See

discussions, stats, and author profiles for this publication at:

https://www.researchgate.net/publication/282258632

Democracies and Dictatorships

in Latin America: Emergence,
Survival, and Fall

Article · January 2013

DOI: 10.1017/CBO9781139047845

CITATIONS READS

63 201

2 authors:

Scott Mainwaring Anibal Perez-Linan

University of Notre Dame University of Pittsburgh
84 PUBLICATIONS 4,877 118 PUBLICATIONS 1,546
CITATIONS CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related
projects:

Ideas, Instituciones y Democracia. Concentración de poder e

inestabilidad política en América Latina View project

All content following this page was uploaded by Anibal Perez-Linan on 11 July 2017.

The user has requested enhancement of the downloaded file.

Democracies and Dictatorships
in Latin America
Emergence, Survival, and Fall

SCOTT MAINWARING
University of Notre Dame

ANÍBAL PÉREZ-LIÑÁN
University of Pittsburgh
32 Avenue of the Americas, New York, ny 10013-2473, usa

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9780521152242
© Scott Mainwaring and Aníbal Pérez-Liñán 2013
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2013
Printed in the United States of America
A catalog record for this publication is available from the British Library.
Library of Congress Cataloging in Publication Data
Mainwaring, Scott, 1954–
Democracies and dictatorships in Latin America : emergence, survival, and fall / Scott Mainwaring,
University of Notre Dame, South Bend, Indiana; Aníbal Pérez-Liñán, University of Pittsburgh.
pages cm
isbn 978-0-521-19001-5 (hardback)
1. Latin America – Politics and government – 20th century. 2. Latin America – Politics
and government – 2st century. 3. Democracy – Latin America – History – 20th
century. 4. Democracy – Latin America – History – 21st century. 5. Dictatorship – Latin
America – History – 20th century. 6. Dictatorship – Latin America – History – 21st
century. 7. Authoritarianism – Latin America – History – 20th century. 8. Authoritarianism –
Latin America – History – 21st century. 9. Political culture – Latin America – History – 20th
century. 10. Political culture – Latin America – History – 21st century. I. Pérez-Liñán,
Aníbal S. II. Title.
jl966.m353 2013
320.98–dc23 2013015859
isbn 978-0-521-19001-5 Hardback
isbn 978-0-521-15224-2 Paperback
Replication datasets and ancillary materials for this book can be found at:
http://kellogg.nd.edu/democracies-materials.shtml
Cambridge University Press has no responsibility for the persistence or accuracy of
urls for external or third-party Internet Web sites referred to in this publication
and does not guarantee that any content on such Web sites is, or will remain,
accurate or appropriate.
1

Introduction

We began this book because we wanted to understand the evolution of political

regimes in Latin America since 1900 and the reasons for the patterns of those
political regimes. What explains why democracies have endured or broken
down? What explains why dictatorships have survived or fallen? What explains
waves of regime change? Even though the literature had many rich case studies,
it was not entirely clear how to cumulate knowledge from these existing studies.
Nobody had previously undertaken a project to explain the emergence, survival,
and fall of democracies and dictatorships for the region as a whole over an
extended period of time.
These empirical issues raised theoretical questions. What theories or theoretical
approaches gave us the most leverage in understanding the emergence, survival,
and fall of democracies and dictatorships in Latin America? From the outset, we
were skeptical that some prominent existing theories would give us much leverage
for explaining these issues for Latin America. Modernization theory, which posits
that more economically developed countries are more likely to be democratic, did
not seem promising as a way of understanding the vicissitudes of democracies and
dictatorships in Latin America. A decade ago, we published an article that showed
a weak and nonlinear relationship between the level of development and democ-
racy in Latin America (Mainwaring and Pérez-Liñán 2003). Our work added to
earlier evidence that modernization theory did not go far toward explaining
political regimes in Latin America (Landman 1999; O’Donnell 1973).
As we worked on some related articles that paved the way to this book, class
theories of democratization enjoyed renewed visibility with the publication of the
works by Acemoglu and Robinson (2006) and Boix (2003). These works see
democratization as a struggle between the poor, who always favor democracy
when it is a viable outcome, and the rich, who prefer dictatorship when stable
dictatorship is feasible. For Latin America (and beyond), these theories are prob-
lematic. In many cases, the poor and the working class strongly supported leftist

1
2 Democracies and Dictatorships in Latin America

and populist authoritarians even when liberal democracy was an alternative out-
come (R. Collier 1999; Germani 1974; Levitsky and Mainwaring 2006; Lipset
1959: 87–126). In other cases, elite actors helped spearhead transitions to democ-
racy (Cardoso 1986; L. Payne 1994). Moreover, contra the assumption of the
class-based theories, for Latin America from the 1980s until 2003, many democ-
racies distributed income from the poor to the wealthy, and none did the opposite.
Nor did Inglehart’s theories of democracy based on mass political culture
(Inglehart 1990, 1997; Inglehart and Welzel 2005) hold much promise as a way
of understanding the rise and fall of democracies and dictatorships in Latin
America. Inglehart’s theories have modernization underpinnings, and modern-
ization theory, as already noted, does not explain regime survival and fall in
Latin America. Moreover, in many Latin American democracies, large numbers
of citizens express indifference about democracy in public opinion surveys. If
large numbers of citizens are not committed to democracy, how can a demo-
cratic public opinion explain the durability of democracy?
Finally, all of the established major theoretical paradigms in comparative
politics focused on within-country variables. Such a focus cannot easily explain
waves of regime change, in which international influences and actors hold sway.
We found theoretical inspiration in the seminal works by Linz (1978b) on
democratic breakdowns and by O’Donnell and Schmitter (1986) on transitions
to democracy, as well as in many case studies about political regimes. We build
on these works, but they did not attempt to develop a theory in the strict sense
(O’Donnell and Schmitter 1986: 3). Linz and O’Donnell and Schmitter focused
on quite proximate questions of regime change and survival and on regime
coalitions, without specifying why different actors join the pro- or anti-
democracy coalitions. Ultimately, our dissatisfaction with existing theories of
regimes and regime change and our desire to provide greater theoretical inte-
gration than Linz (1978b) and O’Donnell and Schmitter (1986) led us to set
forth a new theory of regimes in this book.
We have two primary ambitions. First, we hope to contribute to broader
theoretical and comparative debates about the survival or fall of authoritarian
and competitive (democratic and semi-democratic) regimes. Second, we aspire to
explain regime change and survival1 of dictatorships and competitive regimes in
Latin America from 1945 to 2010, with some glances back at the 1900–44 period.
Because of the inadequacy of existing theories and the advantages that a
theory offers, we concluded that it would useful to elaborate an alternative
theory based on more realistic microfoundations about what motivates political
actors. Our theory looks at systems of actors, posits assumptions about their
preferences and about why regimes fall or survive, and deduces hypotheses from
these assumptions. In a theory, it is not only the individual hypotheses that can

1
Throughout the book, we use the terms “regime survival,” “regime continuity,” “regime durabil-
ity,” and “regime stability” interchangeably. As used here, a stable regime is simply one that
survives even if it faces other forms of upheaval.
Introduction 3

advance social science; it is also the overarching set of integrated and interrelated
propositions (Achen and Snidal 1989). Our theory, which we sketch in this
chapter and present more fully in Chapter 2, integrates the study of transitions
to competitive regimes and of breakdowns of competitive regimes, and by impli-
cation, the study of the durability of dictatorships and of competitive regimes.

a break with the past

Figure 1.1 illustrates the fundamental transformation of regimes in Latin
America, showing the annual percentage of democracies in the region between
1900 and 2010. The first panel depicts the percentage of countries counted as
democracies (as opposed to dictatorships) in the dichotomous classification
developed by Adam Przeworski and his collaborators (Przeworski et al. 2000;
Cheibub and Gandhi 2004). The second panel reflects the percentage of coun-
tries with scores greater than 5 in the Polity scale (Gurr, Jaggers, and Moore

1.1.1. Democratic and Semi-Democratic Regimes 1.1.2. Democratic Regimes

%100 %100

90 90

80 80

70 70

60 60

50 50

40 40

30 30

20 20

10 10

0 0
1900 1950 2000 1900 1950 2000

Democratic, ACLP Democratic, Polity IV

Competitive, MBP Democratic, MBP

figure 1.1 Percentage of Democratic Regimes in Latin America, 1900–2010

Key: ACLP: Classification developed by Alvarez, Cheibub, Gandhi, Limongi, and
Przeworski.
Polity: Countries with scores greater than 5 in the Polity IV scale.
MPB: Mainwaring, Brinks, and Pérez-Liñán trichotomous classification.
Sources: Authors’ elaboration based on Cheibub and Ghandi (2004), Przeworski et al.
(2000), Polity IV 2012 (http://www.systemicpeace.org/polity/polity4.htm), and Table 1.1.
4 Democracies and Dictatorships in Latin America

1990; Jaggers and Gurr 1995; Polity IV Project 2012).2 We also present the
classification of political regimes developed for this project, introduced later.
Figure 1.1 suggests that the Przeworski et al. measure is more lenient than a
classification based on a score of greater than 5 on the Polity IV scale. Yet all
three measures confirm the occurrence of an unprecedented wave of change
between 1978 and 1995. They depict a similar trend for the last part of the
twentieth century, suggesting reliability in the overall picture.3 Democracy
expanded somewhat in the late 1950s, and then hit a nadir in 1976–77, followed
by an unprecedented surge during the 1980s.
Until the wave of democratization that began in 1978, authoritarian regimes
were pervasive in most of the region. Many democracies were short-lived, and
several countries had had no experience whatsoever of competitive political
regimes. The situation changed profoundly between 1978 and 1995. A region
that had previously always been predominantly authoritarian witnessed the
virtual demise of openly authoritarian regimes. Moreover, since 1978, compet-
itive regimes have been far more durable than ever before. Compared to what
occurred in earlier waves of democratization in Latin America, this wave has
lasted much longer and has been broader in scope. This transformation is one of
the most profound changes in the history of Latin American politics.
The increase in the number of democracies and semi-democracies in Latin
America between 1978 and 1995 was dramatic. At the beginning of this period,
Latin America had only three democracies, and the other seventeen countries
had openly authoritarian regimes. By 1990, the only openly authoritarian
governments were those of Cuba and Haiti. By 1995, Cuba was the sole holdout
(although Haiti eroded back into authoritarian rule between 1999 and 2006).
The shift away from authoritarianism was dramatic in speed and breadth. The
trend is even more striking if we consider the total proportion of Latin
Americans living under competitive regimes. In 1900, only 5 percent of the
regional population enjoyed democratic or semi-democratic politics. In 1950,
it was 58 percent. The percentage plummeted to 12 percent of the regional
population by 1977, but it had reached 98 percent by 2006.
Figure 1.1 also displays the evolution of political regimes according to our
own classification. We classify regimes in Latin America using a simple trichot-
omous scale developed with Daniel Brinks (Mainwaring et al. 2001, 2007):
democratic, semi-democratic, and authoritarian. We lump together the demo-
cratic and semi-democratic regimes into a broader category of “competitive

2
The Polity scale ranges between –10 (authoritarian) and 10 (democratic). The threshold of 5 is
conventionally employed to distinguish full democracies from other types of regimes.
3
The Polity score (the only available for the 1900–45 period beside our own classification) does not
consider the extension of voting rights, so it overestimates levels of democracy in the early
twentieth century. These four measures of democracy are strongly correlated. The series for the
proportion of democracies and semi-democracies according to the Mainwaring et al. three-point
scale correlates at .98 with the Przeworski series, at .93 with the Polity index, and at .97 with
Freedom House scores.
Introduction 5

regimes” displayed in panel 1.1.1. We explain our coding of political regimes in

Chapter 3.

the argument in brief

1) Political actors should be at the center of theories of regime survival and
change. Political actors, not structures or cultures, determine outcomes, even
though structures and cultures affect the formation and preferences of actors.
We view presidents and organizations such as parties, unions, business associ-
ations, the military, and organized movements as the most important actors.
These organizations and presidents control political resources and therefore
exercise influence in the competition for power.
We locate our theory between structural or long-term cultural approaches, on
the one hand, and agency and contingent action approaches, on the other. In
many theoretical perspectives, purposeful action is the final step in a long causal
chain that is largely determined by deep structural (e.g., Boix 2003; Skocpol
1979) or cultural (Foucault 1972; Inglehart and Welzel 2005) forces that tran-
scend individual actors. In these structural and cultural accounts, actors’ deci-
sions are largely determined by macro forces. On the other hand, we emphasize
the constraining and structuring of powerful organizations more than
approaches that focus on individual leaders’ decision making.
2) We emphasize the role of political factors that help political regimes survive
or lead them to fail. By “political factors” we refer specifically to the impact of
actors’ normative preferences about democracy and dictatorship, their moder-
ation or radicalization in policy preferences, and international political influen-
ces exercised through external actors. We counterpose an emphasis on these
political factors to analyses that argue that the survival or displacement of
regimes depends largely on structural factors such as the level of development,
the class structure, or income inequalities, or on mass political culture.
These political factors have primacy in determining whether regimes fail or
remain stable. The empirical evidence for Latin America in the twentieth century
supports a primary focus on political factors such as the level of radicalization,
actors’ normative commitment to democracy, and a favorable international
political environment. With a normative democratic commitment on the part
of powerful political players and a favorable international environment, democ-
racy can survive in the face of daunting challenges: poverty, significant ethnic
cleavages, deep social inequalities, high inflation, and low growth (Linz 1988;
Remmer 1996). Indeed, democratic and semi-democratic regimes have survived
in post-1977 Latin America in the face of all these unfavorable conditions. This
capacity of democracy to survive despite seemingly highly adverse conditions
flies in the face of many theoretical expectations before the latest wave of
democracy began.
Other analysts have also focused on political factors in understanding regime
survival and fall. We add to and modify most previous work by presenting these
6 Democracies and Dictatorships in Latin America

ideas in an integrated framework and by testing the theory and specific hypoth-
eses in new ways.
2a) Actors’ normative attitudes about democracy and dictatorship are impor-
tant influences in regime survival or fall. If the most powerful actors have a
normative preference for democracy – if they believe that democracy is intrinsi-
cally the best political regime even if it does not satisfy their other policy
preferences – democracy is more likely to survive.
Our focus on the impact of actors’ normative attitudes on regime outcomes
builds on literatures in political science and sociology that have emphasized the
importance of actors’ beliefs in understanding political outcomes. Actors’ beliefs
influence what they view as desirable and how they pursue their interests
(Berman 1998; Blyth 2002; Finnemore 1998; Goldstein 1993; Hall 1989;
Sikkink 1991, 1993). If powerful actors view liberal democracy as an inefficient,
corruption-plagued obstacle to rapid economic growth, as the Argentine mili-
tary and big business did in the 1960s, when a competitive regime in a poor or
medium income country falters in economic performance, it is vulnerable to
breakdown. If powerful leftist actors believe that liberal democracy is a facade
for bourgeois domination, as most of the Marxist tradition did, they are likely to
mobilize for workers’ gains even if this mobilization endangers the regime.
Conversely, if actors intrinsically value democracy as a “universal value”
(Coutinho 1980), they accept policy sacrifices to preserve democracy, and they
are more likely to view democracy as an intertemporal bargain (Przeworski
1991, 2006) in which they can compensate for today’s sacrifices by gaining
tomorrow. We contribute to the literature on the political impact of actors’
beliefs or preferences by testing this argument in new ways.
2b) Actors’ policy radicalism hinders the probability that a competitive
political regime will survive. Policy moderation facilitates the survival of com-
petitive regimes. Several studies have claimed that the content of the policy
preferences embraced by powerful political actors (for instance, a preference
for or against income redistribution) have important consequences for political
regimes. The intensity of actors’ policy preferences, and not just their substance,
is critical for regime survival and fall. Radical policy preferences make actors on
the left and on the right of the policy spectrum intransigent and thus unlikely to
tolerate the give-and-take of democratic politics.
3) A favorable regional political environment, characterized by the existence
of many democracies in Latin America, increases the likelihood of transitions
from authoritarian rule to competitive regimes and diminishes the likelihood of
breakdowns of existing competitive regimes. Our theory emphasizes the
embeddedness of countries’ political actors and political regimes in a regional
and international context.
Recent work on democratization has emphasized two factors that are at odds
with an exclusive focus on domestic factors. First, democratization occurs in
wave-like processes; what happens in neighboring countries has a significant
impact on a region. Consistent with the arguments of Brinks and Coppedge
Introduction 7

(2006), Huntington (1991), and Markoff (1996) at a global level, change in

political regimes in Latin America has occurred in waves. It would be difficult to
explain wave-like change only on the basis of within-country conditions if there
were no transnational effects. The likelihood that political transformations
regional in scope could be explained solely by the simultaneous change of
domestic conditions in multiple countries is very low. Theories of democratiza-
tion that are based exclusively on country-level conditions are therefore ill
equipped to explain waves of democratization.
Second, these wave-like processes often bring about profound changes in
political regimes in a region in a short time. In Latin America, the change from
a region that was overwhelmingly authoritarian in 1977 to one that is over-
whelmingly democratic or semi-democratic occurred rapidly. Most comparative
politics approaches that explain democratization involve long, slow processes.
Political culture at the mass level, the level of development, the size and strength
of the working class, and income inequality, all of which have been offered as
explanations of democratization, usually changes only over the long run.
Because the domestic factors that have traditionally been used to explain regime
change move relatively slowly, the likelihood that they could account for pro-
found change in a region in a short time is extremely low.
Synchronicity and rapidity of change do not definitively prove that democra-
tization had powerful international causes, but they greatly increase the like-
lihood that international factors were at work. Many recent works have
emphasized the impact of international actors,4 regional influences,5 and inter-
national organizations6 on democratization. Consistent with this burgeoning
literature, we underscore that battles over political regimes involve not only
domestic actors, but also international and transnational actors.
Our work contributes in five ways to the existing literature on international
effects on political regimes. First, we include international effects and actors as
part of a theory of regime change and stability. Little previous work has inte-
grated domestic and international actors in a theoretical understanding of
regime dynamics. Second, an important question has remained unanswered by
the existing literature. Because the wave of democratization was more or less
contemporaneous with an increasing emphasis by U.S. foreign policy on
“democracy promotion,” it is hard to disentangle the effects of regional diffu-
sion per se from the role of U.S. foreign policy. We separate these effects in
Chapter 4. Third, although the literature on international diffusion of political
regimes has burgeoned in recent years, the analysis of the mechanisms behind
diffusion is less developed. We analyze this issue in Chapter 7. Fourth, we show
that international influences have reinforcing dynamics that help explain the

4
Brinks and Coppedge (2006); Gleditsch (2002); Gleditsch and Ward (2006); Markoff (1996);
Pridham (1991, 1997); Starr (1991).
5
Brown (2000); Levitsky and Way (2010); Whitehead (1986b, 1996).
6
Pevehouse (2002a, 2002b, 2005).
8 Democracies and Dictatorships in Latin America

magnitude and pace of waves of democratization and authoritarianism

(Chapters 4 and 7). Finally, in Chapter 8 we show that while international actors
facilitate transitions to democracy and prevent the breakdown of competitive
regimes, they are not effective at promoting the advancement of competitive
regimes once a transition has taken place.

empirical contributions: understanding

political regimes in latin america
Empirically, the book examines democratization and regime change in Latin
America over a long sweep of time.7 We hope to make three empirical contri-
butions. First, we aspire to contribute to understanding the history of political
regimes in Latin America from 1900 to 2010. Along with Daniel Brinks, and
with the help of sixteen research assistants over the course of a decade, we coded
political regimes as democratic, semi-democratic, and authoritarian. We discuss
our coding rules and procedures in Chapter 3. Our classification of political
regimes lays the groundwork for understanding the evolution of democratiza-
tion and authoritarianism in the region and provides a research tool that other
scholars can use.8
Second, this is the first book that tries to explain the emergence, survival, and
fall of political regimes for Latin America as a whole over a long period of time.
There is a huge literature on political regimes in Latin America. However, much
of it focuses on single countries or a few countries. Drake (2009), Hartlyn and
Valenzuela (1994), and P. Smith (2005) offer valuable descriptive histories of
democracy in Latin America, but with little effort to explain regime emergence,
survival, and demise.
Third, this is the first book that has attempted to extend an actor-based
approach to political regimes to the empirical study of a large number of
countries over an extended period of time. Many scholarly approaches agree
that political actors (rather than structures or political culture) offer the most
fruitful perspective to study political regimes. Such approaches claim that actors’
choices determine regime outcomes, and that structures and cultures, even
though they influence the actors that emerge and their behavior, do not deter-
mine their choices. Actor-based approaches to studying political regimes are
common in case studies (Berman 1998; Capoccia 2005; Figueiredo 1993; Levine
1973, 1978; Linz 1978a; O’Donnell 1982; Stepan 1971, 1978; A. Valenzuela

7
By Latin America we refer to the twenty countries in the western hemisphere that were colonized by
Spain, France, or Portugal: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Cuba, the
Dominican Republic, Ecuador, El Salvador, Guatemala, Haiti, Honduras, Mexico, Nicaragua,
Panama, Paraguay, Peru, Uruguay, and Venezuela. We do not include countries colonized by Great
Britain or the Netherlands.
8
Drake (2009) and Smith (2005) also describe the evolution of democracy in twentieth-century
Latin America.
Introduction 9

1978; J. S. Valenzuela 1985; Viola 1982). Theoretical frameworks such as those

of Linz (1978b) and O’Donnell and Schmitter (1986) also posit that actors (or
blocs of actors) are the most useful unit of analysis. Yet given the time-intensive
demands of studying a large number of actors across a long period of time in a
substantial number of countries, there hitherto has been no extensive (i.e.,
involving a large number of cases) empirical testing of theoretical propositions
about the effects of actors’ preferences on regime outcomes.
Working with a different team of nineteen research assistants, we identified
the main actors operating under every presidential administration in the twenty
Latin American countries from 1944 to 2010 and also coded their attitudes
toward democracy and dictatorship and their policy moderation/radicalism.9 If
actors (as opposed to structures or cultures) determine political outcomes,
actually examining their preferences and behavior is essential. Some excellent
studies have followed this precept for one or a small number of countries, but no
previous work has coded actors for so many countries over a long period of time.

why develop a theory?

Scholars working on political regimes confront several choices. In terms of the
overall analytical strategy, the main question has been whether to develop a
theory with an integrated set of hypotheses that is deduced from explicitly
articulated initial assumptions (Acemoglu and Robinson 2006; Boix 2003); a
theoretical framework that provides a general orientation toward studying
political regimes (Linz 1978b; O’Donnell and Schmitter 1986); or a set of
narrower empirical hypotheses (Cutright 1963; Morlino 2008: 47–51;
Przeworski et al. 2000).10
Each of these options has advantages and disadvantages. Theories provide
integrative ways of understanding the world – an advantage, given our objec-
tives. A theory makes explicit who the actors are and how they are constituted,
what motivates their behavior in regime games, and how they form winning
coalitions. Empirical propositions that are not integrated by theories or by
theoretical frameworks such as Linz (1978b) and O’Donnell and Schmitter
(1986) do not explicitly embed their analyses into an understanding of these
issues. In contrast, the empirical propositions that a theory deductively generates
are part of an integrated whole (Bunge 1998: 433–43). Some scholars
(Coppedge 2012: 49–113; Munck 2001) have commented on the lack of theo-
retical integration in most work on political regimes and argued that this

9
To be precise, we coded all presidential administrations that lasted long enough to be in power as
of December 31 in at least one year. If a president began his term in a given year and did not serve
until the end of that year, we did not include that administration in our dataset.
10
These distinctions could be seen as a continuum rather than as three discrete categorical
possibilities.
10 Democracies and Dictatorships in Latin America

constitutes a weakness in this literature. We agree with their judgment; our effort
at building a theory responds to their observations.
Notwithstanding the sophistication of some of the work that has inspired us,
there have been no previous efforts along the lines presented here to develop a
theory of regime survival and fall.11 The insights of the rich literatures on which
we draw do not fully substitute for a theory of regime survival and fall. These
insights are not generally connected to each other in a system of cohesive and
logical relationships. As a result, work on political regimes has accumulated
considerable knowledge, but with less theoretical integration than is desirable.
As Coppedge (2012: 49–113) comments, with loose integration, a research
finding about the importance of certain independent variables could be compat-
ible with a wide range of theories.
Social scientists want to know not only whether some specific independent
variables affect political outcomes, but also what theories hold up (Bunge
1998: 433–43). Because it consists of a system of integrated hypotheses
deduced from explicitly articulated assumptions, a theory helps order and
organize hypotheses.
Our book integrates previous streams of research into a cohesive theory. The
core contribution of our work is not the five discrete hypotheses about regime
survival and fall that we present later. Rather, it is the theory, which links these
hypotheses in deductively logical ways, and the testing of it. A theory is a way of
making sense of the world, of providing an integrated framework. Discrete
hypotheses can also advance understanding in the social sciences, but theories
help stimulate advances in how social scientists think about politics. The devel-
opment and testing of theories is a critical part of social science (Achen and
Snidal 1989; Bunge 1998: 433–43; Coppedge 2012: chapters 3–4; Ferejohn and
Satz 1995; Munck 2001).
Our understanding of “theory” is not restricted to formal models. Our
endeavor is a theory because it starts with some explicitly articulated assump-
tions about the relevant set of actors and the factors that determine their choice
of regime coalition, and then we deduce an integrated set of hypotheses from
these assumptions.12

actors and regime coalitions

The notion of political actors forms the first building block of our theory. We
focus on a parsimonious set of the most important political actors: presidents,

11
Linz (1978b) and O’Donnell and Schmitter (1986) developed theoretical frameworks that have
some of the characteristics of a theory, but without a set of integrated hypotheses.
12
The formal-theory approaches such as Boix (2003) offer tight integrated theories that provide
logical microfoundations for specific macro-hypotheses. Some frameworks (Linz 1978b;
O’Donnell and Schmitter 1986) offer heuristics to guide the inquiry of researchers into cases
or topics. In this regard, our theoretical discussion follows the second tradition more than the
first one.
Introduction 11

powerful organizations, and influential organized movements. In democratic

regimes, the president and the largest parties are important actors. The govern-
ment commands many resources, and because it directs the policy process, it
strongly influences future resources and outcomes. As head of the government,
the president exercises great influence over the government and more broadly
over democratic politics. Parties are the primary route to achieving elected office
in democratic politics. Democratic politics revolve significantly around the
competition among parties.
The military, guerrilla organizations, social movements, nongovernmental
organizations, unions, and business associations are sometimes major actors.13
In authoritarian regimes, the most important actors always include the president
and often include a hegemonic party (if there is one and if it is reasonably
independent with respect to the president), the main opposition party (under
authoritarian regimes with competitive elections), and the military.
In our theory, actors’ purposeful action largely determines regime outcomes.
Actors form preferences about a political regime based on what they see as
desirable outcomes (specifically in terms of policy preferences and regime pro-
cedures) and they act on the basis of those preferences. Political actors are
instrumental, but they are not always only instrumental or narrowly self-
interested. The theory does not deny that actors’ behavior can have unintended
consequences.
Our understanding of who the actors are and what motivates them diverges
from some theories. In class-based accounts, social classes are the actors. In
contrast, in our view, social classes are usually not sufficiently organized and
sufficiently politically cohesive to form political actors. Labor-based political
parties and labor unions are actors, but the working class per se is not unless
labor organizations or political parties forge political unity among most work-
ers. Similarly, capitalists per se are usually not a unified political actor; they have
competing interests and usually lack a single organization that speaks for all of
them.14 When their interests are deeply threatened, business owners might forge
a temporary unity that enables them to function like an actor. For short periods,
social classes can function like actors when they respond almost uniformly to a
political event or process, but such uniformity is the exception. Usually, social
classes face difficult collective action problems (Olson 1965); they are internally
divided both structurally and politically.

13
Congress is an important decision-making arena in competitive political regimes, but it is not
sufficiently united to be an actor. In conflicts about political regimes, legislatures are usually
divided along party lines, so we take the parties, not congress per se, to be the actors.
14
In Chapter 6, we argue that big business in El Salvador usually functioned as a relatively cohesive
actor from 1931 until 1977. This exception to the rule occurred in part because of perceived
powerful threats from radical popular and/or insurgent movements. In addition, many big busi-
ness enterprises in El Salvador were diversified across different sectors. For example, big coffee
producers typically also owned firms in other sectors, thus reducing conflict among different
economic sectors.
12 Democracies and Dictatorships in Latin America

Our treatment of actors also diverges somewhat from that of pioneering

contingent action approaches such as Linz (1978b) and O’Donnell and
Schmitter (1986). Both of these works focus on blocs of actors in a manner
that is akin to our regime coalitions. They conceptualize the actors according to
their positions regarding the existing regime. For example, Linz’s loyal, semi-
loyal, and disloyal oppositions are blocs of actors that share a common orienta-
tion toward the democratic regime.
Public opinion is not an actor because it cannot per se act. However, in
competitive regimes, public opinion is one of the most valuable resources that
actors can employ. It often sways powerful actors one way or the other in regime
battles. For example, it is unlikely that a successful coup could occur in the face
of solid public support for democracy. Conversely, democracy is more likely to
be imperiled if large parts of the public turn against it. In democracies, public
opinion routinely limits what leaders can do (Brooks and Manza 2007).
Likewise, the fate of dictatorships sometimes hinges on whether citizens turn
so obviously against the rulers that it emboldens opposition actors and encour-
ages splits in the ruling coalition. In short, public opinion is important in regime
battles, but it is not an actor. Likewise, electoral support is a hugely important
asset for parties and the president in democracies, but voters do not constitute an
actor; they are divided and are almost never capable of cohesive action.
Actors have different kinds of political resources. “Political resources” are
any assets (including material and human capacities, institutional advantages
that accrue from formal rules of the game, and for the military, arms) that can be
mobilized in the competition for power. Political resources may be highly con-
centrated or widely dispersed. Actors with intense preferences about the political
regime work especially hard to mobilize their resources and to create new ones.
For the government, military and state capacity always represents valuable
resources.
Whether political regimes survive or are replaced depends on how powerful
the coalitions that support them are. Every regime hosts at least two simple
coalitions, one that supports the incumbent regime (for example, a democracy)
and an opposition coalition that supports its displacement (e.g., the author-
itarian coalition). Multiple coalitions (e.g., several opposition blocs pursuing
different forms of authoritarian rule) are not unusual in times of great turmoil.
Many actors remain on the sidelines and join neither coalition. These regime
coalitions are usually not formalized, and the partners in the coalitions shift over
time. Regime coalitions win state power when they control enough resources to
prevail in the competition for power. In the advanced industrial democracies, the
regime coalition that supports the status quo (i.e., liberal democracy) vastly
overpowers any other alternatives, and therefore the probability of breakdown
in the current historical context is virtually zero.
Once in office, the leaders of a regime coalition adopt policies and build,
preserve, or modify the existing political regime. A regime type (a competitive
regime or a dictatorship) survives if the size and leverage of its coalition is greater
Introduction 13

than the coalition working for regime change. The regime changes when the
opposition coalition is more powerful.
Most actors are not intrinsically part of either the democratic or the author-
itarian bloc. They may change regime coalitions depending on how effectively
the existing regime satisfies their instrumental policy preferences and, in some
cases, their normative preferences about the regime itself (i.e., some actors prefer
democracy even if they believe they might get better policy outcomes under
dictatorship). All political actors have policy preferences, and some of them
have value preferences about the political regime itself. They support regime
coalitions that they believe are likely to maximize their policy goals and their
normative preferences about the regime.

between structure and agency: the level

of analysis and the core variables
Another issue in the study of political regimes is where to anchor the proper level
of analysis for explanations of regime change. Some scholars have emphasized
long-run preconditions (e.g., Moore Jr. 1966) while others have emphasized
leaders and contingent action in specific historical contexts (Capoccia 2005;
Hartlyn 1984; Karl 1987; Kuran 1989, 1991; Linz 1978b; Stepan 1978).
Rustow (1970) framed this question as a dilemma between functional theories
and genetic explanations, Przeworski (1986) presented it as a tension between
macro- and micro-oriented perspectives, and Karl (1990) conceived of it as
explanations based on structure and others based on agency. This problem is
related to the substantive distinction between explanations of democratization
based on socioeconomic conditions and those based on political factors, but it is
analytically distinct. Most explanations of regime change based on socioeco-
nomic variables conceptualize long-term processes, but arguments about the
impact of economic performance on political stability (e.g., Merkx 1973) often
imply causal mechanisms operating in the medium or short run. Most arguments
about political factors refer to short-term processes, but claims about political
culture (e.g., Inglehart 1990, 1997; Inglehart and Welzel 2005; Wiarda 2001)
are based on long-term legacies.
We situate our analysis between long-term explanations such as social struc-
tures and short-term explanations based on actors’ contingent decisions in quickly
shifting conditions. Certainly, long-term factors such as the level of development,
the degree of social inequality, and the persistence of profound ethnic, linguistic,
or religious fractures affect the viability of democracy. But a fundamental theme in
this book is that for our universe of cases, these long-term factors have limited
capacity to explain regime survival and fall. To understand why regimes endure or
fall, we need to shift the analysis to more proximate causes of regime change.
We do not deny the role of structural forces (patterns of economic develop-
ment and dependence, class structures, legacies of social inequality, and so on) in
14 Democracies and Dictatorships in Latin America

the constitution of political regimes. Structural conditions powerfully influence

the emergence and development of political organizations and the distribution of
resources. These more distant structural causes play a role in the genesis of
political regimes. But the effect of structural variables is contingent and diffuse;
it ultimately manifests itself in the organization of political actors and in the
relative distribution of their political resources.
At the other end of the spectrum, we emphasize causal factors that are more
distant than those analyzed by Kuran (1989, 1991) and Lohmann (1994) in their
fascinating accounts of the transition to democracy in Eastern Europe and the
former East Germany and O’Donnell and Schmitter (1986) in their landmark
contribution on transitions to democracy. The short term agency based explan-
ations of breakdowns and transitions of Linz (1978b) and O’Donnell and
Schmitter (1986), while richly capturing important processes and interactions,
do not tell us under what less proximate conditions breakdowns, democratic
survival, and transitions are more likely. We need theories and hypotheses that
are situated between structure (or causally distant cultural explanations) and
agency to complement existing knowledge. This is the terrain where our theory is
located.
Because our theory focuses mostly on fairly proximate variables in the
sequence of causation, it is compatible with theories and theoretical frameworks
that examine more distant or more immediate causes. For example, moderniza-
tion theory and our theory could both help explain why democracies and
dictatorships survive or fall. The former focuses on more distant causes, and
ours on more proximate causes. However, for Latin America from 1945 until
2010, as we show in Chapters 4, 5, 6, and 9, the prominent more distant macro
theories have little explanatory power for understanding regime change and
survival.
At the meso level of analysis that inspires our effort, three variables affect
whether regimes remain in power or fall: (1) whether actors have moderate or
radical policy preferences (radical actors tend to be destabilizing in competitive
regimes); (2) whether they have a normative preference for democracy or
authoritarianism; and (3) how supportive the regional political environment is
for competitive and authoritarian regimes. In this section, we introduce the
arguments about these three variables.
Radical Policy Preferences. We define radical policy preferences as those
toward one pole of the policy spectrum (e.g., toward the left or right when the
policy space is effectively unidimensional15) in conjunction with an urgency to
achieve these preferences in the short to medium term where they do not
represent the status quo, or with an intransigent defense of these positions
where these positions represent the status quo.16 They have two main

15
If there is more than one important dimension of competition, the radical/moderation continuum
functions in all of them.
16
In game theoretic terms, these actors have a large discount factor.
Introduction 15

characteristics: (1) because their policy preferences are toward one pole of a
policy spectrum, the unqualified adoption of their preferences imposes impor-
tant costs on other actors; (2) their preferences are very intense, so the actors are
intransigent (i.e., unwilling to bargain) and impatient (i.e., unwilling to wait for
the long term to achieve their policy goals). Radical policy preferences need not
be on the extreme left or extreme right, but they must be far enough from the
policy preferences of other relevant actors to create polarization. The location of
radical policy positions cannot be determined a priori, as it depends on the
nature of the policy space.
The argument about radicalism captures the delicate historical balance
between conservative actors’ demand for security and progressive actors’
demand for policy transformation. Put in Dahl’s (1971) terms, mutual guaran-
tees among actors increase the viability of polyarchy. For a democracy to survive
in poor- and intermediate-income countries, it is helpful that the actors who can
destroy the regime – the military and sometimes the economic elite – not fear the
possibility of major losses in a short time. If they do, they are more likely to join
the authoritarian coalition. At the same time, actors who pursue policy change
should believe that transformations are ultimately viable as a result of demo-
cratic political competition. If the intensity of their policy preferences leads either
conservative or progressive actors to believe that their goals cannot be achieved
under competitive rules, those actors might support an alternative regime able to
impose their most favored policies unilaterally. Their withdrawal from the
democratic coalition often prompts their opponents to do the same, because
uncertainty about policy gains will now turn into the prospect of permanent
losses imposed by the radical group. The fear of major losses in the short term
thus arises when some actors have radical policy preferences.
Actors’ Normative Preferences about Democracy and Dictatorship. Some
actors have strong value preferences about the political regime in addition to
having instrumental policy preferences. These orientations range from a strong
value preference for a particular form of authoritarianism to a strong normative
preference for democracy, with indifference toward regime type in the midpoint
of the scale.
A normative preference for democracy or dictatorship refers to the willing-
ness of political actors to incur policy costs in order to defend or achieve their
preferred regime. It means that an actor prefers a kind of regime on intrinsic
grounds, as the best possible political regime. When candidates acknowledge
their defeat in an election (rather than questioning its results) and gracefully
congratulate their opponents, they are behaving in ways that signal commitment
to the principles of the democratic regime. When government leaders accept
defeat on an important issue that requires a legislative supermajority, even if they
could modify procedural rules to impose the preferred legislation by simple
majority, they are signaling commitment to existing procedures. This commit-
ment is credible to others because the behavior implies a cost to the actor
involved. Observers infer that the player must have a latent normative preference
16 Democracies and Dictatorships in Latin America

(a favorable predisposition) toward the regime, and that this preference must be
strong enough to overcome the short-term losses.
Normative preferences about the regime are part of an actors’ belief system or
view of the world. They are an example of “procedural utility” – the well-being
derived from procedures above and beyond the outcomes they generate (Frey,
Benz, and Stutzer 2004).17 They are consistent with what Max Weber (1978)
called “value rationality.” This argument builds from evidence that individuals
care not only about instrumental gains (outcomes), but also about procedures
(Benz and Stutzer 2003; Frey et al. 2001; Frey et al. 2004; Frey and Stutzer 2005;
Gangl 2003; Levi et al. 2009; Lind et al. 1993; Sen 1995, 1997; Stutzer and Frey
2006), including the ones that constitute a democratic regime.
A strong normative preference for democracy by powerful actors, especially
the president and the major parties, reduces the odds that a competitive regime
will break down. Actors’ normative preference for democracy can help inoculate
competitive regimes from breakdowns. If the key actors are normatively com-
mitted to democracy, a competitive regime can survive bad governing perform-
ance where it might not survive otherwise (Linz 1988; Linz and Stepan 1989;
Lipset 1959; O’Donnell 1986: 15–18; Remmer 1996). Actors with a normative
preference for democracy are not willing to subvert democracy to pursue radical
policies. And – going back to our previous argument – if radical policies are not
on the agenda, it is easier for all actors to accept a competitive regime.
Conversely, actors that normatively prefer a dictatorship readily seize on
opportunities to delegitimize a competitive regime and bolster the authoritarian
coalition. In moments of poor economic performance or radicalism by opposing
forces, actors that are indifferent to democracy can easily be recruited to join the
authoritarian coalition if it is already a force to reckon with (Lipset 1959).
A normative preference for democracy by the main opposition parties and
leaders also signals to leaders of an authoritarian regime and their allies that the
costs of establishing a competitive regime are likely to be bearable. It can help
pave the way for a transition to a competitive regime by assuring the actors that
support the authoritarian coalition that their interests are not likely to be
radically threatened under a competitive regime.
These arguments rest on the assumption that actors’ attitudes toward political
regimes significantly influence political outcomes. Actors’ values about what
political regimes are desirable and feasible affect how they behave politically and
how tolerant they are of policy failures, of dissent on the part of other actors with
strongly opposing preferences, and of political unrest. Normative preferences
create a cognitive map that shapes how actors understand political reality and
their own interests (Blyth 2002; Finnemore 1998).
Most political regimes hit periods of bad government performance. Actors
that are normatively committed to a given regime type accept periods of bad

17
Frey et al. (2004: 381) define procedural utility as “the well-being people gain from living and
acting under institutionalized processes as they contribute to a positive sense of self.”
Introduction 17

performance and blame the administration rather than the regime. In contrast,
actors that are normatively indifferent or hostile to that regime might seize on the
difficult period to attack the regime and join the opposition regime coalition (not
merely the opposition to the government).
Actors’ normative attitudes about democracy and dictatorship are not reduci-
ble to their economic interests or to cultural predispositions. These attitudes,
however, are not perfectly exogenous, a prime mover of political processes. In
order to avoid tautology, an explanation of regime outcomes based on norma-
tive preferences must be willing to inquire into the origins of attitudes toward
democracy and dictatorship, their variance across countries, and their trans-
formation over time. We address this issue in Chapters 2 through 7.
Like Dahl (1971: 124–88) and most authors who have contributed to this
literature, we focus on powerful actors because their beliefs have a more direct
impact on regime outcomes than mass beliefs. We focus exclusively on actors’
value preferences about democracy and dictatorship as opposed to other social
or cultural beliefs. Other scholars have argued that nonpolitical beliefs such as
trust in individuals (Inglehart 1990, 1997) and religious beliefs (Huntington
1984, 1991; Levine 1992; Stepan 2001: 213–53) affect political regimes. These
other beliefs have effects on democracy and authoritarianism, but they are not a
central part of this book.
International Actors and Influences. International actors disseminate new
beliefs about the desirability (or lack thereof) of different kinds of political
regimes and policies, and they prove by example that some political projects
are feasible (or not). They provide resources to empower some domestic regime
coalitions, and they offer incentives to domestic actors, thereby altering the costs
and benefits of different options in the domestic regime game. Where the
regional political environment and the U.S. government are favorable to com-
petitive political regimes, the costs and benefits of the regime game shift for
domestic political actors, creating stronger incentives for transitions to compet-
itive regimes. Where the United States and the Organization of American States
(OAS) adamantly oppose the breakdown of competitive regimes, potential coup
leaders and their supporters face higher costs.
International actors exercise indirect as well as direct effects on regime
change. For example, external influences may affect domestic actors’ radical-
ization and commitment to democracy, which in turn affect regime outcomes.18
International actors also influence domestic actors’ calculations about their
policy benefits under different regimes. For example, if international actors
threaten to impose sanctions against dictatorships, most domestic actors will
typically lower their expectations regarding their policy benefits under author-
itarian rule.

18
There is a related body of work on the impact of the international diffusion of ideas on social
policy. See Meseguer (2002) and Weyland (2006).
18 Democracies and Dictatorships in Latin America

In addition to operating indirectly, international actors sometimes have direct

impacts on political regimes. For example, U.S. or OAS military actions led to
transitions to competitive regimes in Panama in 1990 and Haiti in 1995 and
2006. The U.S. invasion of the Dominican Republic in 1965 helped maintain an
authoritarian regime in power. On several occasions, including Honduras in
1983, Bolivia in 1984, and Peru in 1989, the United States lobbied against
military coups and might have thereby directly influenced regime outcomes.
We summarize these five core empirical arguments as follows:
1. Policy radicalization makes a breakdown of a competitive regime more
likely.
2. A normative preference for democracy by important actors (e.g., parties,
leaders, the government) makes a transition to a competitive regime more
likely.
3. A normative preference for democracy by important actors makes a break-
down of a competitive regime less likely.
4. A regional political environment favorable to democracy makes transitions
to competitive regimes more likely.
5. A regional political environment favorable to democracy makes break-
downs of competitive regimes less likely.
None of these empirical arguments is surprising or counterintuitive. The origi-
nality of our work rests in an attempt to integrate these arguments through an
actor-based theory on regime change and stability and on how we develop and
test the theory.
Like all theories about highly complex political realities, ours simplifies
reality. Its purpose is not to capture all the complexities of regime change and
survival, but rather to call attention to a few highly important issues within an
integrated theoretical framework.

testing the theory

Most work on political regimes has chosen between extensive and intensive
testing. We undertake both kinds of testing because both give us different
kinds of leverage for understanding the emergence, stabilization, and fall of
democracies and dictatorships.
We followed two overarching principles about testing the theory. First,
quantitative evidence across a broad range of cases should support the theory.
When it is possible to measure theoretically important independent and depend-
ent variables in a reasonably efficient and valid manner, quantitative analysis is a
useful beginning point to assess the causal impact of the independent variables.
Otherwise, there is no good way of knowing how extensively a theory travels. In
addition, the quantitative analysis tests a wider range of alternative explanations
more rigorously for a broader range of countries than our qualitative evidence.
Introduction 19

The quantitative analysis in Chapter 4 provides this extensive test of our

theory. It tests whether our theory holds up for a large number of observations
(twenty Latin American countries for 1945–2005, for a total of 1,220 country-
years). In the quantitative analysis, the dependent variable is whether a regime
breaks down or survives in a given year. Over the course of a decade of research,
we collected information on a wide variety of independent variables to test our
theory and several competing theoretical approaches to regime change and
survival. The dataset contains several original variables (including our regime
classification, a novel indicator of U.S. policies toward Latin America, and new
indicators of actors’ radicalism and normative regime preferences) with varying
time coverage beginning in 1900 and ending in 2010. Because of data limita-
tions, our quantitative explanation focuses exclusively on the period since 1945.
The quantitative testing is indispensable for seeing how far in space and time a
theory travels – that is, for assessing its generality and its scope conditions
(Goldthorpe 1991; King et al. 1994).
Our second principle for testing is that structured case studies must fit the
theory. Theory building is facilitated by detailed case knowledge (Capoccia and
Ziblatt 2010). With large macro processes such as the rise and fall of political
regimes, it is not sufficient that quantitative evidence line up behind a hypothesis
or a theory. With such processes, several competing accounts could explain the
same quantitative findings. Theory that is not informed by the reality of cases is
therefore more prone to misunderstand large macro causal processes. The
combination of quantitative and qualitative analysis is far better than either
alone.
In Chapters 5 and 6 we employ qualitative case studies of Argentina and El
Salvador to provide intensive testing of our theory. Structured case analyses are
an essential part of our testing process for five reasons. First, because structured
qualitative case analysis allows for attention to sequences, it is useful for assess-
ing causal mechanisms. Sequences and actors’ interactions can help disentangle
mechanisms that are not clear on the basis of regression analyses. We can
analyze what precipitated regime change or “reequilibration” (Linz 1978b) at
crucial historical moments.
Second, the structured case studies enable us to examine interactions among
actors. Such interactions are decisive in regime outcomes. Although the quanti-
tative analysis in Chapter 4 provides an essential test of important parts of the
theory and of competing explanations, it does not test hypotheses about inter-
actions among actors. For example, in Argentina, from 1930 until 1976, the lack
of a normative preference for democracy was mutually reinforcing among
actors. President Juan Perón’s (1946–55) authoritarian proclivities and radical
tendencies generated deep hostility and reinforced radicalism in much of the
anti-Peronist camp from 1946 to 1966. Likewise, after 1983, the building of a
normative preference for democracy was mutually reinforcing among actors. It
is very difficult to capture such interactions in a quantitative analysis involving
twenty countries over a long period of time.
20 Democracies and Dictatorships in Latin America

Third, part of our theory addresses the formation and dissolution of regime
coalitions. The structured qualitative cases help illuminate and test this part of
our theory. The case studies revolve centrally around the formation of winning
regime coalitions and the stability or lack thereof that results from those coali-
tions. This key part of the theory is difficult to test quantitatively.
Fourth, the variables for actors’ normative regime preferences present challeng-
ing problems of endogeneity. Do actors’ normative preferences cause regime
change, or does regime change causes actors’ normative preferences? These prob-
lems are both statistical and substantive. We address the econometric problems in
Chapters 3 and 4, and the structured qualitative cases in Chapters 5 and 6 also
help untangle these problems of endogeneity. They also illustrate more clearly
than the quantitative analysis why normative preferences for democracy or dicta-
torship are important in understanding regime change and stability.
Fifth, the structured case studies allow us to scale down to the level of political
actors in each historical period. We can then study actors’ attitudes toward
democracy and dictatorship and their radicalism or moderation in more detail.
These issues create questions of internal validity for which a case study can be
particularly enlightening (Gerring 2007: 43–48). The case studies also enable us
to explore the actors’ reasons for a low normative preference for democracy and
radicalization. Such information allows us to reconstruct historical causal
sequences that lead to regime breakdown or stability.
We draw on the rich tradition of qualitative research that has enriched the
analysis of why democracies emerge (R. Collier 1999; Huntington 1991; Levine
1973, 1978, 1989; O’Donnell, Schmitter, and Whitehead 1986; Rueschemeyer,
Stephens, and Stephens 1992; J. S. Valenzuela 1985; Yashar 1997), consolidate
or fail to (Linz and Stepan 1996), and stabilize or break down (Capoccia 2005;
D. Collier 1979; Figueiredo 1993; Linz and Stepan 1978; O’Donnell 1973;
Potter 1981; Santos 1986). We part paths from most of this tradition by (1) try-
ing to be more systematic in coding actors and our core independent variables;
(2) working with a larger number of country cases (twenty) than most qualitative
studies; and (3) using quantitative analysis to test the extension of our theory
beyond the qualitative cases.
The number of countries that we study – the twenty countries of Latin
America – occupies an uncommon intermediate niche in regime studies.
A majority of the work on political regimes involves a small number of countries,
most often one or two, and most of the rest is quantitative work based on a larger
number of countries. One of the least developed strategies in studies on political
regimes is the intermediate-N strategy (in terms of the number of countries) that
we pursue. Region-wide studies of democratization that are sensitive to intra-
regional differences are uncommon (for an exception, see Bratton and van de
Walle 1997).19 Both the intermediate-N strategy and the regional research

19
Many works focus on differences across a few cases in a given region, but few simultaneously take
a region as a whole and evince a strong interest in intra-regional differences.
Introduction 21

design, which in principle are discrete but in our case are combined, are useful
compliments to the large-N and small-N studies that dominate regime studies.
This intermediate niche has distinctive advantages. The much larger number
of countries and observations than single-country case studies enables us to test
hypotheses in a more systematic and extensive manner than a single country or a
few countries would allow. The twenty countries display considerable variance
in regime types across countries and over time, and offer a broad range of
conditions in terms of the independent (and control) variables for this study.
At the same time, the number of countries is sufficiently small that we know a
reasonable amount about regime dynamics in a majority of them. This knowl-
edge helps generate hypotheses and informs the understanding of causal mech-
anisms. The mixed quantitative/qualitative, intermediate-N strategy pursued
here is not superior to other alternatives, but it is an underutilized strategy that
yields distinctive benefits. We try to bridge the gap between qualitative area
studies and large-N research through close knowledge of some cases for inten-
sive testing and a more extensive test of hypotheses provided by a quantitative
design.

case selection for qualitative cases

In this section, we first discuss why we chose two countries for our qualitative
cases as opposed to looking at a similar number of transitions, breakdowns, and
regime survivals in a larger number of countries. We then explain the logic for
choosing Argentina and El Salvador.
The need to examine actors’ interactions and use structured case studies to
understand sequences and causal mechanisms precluded a qualitative analysis of
more than a small number of cases and dictated a strategy of treating these cases
in enough detail to support our primary claims. In light of these considerations,
we focus on two country studies over time rather than selecting breakdowns and
transitions from a larger number of countries.
The logic of our qualitative analysis of Argentina and El Salvador rests
primarily on understanding interactions among actors, processes, and sequences
to understand regime outcomes. Because within-country observations allow for
close examination of processes, interactions, and sequences, it is generally easier
to identify causal mechanisms than in cross-country comparisons. Within-
country observations have far less variance in most control variables than
observations across countries and thus help clarify which independent variables
account for the change in the dependent variable. Finally, given the extensive
historiography, focusing on two countries allowed for greater coverage of the
secondary literature and for better case knowledge than would have been
possible had we chosen the same number of breakdowns, transitions, and
stabilizations but with a larger number of countries.
Within-country observations are ideal for process tracing – for close attention
to sequences and causal mechanisms (D. Collier 1993; Collier, Brady, and
22 Democracies and Dictatorships in Latin America

Seawright 2004: 250–64; Collier, Mahoney, and Seawright 2004; George and
Bennett 2004: 204–32; Mahoney 2003: 360–67). Within-country analysis
reduces the number of explanatory variables because many change slowly and
hence do not explain short-term variations in the dependent variable. We
increase the number of observations by looking at multiple administrations
within each country. This combination of a smaller number of explanatory
variables and multiple within-country observations ameliorates the well-
known concern about the indeterminate research design in many small-N stud-
ies: many variables, few cases (Lijphart 1971: 685–91). In within-country
qualitative analysis, the logic of causal inference is not reducible to a cross-
country comparative method based on a small number of observations – a
method that is vulnerable to deep weaknesses in causal logic. Unless it is
accompanied by within-country process tracing, it is difficult in small-N cross
national comparison to weigh competing explanations (Collier, Brady, and
Seawright 2004; George and Bennett 2004: 153–66; Goldthorpe 1991; King
et al. 1994: 199–207).
In the post-1977 wave of democratization in Latin America, there have been
two dramatic changes relative to earlier periods. First, many countries that
earlier went through cycles of democratic breakdowns and transitions back to
competitive political regimes become stable democracies. Eight countries in the
region had at least three breakdowns since 1900: Peru (with seven transitions
and six breakdowns), Argentina (six transitions and five breakdowns), Panama
(five transitions and five breakdowns), Ecuador, Honduras (five transitions
and four breakdowns each), Uruguay, Costa Rica (four transitions and three
breakdowns each), and Chile (three transitions and three breakdowns).
Notwithstanding the breakdowns in Peru in 1992 and Honduras in 2009, as a
group, these countries have been vastly less prone to breakdowns of competitive
regimes since 1978 than they were before then.
Second, eight countries have shifted from deep authoritarian pasts, with little
(and short-lived) or no prior experience with competitive regimes, to having
stable competitive regimes in the post-1977 period. This includes Bolivia, whose
experience of competitive regimes before 1978 was limited to the 1956–64
period; the Dominican Republic, which was semi-democratic from 1924 to
192820; El Salvador, which had no experience of a competitive regime until
1984; Guatemala, which was semi-democratic from 1926 until 1931 and from
1945 to 1954; Haiti, which never had a competitive political regime until the one
that broke down after a few months in 1991; Mexico, which was semi-
democratic from 1911 to 1913 but otherwise authoritarian until 1988;
Nicaragua, which was semi-democratic from 1929 to 1936 but then had author-
itarian regimes until 1984; and Paraguay, which had dictatorships steadily until

20
The Dominican Republic also had a very short-lived competitive regime for seven months from
February to September 1963, but it did not reach our threshold of surviving until December 31 of
the year in which it was inaugurated.
Introduction 23

1989. Except for Haiti, these countries have gone from largely unchecked and
often brutal histories of dictatorship before the third wave of democratization to
competitive regimes after the transitions.
The story of the third wave is largely the story of these two sets of countries.21
Accordingly, we chose two countries that together exemplify the most common
regime patterns in twentieth-century Latin America: one (Argentina) that had
many breakdowns before the third wave and has been steadily democratic
during the third wave, and one (El Salvador) that has shifted from persistent
authoritarianism before the third wave to a durable competitive regime. Sixteen
of the twenty countries in Latin America squarely fit one of these two patterns.
Argentina had experienced chronic instability of both competitive and
authoritarian regimes between 1930 and 1983, including five breakdowns of
competitive regimes during this period. We address two questions. First, why did
competitive regimes consistently break down before 1983 despite many favor-
able social and economic conditions? Second, what explains the dramatic
change from the chronic breakdown of competitive regimes until 1976 to
democratic survival in the period since 1983?
Chapter 6 focuses on El Salvador and asks the opposite questions. What
explains persistent authoritarianism for almost the entire twentieth century until
1984? How did a country with a history of consistent and often brutal authori-
tarianism overcome daunting obstacles and experience a transition to a com-
petitive political regime? Why did this regime fend off threats and become
stable? Whereas Chapter 5 explains repeated breakdowns in Argentina during
much of the twentieth century and the absence of breakdowns after 1983,
Chapter 6 explains the absence of transitions in El Salvador during most of the
twentieth century and the occurrence of a transition after 1984.
Although we present detailed qualitative evidence about only two country
cases, our analysis was informed by reading about a much larger number of
countries and by doing some fieldwork at some point in our careers in twelve
countries in the region: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica,
Ecuador, Mexico, Paraguay, Peru, Uruguay, and Venezuela. This fieldwork
enhanced our understanding of these national realities.

latin america and theory development

If we cast our argument as a somewhat general theory of regime change and
survival, why should we focus on a single region of the world? We have two
theoretical and one pragmatic reason for following this strategy.

21
The remaining four countries are Brazil, Colombia, Cuba, and Venezuela. Brazil and Colombia
had only one breakdown, so they did not follow the more common pattern of multiple break-
downs. Cuba and Venezuela are exceptions because as of this writing they have authoritarian
regimes.
24 Democracies and Dictatorships in Latin America

First, as we argued elsewhere, regions have particular dynamics and polit-

ical processes that are specific to those regions (Mainwaring and Pérez-Liñán
2007). Social science generalizations that are based on large-N, cross-regional,
or worldwide units of analysis must be attentive to these regional specificities
(Bunce 1995, 1998, 2000). Otherwise, social scientists will generalize where
they should not. Causal inferences based on a worldwide sample could lead to
a misleading understanding of what factors promote democratization in some
regions. Different regions may present distinctive and systematic causal pat-
terns that an assumption of worldwide causal homogeneity would obscure.
The effect is more substantial and hence the need for caution is greater when
entire regions of the world rather than simply a few countries are exceptions to
a generalization.
Consider the finding in Chapter 4 that the level of development does not affect
the probability of transitions to or breakdowns of competitive regimes. The fact
that modernization theory does not hold for a wide income range in Latin
America between 1945 and 2005 is important, and it suggests a likely pattern
of causal heterogeneity by region. Even though on average, wealthier countries
are more likely to transition to democracy and less likely to establish dictator-
ships (Przeworski et al. 2000), the causal effect of economic development may
differ across regions in the same broad band of levels of development. Particular
causal factors may have heterogeneous effects in different regions of the world.
Therefore, an analysis that overlooks regional patterns may impose a misleading
assumption of causal homogeneity (Mainwaring and Pérez-Liñán 2003; 2007).
A conventional response to this argument is that regions represent “proper
names” that should be replaced by “variable names” in the analysis. In principle
we agree, but until all the variables that define regional patterns in world politics
are thoroughly identified (which is an extraordinarily difficult task), an assump-
tion of causal homogeneity across regions may induce greater bias in the results
of an empirical analysis than the assumption of causal heterogeneity at the
regional level.
Second, as we emphasize throughout this book, political developments in one
country influence regimes in other countries of the same region. Regions are
more than labels for arbitrary sets of countries; they identify geographic net-
works defined by spatial and cultural proximity. In Chapters 4 and 7, we show
that it is impossible to understand regime outcomes without emphasizing region-
wide factors. Analyses that fail to consider regional influences would overstate
the importance of domestic factors, conclude that regime change and survival
are highly idiosyncratic processes, or perhaps commit both mistakes.
Regime change has occurred in region-wide waves: a first wave of democra-
tization from 1902 to 1911; a second wave from 1938 to 1946, a counterwave
from 1948 to 1955; a third wave of democratization from 1956 to 1958, another
counterwave from 1962 to 1977; and finally, the post-1977 wave of democra-
tization. In Chapters 4 and 7, we show that region-wide influences account for this
wave-like behavior. To explain the vicissitudes of democracy and authoritarian
Introduction 25

regimes, idiosyncratic factors come into play in every country, but there never-
theless have been distinctive region-wide trends, including the post-1977 trend
toward democracy. To understand political regimes, we therefore must examine
both region-wide trends and explanations and country-specific processes.
It is impossible to understand regime outcomes by focusing only on individual
countries or only on global trends. Political regimes were traditionally a subject
matter for comparative political scientists who focused on domestic processes,
but regime dynamics are not exclusively domestically driven. Both because of
regional specificities and because of distinctive intra-regional influences, social
scientists and historians must be attentive to the importance of regions in
politics. International influences on political regimes are especially important
within regions (Gleditsch 2002). If we always treat countries as the unit of
analysis and fail to pay attention to regional effects and dynamics, we will
miss these regional effects and as a result will fail to understand causal processes.
While advocating the importance of regions in comparative politics, we reject
the assumption that Latin America is relatively homogeneous in a descriptive
sense (i.e., that variance in fundamental conditions across countries in the region
is small), and we reject gross generalizations about regions as a whole unless
there is empirical evidence to support them.22 Our approach looks at regional
influences, but it treats the countries within the region as distinct. In Chapter 4,
we treat each country differently by virtue of assigning each one a different score
for most independent variables and for the dependent variable for a given year.
We believe that this is the way that regions of the world should be studied. Latin
America has important common trends and influences, but it also has huge
cross-country differences in everything from political regimes to the level of
development. For example, in 2005, Argentina had a per capita GDP of
$5,721 in 2000 dollars, more than fifteen times greater than Haiti’s ($379),
which was one of the lowest in the world outside of sub-Saharan Africa (World
Bank 2007). Similarly, seven countries (Argentina, Brazil, Chile, Colombia,
Costa Rica, Uruguay, and Venezuela) had lengthy experiences of democracy
before 1978 while a handful of others had histories of continuous or nearly
continuous dictatorships late into the twentieth century (El Salvador, Haiti,
Mexico, Nicaragua, and Paraguay). Our research design is predicated on rec-
ognizing these differences across cases and within cases over time.
Our empirical focus on one region does not entail a position against broader
generalizations in social science research. We adopt an intermediate position:
generalizations are important, but there are few truly universal findings in
analyses of political regimes.23 Most generalizations in social science are

22
Broad generalizations about Latin America as a whole characterize some works that emphasize
Iberian political culture.
23
Universal findings are expected to hold for most representative samples of the same population,
but the definition of the population is itself an analytical task (Ragin 2000: 43–63). For instance,
“universal” may simply refer to all U.S. voters in the second half of the twentieth century.
26 Democracies and Dictatorships in Latin America

bounded by geographic or historical contexts.24 Regional specificities are not the

only way to bound generalizations in social science, but because regions are large
parts of the world with distinctive dynamics and intra-regional influences,
delimiting some analyses and generalizations by regions is useful. We do not
claim that regions should be the primary unit of analysis in comparative politics
or that analysis of regions is superior to other research designs. But regions are
substantively important, and the reasons for this importance have been under-
articulated in political science. For developing and testing theories about regime
change, it is substantively useful to examine regions.
We also have a pragmatic reason to focus on Latin America. Even though our
theory of regime change should travel beyond Latin America, a focus on one
region allows for testing hypotheses using better-quality data for a longer
historical period, without assuming that some indicators (e.g., U.S. policy
toward democracy) would have an equivalent effect in other regions of the
world. We coded political regimes in the twenty Latin American countries
between 1900 and 2010, and also identified and coded the normative and policy
preferences of 1,460 political actors throughout the region from 1944 to 2010.
Much of the critical information collected for this project involved labor-
intensive coding of political regimes and actors. Nothing remotely similar to
this coding of actors is available for other regions. The use of more conventional,
readily available indicators would have allowed us to expand the geographic
scope of our tests, but would have undermined the validity of the indicators and
thus the interpretation of the results.
Consistent with a perspective that emphasizes regional influences and dynam-
ics while underscoring the specificity of individual countries, we deal with two
different levels of analysis: countries and Latin America as a region. Our primary
analysis of the rise, survival, and fall of political regimes takes place at the
country level. However, region-wide actors and influences affect country level
actors, processes, and regime outcomes. At the country level of analysis, our
theoretical puzzle is to explain the rise, survival, and fall of regimes. At the
regional level, it is to explain waves of democratization and authoritarianism.
The regional trend is the mere aggregation of country outcomes, but country
patterns in turn are influenced by what takes place in the region.
Analysts have used a variety of different theoretical approaches and inde-
pendent variables to explain why democracy exists in some countries but not
others. Many factors affect the likelihood that democracy will exist. One final
advantage of focusing on Latin America is that it holds constant a few such
factors: predominant religious preference, presidential systems, and Iberian
colonial experience (except for Haiti). These commonalities reduce the number
of independent variables and thus facilitate the explanatory process.

24
For an excellent example of how presumably universal findings may be historically bounded, see
Boix and Stokes (2003) on the historically changing relationship between the level of development
and democracy.
Introduction 27

plan for the book

Chapter 2 outlines our theory of regime change and durability in more detail.
Chapter 3 discusses waves of regime change in Latin America from 1900 to
2010. It describes the evolution of our dependent variable over time, focusing on
periods of expansion and contraction of democracy. This chapter also addresses
the measurement of our main independent variables. We introduce novel indi-
cators of normative regime preferences, radicalism, and international condi-
tions, and discuss the historical evolution of those factors for our sample of
twenty countries. In the last part of the chapter we treat actors’ normative
preferences as an endogenous explanatory variable, showing that dominant
preferences may be influenced by incumbent regimes, but they are not a mere
reflection of structural conditions.
Chapters 4, 5, and 6 test the theory. Chapter 4 presents a quantitative analysis
of the twenty Latin American countries for the 1945–2005 period. A set of
survival models allow us to reconstruct the probability of transitions and break-
downs in particular countries and years, and also the overall wave of democra-
tization experienced by the region after 1977.
Chapters 5 and 6 present intensive tests of the theory through qualitative case
studies. Chapter 5 examines Argentina, which had chronic breakdowns of
competitive regimes before 1978 despite many favorable circumstances and
has enjoyed a democracy without breakdown since 1983 despite many unfav-
orable circumstances. Decreased radicalism, an increase in commitment to
democracy, and a more favorable international environment have been crucial
in Argentina’s post-1983 political transformation. In Chapter 6, we trace the
reasons for the breakdown of the traditional very powerful authoritarian coali-
tion and the emergence of a democratic coalition in El Salvador in the 1980s and
early 1990s.
Chapter 7 further explores the mechanisms behind our finding that interna-
tional actors and influences are an important explanation of regime outcomes.
This finding has become common in regime studies since 1986, but the mecha-
nisms behind it are not clear in the existing literature. We discuss six mechanisms
that help explain the impact of international actors on regime outcomes: (1) the
preferences of actors regarding political regimes and policy diffuse across coun-
try borders to domestic actors, generating an indirect mechanism of influence on
political regimes; (2) domestic actors in one country draw inspiration from
events in another country (demonstration effects); (3) international actors
sway domestic actors to join a regime coalition; (4) international actors provide
resources to strengthen some actors; (5) international actors such as the Catholic
Church simultaneously function as domestic actors, and as domestic actors they
influence regime outcomes; and (6) international military interventions. The
combination of quantitative testing in Chapter 4 and qualitative analysis in
Chapters 5, 6, and 7 enable us to add to the existing literature on regional
influences on political regimes.
28 Democracies and Dictatorships in Latin America

Chapter 8 explores the implications of our theoretical conclusions for the

current (and future) direction of Latin American regimes. It describes regional
tendencies in the evolution of political regimes after the third-wave transitions.
To an unprecedented degree, competitive regimes have survived during this time.
However, an analysis of the levels of democracy achieved by those competitive
regimes after 1978 indicates, alongside many striking advances, two different
problems in the region: democratic stagnation and democratic erosion. Some
countries had relatively low levels of democracy after their transitions to com-
petitive politics, and they have been unable to improve significantly over the past
three decades. Other countries are experiencing an erosion of political rights and
civil liberties.
Current democratic stagnation is partially anchored in historical legacies. In
countries where political actors lacked a normative preference for democracy
before 1978, they failed to invest in the development of institutions (competitive
parties, independent courts, and civic-minded security forces) important for
building high-quality democracy. By contrast, democratic erosion is related to
more recent trends. Governments in Venezuela (since 1999), Bolivia (2006),
Ecuador (2007), and Nicaragua (2007) have revitalized the somewhat radical
forces in the region, and they have fostered intransigence in some sectors of the
right. Regional political influences are more supportive of leftist radicalism that
does not embrace (and often even opposes) liberal democracy. During most of
the period since 2001, the United States emphasized antiterrorism more than
human rights as the focus of its foreign policy, weakening the credibility of
democracy promotion efforts. Because of the opposition of some leftist presi-
dents with dubious democratic credentials to OAS interventions, the OAS
cannot act as coherently as it did during the 1990s on behalf of democracy.
We test those arguments using a latent growth curve model for levels of democ-
ratization in nineteen post-transition countries during the contemporary period.
Chapter 9 discusses the implications of our findings for alternative theoretical
approaches to explaining regime change and survival. The Latin American
experience creates doubts about some prominent existing theoretical approaches
to political regimes: modernization theory, class theory, theories based on
economic performance, and theories based on political culture. The evidence
presented in this book suggests that some prominent theories of political regimes
are not convincing. We argue that a theory can help integrate some of the most
important lessons about the emergence and fall of political regimes in ways that
are consistent with the historical evidence about Latin America.
9

Rethinking Theories of Democratization in Latin

America and Beyond

We began this book because we wanted to understand regime survival and fall
in twentieth-century Latin America. As we studied these issues, we developed
doubts about many theoretical approaches to understanding political regimes.
It became essential to engage in a broader effort to theorize about the rise and fall
of democracies and dictatorships.
Therefore, we developed a theory to explain the survival or fall of democracies
and dictatorships. Starting from assumptions about how actors are constituted
and what motivates them to join regime coalitions, we deductively derived five
hypotheses about regime survival or fall. We particularly drew on three literatures:
(1) transitions, breakdowns, and the survival of political regimes; (2) international
factors in regime change and survival; and (3) the impact of ideas and beliefs
on political outcomes. But we go beyond most of the existing work in these
literatures by articulating an integrated theory and testing it in new ways. We
believe that this theory is more realistic than competing theories; that there are
benefits to systematizing it as a theory; and that it explains regime change and
survival in twentieth-century Latin America better than alternative theoretical
explanations.
This chapter undertakes three main tasks. First, we summarize our theoretical
arguments and contributions. Our theory is based on more realistic microfoun-
dations than most alternatives, and it has stronger empirical support. In addition,
we devised an original research strategy to test hypotheses about actors across
a much broader range of countries and time than previous actor-based theories.
We also articulate our contributions to the literatures on actors’ normative regime
preferences, their policy radicalism or moderation, and international influences
on regime outcomes.
Second, we briefly argue that the theory could fruitfully be extended beyond
Latin America. As examples of this potential, we claim that prominent analyses
of the breakdowns of democracy in Spain (1936) and Germany (1933) and of the
transition to democracy in Spain (1977) are fully consistent with our approach.

269
270 Democracies and Dictatorships in Latin America

We then use the Latin American experience and some broader evidence to
reflect on the theoretical approaches commonly employed to understand the
emergence and fall of democracies and dictatorships. We argue that the Latin
American experience in the twentieth century is not consistent with moderniza-
tion theory, class theories, works based on economic performance, mass poli-
tical culture approaches, works based on formal institutions, and theories that
strongly emphasize leadership and agency. We do not question all the results
that have stemmed from these theoretical approaches, but the evidence in this
book suggests modifications, boundaries, and nuances to these theories.

rethinking regime survival and change

We offer an alternative theoretical framework with more realistic microfounda-
tions than existing theories. In agreement with Coppedge (2012) and Munck
(2001), we believed that it could be fruitful to articulate a theory that makes explicit
how we conceptualize the actors and what drives their behavior. Articulating how
to conceptualize the actors, making explicit assumptions about what motivates
them, and linking different levels of analysis (actors and coalitions at the country
level, as well as international forces) integrates different elements in potentially
useful ways. Hypotheses that could otherwise seem ad hoc instead are grounded in
the theory.
Our theory integrates the analysis of actors’ normative preferences about
political regimes and policy moderation or radicalism in an explicitly articulated
way rather than just presenting discrete hypotheses. Our primary contribution is
not the discrete hypotheses, but rather the theory and the testing of it.
Likewise, we followed many works that established the empirical impor-
tance of international factors in regime change and survival, but with few
exceptions, this work did not integrate the findings about international factors
into a theory of regime change and survival. In order to advance understanding
of how international forces and influences in democratization function, we
concluded that it would be useful to connect domestic and international factors
into a theory.
A second contribution of this book is our effort to test an actor-based theory
of political regimes across a wide time period and over a sizable number of
countries (twenty), using both quantitative and qualitative tests. Actors make
history, although structures, cultures, and formal rules of the game condition
those actors. Many authors have enriched the understanding of democratic
transitions and breakdowns through case studies, usually using one or a small
number of countries or through inductive generalizations such as those pre-
sented by Linz (1978b) and O’Donnell and Schmitter (1986). The work in this
qualitative tradition has bumped up against some limits. How can we go beyond
case studies and examine how far in time and space these insights travel? How
can we aggregate and test the knowledge built through these case studies? How
can we test hypotheses about actors across a broader range of cases?
Rethinking Theories of Democratization 271

To build on insights of previous studies and to test them, we integrated some

hypotheses through a theory and then coded actors in the twenty Latin American
countries for each presidential administration from 1944 to 2010. This endeavor
required developing clear and explicit coding rules, ensuring consistency across
different coders, and undertaking substantial historical research on political
regimes and actors in the twenty Latin American countries. In order to test
hypotheses about the impact of actors’ preferences on regime outcomes, it
was essential to actually look at actors’ preferences rather than using inadequate
aggregate proxies. The coding of actors enabled us to test hypotheses about
radicalism and actors’ normative preferences with greater validity, and much
greater extension across time and space, than we otherwise would have been
able to. This kind of historical qualitative coding grounded in explicitly articu-
lated coding rules could be useful for other research projects.

International Actors and Regime Outcomes

Prior to 1986, theoretical works on regime change that emphasized the role of
political factors focused largely, and usually exclusively, on domestic processes.
Research beginning with Stepan (1986) and Whitehead (1986b) established that
this traditional approach had neglected international influences. Over the last
twenty-five years, a substantial body of literature has emphasized international
influences in regime outcomes.
Our book adds to the literature that has shown that it is essential to examine
regional political and ideological trends, the policy of hegemons, and the inter-
connection between domestic and transnational actors. Transnational trends
and actors profoundly influence domestic regime outcomes. It is impossible to
understand regime dynamics exclusively in terms of the cumulative effect of
isolated political processes in individual countries. What happens in one country
affects others. Moreover, developments among transnational and international
actors affect political regimes in multiple countries.
We drew on many insights from this literature, but only a few works in this
field are richly theoretical, and none integrated their emphasis on international
factors into a theory of regime change and survival.1 Our treatment of this issue
added to the existing literature on international influences in democratization in
four ways. First, we incorporated international influences into a theory of regime
change that links domestic and international factors. Second, we quantitatively
disentangled different international effects (U.S. policy toward authoritarian
and democratic regimes in Latin America, the regional political environment,
and the average world level of democracy) more than most previous work has.

1
The excellent work of Levitsky and Way (2010) is a partial exception, but their dependent variable
is different from ours. They analyzed whether competitive authoritarian regimes that existed in the
early 1990s became democratic or remained authoritarian subsequently.
272 Democracies and Dictatorships in Latin America

Third, our analysis in Chapters 4 and 7 documented the dynamic consequences

of international influences. International effects not only operate across coun-
tries; they also sustain democratization trends over time. Regional influences
affect domestic actors in individual countries, and those actors in turn refract
those influences back into the regional arena. Waves of democratization as well
as counterwaves of authoritarianism are hard to explain without understanding
such dynamic effects.
Fourth, this literature has not adequately theorized the mechanisms by which
international actors influence regime outcomes. International actors may influ-
ence some regime outcomes directly, but their effects are often mediated by other
variables. In Figures 2.1 and 7.3, international actors influence domestic actors’
attitudes toward democracy and dictatorship, their policy preferences, and their
political resources. These variables in turn shape domestic actors’ decisions about
which regime coalition to join and how many resources those coalitions have.
International actors can influence regime outcomes by (1) generating policy
preferences and attitudes about dictatorship and democracy that disseminate
across country borders; (2) creating demonstration effects; (3) swaying domestic
actors’ decisions to join one of the competing regime coalitions; (4) providing
resources to domestic actors and thereby influencing the power of the two
competing regime coalitions; (5) joining one of the competing regime coalitions
(e.g., the Catholic Church) and thereby swaying the regime outcome; and
(6) undertaking a military invasion that topples or preserves the political regime.
We also emphasized that there are limits to the explanatory power of interna-
tional variables. International actors usually exercise their influence indirectly,
by affecting domestic actors’ calculus of policy benefits under the competing
regime coalitions and their normative preferences about the political regime.
They explain change over time better than change across countries at a given
point in time, and in Latin America they have infrequently been the main cause
of a regime change. Moreover, international support does little to enhance the
quality of democracy in contexts where it is low. The international community
has devised mechanisms to deal with overt attempts to impose authoritarian rule,
but it is ill equipped to deal with more subtle or gradual authoritarian regressions.

Normative Preferences for Democracy and Dictatorship

Our emphasis on the central role of actors’ normative preference for democracy
or dictatorship in explaining regime outcomes theoretically resonates with
works by Berman (1998), Capoccia (2005); Dahl (1971: 124–88), Levine
(1973), Lijphart (1977), Linz (1978a, 1978b), and O’Donnell (1986: 15–18).2

2
Few of these scholars argued that attitudes toward democracy affect its odds of survival, but they
made related points. For example, Dahl (1971: 124–88) argued that activists’ beliefs influenced
regime outcomes. His discussion of beliefs included the legitimacy of polyarchy (pp. 129–40),
which coincides with our focus on normative commitment to democracy. Lipset (1959: 90) claimed
Rethinking Theories of Democratization 273

We added to this literature in four ways. First, we brought together two bodies
of literature that have been largely divorced from one another: work that
emphasizes the impact of actors’ beliefs on different political outcomes and the
scholarship on political regimes. Little of the expanding literature on actors’
beliefs focuses on regime outcomes (Berman 1998 is an exception), and little of
the work on political regimes emphasizes the importance of actors’ beliefs (Dahl
1971: 124–88, Linz 1978b, and Stepan 1971 are exceptions).
Second, we tested arguments about the impact of actors’ normative prefer-
ences on regime survival or fall in new ways. The coding of 1,460 actors across
290 presidential administrations in Latin America from 1944 to 2010 enabled
us to undertake a more extensive test of the impact of actors’ beliefs on regime
outcomes than any previous work. The qualitative analysis in Chapters 5 and 6
enabled us to look at causal mechanisms intensively.
Third, we confronted in new ways some challenges that causal claims about
normative preferences must address (see the discussion in Chapter 2). Other
scholars have devised strategies for assessing the causal impact of beliefs in
qualitative small-N studies. We add to this discussion by confronting these
challenges for an intermediate number of countries over a long period of time.
Our strategy includes devising careful coding rules to distinguish between
sincere and strategically stated preferences and to ensure a clear separation
between the independent and dependent variables; undertaking statistical tests
for reverse causality; looking at reverse causality and examining causal mecha-
nisms in the qualitative case studies; ensuring that normative preferences are
not reducible to structural or broader cultural variables; and verifying in the
qualitative case studies that actors’ regime choices cannot be readily explained
by their material gains.
Fourth, we added to the discussion of why actors’ normative preferences
sometimes change. Actors’ preferences are not static (Bowles 1998), but social
scientists have not often systematically addressed why they change.
We do not claim that democracy emerges or stabilizes because political actors
have the “right values.” Actors derive procedural utility from political regimes
(Frey et al. 2004), and they measure the performance of incumbent regimes
against their normative preferences. If actors are normatively committed to
democracy, they are willing to tolerate disappointing policy outcomes that
might tip uncommitted actors to join the authoritarian coalition. Actors that are
committed to democracy are less likely to understand policy failures as a regime
failure. Instead, they might accept policy failures as a consequence of negative
legacies inherited from a previous regime, of negative trends in a country’s terms
of trade, of a poor leader who can be replaced through the democratic process, of
a difficult time in the world economy, or of policies that are not tightly condi-
tioned by the political regime and therefore might not change even if the regime

that citizen beliefs in democratic legitimacy help protect the regime from the destabilizing con-
sequences of low effectiveness (i.e., poor performance).
274 Democracies and Dictatorships in Latin America

changed. Given this reasoning, a change of regime would not necessarily produce
better policy outcomes (Linz 1988; Remmer 1996). A normative preference to
democracy extends actors’ time horizons.
It is impossible to understand regime fall and survival in Latin America without
examining changing normative views about democracy and dictatorship. For
example, the Cuban revolution inspired a generation of revolutionary struggle
in the region based on the belief that socialist revolution was desirable and
possible, with negative consequences for democracy including a powerful coun-
terreaction from conservative forces, leading to many military coups. Similarly,
the embracing of liberal democracy as an ideal by actors across the political
spectrum in the 1980s and the 1990s facilitated the establishment and survival
of competitive regimes in bad economic times.

Radicalism and Moderation

Actors’ radicalism or moderation is another important determinant of regime
outcomes. Greater radicalism makes it more difficult to sustain a competitive
regime. This argument builds on but refines the insights of Bermeo (1990, 1997,
2003), Figueiredo (1993), Przeworski (1991), Sani and Sartori (1983), Santos
(1986), and Sartori (1976), among others. We modify this literature by how
we conceptualize and define the continuum from moderation to radicalism.
Whereas Sartori (1976) focused exclusively on ideological distance among
actors, we define radicalism as policy positions toward the left/right pole in
combination with urgency to achieve these positions in the short to medium
term where these positions do not represent the status quo or with an intransi-
gent defense of these positions where these positions represent the status quo.
Although we build on Sartori’s insights, his formulation overstates the destabi-
lizing effects of polarization on competitive regimes when leftist or rightist actors
perceive their projects as long term.
Sartori’s formulation also misses the deleterious impact of actors that are not
extreme in ideological terms, yet whose policy impatience coupled with policy
preferences toward the left or right of center makes them threatening to other
actors and contributes to regime breakdown. Chapter 5 discussed two some-
what radical actors par excellence: Juan Perón in the period from 1946 to 1970
and Argentina’s labor unions from 1955 through 1976. Although most unions
were not ideologically extremist, they combined policy preferences to the left
of center (statist, nationalistic, pro-union, etc.) with considerable policy impa-
tience, as manifested by factory takeovers, general strikes, and repeated willing-
ness to support military coups to achieve policy gains. Similarly, although Perón
rejected socialism, his confrontational discourse and behavior, his somewhat
left-of-center policies, and his willingness to run roughshod over the opposition
in order to achieve his policy goals made him a somewhat radical and deeply
polarizing figure.
Rethinking Theories of Democratization 275

The analysis of radicalism poses questions for bargaining models of policy

making and regime change. Conventional representations of impatience assume
positive payoffs and thus identify a first-mover advantage: impatient actors
are willing to accept a discounted offer today rather than wait until tomorrow
(Rubinstein 1982; Sutton 1986). Our historical analysis of radical actors in
Latin America identified a very different pattern of behavior that calls for a
different formal setup (Primo 2002). If the incumbent regime promises to deliver
the player’s ideal policy in the future but demands some policy loss or patience
today, a radical actor will endure large immediate loses and may prefer to
gamble on an alternative regime. In this sense, radical players display a behavior
inclined toward risk (Kahneman and Tversky 1979). This reverses the first-
mover advantage, as governments need to offer additional benefits in the short
run to appease radical players, who appear to be disloyal opponents exercising
blackmail (Linz 1978b; Sartori 1976).

can the theory travel beyond latin america?

For reasons outlined in Chapter 1, we focused on Latin America rather than a
broader set of countries. This raises a question: Can the theory travel beyond
Latin America? We cannot extend the theory empirically to other regions of the
world here except by way of brief illustration, but two observations are in order.
First, our theory is compatible with several extant theoretical traditions
in democratization studies. For example, it is compatible with Linz’s (1978b)
emphasis on orientations toward the democratic regime (loyal, semi-loyal, and
disloyal oppositions); with O’Donnell and Schmitter’s (1986) and Przeworski’s
(1991) delineation of actors into blocs depending on their orientation toward
the political regime; and with Berman’s (1998) argument that Social Democratic
parties’ programmatic beliefs strongly affected their behavior in interwar
Sweden and Germany, which in turn affected the survival or breakdown of
democracy in the 1930s. It is also compatible with the extensive literature that
has documented international influences in regime outcomes. These compatibi-
lities suggest that our theory can travel beyond Latin America.
In addition, analyses of many non–Latin American cases of breakdowns of
competitive regimes, transitions from authoritarian rule to competitive regimes,
stabilizations of authoritarian regimes, and democratic stabilizations are fully
consistent with our theory. We illustrate this point by briefly indicating the
strong compatibility of our theory with existing analyses of (1) the breakdown
of democracy in Spain in 1936; (2) the breakdown of democracy in Germany in
1933; and (3) the stabilization of democracy in Spain after 1978.
Without using exactly the same concepts as we do, many scholars have
argued that democracies broke down because actors had weak normative pref-
erences for democracy or had normative preferences for dictatorship, or because
radicalism made it impossible for some actors to be willing to abide by democ-
racy. Most interpretations of the breakdown of democracy in Spain in the 1930s
276 Democracies and Dictatorships in Latin America

mesh with our theoretical approach. Casanova (2010) and Linz (1978a: 144,
151, 169) emphasized the negative impact of international influences, especially
the rise of fascism in Germany and Italy, on Spanish democracy. No powerful
actors had a steadfast normative preference for democracy (Casanova 2010:
95, 111, 116, 122; Linz 1978a: 149, 160–68, 180–81; S. Payne 2006: 41–45,
346–47, 350–543). When this is the case, especially in a polarized high-stakes
environment, actors easily turn against democracy. Powerful radical actors from
the far left to the far right were willing to use violence to pursue political ends
(Casanova 2010; Linz 1978a: 145, 153–54, 157–58, 187–94; Malefakis 1996:
644–46; S. Payne 2006; Preston 2006: 53–64). They were decisive in the spiral
of violence, revenge, and hatred that led to the breakdown. Right-wing radica-
lism fueled left-wing radicalism, and vice versa. No actors were willing to make
significant policy sacrifices in order to save democracy. By the time Franco
launched his coup in July 1936, several powerful actors on the right had a
normative preference for dictatorship (Casanova 2010: 124, 137).
Many scholars have also analyzed the German breakdown of democracy in
1933 along the lines that are fully consistent with our theory. Some extremely
radical actors, no actors with solid normative preference for democracy, several
(including the Nazis, the Communists, and some traditional right-wing parties)
with a normative preference for dictatorship, and an inhospitable international
political environment – all in the context of a severe economic crisis – led to the
breakdown. The German Social Democrats (SPD), the largest party during much
of the Weimar Republic, embraced some radical policy preferences including
orthodox Marxism (Berman 1998: 77–95, 123–31, 180–98). They did not have
a clear normative preference for democracy (Berman 1998: 85–88, 130–31,
180–81). Berman argues that if the SPD had been more flexible, less radical,
and more oriented toward preserving democracy, it could have undercut the
Nazis’ appeal. Chancellor Heinrich Brüning (1930–32) and the Center Party
were willing to sacrifice democracy in order to achieve other policy goals
(Berman 1998: 187; Weitz 2007: 122–23). The Communists and the Nazis had
very radical policy preferences and a normative preference for different kinds
of totalitarian dictatorship. The rightist German People’s Party (Deutsche
Volkspartei) was somewhat hostile to democracy, and the German National
People’s Party (Deutschnationale Volkspartei) combined radical right-wing policy
preferences with a normative preference for authoritarian and monarchical rule
(Lepsius 1978: 37, 43, 45; Weitz 2007: 92–97). The Landvolkbewegung was a
right-wing peasant movement with radical policy preferences and antidemocratic
normative preferences (Lepsius 1978: 53–54). The army and the Protestant and
Catholic churches were hostile to democracy (Weitz 2007: 115–21).
A third case that has often been analyzed in ways entirely consistent with our
theory is the stabilization of democracy in Spain after 1978. Many conditions

3
Echoing our terminology, S. Payne (2006: 354) wrote that “[m]ost major actors had limited or no
commitment to democracy.”
Rethinking Theories of Democratization 277

favored democratic consolidation in Spain, but this favorable outcome was by

no means a foregone conclusion (Agüero 1995: 18–22; Linz and Stepan 1996:
87–115). Democratic consolidation was facilitated by strong support from
the European Union (Linz and Stepan 1996: 113). De-radicalization of the
Socialist Party (Tussell 2005: 285, 327), the Communist Party, the labor move-
ment (Fishman 1990b), and the right (Share 1986) also facilitated democratic
survival. By the time of the transition in 1978, actors with a normative prefer-
ence for dictatorship were weak (G. Alexander 2002: 138–81). Finally, by 1978,
most actors either already had a normative preference for democracy or quickly
developed one. This includes the right (G. Alexander 2002: 138–81); Prime
Minister Adolfo Súarez (1976–81) and his party, the Union of the Democratic
Center (Unión de Centro Democrático, UCD) (Share 1986: 86–153); and labor
(Fishman 1990b). The fact that most powerful actors have had an unambiguous
normative preference for democracy and the highly supportive international
political environment enabled the democratic coalition to thwart difficult chal-
lenges including the Basque terrorist organization ETA (Euskadi Ta Askatasuna,
Basque Homeland and Freedom), a military that had strong golpista factions
until 1981 (Agüero 1995), and a bruising depression in recent years.

alternative theories of democratization

Throughout this book, especially in Chapters 4 through 6, we have compared
our theory with alternatives, but without an extended discussion of the impli-
cations of our results for these alternative theories. In this section, we examine
the implications of our analysis and empirical results for modernization theory,
class theories, work on the impact of economic performance on regime fall or
survival, mass culture theory, work on the impact of formal institutions on
regime fall or survival, and work that strongly emphasizes political agency and
leadership.
Some of our observations in this section focus on inconsistencies between the
evidence for Latin America and broader research findings. Other arguments hold
beyond Latin America. For example, our criticisms of class and mass cultural
theories of democratization, while initially inspired by the Latin American expe-
rience, go beyond it.

Modernization Theory
One of the most influential theoretical approaches to studying democracy is
modernization theory, which was famously formulated by Lipset (1959, 1960:
27–63) and subsequently empirically supported by many other scholars.4
Modernization theory claims that more economically developed countries are

4
The second part of Lipset’s classic (1959) article made a different claim about the effects of regime
legitimacy and efficacy that anticipated some points in our book.
278 Democracies and Dictatorships in Latin America

more likely to be democratic, and proposes causal mechanisms to explain this

relationship.
Lipset argued that wealthier countries are more likely to be democratic for
several reasons. Higher education, which is associated with greater wealth,
promotes more tolerant worldviews. Greater wealth tends to reduce political
extremism and to increase the size of the middle class, which he claimed tended
to be more prodemocratic than the poor. In developed countries, the lower strata
tended to have more reformist political perspectives (Lipset 1959: 83), and the
upper strata tended to have more democratic values (Lipset 1959: 83–84).
Many subsequent authors have demonstrated that higher levels of development
are strongly associated with a greater likelihood of democracy (Bollen 1980, 1983;
Bollen and Jackman 1985a,1985b; Burkhart and Lewis-Beck 1994; Coppedge
1997; Dahl 1971: 62–80; Diamond 1992; Epstein et al. 2006; Huntington 1984,
1991: 59–72; Jackman 1973; Lipset et al. 1993; Londregan and Poole 1996;
Przeworski et al. 2000).5 Recent work has demonstrated that the likelihood of
democratic breakdowns diminishes and the likelihood of democratic transitions
increases at higher per capita income (Epstein et al. 2006).
However, the seemingly robust association between income and democracy
does not hold for Latin America for the lengthy period from 1945 to 2005
(Landman 1999; Mainwaring and Pérez-Liñán 2003; O’Donnell 1973). During
this period, competitive regimes were as vulnerable to breakdown at a higher level
of development as at lower levels (Tables 4.4 and 4.5). The level of development
likewise had no impact on the probability of a transition from authoritarianism
to a competitive regime (Tables 4.2 and 4.3).6 As we discussed in Chapters 5 and
6, modernization theory does not go very far toward explaining regime outcomes
in Argentina and El Salvador.
Building democracy in poor countries is difficult, and yet as the experience of
poor countries in Latin America shows, the obstacles are not insurmountable.
We reject Lipset’s (1960: 40) argument that a certain level of development is
a requisite for democracy or that “a high level of education . . . comes close
to being a necessary” condition for democracy. Costa Rica was a relatively
poor country with a relatively low level of education at the inauguration of its
competitive regime in 1949, with a per capita GDP of $1,546 in 2000 constant
dollars, below the mean value of $1,846 for our entire dataset. Yet this regime
has now lasted for more than six decades, becoming the longest-lasting democracy
ever outside countries that today are part of the advanced industrial democracies.

5
Acemoglu et al. (2008) argue that using a proper model specification, the level of development does
not affect regime outcomes across all countries for which data was available. They advocate fixed
effects models.
6
In model 4.3.3 with the Gini index of income inequality, a higher per capita GDP seemed, against
conventional expectations, to lower the probability of a transition. However, because of the large
number of gaps in the data on income inequality, the number of observations fell from 576 to 222.
Given the consistency of the results across many model specifications with 576 observations, it
seems very likely that the result in model 4.3.3 stems from the reduced number of observations.
Rethinking Theories of Democratization 279

A low per capita income did not preclude building what has become a high-quality
democracy. Competitive regimes have also endured at fairly low levels of
development in countries such as Ecuador since 1979 and Nicaragua and El
Salvador since 1984, or (a non–Latin American example) India from 1947 until
Indira Gandhi’s declaration of a state of emergency from 1975 until 1977, and
then again since 1977.
We do not claim that modernization theory is wrong, but the relationship
between the level of development and democracy has been far from determinate
in Latin America until a high level of development makes radicalization unlikely.
At a high level of development, democracy has historically never broken down
(Przeworski et al. 2000; Epstein et al. 2006). It therefore seems that a high level
of development is a sufficient condition to ensure the survival of a competitive
political regime. It is possible, as Przeworski (2006) suggests, that the reason is
that at high levels of development, few actors are radical, and radical actors are
isolated. If this argument is correct, then the core causal mechanism linking high
income to democracy is de-radicalization. Below that high level of wealth, for
Latin America, the relationship between the level of development and democ-
racy has been overpowered by the political factors to which we call attention.
High levels of poverty and glaring inequalities provide grist for radicalism
and dampen the likelihood of strong normative commitments to democracy. Yet
as the examples of the southern cone suggest, this effect is far from linear. As
Lipset (1959: 90–91) himself recognized, poverty and inequality do not directly
produce radicalization and do not automatically suppress normative preferences
for democracy (see also Dahl 1971: 81–104; Moore 1978; Portes 1971; Powers
2001; Weyland 2002).

Class, Inequality, and Democratization

The Latin American evidence and broader evidence are largely at odds with class
theories of democratization. These theories see some social classes as being
consistently prodemocratic when democracy is possible and others as consis-
tently supporting authoritarian regimes when stable dictatorship is feasible. The
most prominent class theories include Acemoglu and Robinson (2006), Boix
(2003), Moore (1966),7 and Rueschemeyer, Stephens, and Stephens (1992).
Rueschemeyer et al. (1992) argued that prospects for democracy rest on the
balance of power among social classes. In their general theoretical statements,
they claimed that the working class is the agent of democratization par excellence
(p. 8). Therefore, a strong working class is favorable to democracy. Conversely,

7
Moore (1966) argued that a historical coalition of a strong landed aristocracy, a relatively weak
bourgeoisie, and a modernizing state produced fascism; the combination of a recalcitrant aristoc-
racy and an absolutist state triggered socialist revolutions; and the hegemony of the bourgeoisie
over the aristocracy, the agricultural labor force, and the state led to the establishment of liberal
democracy. See J. S. Valenzuela (2001) for a compelling critique.
280 Democracies and Dictatorships in Latin America

they saw the landed elite (pp. 60–61) and the bourgeoisie as usually favorable to
the status quo before democracy and as resistant to democratization.
Acemoglu and Robinson (2006) and Boix (2003) assume that classes try to
maximize income and choose a political regime accordingly. They posit that
democracy will economically benefit the poor and redistribute away from the
rich. They conclude that the poor favor democracy over any nonrevolutionary
authoritarian regime, whereas the wealthy concede democracy only to avoid
revolution. The wealthy have more to lose with democracy in more inegalitarian
societies. According to Boix, the rich block the emergence of democracy in
unequal societies unless the cost of repression is high, but they accept democracy
if capital mobility prevents high taxation.
Although they differ in many ways,8 these class theories share four assump-
tions: (1) classes are the most important political actors; (2) members of social
classes value political regimes exclusively for economic reasons; (3) democracies
redistribute income in favor of the poor; and (4) given this outcome, the working
class and the poor are strong supporters of democratization while the bour-
geoisie or the rich concede democracy only reluctantly.9 In addition, Acemoglu
and Robinson (2006) and Boix (2003) assume that (5) high inequality reinforces
resistance among the rich, making the establishment and survival of democracy
unlikely.
These assumptions are not consistently realistic, and shortcomings of class
theories result. First, classes as Boix and Acemoglu and Robinson conceptualize
them (i.e., the poor and the rich) do not form cohesive political actors. Members
of the same class are divided by religious, national, ethnic, and other value
questions. These divisions make it difficult to act cohesively, and rich and poor
face daunting collective action problems (Olson 1965). Moreover, in the strug-
gles for and against democracy in most countries, political parties, militaries,
and other nonclass organizations are key actors. This is clear in our analysis
of Argentina (Chapter 5) and El Salvador (Chapter 6). The history of both
countries involved important class-related actors. But in both countries, political
parties, militaries, churches at some periods, and guerrillas in others were power-
ful actors whose behavior was not reducible to class interests. Throughout the
region, actors other than classes have been powerful.
Second, class theories assume that the only issue that drives political conflict
in all countries is income distribution and resource allocation. Classes prefer

8
Rueschemeyer and colleagues and Moore employ class categories, and they delve into the historical
development of democracy in different parts of the world. Acemoglu and Robinson (2006) and
Boix (2003) base their analysis on income categories (poor, middle sector, rich) rather than class
understood structurally, and Boix’s evidence is largely quantitative. For Boix, the relationship
between inequality and democracy is linear: more inequality generates a lower probability of
democracy. In contrast, Acemoglu and Robinson posit an inverted-U-shape relationship; democ-
ratization is very unlikely at high or low levels of inequality.
9
Rueschemeyer and colleagues make these assumptions in their general theoretical propositions, but
their analysis of Latin America clearly breaks from the first and fourth.
Rethinking Theories of Democratization 281

political regimes for purely instrumental material reasons – to advance their

economic interests. However, an extensive literature has argued otherwise
(Haggard and Kaufman 2012). Again, it is difficult to understand the history
of political regimes in Argentina (especially) or El Salvador as a battle exclusively
over material goods.
As we have argued throughout this book, value divides about democracy and
authoritarianism are not reducible to economic issues, and they often strongly
influence regime outcomes (Berman 1998; Dahl 1971: 124–89; Ollier 2009;
Viola 1982). So do conflicts over religion (Casanova 2010; Huntington 1996;
Levine 1973; Linz 1991; Pérez-Díaz 1993; Stepan 2001; J. S. Valenzuela 2001),
urban/rural and regional cleavages, nationalism (Linz 1997; Linz and Stepan
1996: 24–33; Stepan 1994, 2001: 181–212; 323–28; Stepan et al. 2011),
and ethnicity (Diamond 1988; Rabushka and Shepsle 1972; Snyder 2000).
Empirical analyses of post-Soviet countries in the 1990s underscored that
citizens value democracy independently of economic results (Hofferbert and
Klingemann 1999; Rose and Mishler 1996). In addition, international ideational
currents – the Zeitgeist of an epoch – affect the survival and fall of political
regimes. For most countries, it is impossible to understand the survival or fall
of political regimes by looking exclusively at class conflict related to income
distribution. A range of policy issues, not just economic distribution, affects
regime outcomes (Haggard and Kaufman 2012). By emphasizing that actors
do not join or defect from regime coalitions based exclusively on the regime’s
material payoffs, we are better able to understand the survival of competitive
regimes in the third wave despite dismal economic performance in the 1980s,
1990s, and early 2000s.
Third, these class approaches rest on the questionable assumption that
democracy is consistently good for the working class’s material interests and
that noncommunist authoritarian regimes are detrimental to the income of the
poor. The empirical evidence however, is mixed, as our book and other works
show. Huber and Stephens (2012) made a compelling argument for why com-
petitive regimes, by allowing left parties to organize and gain office, may pro-
mote redistribution over the long run. They also documented that in the short
run these effects are subject to constraints imposed by international factors
(Chapters 6–7), and that Latin America has also experienced an authoritarian
path to redistribution illustrated by Argentina and Brazil in the twentieth century
(Chapter 4; Segura-Ubiergo 2007: chapter 2).
Whether we look at changes in real wages or income distribution, it is far
from evident that Latin American workers have fared better materially under
competitive regimes than under authoritarianism. Real wages fell in most coun-
tries after the establishment of competitive regimes in the third wave of democ-
ratization, including quite dramatically in Argentina after re-democratization in
1983. In Peru, the real urban minimum wage declined an astonishing 84 percent
between 1980 and 1992. Real urban minimum wages fell drastically after the
transitions to democracy in five of the six countries for which ECLAC reported
282 Democracies and Dictatorships in Latin America

data in 1992.10 Conversely, real wages increased under many authoritarian

regimes between 1945 and 1980. In the 1970s, real income improved at least
30 percent in all deciles of the Brazilian population under the military dictator-
ship. Real mean income for the poorest decile increased 50 percent from 1970 to
1980 (Skidmore 1988: 287).
For Latin America, the average currently existing competitive regime has not
promoted income distribution in favor of the poor. For the seventeen countries
for which data are available (all but Cuba, Haiti, and Nicaragua11), mostly from
the World Bank inequality dataset, from the year of a transition to a competitive
regime until 2010,12 income distribution improved on average by a trivial
1.2 points (from 52.0 to 50.8) on the 100-point Gini index (weighting every
country equally). Conversely, some populist and nationalist left-of-center (but
not revolutionary) authoritarian regimes have redistributed income to the poor –
a possibility that these works often neglect.13
The evidence beyond this book about the impact of regime type on income
distribution and social policy is mixed. Huber and Stephens’s (2012) analysis
of eighteen Latin American countries between 1970 and 2007 suggests that a
longer history of democracy may lead to greater investments in social programs
and to a reduction in income inequality over time (chapter 5). On the other hand,
Mulligan et al. (2004) show that on average, democracies do not spend more
than dictatorships on social programs, that they tax less than dictatorships, and
that they promote less income redistribution than dictatorships (p. 60). Bollen
and Jackman (1985b) also showed that democracies are not more redistributive
than dictatorships. Nelson (2007) reports converging findings about the impact
of democracy on health and education. Democracies do not have demonstrably
better results than dictatorships do in these social domains. Burkhart (1997)
shows that democracies improve income distribution only at a fairly low level of
democracy.
The fourth shortcoming of these class theories is that the empirical evidence
to substantiate their claims about the relationship between class position and
support for democracy is thin. Boix and Acemoglu and Robinson present little
evidence to support the claim that the poor actually prefer democracy and have
fought for it on a consistent basis. Rueschemeyer et al. (1992: 8) note that the

10
ECLAC 1992, tables 6 and 7, pp. 44–45. ECLAC also reported data for urban real minimum
wages for Rio de Janeiro and São Paulo, but did not give an average figure for Brazil. ECLAC
1995: 131–34, and ECLAC 1994: 127–28, also report figures for urban minimum wages. In most
third-wave democracies, urban minimum wages fell after the transitions to competitive regimes.
11
For Nicaragua, there are no data points close to 1984, the year of the transition to semi-democracy.
For Cuba and Haiti, there are none whatsoever.
12
For Colombia, Costa Rica, and Venezuela, no data were available for the transition years (1958,
1949, and 1959, respectively). We used the earliest available data points: 1970, 1961, and 1962,
respectively.
13
Along similar lines, Albertus (2011) showed that in Latin America, authoritarian regimes have
undertaken more agrarian reform than democracies have.
Rethinking Theories of Democratization 283

Latin American cases do not fully conform to their general theory. R. Collier
(1999: 33–76) argues that their theory does not work for many Latin American
and Western European cases because elites and middle sectors rather than the
working class were primarily responsible for establishing democracy.
The relationship between class and support for democracy is more mediated
and less linear than class theories suggest. The historical evidence about which
classes were more likely than others to support democratization is more mixed
than class theorists claim (R. Collier 1999; Levitsky and Mainwaring 2006;
J. S. Valenzuela 2001). In many cases, some sectors of the elite were at the
forefront of democratization even in the absence of a credible revolutionary
threat, and in some cases, the poor actively preferred a nonrevolutionary autho-
ritarian regime to democracy.
Until recent decades, organized labor in most Latin American countries did
not consistently support democratic regimes. As an illustration, in Argentina
(Chapter 5), organized labor supported Peronism from 1945 on, notwithstand-
ing its frequently authoritarian character. In 1962 and 1966, labor supported
military coups against competitively elected governments. In Latin America,
populist leaders with radical policy preferences and authoritarian proclivities
often captured organized labor’s support because of their promises or delivery
of benefits for workers and their symbolic appeals to the poor (Germani 1974:
169–92, Lipset 1960: 87–126; Ostiguy 2009).
Rather than understanding democratization in terms of consistent democratic
or authoritarian proclivities of class actors (Rueschemeyer et al. 1992) or of
consistent first choice preferences that shift only if the first choice regime is
not feasible (Boix 2003), we see classes as being conditional authoritarians
and conditional democrats (Bellin 2000). As the Argentine case discussed in
Chapter 5 showed – and our coding of actors in other countries confirmed –
under some circumstances, organized labor will support authoritarian leaders,
movements, parties, and regimes even if democracy is feasible. Whether labor
supports democracy depends on (1) its normative preferences regarding the
political regime and (2) whether it believes authoritarian or democratic leaders
and parties better serve labor’s policy goals. The fact that the working class does
not consistently support democracy helps explain why the size of the working
class had no impact on reducing the probability of democratic breakdowns in
our quantitative analysis in Chapter 4 (Tables 4.4 and Tables 4.5).14
Voting patterns and public opinion surveys also show a mixed relationship
between class position and support for democracy. For example, in Mexico,
during the democratization process from 1988 to 1997, the poor and least
educated solidly supported the PRI (the ruling authoritarian party). The middle
and upper classes and the most educated were more likely than the poor to
support the largest democratic opposition party, the PAN (Domínguez and

14
A large working class was favorable to democratic transitions in the regressions in Table 4.4 but at
most weakly favorable to transitions in Table 4.3.
284 Democracies and Dictatorships in Latin America

McCann 1996: 99–100, 203–04; Klesner 2004: 103–07, 112, 116; Magaloni
1999: 228–31; 2006: 122–50; Magaloni and Moreno 2003: 268–69).
An analysis that sees the poor as the bearers of democracy and the rich as its
opponents must also confront the fact that in public opinion surveys, respond-
ents with lower income usually evince less democratic attitudes than those with
higher income. In eleven of the nineteen Latin American countries included
in the 2008 AmericasBarometer,15 wealthier respondents displayed stronger
pro-democracy attitudes (at p < .05) than poor respondents in response to the
statement “Democracy has problems, but it is better than any other form of
government.”16 Interestingly, in light of the 2009 coup, Honduras was the only
country in which higher-income respondents gave less democratic answers.
In the remaining seven countries, income did not have a statistically significant
effect on responses to this question. The 2010 AmericasBarometer surveys
confirmed this finding. The correlation between household income and support
for democracy was positive and significant (p < .05) in fifteen of the nineteen
countries, positive but insignificant in two cases (Brazil and Nicaragua), and
negative but insignificant in only two countries (Bolivia and Honduras). The
results of the bivariate correlations do not prove that poor citizens are generally
less supportive of democracy, but they call into question a fundamental assump-
tion of class theories of democratization.17
Fifth, most of the empirical evidence does not support the core claim that
inequalities have a powerful impact on regime survival and change. Teorell
(2010: chapter 3) finds no impact of inequality on democracy. According to
Muller (1988: 61), the level of inequality had no impact on the probability of
a democratic transition, although high inequalities made democracies more
vulnerable to breakdown (pp. 61–65). Burkhart (1997) found that high inequa-
lity lowered the level of democracy (a different dependent variable than we use
in this book), but the effect was modest.
The evidence in this book is consistent with these broader findings. For Latin
America, income inequality had no statistically significant impact on the survival
or fall of democracies or dictatorships (see Tables 4.3 and 4.5). According to some
class theories, the deterioration of the already skewed income distributions during
the 1980s and 1990s should have made competitive regimes more vulnerable
and wealthy elites more resistant to democratization. In fact, competitive regimes

15
The AmericasBarometer is conducted by the Latin American Public Opinion Project (LAPOP) at
Vanderbilt University. All countries in our sample, with the exception of Cuba, were covered by
the 2008 and 2010 waves of the project.
16
Responses to the statement are captured by a seven-point scale, ranging from “Disagrees a lot” to
“Agrees a lot.” We ran a bivariate OLS regression for each country using this item as the dependent
variable. The income variable is calibrated for local currency and coded using an eleven-point scale
in all countries.
17
In a study of mass attitudes in eight Latin American countries, Booth and Seligson (2009) found
that household wealth is uncorrelated with support for core principles of democracy or demands
for democracy, in statistical models that also control for educational levels (tables 4.3 and 7.1).
Rethinking Theories of Democratization 285

became far less susceptible to breakdown during the third wave. Even in the
absence of a revolutionary threat, wealthy elites were critical actors in supporting
democratization in many countries, including Chile in the nineteenth and early
twentieth centuries (J. S. Valenzuela 1985, 2001), Brazil in the 1970s and 1980s
(Cardoso 1986; L. Payne 1994), El Salvador in the late 1980s and early 1990s
(Johnson 1993; Wood 2000a, 2000b; Chapter 6 in this book), and Mexico in
the 1980s and 1990s. Bad income distribution did not prevent a large number
of transitions to competitive regimes from occurring, and the further exacerbation
of glaring inequalities did not lead to the breakdowns of competitive regimes
after 1978.
Boix’s (2003) own results provide weak support for the idea that better income
equality increases the likelihood of transitions to democracy and decreases the
likelihood of democratic breakdowns. In only one of four models for all countries
(Model 3A) in his book did income distribution affect the likelihood of transitions
to democracy at p < .10 (Boix 2003: 79–81). Income inequality had a significant
impact on democratic breakdowns in three of the four models for all countries,
but in one of the three (Model 1A), contrary to the theory, inequality facilitates
democratic survival. Additional interactions of income inequality with other
variables in the model do not provide unequivocal support for Boix’s theory.
Boix qualifies his argument by asserting that high capital mobility (or high
asset specificity) makes it easier for the rich to invest outside their country, and
hence lowers the probability of major redistributive efforts. He argues that
in contexts of high capital mobility, governments are forced to keep taxes low;
otherwise, capital flight will result (pp. 12, 19, 25, 39). Because taxes are low,
elite resistance to democracy will diminish.
In Latin America, however, increasing capital mobility after 1985 coincided
with notable increases in tax collection in most countries. According to ECLAC
data, between 1990 (the earliest data point) and 2010, total central government
tax revenue increased substantially (at least 5 percent of GDP) in nine Latin
American countries (Nicaragua, +14 percent; Bolivia, Ecuador, and Argentina,
+10 percent; Colombia, +7 percent; Brazil and the Dominican Republic, +6
percent; Paraguay and El Salvador, +5 percent) under competitive regimes.
In most other countries, tax revenue increased somewhat. Only in Venezuela
(–7 percent) did central government tax revenue decrease at least 5 percent of
GDP during this period of increasing capital mobility.18 Therefore, for Latin
America greater capital mobility did not reduce the capacity of democratic
governments to collect taxes. A cross-regional comparison between Western
Europe and Latin America further underscores the problematic nature of this
argument. Both capital mobility and tax collection are higher in Western Europe
than Latin America. Circa 2003, the average total tax revenue for fifteen EU
countries was 41 percent of GDP, while according to 2005 estimates, nine Latin
American countries had central government tax revenue of less than 15 percent

18
ECLAC’s data on central government tax revenue are online at http://www.eclac.org/estadisticas/
286 Democracies and Dictatorships in Latin America

of GDP (ECLAC 2008). No Latin American democracy approached the level of

the average EU country.
The impact of inequalities on democratic regime stability depends on what is
going on in different countries at different times (Frey et al. 2004: 389–90).19
In the 1990s, in times of deep economic crisis, poor people accepted the exacer-
bation of inequalities in order to achieve macroeconomic stability (Powers 2001;
Stokes 2001; Weyland 2002).
We believe that our theory better interprets the survival and fall of democ-
racies and dictatorships in and beyond Latin America than class theories. The
main differences are as follows: (1) we view organizations, not classes, as the
most important actors – some but not all organizations primarily defend class
interests; (2) we assume that actors are interested in a broader range of policy
outcomes than just material and distributive issues; conflicts over religion,
ethnicity, and nationality, among others, influence regime outcomes; (3) we
believe that many actors have normative preferences about the political regime
in addition to policy preferences; (4) we situate our theory in an international
context more than most class theories; and (5) we see the relationship between
class and regime preference as highly conditional.

Economic Performance
Some authors have shown that democratic and authoritarian regimes are more
likely to survive if their economic performance is better. Most of this literature is
empirical and does not invoke strong theoretical claims about the relationship
between economic performance and regime stability. We do not dispute the
empirical assertions made by these authors. Among well-known works that peg
democratic stability to economic performance are Gasiorowski (1995), Haggard
and Kaufman (1995) and Lipset (1960: 64–70).20
The general theoretical proposition that government performance affects
regime stability in developing countries is sensible. Consistent with this litera-
ture, we expected the regime’s economic performance to affect actors’ adhesion
to the incumbent regime – but we expected this impact to be modest, especially in
competitive regimes.
By the logic of our theory, poor economic performance creates a threat to the
survival of democracy only if (1) some actors conclude that authoritarianism
offers net policy advantages to them – that is, they believe they would be better

19
Moore (1978: 41) comments that in popular perception, “a high degree of inequality may not only
be acceptable but even regarded as very desirable, as long as in the end it somehow contributes to
the social good as perceived and defined in that society.”
20
Lipset argued that regimes needed a combination of good performance and legitimacy. A reservoir
of legitimacy can enable a democracy to remain stable despite poor performance. Thus, his was
not a simplistic performance-based argument. See also Linz (1978b: 16–23) on the relationship
between legitimacy and performance.
Rethinking Theories of Democratization 287

off under an authoritarian regime; (2) this net policy advantage is not offset by
a normative commitment to democracy; and (3) the authoritarian coalition is
powerful enough to consider overthrowing a democratic regime. Actors’ deci-
sions about whether to work to overthrow a competitive regime hinge on all of
their policy preferences and their normative preferences about the political
regime, as well as a strategic calculation about the odds of successfully subvert-
ing the regime. Democratic regimes can win support on bases other than regime
performance (Linz 1988; Remmer 1996). Citizens do not necessarily attribute
performance failures to the regime; they normally blame particular administra-
tions or parties in office.
Consistent with our expectation, the most democratic period in the history of
Latin America (since the mid-1980s), and the period with by far the highest-ever
rate of survival of competitive regimes (since 1978), coincided with a prolonged
period of dismal economic and social performance in most countries (1982–2002).
The logic of our theory correctly predicts that actors’ normative preferences
for democracy, low radicalism, and strong regional support for democracy
could protect competitive regimes in times of bad performance. Bad performance
had adverse effects on democracy, but it has rarely led to regime breakdown in
the post-1977 period.21 For a generation, regime survival has not depended on
economic performance, suggesting that the impact of bad economic performance
on political regimes is mediated by citizen expectations, which vary over time; by
the way political leaders do or do not politicize bad economic performance; and
by actors’ normative commitment to democracy.
In Latin America, the rate of economic growth had little or no impact on the
survival of competitive (Table 4.4) or authoritarian regimes (Table 4.2). Inflation
also had no impact on regime change (Tables 4.3 and 4.5). Competitive regimes
have been vastly less vulnerable to breakdown since 1978 compared to 1945–77,
even though the median regime’s economic performance fell from solid in the
earlier period to poor. The average per capita GDP growth rate of competitive
regimes was 1.9 percent for the 1945–77 period and a meager 1.1 percent for the
1978–2005 period, and the mean inflation rate jumped from 19 percent in the
earlier period to 257 percent in the later years. Yet the breakdown rate of these
regimes was more than ten times greater (9.3 percent in the earlier period versus
0.8 percent in the post-1977 period).
The Latin American experience since 1978 shows that the impact of economic
performance on regime survival is mediated by actors’ understanding of what
is possible in a given moment (i.e., their view of constraints and opportunities)
and can be overcome by their normative attitudes about political regimes.
Democracy in Latin America would be in better shape in many countries if

21
There have been only six breakdowns since 1978: Bolivia in 1980; Peru in 1992; Haiti in 1991 and
1999; and Honduras and Venezuela in 2009. Because the Haitian regime of 1991 lasted only a few
months before giving way to a coup before the end of the year, our regime classification registers
only the other five breakdowns.
288 Democracies and Dictatorships in Latin America

economic performance had been better during the third wave. Nevertheless,
competitive regimes survived despite economic and social disappointments, a
deterioration of public security, and rampant corruption in many countries.
Although poor economic performance has weakened many competitive regimes,
it has doomed few. Poor governing performance has bred citizen disaffection
and paved the way to populist politicians with dubious democratic credentials,
but it has rarely caused regime breakdowns during the third wave.
At some historical junctures, because of ideological currents, some actors
might conclude that an authoritarian regime is more likely to be efficient and
therefore more effective at fostering growth. This was the case in Argentina
in 1965–66 (Chapter 5), when many actors concluded that democracy was
inefficient and suboptimal despite the Illia government’s respectable record in
economic growth. However, even if government performance is deficient,
actors might doubt that an authoritarian regime would be better for them.
In the aftermath of bad economic performance and the accumulation of huge
foreign debts under authoritarian regimes in the 1970s and early 1980s,
citizens in most Latin American countries gave competitive regimes great leeway
in managing the economy until the late 1990s (Powers 2001; Stokes 2001;
Weyland 2002).
In many countries, citizens and elites had little reason to believe that a new
round of authoritarianism would ease their economic troubles. The new com-
petitive regimes inherited challenging and in several cases ruinous economic
legacies. The dismal economic performance of these antecedent authoritarian
regimes helps explain the disappearance of actors that have a normative prefer-
ence for dictatorship and the high tolerance for poor economic performance
under competitive regimes in most of Latin America from 1982 to 2002 (Powers
2001; Remmer 1996; Weyland 2002).
Assuming that some actors anticipate a net policy advantage under some form
of authoritarian rule, policy preferences may still be offset by a normative
commitment to democracy (Frey et al. 2004). Even where past achievements
have not built a cushion to buffer democracies from poor performance, good
economic performance might not be necessary for regime stability at some
historical moments. Actors’ policy expectations and their normative preferences
about the regime mediate the relationship between government performance
and regime stability. Actors that are committed to democracy have a reservoir of
goodwill toward competitive regimes; they do not readily jump ship to further
their policy goals.
Finally, even if some actors anticipate net gains from authoritarianism and
lack a strong normative preference for competitive politics, the authoritarian
coalition must be powerful enough to overthrow a democratic regime. In con-
texts where international actors might impose sanctions against coup leaders,
only actors unusually concerned with economic growth are likely to believe
that the growth advantage they presume an authoritarian regime would offer
is sufficient to offset the risk of supporting a coup.
Rethinking Theories of Democratization 289

We do not claim that Latin American democracies have been permanently

inoculated against instability resulting from bad performance. Citizen tolerance
for poor economic performance under competitive regimes appears to have
dropped somewhat in many countries in the late 1990s. At that time, a new period
of prolonged stagnation (1998–2002) in the region as a whole and of increased
poverty in many countries fueled growing disgruntlement in Argentina, Bolivia,
Ecuador, and Venezuela. The theoretical lesson is that citizen and elite sensitivity
to poor economic performance varies widely across time and space (Kapstein and
Converse 2008).

Mass Political Culture and Democratization

Political culture studies based on individual attitudes see democracy as emanat-
ing from democratic values among the citizenry; where citizens have democratic
values, democracy flourishes (Almond and Verba 1963; Inglehart 1990; 1997:
160–215; Inglehart and Welzel 2005). We agree that mass support for democ-
racy is a powerful resource for democratic actors. But our work diverges from
political culture approaches based on mass surveys in several ways. First, we
emphasize the role of leaders and organizations, not of ordinary citizens, in
determining regime outcomes The beliefs of leaders and organizations usually
have more weight than citizen views in determining regime outcomes (Berman
1998; Bermeo 2003; Dahl 1971: 124–88; Linz 1978b).
Second, whereas our theory calls for analyzing specific actors and coalitions
that trigger regime change or stabilize the incumbent regime, mass political
culture approaches usually do not establish convincing mechanisms by which
mass attitudes determine regime outcomes. They usually lack a sense of agency –
that is, of specific actors or mechanisms through which mass beliefs about
politics affect regime change. Inglehart and Welzel (2005) discuss this issue in
greater detail than most work on mass political culture. They argue that political
regimes confront pressures for change when mass values are incongruent with
the regime (pp. 158, 174, 186–91).The variable “demand for freedom” is at the
core of their theory that mass values are the most important long-term determi-
nant of democracy. However, based on how the variable is constructed, it is
difficult to see why it represents citizen demands for democracy. It is based on a
factor analysis that combines five items: postmaterialism, personal happiness,
tolerance of homosexuality, willingness to sign a petition, and interpersonal
trust. None of these five survey items constitutes a demand for democracy, and
it is not clear how any of them facilitates a transition to democracy.22

22
By contrast, Mattes and Bratton (2007) measured demand for democracy using a battery of indicators
that capture whether respondents reject one-man rule, reject military rule, reject one-party rule, and
prefer democracy above other forms of government. Booth and Seligson (2009) measured demand for
democracy using a dichotomous indicator that captured if respondents preferred an elected leader to a
strong but unelected leader (Chapter 7).
290 Democracies and Dictatorships in Latin America

Booth and Seligson (2009: chapter 8) theorized a more specific causal mech-
anism that is consistent with our approach, arguing that elites with a low commit-
ment to democracy find it easier to curtail civil liberties and political rights when
large segments of the population simultaneously present low levels of support
for democratic principles, national political institutions, and regime perform-
ance. However, their comparison of those “triply dissatisfied” citizens against
satisfied citizens showed only modest differences in terms of support for con-
frontational politics, military coups, and unelected governments (Figure 8.3). The
evidence supports their arguments but does not sustain more sweeping claims
about the impact of mass political culture on political regimes.
In contrast to theories that claim that mass political culture determines regime
outcomes through some difficult-to-specify mechanisms, we begin with concrete,
identifiable historical actors. Citizen opinion affects these actors, but the rela-
tionship between citizen opinion and actors’ behavior is very far from linear
(Bermeo 2003).
Third, mass political culture approaches generally do not attempt to explain
regime change, which is one of our primary concerns. They can attempt to explain
regime stability on the basis of patterns of association between mass attitudes
and regime type, for example, that authoritarian mass attitudes are conducive to
authoritarian regimes. But because mass attitudes are putatively relatively stable
over the medium term, they are less successful at explaining dramatic change.
Inglehart and Welzel (2005) assert that self-expression (which is exactly the same
variable as “demand for freedom”) values explain political regimes. However, their
own data indicate that their cultural explanation of regimes based on self-
expression values works only modestly for the 1995–2002 period and not well
for the 1978–89 period. They report modest country-level correlations, ranging
from about .32 to about .39, between self-expression values measured between
1990 and 1995 and levels of democracy (measured by Freedom House scores) from
1995 to 2002 (figure 8.3, p. 184). Even more problematic for their argument, the
correlation between self-expression values (again measured between 1990 and
1995) and the level of democracy from 1978 to 1989 is consistently low, ranging
from about .01 to about .16. Because they claim there is very high stability over time
in self-expression values, the correlation between these values from 1978 to 1989
and democracy in those years must also be low. At best, their theory is valid to a
very modest extent for the 1995–2002 period and generally not valid for a longer
time period (1978–89).
Fourth, mass political culture approaches usually disregard the problem of
reverse causality – that is, the possibility that a democratic political regime
fosters a democratic political culture (Barry 1978: 47–74; Muller and Seligson
1994; Seligson 2002). For instance, Booth and Seligson (2009) showed that
respondents in countries with a longer history of democracy tend to express
stronger support for democratic principles (chapter 4). Inglehart and Welzel
(2005: 176–209) explicitly addressed reverse causality, claiming that a demo-
cratic political culture causes democracy and not vice versa. They correctly noted
Rethinking Theories of Democratization 291

that “[i]f self-expression values cause democracy, they must be in place before
democracy” (Inglehart and Welzel 2005: 178). Their statistical work thus
implicitly assumes that all democracies in their sample transitioned to democ-
racy after their measurement of self-expression values (i.e., 1990 or 1995,
depending on the country), but this is not the case. Twenty-three of the sixty-
one countries in their sample were democracies for generations before their
measurement of the independent variable.23 Moreover, the history of democra-
tization in these countries raises serious doubts about an argument that invokes
self-expression values as the cause of democracy. Inglehart’s (1990, 1997) own
work indicates that self-expression values emerged in recent decades, which
means that they cannot explain the emergence of democracy in many countries
before then.
Many other countries in their sample (e.g., Argentina, Brazil, the Dominican
Republic, Greece, Portugal, the Philippines, South Korea, Spain, Turkey,
Uruguay) transitioned to democracy before their measurement of self-expression
values. Most of the countries that underwent transitions to competitive regimes
at the time that fits their argument (between 1989 and 1996 – see Inglehart and
Welzel 2005: 176–80) were in the Soviet bloc. In this region, international
influences, in particular Gorbachev’s willingness to accept growing autonomy
of countries dominated by the Soviet Union, followed by demonstration effects
that spread across the region and later by the dissolution of the Soviet Union,
were hugely important (Brown 2000; Kuran 1991; Lohmann 1994).
Finally, the empirical predictions of mass political culture approaches are not
demonstrably fruitful for explaining regime patterns in Latin America. There is
no convincing empirical basis for claiming that a change in mass attitudes was
primarily responsible for transitions to competitive regimes after 1977 or for
democratic stability in the third wave. In contemporary Latin America, mass
attitudes are far from unequivocally supportive of democracy. In the 2011
Latinobarómetro, for eighteen Latin American countries (all but Haiti and
Cuba), only 58 percent of respondents agreed that “Democracy is better than
any other form of government.” Seventeen percent agreed that “Under some
circumstances, an authoritarian government can be preferable to a democracy,”
and 18 percent agreed that “For people like us, it does not matter whether the
regime is democratic.” Another 7 percent did not know or did not respond.24
This distribution of responses does not support the hypothesis that democratic
mass values explain stable democracy. Conversely, the available empirical evi-
dence does not support the idea that mass attitudes caused earlier breakdowns
(Bermeo 2003).

23
Australia, Austria, Belgium, Canada, Colombia, Denmark, Finland, France, Great Britain, Iceland,
India, Israel, Italy, Japan, Luxembourg, the Netherlands, New Zealand, Norway, Sweden,
Switzerland, the United States, Venezuela, and West Germany.
24
Corporación Latinobarómetro, Informe 2011, p. 40. Online at http://www.latinobarometro.org/
latino/LATContenidos.jsp
292 Democracies and Dictatorships in Latin America

In short, mass political culture (or public opinion) influences whether democ-
racies and dictatorships survive or fall. But the empirical evidence does not
support strong causal claims about the impact of public opinion on the survival
and fall of political regimes (Bermeo 2003).
Mass political culture could determine regime types if elites were “sampled”
from the larger population or if, in order to mobilize followers as a political
resource, elites needed to embrace the policy and normative regime preferences
of mass publics. These two statements are partially true, but elites do not
faithfully reflect mass preferences, for two reasons. Given their location in the
social structure, elites usually differ from the larger population in terms of
preferences (Dalton 1985; Iversen 1994). Even when elites claim to represent
mass publics, there are serious monitoring problems (Przeworski et al. 1999).
Elites have significant autonomy and preferences of their own, and elections do
not suffice to induce them to mirror mass preferences. Elites frame the menu of
feasible policy and regime options for their followers, and in this way they also
shape mass preferences (Chhibber and Torcal 1997; Przeworski and Sprague
1986; Sartori 1969; Torcal and Mainwaring 2003).
We expect a correlation between elite and mass attitudes at the national level,
but this correlation might be modest, and the causal direction of the association
is not obvious. Because elites play a critical role in all episodes of regime change
while mass publics play an important role only in some episodes (mass actors
are mostly absent from processes based on elite pacts or imposition), it is safer to
assume that the main explanatory variable behind regime outcomes is the elites’
normative and policy preferences rather than mass attitudes per se.

Agency and Democratization

Some social scientists have underscored the role of elite values and strategies
in regime breakdowns and transitions, emphasizing the importance of leaders’
decisions (Capoccia 2005; Di Palma 1990, Linz 1978b, O’Donnell and Schmitter
1986, Stepan 1978). Following these scholars, we acknowledge the importance
of leaders’ decisions, perhaps especially in moments of regime crisis. Whereas
structural and cultural theories such as Inglehart’s focus on causally more distant
explanations, these works that focus on agency highlight more causally prox-
imate explanations.
Our theory, which is situated in the causal chain between structural approaches
and agency explanations, is compatible with an emphasis on leaders and agency.
We almost always consider the president an actor, which is consistent with an
emphasis on individual leadership. The core of our theory, however, emphasizes
political factors that, although amenable to being influenced by agency, are not
primarily a result of individual leaders’ decisions. In the short term, the political
variables that we highlight are key parts of the landscape that political leaders
confront. In this respect, our theory is not primarily about agency or political
Rethinking Theories of Democratization 293

leadership. Organizations and movements more than individual leaders are at

the core of our theory and empirical analysis.
For example, after the United States began to emphasize democracy in its
Latin American policy and after the OAS institutionalized a system of sanctions
to support competitive regimes and reduce the incentives for coups in 1991,
open coups in the western hemisphere have been a rare exception. Even rela-
tively inept leaders of competitive regimes have rarely fallen to coups. The
development of international mechanisms to sanction overt coups has had
greater weight in determining broad regime outcomes than the quality of poli-
tical leadership. This is a contrast to what occurred in the past, when presidents
who exercised poor decisions could trigger a coup (Stepan 1978). In the current
inter-American system, the effect of poor presidential leadership on regime
survival is circumscribed by norms about the desirability of democracy and
sanctions.

Formal Institutional Rules and Democratic Stability

Our theory also differs from those that focus on the impact of formal political
institutions on regime continuity and change. Linz (1994) famously argued that
presidential systems are more vulnerable to breakdown than parliamentary
systems. However, other scholars have questioned this argument (Cheibub
2007; Shugart and Carey 1992). Presidentialism might help explain democratic
breakdowns before the third wave, but during the third wave, the breakdown
rate of competitive regimes has been very low with presidential systems still in
place throughout Latin America. Presidentialism does not help explain variance
across the twenty countries of Latin America or over time in the region because
presidential systems have been a constant.
In another well-known argument based on formal institutional rules, Shugart
and Carey (1992) theorized that systems with strong constitutional powers
for the president might be more vulnerable to breakdown than those with
more balance between the executive and legislature. In a converging argument,
Fish (2006) argued that strong legislatures are good for democracy. Although
these arguments about the impact of formal institutions are intuitively sensible,
and although Fish’s claim has solid empirical support for a broad sample of
countries, this hypothesis does not hold up for Latin America. On this point,
our skepticism is grounded in empirical observations rather than theoretical
conviction. In Latin America, the constitutional powers of presidents actually
expanded during the third wave of democratization (Negretto 2009). As a result,
in the models presented in Chapter 4, greater presidential powers enhanced
democratic survival.
Although formal rules shape actors’ incentives and behavior, their impact on
regime survival or fall is mediated by many other factors that seem to have more
weight than the formal rules. The willingness of actors to accept policy losses
does not depend directly on the formal rules of the game, and it has an important
294 Democracies and Dictatorships in Latin America

impact on the capacity of competitive regimes to survive. Intransigent actors

stretch their legal prerogatives to the limit (and beyond it) in order to impose
their preferred policies, and they seek to undermine the power of veto players by
casting them as illegitimate institutions. By contrast, non-radical players accept
the existing institutional design as exogenous and bargain to achieve their policy
preferences within the constraints imposed by those rules. In its focus on actors,
our theory is fully consistent with institutional approaches to regime change and
survival. But we focus on organizational actors (parties, militaries, unions, etc.),
not formal institutional rules.

looking ahead
The inability of these alternative theoretical approaches to account for the
historical transformation of political regimes in Latin America may portend
well for the region. By 2010, at least ten of the twenty Latin American countries
remained below the income level of Argentina in 1976, identified by Przeworski
et al. (2000: 98) as the threshold above which “no democracy has ever been
subverted.”25 If modernization were the main source of inoculation against
coups, most Latin American competitive regimes would still be at risk.
Latin America also remains one of the most unequal regions in the world.
Data compiled by the United Nations Development Program in its 2011 Human
Development Report indicated that the richest 20 percent of the population in
the typical Latin American country earns sixteen times more than the poorest
20 percent. As a comparative reference, the mean ratio between the richest
and the poorest quintiles of the population is about nine times for countries at
high levels of human development, eleven times for countries at medium levels
of development, and ten times for countries at low levels of development.
On average, Latin American countries lost nine positions in the international
ranking of human development once income inequality was taken into
account.26 Even though a combination of social policy, leftist governments,
and commodity booms led to an improvement of income distribution in the
last decade covered by our study (Gasparini and Lustig 2011), prospects for
democracy in Latin America would be bleak if inequality was an insurmountable
threat to competitive politics.
Most of Latin America remained shielded from the recession that undercut
the U.S. and EU economies in the years after 2008. Estimates by the Economic

25
Nominal GDP has risen over time, but the comparison refers to income measured in constant
(2005) purchasing power parity dollars (data from the Penn World Table 7.0 for 2009). Using the
figures in our dataset (in constant dollars, but not PPPs), some eighteen countries still remain
below the threshold.
26
Adjustments for income inequality in the HDI world ranking ranged from a loss of twenty-four
positions for Colombia to a moderate gain of three positions in the case of Nicaragua (http://hdr.
undp.org/en/media/HDR_2011_EN_Table3.pdf).
Rethinking Theories of Democratization 295

Commission for Latin America and the Caribbean (ECLAC) indicate that the
economy of the average Latin American country grew by one-third between 2004
and 2010, and by 8 percent even in the difficult global environment experienced
between 2008 and 2010.27 Yet at the turn of the decade, Latin American growth
often remained volatile and dependent on primary export booms, inflation
emerged as a pressing issue in several countries, and the typical unemployment
rate fluctuated around 8 percent.
There is no clear evidence that Latin American leaders were savvier, more
prudent, or more inclined to act as statesmen by 2010 than they were two
decades earlier. The legacy of past leaders who navigated the stormy waters of
democratic transitions, such as Raúl Alfonsín in Argentina or Patricio Aylwin in
Chile, or those who tamed hyperinflation, such as Fernando Henrique Cardoso
in Brazil, reminds us that Latin American leadership has always included a
good measure of vision and talent, as well as – in more unfortunate instances –
short-sightedness and negligence.
Presidential institutions will remain a feature of Latin American politics for
years to come. Some constitutional rules that presumably compound the effects
of presidentialism have even expanded over time. Repeated constitutional
reforms have extended the legal prerogatives of Latin American presidents
(Negretto 2013). Constitutional amendments (or acts of judicial review) have
also relaxed restrictions on presidential reelection to accommodate the ambi-
tions of popular incumbents in Argentina (1994), Bolivia (2008), Brazil (1997),
Colombia (2005), Costa Rica (2003), the Dominican Republic (2002), Ecuador
(2008), Nicaragua (2009), Peru (1993), and Venezuela (1999, 2009). If extra-
ordinary leaders or particular institutions were necessary to sustain democracy,
the future of competitive regimes in the region would be uncertain.
By contrast, normative regime preferences, policy orientations, and interna-
tional forces changed over the long run in ways that made Latin American
political actors more willing to accept democracy by 2010 than at any previous
point. If the argument presented in this book is correct, this fundamental trans-
formation involving organizational ideas and collective goals, transnational
networks, and international organizations anticipates a more promising future
for democrats in the region than most alternative theories would predict.
At the same time, there are reasons to temper this optimism with caution.
Chapter 8 documented a slight increase in radical policy preferences and a modest
decline in normative commitments to democracy since the late 1990s. It also
showed that investments in the construction of democratic institutions (or the
lack thereof) have lasting consequences for the quality of competitive regimes
over the long run. In this context, democratic stagnation and erosion have been
common phenomena. A surge in radicalism could have deleterious effects for the

27
Anuario Estadístico de América Latina y el Caribe (2011), table 2.1.1.1 (http://www.eclac.cl/
publicaciones/xml/7/45607/LCG2513b.pdf).
296 Democracies and Dictatorships in Latin America

strength of twenty-first-century democratic coalitions in the countries plagued by

weak states, bad governance, and social exclusion.
These findings open an exciting research agenda that we can only begin to
sketch. Studies of democratization have usually relied on theories that invoke
the power and motivations of specific actors, but they have tested their theories
using aggregate cross-national data at the country level. This inconsistency
between the level of analysis invoked by the microfoundations of the theory
and the units of analysis employed for hypothesis testing is common in other
subfields – including comparative studies of political economy, conflict, institu-
tions, and policy making – and it poses three important challenges.
The first challenge that future studies of democratization (and other subfields)
must confront is the generation of systematic indicators to portray political
actors in multiple countries and different historical periods. We addressed such
measurement issues in ways described in Chapter 3. Large collaborative efforts
would be required to develop accurate measures for political actors – their
preferences, goals, and resources – worldwide. Yet the payoff of such large-
scale undertaking for the social sciences could be great.
Previous chapters have shown that normative orientations and policy prefer-
ences have powerful consequences for regime change and stability. But once
those preferences are identified, several questions emerge. Where do these pref-
erences originate? How do they change? Under what conditions certain prefer-
ences spread in society? The second challenge for comparative politics is to take
those questions seriously. We explored the origin and changes of normative
preferences in Chapters 2 through 6, partly to dispel concerns about endoge-
neity, but a full treatment of this issue transcends the scope of this book. This is
an area in which interpretive and positivist approaches in political science will
need to engage in a joint effort (Bowles 1998).
Third, we need to extend our models of how actors’ preferences aggregate
into collective outcomes. The theoretical literature has addressed this issue in
many ways (for a classic example, see Schelling 1978), but empirical estimators
to model such processes have lagged behind. Hierarchical models conventionally
assume that variance in the outcome variable takes place at a lower level of
aggregation than variance in explanatory factors (Raudenbush and Bryk 2002).
Yet the combination of actors’ preferences in social outcomes presents the
opposite situation. We handled this issue by averaging actors’ preferences at
the country-year level in order to create summary measures. More powerful
estimators of aggregate choices may become available in the future.
These analytical challenges comprise an agenda that transcends the study of
regime change and has broader implications for the field of comparative politics.
We started this book in search of an explanation for the emergence, survival, and
fall of democracies and dictatorships in the past. We end this book by looking
ahead – to the perils to be met by future democratic actors in Latin America, and
to the questions to be met by future social scientists seeking to understand them.

View publication stats