relief to the

mixed profiles’ ambivalence.14 And the four mixed profiles hold distinct orientations to the

exercise of political power. Power constrainers and checkers seek to reign in the executive

through checks and balances, whereas power delegators and refrainers each accept one form of

checks and balances but not the other.

Competing Theories of Polyarchy Support

Widespread dissention from polyarchy’s ideals raises the question of why these groups

abandon them. Of particular interest is how well modernization theory, psychological factors,

ideology, and performance considerations differentiate polyarchs from citizens with mixed

support profiles. An extensive literature offers some guidance.

Modernization theory predicts economic development raises education and income levels

and encourages trust and collective action. In turn this promotes the moderation,

accommodation, cooperation, and bargaining norms that allow citizens “to cope with one of the

central dilemmas of democracy: to balance cleavage and conflict with the need for consensus”

(Diamond 1999, 166). Development also erodes traditional gender norms that impede women’s

participation in politics and the labor force and, thus, exposure to evolving values and resources.

Despite the theory’s appeal, evidence from new democracies regarding the relationship between

democratic values and education, income, and gender is mixed (e.g. Gibson et al. 1992; Booth

and Seligson 2009; Evans and Whitefield 1995; Seligson 2001, 2007; Mishler and Rose 2002;

Moreno 2001; Bratton et al. 2005; UNDP 2004).

Modernization theory also suggests age may be linked to polyarchy support. Inglehart

and Welzel (2005) demonstrate the growth of self-expression values they consider crucial for

democracy among successive young cohorts as material needs are increasingly met. In contrast,

14

there is some evidence that Latin Americans socialized under authoritarianism may be more

open to authoritarianism than cohorts from the transition and post-authoritarian periods (UNDP

2004, Lagos 2008).

The Civic Culture theorizes that democracy depends on a judicious number of citizens

who are psychologically invested in the political system (Almond and Verba 1963). Are

polyarchs more politically interested and sophisticated than their counterparts? Preference for

democracy increases with cognitive engagement in Africa (Bratton et al. 2005) and Chile

(Huneeus and Maldonado 2003) but not in Peru (Seligson and Carrión 2002) or Russia (Gibson

et al. 1992). Political engagement may be positively associated with democratic norms even if it

leads citizens to criticize practices that fall short of that ideal (Booth and Seligson 2009).

Democratization studies beyond Latin America examine the roles of race and ethnicity in

democratic culture (e.g. Dowley and Silver 2002, Bratton et al. 2005) while within Latin

America these matters have received far less attention. In their work on Mexico, Costa Rica, and

Chile, Basáñez and Parás (2001) summarize important racial effects as follows: “The darker the

person’s color, the less democratic their views” (150). A recent study finds ethnicity unrelated to

national pride, though ethnic minorities “are less persuaded of the existence of common values

among citizens in their countries” (Moreno 2008).

Are profiles of support for polyarchy a matter of ideology? With some exceptions, the

belief that democracy is the best system, rejection of unelected rulers, and political intolerance is

hypothesized to increase from Left to Right in Latin America (Seligson 2007). Gunther and

Hsin-chi (2007) find Left-Right placement and party preference correlate with a host of values

that define ideological cleavages in new and old democracies alike.

Citizens’ electoral experiences may inform their attitudes toward democratic principles.

15

In developed democracies, voting for losing candidates leaves citizens less satisfied with and less

supportive of democracy voting for election winners (e.g. Anderson et al. 2005). Electoral

“losers” in Latin America are less likely to perceive existing democratic institutions as legitimate

(Booth and Seligson 2009). Electoral winners, however, may be less likely to oppose an

executive’s attempts to suppress opposition from the public or other branches of government

(Moehler 2009). Thus it is unclear whether electoral winners or electoral losers are most likely to

endorse polyarchy’s principles of inclusive contestation and checks on executive power.

Presidential popularity poses a similar conundrum. If executives are responsive to the

interests and priorities of citizens, then the democratic values behind executive support may help

determine the processes of democratization. Incumbent approval translates into a rejection of

coups d’état in Peru (Seligson and Carrión 2002) and preference for democracy in Chile

(Huneeus and Maldonado 2003) and Africa (Bratton et al. 2005). Yet supporters of the president

may reject authoritarianism while eschewing democratic ideals. Instead they want few limits on

the president they cheer (Carrión 2008). The models below will, thus, test the extent to which

profiles of support for polyarchy, what Easton (1975) might term “diffuse political support,”

reflect incumbent approval, a form of “specific political support”, in the Americas.

A final question is whether support for polyarchy’s main tenets fluctuates with short-run

regime performance. Democratic support may be less tied to performance than support for

specific institutions, satisfaction with democracy, or incumbent popularity (e.g. Duch 1995).

Nevertheless, poor performance can lead citizens to question democracy’s basic premises

(Przeworski et al. 2000, 109; Feng 1997), as can poor governance in key areas like corruption,

discrimination, and crime (Evans and Whitefield 1995, Bratton et al. 2005, Seligson 2002,

Espinal et al. 2006, Weitz-Shapiro 2008, Booth and Seligson 2009). Yet we know little about

16

how performance affects more bedrock democratic principles like free expression, political

tolerance, and civil liberties.

Variables and Methods

The analysis tests which of the most prominent predictors in the literature distinguish

polyarchs from the four mixed profiles—power constrainers, power checkers, power delegators,

and power refrainers. The most theoretically relevant modernization/demographic variables are

age, gender, education, income, religion, and urban/rural residence (Appendix 1 reports full

question wordings and coding rules). Since surveys in several countries did not inquire about

ethnic identity, a dummy variable – scored 1 for respondents whose dominant language is

Spanish and 0 for those who primarily speak an indigenous language – is the proxy.

Two indicators tap psychological engagement in politics: expressed interest in politics,

and political knowledge, based on a battery of five factual questions about politics consistent in

each survey. Ideology is operationalized by dividing a 10-point Left-Right self-placement scale

into three dummy variables: Left Ideology, Center Ideology, and No Ideology (for those who do

self-identify with any point on the scale). Right Ideology is the reference category.

Political variables first distinguish those who voted for a candidate in the previous

presidential election from citizens who abstained or cast a blank/null vote (43% of the sample).

Among the voters, other variables identify those who voted for a losing candidate for president

whose party has won the presidency in the past (13.9%) and those who voted for a losing

candidate whose party has never controlled the executive (11%). Another variable gauges

evaluations of the president’s job performance. Finally, evaluations of the economy, personal

finances, and recent victimization of crime, corruption, and discrimination are included.15

Aside from these individual-level variables, country dummies are included to control for

17

the different country trajectories. These country fixed effects also reduce heteroskedasticity due

to clustering across survey units that could result in overly small standard errors (Steenbergen

and Jones 2002) (see Appendix 3 for country-dummy results).

The first model compares polyarchs to each of the four mixed support profiles by making

polyarchs the base category in a multinomial logistic regression (Table 2). It is worth noting, that

certain variables that do not distinguish polyarchs from citizens in any of the four mixed profiles

do effectively distinguish among the four mixed profiles. In other words, some predictors that are

not statistically significant in Table 2 would be significant if another support profile were the

base category. The most theoretically relevant of these instances appear in the discussion below.

Why Do Citizens Support Polyarchy?

The two variables that have received the most attention by modernization theorists,

education and income, are both substantially related to support for polyarchy. The more years

citizens spend in school, the more likely they endorse the principles of public and inclusive

contestation and accountability. The predicted effect of increasing education from its minimum

to its maximum level is to increase the probability of fitting the polyarch profile by 88%.16

Although education’s effect on polyarchy support is consistent with modernization theory,

education plays almost no role in distinguishing between alternatives to full support for

polyarchy. The only significant difference among the four mixed profiles is that power checkers

are slightly less educated than power constrainers. This suggests that although a lack of

education makes it unlikely that citizens are ideal-type polyarchs, it does not determine the

specific areas of polyarchy they reject.

(Table 2 about here)

Income has an effect similar to education. Wealthy citizens are 74% more likely to

18

adhere to the principles of public and inclusive contestation and to endorse checks and balances

than are the poorest citizens in the sample. Just as with education, though, income levels do not

significantly discriminate between the four less-than-polyarchic support profiles.

Modernization theories are sometimes invoked to explain generational differences.

Young citizens are significantly less likely to be power checkers than power refrainers.

However, age does not distinguish polyarchs from any other support profiles nor does it

significantly differentiate between any other mixed profiles.17

Taken together, then, education and wealth set polyarchs apart from citizens with mixed

support profiles as modernization theories predict. But these variables poorly predict which

specific mixed profiles citizens are likely to fit. One possible implication is that education and

wealth instill tolerance towards the participatory rights of political dissidents. Thus

modernization theory provides little leverage on how citizens hold illiberal orientations toward

public contestation and the legitimate authority of elected and constitutionally defined

institutions. Other theoretical perspectives must be brought to bear on these questions.

Moving to cognitive traits, political engagement – both in terms of expressed interest in

and observed political knowledge – is associated with support for the core tenets of polyarchy.

Increasing political interest over its range boosts the probability of an average respondent being a

polyarch by 30%. A similar increase in political knowledge bolsters the chances of being a

polyarch by 59%. Unlike education and income, however, political interest and knowledge

distinguish among the mixed support profiles. Power constrainers, delegators, and refrainers all

tend to have low levels of political interest and moderate levels of political knowledge. In

contrast, power checkers tend to be just as politically interested as polyarchs (and are

significantly more interested in politics than citizens with the three other profiles) but are the

19

least knowledgeable about politics. As political interest reaches its maximum, the probability of

being a power checker increases by 12%; but as political knowledge reaches it maximum the

odds of being a power checker fall by 30%. This implies a link between political ignorance the

denial of popular participation as an important check on authorities.

Ideology also relates to support for polyarchy, though in nuanced ways not previously

considered. Compared to right-wingers, self-identified leftists are more likely to be polyarchs

than power checkers. An average citizen identified with the ideological left is 21% more likely to

be a polyarch than the average self-identified rightist. Otherwise, left ideology is uncorrelated

with alternative support profiles, a major non-finding given the sample size. Rather, polyarchs

tend not to identify with any part of the Left-Right ideological spectrum. Non-identification with

the ideological scale increases the probability of fitting the polyarch profile by 48%. In fact, the

citizens least likely to be polyarchs locate themselves at the center, not on the right, of the

ideological scale. Centrists, for their part, tend to be power constrainers or power delegators.

In addition to theories of modernization theory and cognitive engagement, the analysis

supports theories of democratic values that focus on short-term performance. Incumbent

approval correlates with ambivalence towards restraining presidential power and opposition

rights. Citizens who support the president may view opposition groups as a threat to their

interests and prefer a president with few checks on power (Carrión 2008). Hence strong

presidential approval could lead citizens to value autocracy, delegative governance (Walker

2009), or hybrid regimes. Polyarchs are the least adulating of their presidents of any profile,

especially compared to their power refrainer, power checker, and power delegator counterparts.

Polyarchy places the burden on rulers to limit their actions and to allow their opponents free

expression and organization. Specifically, moving presidential approval from its minimum to its

20

maximum values decreases the probability of an average citizen supporting polyarchy by 52%.

Not surprisingly, power delegators (who are willing to dramatically increase the president’s

power) tend to approve most strongly of presidents. Indeed, the staunchest supporters of the

president are twice as likely to be power delegators as the staunchest critics.

An interesting pattern emerges regarding previous electoral experience.18 There are no

significant differences in democratic values between those whose favored candidate won the last

election, those who abstained, and those who supported a losing candidate whose party has won

in the past. But citizens who support perennially losing parties have different profiles of support

for democracy than the rest of society. In contrast to the findings in Anderson et al. (2005) but

consistent with work on Africa (Moehler 2009), Latin America’s perennial losers most value

polyarchy’s protections for political dissidents and free speech and its limits on executive power.

Losers are not withdrawing from democracy, they are mobilizing for its protections. So election

winners in Latin America may trust democratic institutions (Booth and Seligson 2009), but

access to power dampens their enthusiasm for polyarchy because polyarchy imposes limits on

election winners as forces them to respect opposition.

Finally, support for polyarchy’s principles is partially contingent upon short-term

performance. Much of this effect runs through presidential approval. If presidential approval is

excluded from the model, citizens who judge the national economy as strong or who report an

improved personal situation are significantly less supportive of polyarchy and the accompanying

restrictions on authority. Economic evaluations are also the main difference between power

delegators and power constrainers. By implication, executives who deliver a strong economy will

face less public opposition to rolling back freedoms, and less clamoring to restrict their authority.

Viewed together, the findings for presidential approval, previous electoral choices, and

21

economic performance suggest orientations to polyarchy are consistently shaped by short-term

political experiences. Positive evaluations of the sitting executive, especially in response to

success in stimulating the economy, open citizens to deviations from polyarchy’s ideals public

contestation, inclusive participation, and citizen control of the agenda. Presidents’ supporters

may be more willing to delegate presidents the authority to go beyond the restrictions polyarchy

idealizes. The good news, however, is that election losers are not rejecting democratic norms.

Instead, losing fosters support for government self-restraint and opposition rights.

Unlike economic outcomes, political outcomes (crime, corruption, and discrimination) do

not differentiate polyarchs from mixed profiles, even without presidential approval in the model.

But these variables do differentiate between the mixed profiles. Corruption and discrimination

victims are significantly more likely to fit the power refrainer profile—which balks at granting

executive extensive powers but remains wary of other formal institutions—than any other mixed

clusters. Yet power refrainers are not significantly more likely to think the national economy is

weak. This suggests that specific political failings undermine support for all major political

institutions, whereas economic failings focus citizens’ attention on the executive.

In summary, the analysis bolsters the validity of multidimensional profiles of polyarchy

support. In line with long-standing theories, support for polyarchy correlates consistently with

education, wealth, and political engagement. It is also related to a lack of ideological

identification and disapproval of the president. However, political interest, presidential approval,

and economic and political performance are what distinguish amongst the various sub-types of

polyarchy and help explain the specific deviations in support for polyarchy’s principles.

Discussion

Latin Americans’ regime preferences are not neat, and often defy simple categorizations

22

like “democrat” and “authoritarian.” While previous studies have discussed this possibility, the

corroborating evidence here it is based on a novel approach that incorporates the collective

wisdom of previous attempts to measure the multiple forms of democratic support in Latin

America. Like UNDP (2004), it begins with a rigorous conceptualization of the background

concept—in this case, polyarchy—and assumes orientations to it are multidimensional. Like

Schedler and Sarsfield (2007, 2009), the measurement approach is exploratory, and, like Carrión

(2008), it incorporates support for liberal rule. Unlike these previous approaches, it makes no

reference to “democracy” thereby enhancing the comparability of its findings across individuals

and countries. The results suggest citizens of the Americas fit one of five belief profiles in which

orientations to four dimensions of polyarchy are bunched categorically not arranged on a linear

continuum. While true polyarchs exist, their ranks are dwarfed by citizens who reject or are

ambivalent towards at least one of polyarchy’s core tenets.

Compared to their counterparts across the Americas, polyarchs stand out for their

vigorous support for contestation. But the four mixed profiles match polyarchs on support for

other dimensions of polyarchy. For example, power checkers are the only citizens who do not

embrace inclusive participation; only power delegators reject limiting executive authority; and

power refrainers alone deny the sanctity of horizontal institutions and processes. So while

polyarchs do not predominate in Latin America, they can potentially rally other citizen sub-types

in defense of democratic principles and norms.

The statistical analysis clarifies why individuals reject some of polyarchy’s principles.

All the mixed profiles lack the levels of education and wealth which seem to make polyarchs

embrace democracy. Orientations to public contestation and inclusive participation are driven by

political engagement and ideology. Hence power constrainers, who are ambivalent towards

23

participation, are not interested in politics, and power checkers, who support neither participation

nor contestation, tend to be politically ignorant and right-wingers. Orientations to political

institutions, by contrast, tend to reflect short-term political considerations and regime

performance. Thus power delegators, who strongly support the president, prefer to erase formal

boundaries to executive power and remain lukewarm on participation while those citizens who

are less enamored with the executive and especially his or her economic record support prefer

more limited government. Finally, as victims of corruption and discrimination, power refrainers

both reject the legitimacy of horizontal institutions and seek to limit presidential power.

The results allow us to revisit some central predictions of democratization theory.

Education and income strongly influence support for polyarchy, thus aligning with theoretical

expectations extending back to Lipset (1959, 1981) and Almond and Verba (1963). Yet short-

term political and performance concerns can weaken citizens’ support for these principles in

systematic ways. Most worrisome from the perspective of democratic theory is the relationship

between presidential approval and rejection of polyarchy. Successful presidents may use their

perceived efficacy to undermine the legitimacy of other democratic institutions. This suggests a

perverse cycle in which specific support for the president, grounded in mixed support for

polyarchy, undermines diffuse support for free contestation and participation. Whereas

presidential supporters appear to have lower resistance to executive consolidation of authority,

presidential detractors, and those whose candidate has yet to win power, seem to seek shelter in

rights and institutions that can prevent the growth of delegative democracy.

In light of this reality, the classification strategy employed here—analyzing profiles of

support for polyarchy—holds key advantages over existing approaches to measuring democratic

support. Conceptually, the measures portray polyarchs as citizens who support the values, norms,

24

25

procedures, and institutions undergirding polyarchy. Operationally, the profiles exclude direct

references to “democracy” in order to avoid the interviewer effects and problems stemming from

disparate definitions of “democracy” that taint existing multidimensional measures. Instead they

intertwine support for four dimensions polyarchy: public contestation, inclusive participation,

limits on executive authority, and horizontal institutions and processes. The goal of this approach

is to create measures of support for polyarchy that adhere closely as possible to Dahl’s original

efforts to conceptualize polyarchy (1971, 1-14) and operationalize it with existing measures

(ibid, Appendix A). Empirically, support for polyarchy is not assumed to be all of a piece, but

rather viewed as belief profiles in which support for the four dimensions of polyarchy cluster

categorically, not linearly. Confirmatory factor analysis clearly violates the linearity

assumptions, as do the subsequent inferential models. Thus polyarchy support profiles are well

suited to testing hypotheses about the cultural foundations of polyarchy and to advance the

debate over the measurement of democratic support than existing measures.

Nevertheless, a key implication of the present analysis is that if support for democracy is

multidimensional, then our theorizing about democratic support must take on a multidimensional

character as well. Moreover, this study demonstrates that theories of democratic support need to

specify the causal factors that lead citizens to reject specific democratic principles (or particular

sets of democratic principles) while clinging to others. Further work is thus needed to verify and

clarify these factors and to refine their theoretical linkages to the formation of mass democratic

values.

References

Almond, G. A. & S. Verba. (1963). The Civic Culture: Political Attitudes and

Democracy in Five Nations. Scranton, PA: Princeton University Press.

Anderson, C. J. A. Blais, S. Bowler, T. Donovan, and O. Listhaug. (2005). Losers’ Consent:

Elections and Democratic Legitimacy. Oxford: Oxford University Press.

Basáñez, M. & P. Parás. (2001). Color and Democracy in Latin America. In R.A. Camp (Ed.),

Citizen Views of Democracy in Latin America (pp. 139-153). Pittsburgh, PA: University

of Pittsburgh Press

Baviskar, S. & M. Malone. (2004). What Democracy Means to Citizens – and Why It Matters.

Revista Europea de Estudios Latinoamericanos y del Caribe 76: 3-23.

Booth, J. & M. A. Seligson. (2009). The Legitimacy Puzzle: Democracy and Political Support in

Eight Latin American Nations. Cambridge: Cambridge University Press.

Bratton, M., R. B. Mattes, E. Gyimah-Boadi. (2005). Public Opinion, Democracy, and Market

Reform in Africa. New York: Cambridge University Press.

Camp, R. (2001). Democracy through Latin American Lenses. In R.A. Camp (Ed.), Citizen

Views of Democracy in Latin America (pp. 3-23). Pittsburgh, PA: University of

Pittsburgh Press.

Carnaghan, E. (2007). Out of Order: Russian Political Values in an Imperfect World.

University Park, PA: Pennsylvania State University Press.

Carrión, J. F. 2008. Illiberal Democracy and Normative Democracy: How is Democracy Defined

in the Americas? In M. A. Seligson (Ed.), Challenges to Democracy in Latin America

and the Caribbean: Evidence from the AmericasBarometer (pp. 21-46). Washington,

D.C.: USAID.

26

Collier, D. & S. Levitsky. (1997). Democracy with Adjectives: Conceptual Innovation in

Comparative Research. World Politics 49 (3): 430-451.

Converse, P. (1964). The Nature of Belief Systems in Mass Publics. In D. E. Apter (Ed.),

Ideology and Discontent. Ann Arbor: University of Michigan Press.

Coppedge, M. (n.d.). Political Culture and Democratization (Chapter 8), in Approaching

Democracy: Research Methods in Comparative Politics.

Coppedge, M., A. Alvarez, and C. Maldonado. (2008). Two Persistent Dimensions of

Democracy: Contestation and Inclusiveness. Journal of Politics 70 (3): 632-647.

Dahl, R. A. (1971). Polyarchy: Participation and Opposition. New Haven, CT: Yale University

Press.

Dahl, R. A. (1997). Procedural Democracy. In R. A. Dahl (Ed), Toward Democracy—A Journey,

Reflections. Berkeley, CA: Institute of Governmental Studies Press.

Diamond, L. (1999). Developing Democracy: Towards Consolidation. Baltimore, MD: Johns

Hopkins University Press.

Diamond, L. (2008). The Democratic Rollback: The Resurgence of the Predatory State.

Foreign Affairs 87 (1): 36-48.

Dowley, K. & B. Silver. (2002). Social Capital, Ethnicity and Support for Democracy in the

Post-Communist States. Europe-Asia Studies 54 (Jun): 505-527.

Dryzek, J. S. & L. Holmes. (2002). Post-Communist Democratization. New York:

Cambridge University Press.

Duch, R. M. (1995). Economic Chaos and the Fragility of Democratic Transition in Former

Communist Regimes. Journal of Politics 57(1): 121-158.

Duda, R. & P. Hart. (1973). Pattern Classification and Scene Analysis. New York:

27

Wiley-Interscience.

Easton, D. (1975). A Reassessment of the Concept of Political Support.” British Journal of

Political Science 5 (4): 435-457.

Espinal, R., J. Hartlyn & J. Morgan Kelly. (2006). Performance Still Matters: Explaining Trust in

Government in the Dominican Republic. Comparative Political Studies 39 (2): 200-223.

Evans, G. & S. Whitefield. (1995). The Politics and Economics of Democratic Commitment:

Support for Democracy in Transition Societies. British Journal of Political Science 25

(4): 485-514.

Feng, Y. (1997). Democracy, Political Stability and Economic Growth. British Journal of

Political Science 27 (3): 391-418.

Gallie, W. (1956). Essentially Contested Concepts. Proceedings of the Aristotelian Society 56:

167-198.

Gibson, J. L, R. M. Duch & K. L. Tedin. (1992). Democratic Values and the Transformation of

the Soviet Union. Journal of Politics 54 (2): 329-371.

Gunther, R. & K. Hsin-chi. (2007). Value Cleavages and Partisan Conflict. In R. Gunther, H.

Puhle & J. Montero (eds.), Democracy, Intermediation, and Voting on Four Continents.

Oxford: Oxford University Press.

Huneeus, C. & L. Maldonado. (2003). Demócratas y nostálgicos del antiguo régimen. Revista

Española de Investigaciones Sociológicas 10 (3): 9-50.

Inglehart, R. & C. Welzel. (2005). Modernization, Cultural Change, and Democracy: The

Human Development Sequence. New York: Cambridge University Press.

Kapstein, E. B. & N. Converse. (2008). The Fate of Young Democracies. New York: Cambridge

University Press.

28

King, G., R. O. Keohane & S. Verba. (1994). Designing Social Inquiry. Princeton, NJ: Princeton

University Press.

Lagos, M. (2008). The Democracy Barometers: Latin America's Diversity of Views. Journal of

Democracy 19 (1): 111-125.

Lipset, S. M. (1959). Some Social Requisites of Democracy: Economic Development

and Political Legitimacy. American Political Science Review 53 (1): 69-105.

Lipset, S. M. (1981). Political Man. Baltimore, MD: Johns Hopkins University Press.

McCoy, J. (2006). International Response to Democratic Crisis in the Americas, 1990-2005.

Democratization 13 (5): 756-75.

Miller, A. H., V. L. Hesli & W. M. Reisinger. (1997). Conceptions of Democracy Among Mass

and Elite in Post-Soviet Societies. British Journal of Political Science 27 (2): 157-190.

Mishler, W. & R. Rose. 2002. Learning and Re-learning Regime Support: The Dynamics of

Post-Communist Regimes. European Journal of Political Research 41 (1): 5-36.

Moehler, D. 2009. Critical Citizens and Submissive Subjects: Election Losers and Winner in

Africa. British Journal of Political Science. 39 (2): 345-366.

Moreno, A. 2001. Democracy and Mass Belief Systems in Latin America. In R.A. Camp (Ed.),

Citizen Views of Democracy in Latin America (pp. 27-50). Pittsbugh, PA: University of

Pittsburgh Press.

Moreno, D. (2008). National Identity and Ethnic Minorities in the Americas. In M. A. Seligson

(Ed.), Challenges to Democracy in Latin America and the Caribbean: Evidence from the

AmericasBarometer (pp. 177-94). Washington, D.C.: USAID.

29

Norris, P. (1999). Conclusions: The Growth of Critical Citizens and its Consequences. In

Critical Citizens: Global Support for Democratic Governance (pp. 257-272). Oxford:

Oxford University Press.

Ottemoeller, D. (1998). Popular Perceptions of Democracy: Elections and Attitudes in Uganda.

Comparative Political Studies 31 (1): 98-124.

Pérez-Liñán, A. (2007). Presidential Impeachment and the New Political Instability in Latin

America. New York: Cambridge University Press.

Powers, N. (2001). Grassroots Expectations of Democracy and Economy: Argentina in

Comparative Perspective. Pittsburgh, PA: University of Pittsburgh Press.

Prothro, J. W. & C. M. Grigg. (1960). Fundamental Principles of Democracy: Bases

of Agreement and Disagreement. Journal of Politics 22 (2): 276-294.

Przeworski, A, M. E. Alvarez, J. A. Cheibub & F. Limongi. (2000). Democracy and

Development: Political Institutions and Well-Being in the World, 1950-1990. New York:

Cambridge University Press.

Puddington, A. (2007). The Pushback against Democracy. Journal of Democracy 18 (2): 125-

137.

Royston, P. (2007). Multiple Imputation of Missing Values: Update of ice. Stata Journal 5 (4):

527-536.

Rubin, D. B. (1987). Multiple Imputation for Nonresponse in Surveys. New York: John

Wiley and Sons, Inc.

Schedler, A & R. Sarsfield. (2007). Democrats with Adjectives: Linking Direct and Indirect

Measures of Democratic Support. European Journal of Political Research 46 (5): 637-

659.

30

31

Schedler, A & R. Sarsfield. (2009). Demócratas Iliberales: Configuraciones Contradictorias de

Apoyo a la Democracia en México. Espiral 44 (XV): 123-159.

Seligson, M. A. (2001). Costa Rican Exceptionalism: Why the Ticos Are Different. In R.A.

Camp (Ed.), Citizen Views of Democracy in Latin America (pp. 90-106). Pittsburgh, PA:

University of Pittsburgh Press.

Seligson, M. A. (2007). The Rise of Populism and the Left in Latin America. Journal of

Democracy 18 (3): 81-95.

Seligson, M. A. & J. F. Carrión. (2002). Political Support, Political Skepticism, and

Political Stability in New Democracies: An Empirical Examination of Mass Support for

Coups d`Etat in Peru. Comparative Political Studies 35 (1): 58-82.

Steenbergen, M. R. & B. S. Jones. (2002). Modeling Multilevel Data Structures. American

Journal of Political Science 46 (1): 218-237.

Tomz, M., J. Wittenberg, & G. King. (2001). CLARIFY: Software for Interpreting and

Presenting Statistical Results. http://gking.harvard.edu

UNDP. (2004). La Democracia en América Latina: Hacia una Democracia de Ciudadanas y

Ciudadanos. New York and Buenos Aires: Aguilar, Altea, Taurus, Alfaguara, S.A.

Walker, L. D. (2009). Delegative Democratic Attitudes and Institutional Support in Central

America. Comparative Politics 44 (1).

Weitz-Shapiro, R. (2008). The Local Connection: Local Government Performance and

Satisfaction with Democracy in Argentina.” Comparative Political Studies 41 (3): 285-

308.

Table 1: Profiles of Support for Polyarchy

Variable Polyarch Power Constrainer Power Checker Power Delegator Power Refrainer Public Contestation Liberal Liberal Ambivalent Liberal Liberal Inclusive Participation Liberal Ambivalent Ambivalent Ambivalent Ambivalent Limits on Executive Authority

Liberal Liberal Liberal Illiberal Liberal

Institutions and Processes

Liberal Liberal Liberal Liberal Illiberal

Percentage of Sample 17.89% 24.19% 19.36% 14.3% 24.25%

32

Table 2: Support for Alternatives to Polyarchy

Power Constrainer

Power Checker

Power Delegator

Power Refrainer

Education -0.044*** -0.066*** -0.056*** -0.045** (0.013) (0.012) (0.013) (0.014)Income -0.076*** -0.088*** -0.082*** -0.056** (0.020) (0.015) (0.016) (0.018)Age -0.002 -0.005 -0.003 0.001 (0.002) (0.003) (0.003) (0.002)Political Interest -0.161*** -0.051 -0.115*** -0.111*** (0.028) (0.032) (0.034) (0.028)Political Knowledge -0.087*** -0.171*** -0.088*** -0.105***

(0.025) (0.027) (0.028) (0.025)Speaks the Dominant Language 0.179 -0.349 -0.096 -0.246 (0.227) (0.237) (0.258) (0.217)Left Ideology -0.093 -0.373*** -0.119 -0.307*** (0.083) (0.098) (0.101) (0.090)Center Ideology 0.151* 0.123 0.191* 0.030 (0.071) (0.085) (0.081) (0.074)No Ideology -0.308** -0.508*** -0.434*** -0.598*** (0.113) (0.104) (0.110) (0.098)Voted for a Losing Candidate that Has Held Office 0.024 0.054 -0.166 0.026

(0.100) (0.128) (0.133) (0.107)Voted for a Losing Candidate that Has Never Won -0.227* -0.112 -0.344** -0.086

(0.102) (0.110) (0.111) (0.102)Abstained in the Last Election -0.064 0.002 -0.127 0.064

(0.078) (0.098) (0.101) (0.083)Presidential Approval 0.123*** 0.268*** 0.378*** 0.233*** (0.037) (0.036) (0.036) (0.037)Evaluation of the National Economy -0.151** -0.092 -0.054 -0.012 (0.044) (0.048) (0.038) (0.046)Evaluation of Personal Finances -0.023 -0.022 -0.117** -0.024

(0.037) (0.044) (0.043) (0.041)Victim of Crime 0.050 0.050 -0.107 -0.053 (0.067) (0.080) (0.082) (0.075)Victim of Corruption -0.087 -0.027 -0.068 0.191 (0.094) (0.087) (0.126) (0.109)Victim of Discrimination -0.215* -0.213* -0.079 0.030 (0.092) (0.091) (0.097) (0.082)Catholic 0.086 0.044 0.061 0.093 (0.077) (0.084) (0.085) (0.062)Female 0.158** 0.166* 0.080 0.010 (0.059) (0.069) (0.073) (0.066)Urban -0.157* -0.152 -0.115 -0.015 (0.073) (0.083) (0.086) (0.075)Constant 1.622*** 2.259*** 1.396** 0.935**

(0.435) (0.517) (0.475) (0.408)N 15342

* p<0.05, ** p<0.01, *** p<0.001, Standard Errors in Parentheses, Country Dummies Omitted

33

34

1 The Honduran case reaffirms democracy’s continued frailty when the military aligns with

factions within a regime.

2 For excellent and detailed reviews, consult Carrión (2008), Schedler and Sarsfield (2007),

Carnaghan (2007), and the papers presented Candidate Indicators for the UNDP Democracy

Support Index (DSI), Center for the Americas at Vanderbilt University, May 5-6, 2006.

3 Support for democracy could be measured based on any definition. Polyarchy is chosen to

since regime-level measures of democracy chiefly tap variation on its two axes, contestation and

participation (Coppedge et al. 2008).

4 For additional information see http://www.vanderbilt.edu/lapop/.

5 Scale reliability is calculated with Cronbach’s α: Public Contestation (0.77); Inclusive

Participation (0.87); Limits on Executive Authority (0.55); and Institutions and Processes (0.77).

6 Since the survey only asks about the participation of regime opponents rather than political

opponents or other less tolerated groups, the scope of the inclusiveness measure is somewhat

limited. Nevertheless, regime opponents serve as a nice reference point from which to measure

the depth of citizens’ tolerance and commitment to open elections.

7 These components jointly explain 68% of the variance in these variables.

8 See fn. 15 for details on the imputation method and model.

9 An alternative approach is to lay out all possible belief profiles from these four dimensions,

discuss their theoretical properties, and use a confirmatory analysis to attempt to identify them.

But with 81 possible combinations of these four variables, this is not tractable nor does the extant

theoretical literature provide strong guidance about which combinations are to be expected. Thus

exploratory methods are preferable. Observing the same clusters in the two sets of countries

analyzed, however, suggests these results are generalizable.

35

10 These 6 countries are grouped because they have the largest populations.

11 Dividing the countries into two sub-samples was also necessary because the cluster analysis of

the unified sample outstripped the constraints of the software package.

12 In addition to the four variable in the cluster analysis, the imputation model includes: income,

education, age, gender, urban-rural, Catholic; socio- and ego-tropic evaluations of the current

economic situation; protest and civic group participation; civic activism index; life satisfaction;

interpersonal trust; system support; trust in the police, church, congress, judiciary, supreme

court, and political parties; presidential approval; political interest; Churchillian and Linzian

democratic support measures; support for a strong unelected leader; authoritarianism index.

13 This convention borrows from Schedler and Sarsfield (2007, 2009).

14 Means tests on contestation and inclusivness confirm polyarchs score higher than the mixed

groups (p < 0.01). The mixed groups are not statistically different from each other (95% level).

15 The survey did not contain questions asking citizens to look beyond their personal experiences

with corruption, crime, or discrimination and to evaluate trends at the national level.

16 Simulations performed with Clarify (Tomz et al. 2001), with all variables set at their mean.

17 Less theoretically relevant demographic variables have even smaller and more inconsistent

effects. Power constrainers and refrainers are 8-9% more likely to be female and polyarchs are

10% more common in urban areas, though this difference is significant only when compared to

power constrainers. Religion has no significant impact.

18 The correlation between presidential approval and previous vote choice is less than 0.13 and

the results do not change if only one of the two sets of variables is included in the model.