ABSTRACT
Title of Dissertation: DICTATORSHIP IN TIMES OF DEMOCRACY:
THE ROLE OF ELITES, WEALTH AND INSTITUTIONS
IN REGIME CONSOLIDATION AND CHANGE
Joan Carreras Timoneda
Doctor of Philosophy, 2019
Dissertation directed by: Professor Ernesto Calvo
Government and Politics
The recent revival of authoritarianism and worldwide losses in democratic quality merit a
reassessment of current theories of regime consolidation and change. In this dissertation, I
provide four interlinked theories of authoritarian consolidation, authoritarian replacement,
democratization and democratic erosion. First, I argue that dictators consolidate power after
crises by expanding the ruling coalition and using formal institutions as signaling devices.
Doing so weakens the coordination capacity of the plotters and reduces the likelihood of a
coup. I take a political economy perspective in the remaining chapters, showing first that
authoritarian turnover increases when wealth levels in the elite are low. This generates in-
centives for elite groups to control the state to capture rents. However, as wealth increases,
elites become concerned with protecting their wealth instead of fighting each other over
state rents. Democracy ensues as it provides a system of rule of law that guarantees prop-
erty rights. Lastly, I provide a theory of democratic erosion based on the notion of a pro-
ductivity gap between economic sectors: democracy’s equilibrium breaks down when low
productivity sectors seek to control the state and extract rents over reinvesting in more pro-
ductive enterprises. This productivity imbalance creates the space for a ‘Caesarian’ leader
to emerge. To probe these theories, I use a mixed methods approach that includes formal
modeling, various quantitative techniques and case studies of the Dominican Republic, the
United States and Spain.
1
DICTATORSHIP IN TIMES OF DEMOCRACY:
THE ROLE OF ELITES, WEALTH
AND INSTITUTIONS IN REGIME
CONSOLIDATION AND CHANGE
by
Joan Carreras Timoneda
Dissertation submitted to the Faculty of the Graduate School
of the University of Maryland, College Park, in partial
fulfillment of the requirements for the degree of
Doctor of Philosophy in Political Science
2019
Advisory Committee:
Professor Ernesto Calvo, Chair
Professor Mark I. Lichbach
Professor Isabella Alcañiz
Professor Calvert W. Jones
Professor David M. Waguespack (DR)
©c Copyright by
Joan Carreras Timoneda
2019
Preface
I didn’t enter graduate school planning to write on regime change, democratization, or why
democratic quality is seemingly eroding everywhere. Rather, I wanted to stay intellectually
stimulated and, most importantly, have complete power to decide which issues I would ex-
plore. This freedom came at the cost of a modest graduate school stipend –which shrinks
even further in an expensive area like Washington, DC. Any rational cost-benefit analysis
would discourage pursuing the path I (and so many others) chose, but as I found out before I
started my Ph.D., the price of intellectual freedom is difficult to quantify and hard to appreci-
ate. In Barcelona, after I finished my previous studies, I started and ran a satirical magazine
in Catalan for over three years. We didn’t change the world but we learned to appreciate the
true value of our freedom of thought and expression, of being able to form criticisms, pub-
lish them and engage in discussions with close friends and complete strangers. Sadly, many
of the articles and cartoons we published then could land us in trouble now –the Spanish
government has passed new laws and began to enforce old ones that severely restrict cer-
tain expressions of dissent, primarily those attacking the crown. This downward trend only
reinforces my belief that, in the end, there is no price to freedom. For me, academia is a
refuge, a cocoon where we can express ourselves to the fullest extent in a rigorous way (at
least we genuinely try). Its greatest gift is that it helps us preserve our intellectual freedom
like no other sector does.
From this I think I derive my greatest fascination: to borrow from the great John Oliver,
why is autocracy still a thing in 2019? If, as I believe, there is no price for freedom, why do
authoritarian rulers continue to survive and why do people accept them? There may not be
a clear answer to this question, and I certainly don’t provide a good one in this dissertation.
The way I have decided to approach answering this question is to study regime change. What
makes dictators stay longer in power? What brings about a transition to democracy? What
makes democracy weaker, shrinking my ability to, for instance, openly mock the Spanish
crown in print? This dissertation is my first stab at answering some of these questions.
I wouldn’t be mindful about the importance of my intellectual freedom and my liberty
to speak out without people that did the same before me. In retrospect, the greatest gift I
ever received was the daily political discussions I had with my family since I was a kid.
We all know at home that my grandfather is mostly at fault for that. He may not be able
to read it, but to him I dedicate this imperfect dissertation, of which I am sure he is proud.
Another special dedication goes to my mother, my rock and staunchest advocate –even in
ii
times when she shouldn’t be. Brilliant and unwavering in her beliefs, she was the first Ph.D.
in the family, and I am proud to say I am now the second one. My grandmother, sweet and
much more street smart than me; my uncle, as headfirst, intelligent and caring as they come;
and Martı́, supportive and generous in ways I can only aspire to be, also deserve a heartfelt
dedication and a sincere thank you.
Academically, I owe my deepest gratitude to Ernesto Calvo and Mark Lichbach, who
dedicated their time and energy to my cause with a level of commitment that is any gradu-
ate student’s dream. Their unwavering support and constant guidance helped me grow pro-
fessionally and personally in ways I will only understand with the passage of time. I am
indebted to Michael Hanmer, who was instrumental to my progress and well-being as a
Ph.D. student. I also extend a heartfelt thank you to Calvert Jones and Isabella Alcañiz for
their feedback and thoughtful contribution to this dissertation; to Bill Reed for his help with
game theory; to Joe Wright for his support, feedback and thoughtful conversations during
my time at Penn State; to Patrick Wohlfarth for being wonderful to his TAs; to Mike Spivey
for the camaraderie and good times teaching undergraduate students; and to Ann Marie for
her endless patience as I navigated the bureaucracy. Last but not least, a note of appreci-
ation to my fellow grad students for endless conversations, for coming to my rescue when
I needed it, and for our evenings together in DC and College Park. A special thank you to
Sebastian Vallejo, friend and co-author, for being an example in academia and in life; to Jori
Breslawski for being a friend in good and bad times; and to Patrick Tiney for getting me out
of my office when I worked too late on Fridays. And, finally, a thank you to someone who
enriched my life more than anyone else and without whom this dissertation would never
have been completed. My accomplishments now and in the future will continue to be partly
yours.
iii
Contents
Preface ii
List of Tables vi
List of Figures vii
1 Introduction 1
2 Institutions as Signals: How Dictators Consolidate Power In Times of Crisis 11
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Extant Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Institutions Under Trujillo In The Dominican Republic . . . . . . . . . . . 33
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3 Et tu Brute? Wealth Inequality and Authoritarian Replacement 41
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3 Empirics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.6 Robustness Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.7 The Rise of Trujillo in the Dominican Republic . . . . . . . . . . . . . . . 72
3.8 Implications and Concluding Remarks . . . . . . . . . . . . . . . . . . . . 76
iv
4 Cashing In: Elite Cleavages, Wealth Inequality, and Democratization
with Sebastián Vallejo 80
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.2 Democracy as insurance . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.3 Inequality, Repression and Democratization . . . . . . . . . . . . . . . . . 87
4.4 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.5 Empirical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.6 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.7 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.8 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.8.1 Alternative Hypotheses and Robustness Checks . . . . . . . . . . . 111
4.9 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5 The Productivity Paradox: How Sharp Increases in Productivity Motivate Rent-
seeking and Erode Democracy 115
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.2 Democratic erosion in developed economies . . . . . . . . . . . . . . . . . 119
5.3 Shocks, economic inequality and the rise of a political entrepreneur . . . . 124
5.4 Economic elites and why Caesarian leaders triumph . . . . . . . . . . . . . 127
5.5 Wealth, the productivity gap, and elite polarization . . . . . . . . . . . . . 130
5.6 Why Caesarian leaders subvert democracy . . . . . . . . . . . . . . . . . . 132
5.7 Empirical Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.8 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
5.9 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.10 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
5.11 Wealth Inequality and Democratic Erosion in Spain . . . . . . . . . . . . . 140
5.12 Wealth Inequality and Democratic Erosion in the United States . . . . . . . 146
6 Appendices 151
6.1 Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
6.2 Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
6.3 Appendix C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
References 170
v
List of Tables
3.1 Effects of Wealth Inequality on All Transitions Conditional on Levels of Wealth . . 69
4.1 Effect of Wealth Accumulation and Income Inequality on Democratization . . . . 108
5.1 Effect of the productivity gap and inequality on democratization . . . . . . . . . 137
5.2 Effect of the productivity gap and inequality on democratization using Solt’s Gini. 141
A1 Snippet of the Data with Examples from The USA and Cuba (after expanding)152
A2 Summary statistics (before expanding) . . . . . . . . . . . . . . . . . . . . 153
C1 Summary statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
C2 Dynamic Probit results using Houle’s Capital Shares variable . . . . . . . . 162
C3 MI Estimate results using Solt’s disposable income Gini. . . . . . . . . . . 164
C4 Replication of Acemoglu and Robinson’s (2001) Table 4 using BMR’s di-
chotomous measure of democracy as DV. . . . . . . . . . . . . . . . . . . 165
C5 Replication of Acemoglu, Johnson and Robinson’s (2001) Table 4 using
VDem’s continuous liberal democracy index as DV. . . . . . . . . . . . . . 166
vi
List of Figures
1.1 All possible transitions between democracy and authoritarianism, plus sta-
bility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.1 Signaling Model In Extended Form . . . . . . . . . . . . . . . . . . . . . 26
3.1 All Possible and Recursive Transitions . . . . . . . . . . . . . . . . . . . . 61
3.2 Observed probability in the data of replacement at different levels of wealth 68
3.3 Joint effect of wealth and inequality on authoritarian replacement (a). Sur-
vival curves at high levels of inequality (b). . . . . . . . . . . . . . . . . . 70
4.1 Democratization as a function of wealth and inequality . . . . . . . . . . . 98
4.2 Frequency of Transitions to Democracy at Different Quintiles of Wealth and In-
equality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.3 Marginal effects of inequality on democratization at different levels of wealth ac-
cumulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.1 Evolution of V-Dem scores by country since 1995. . . . . . . . . . . . . . . . . 119
5.2 Evolution in the number of democracies from 1950 and 2005. . . . . . . . . . . 121
5.3 Effects of the productivity gap on democracy conditional on inequality. . . . . . . 138
5.4 Evolution of V-Dem scores by country since 1995. . . . . . . . . . . . . . . . . 142
A1 Marginal effect of decreasing levels of wealth on replacement at different
levels of wealth inequality, no controls. . . . . . . . . . . . . . . . . . . . . 154
A2 Joint effect of wealth and inequality on authoritarian replacement. Displays
the . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
A3 Joint effect of wealth and inequality on authoritarian replacement. Displays
the . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
A4 Joint effect of wealth and inequality on authoritarian replacement. Displays
the . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
vii
B1 Democratization as a function of wealth and inequality . . . . . . . . . . . 158
C1 Marginal effect of inequality on democratization at different levels of wealth,
using Houle’s Capital Shares data. . . . . . . . . . . . . . . . . . . . . . . 163
C2 Marginal effect of inequality on democratization at different levels of wealth,
using Solt’s Disposable Income Gini Index. . . . . . . . . . . . . . . . . . 165
C3 Evolution of Piketty’s elite wealth variable and our proxy by country. . . . . 168
C4 Marginal effect of wealth on democratization at different levels of inequal-
ity, using UNIDO inequality data as in the main chapter. Table 1 is from
the main text. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
viii
Chapter 1
Introduction
History is not yet over. Autocracies continue to stabilize and strengthen, without any indic-
ation of a growing willingness to democratize. Worryingly, democracy is eroding across
the globe. We are observing a degradation of civil and political rights in advanced and new
democracies alike, and the likelihood of authoritarian reversals increases with every new
election cycle. Indeed, democracy had not been so fragile since the end of the cold war at
the very least, and neither had reformed, economically-open dictatorship been so en vogue.
In twenty years, we may be living in a world with strong democracies that overcame a period
of hesitation to solidify and strengthen citizen’s civil and political rights. Or, perhaps, a ma-
jority of humanity may be governed by illiberal regimes that compromise on citizen rights in
order to stay in power, using the need for law and order and an old, unimaginative nationalist
discourse to rally a majority political coalition.
This predicament will be object of study in comparative politics for years to come, as it
affects many of its debates and subareas of study. For instance, regionally, Europe and Latin
1
America are particularly hard hit by the recent downtrend in democracy: the oversize gains
made in the 1980s and 1990s, a period during which virtually every country democratized,
are now called into question in Brazil or Poland, key economic and geopolitical centers. A
reversal in Poland would bring a new iron curtain right at the German border; the loss of
Brazil would create a dangerous precedent for neighboring countries dealing with weakened
institutions, such as Argentina. In political economy, democratic erosion is challenging the
very notion that democracy is impregnable at high levels of wealth, which has been more or
less considered a stylized fact in the discipline for twenty years. Similarly, autocracies are
gaining stability even as they modernize and develop a business class, which runs counter
to modernization theory and old arguments about state development and democratization.
The list here goes on: the study of civil and political rights, protest and social movements,
institutions, voting behavior, and so on.
In this dissertation I provide a theory of elite driven regime change that helps answer
some of these puzzles. I take a strong stance that regime transitions are driven by elites,
whose conflicts and agreements shape the outcome of events such as democratization, au-
thoritarian consolidation and replacement, and democratic erosion. This idea was at the
core of the transition paradigm of the 1980s and 1990s, which described democratization
processes as originating in a cleavage between hardliners and softliners within the heavy-
handed ruling elite of Latin American military dictatorships. A push by civil society led
to democratization, but the key cog in the democratization engine was the receptiveness
of softliners to liberalization and openness.1 The literature on inequality and democratiza-
1Carothers, 2002; O’Donnell and Schmitter 1986; Przewroski 1991.
2
tion, on the other hand, has focused primarily on the relationship between the elite and the
masses, arguing that transitions are a result of redistributive class conflict.2 In this disserta-
tion I bring back elite conflict into the debate around regime change and apply to a broader
spectrum of political transitions: authoritarian replacement and stability, democratization,
and democratic erosion. I also incorporate the concept of wealth into the regime change
discussion, and prod the interactive effect of inequality with levels of wealth on different
transition outcomes. These constitute the main theoretical contributions of this disserta-
tion.
Each of the chapters of this dissertation develops the theory as it relates to a particular
type of transition. In Chapter 1, I explore how dictators consolidate power in times of
crisis by expanding their coalition and using formal institutions. In chapters 2 and 3, I
take a political economy approach that tries to challenge some of the traditional arguments
in the inequality/democratization debate. In Chapter 4 I provide a theory of democratic
erosion based on the productivity across sectors and polarization within the economic elite
–as opposed to classic arguments that focus on polarization among voters or the political
elite. The common thread in this dissertation is the study of regime stability and change.
Figure 1.1 illustrates all possible transitions and states. This dissertation focuses on four
of these: authoritarian stability, authoritarian replacement, democratization and democratic
erosion. Extensions to democratic stability and breakdown will follow in the future.
2Boix 2003, Acemoglu and Robinson 2006.
3
Stability
Democracy1 Democracy2
Erosion
Democratic
Breakdown Democratization
Stability
Authoritarianism1 Authoritarianism2
Replacement
Figure 1.1: All possible transitions between democracy
and authoritarianism, plus stability.
In Chapter 1, I introduce the reader to a discussion in comparative politics regarding the
role of political institutions in dictatorship and their effect on authoritarian stability. The
literature has tended to agree that formal and seemingly democratic institutions, such as
political parties and parliaments, improve dictatorial stability by fostering power-sharing
agreements that place some constraints on autocrats.3 Dictators use them to co-opt op-
position groups and share governance with them.4 Indeed, power-sharing has been found
to increase authoritarian stability.5 As for political parties, Brownlee has shown that large
parties tend to have a binding effect on elite coalitions, which otherwise could fractionalize
and destabilize regimes.6 Singe party regimes are also the longest-lasting.7
I take a different view from the general consensus in the literature. I contend that formal
institutions are used as signaling devices by autocratic leaders in times of weakness. They
often emerge as a consequence of the leader expanding his ruling coalition to dissuade
3Wright 2008.
4Gandhi 2008.
5Gandhi 2008.
6Brownlee 2007.
7Geddes 1999.
4
plotters from launching a coup. In times of weakness, such as the outset of tenure of after
an important shock, incumbent autocrats face an increased threat of a coup.8 This entails
that, within their ruling coalition, a certain faction usually develops that plots to overthrow
the ruler. The success of a coup depends on having a majority of the coalition in favor of
it9 and on how information propagates during the brief period between the start and end of
a coup.10 One of Singh’s major insights is that there exist major informational asymmetries
within the dictator’s winning coalition, and that the fate of a coup often depends on how the
leader is perceived while the coup unfolds. New information may sway sufficient numbers
to join the coup even though they would not have done so before. The implication of this is
that controlling the discourse and information channels is as important, at least, as having
the support of at least half of the ruling coalition.
I argue that dictators oversize their coalition to break down the coordination capacity of
the plotters, thus attempting to reduce support below fifty percent –or, at least, increase un-
certainty regarding members’ allegiances. Institutions, for their part, alter the information
game with the broader coalition and broadcast support to the rank-and-file. The latter ob-
serve greater commitment by key elites to the current leader, making it less likely that they
will join a coup if launched. I introduce a formal model to show the logic and implications
of this argument.
In the remaining theoretical chapters –2, 3 and 4–, I investigate the role wealth inequal-
ity plays in authoritarian replacement and democratization from the point of view of elite
8Little 2017.
9Little 2017.
10Singh 2014.
5
cleavages. I theorize that elites decide, at each moment in time, which system best suits their
long term prospects for wealth accumulation. This is based on an intuition that economic
elites are at the front and center of regime change, and that their political affiliations tend
to shift with changing circumstances. Indeed, Barone and Mocetti have recently found that
the wealthiest families in the Italian city of Florence are descendants of wealthy families
600 years ago.11 These families have gone through major political events: the vagaries of
Medici rule, inclusion the Austrian crown, French annexation, and finally the unification of
Italy. How do the wealthy stay wealthy, and even expand their wealth, through long periods
of time and political change? I argue that elites have one goal, to accumulate wealth in the
long run. In the short and medium term, however, they adapt their political preferences and
favor the institutional arrangement that best serves their long-term objectives. They do so
even if that means sacrificing certain immediate types of wealth and power. I call this logic
cashing in: elites play nested games of wealth maximization and may appear, at times, to
make a suboptimal choice in the short term, such as allow greater redistribution or political
instability. This is, however, perfectly rational, as it yields the optimal payoff in the long
run. Elites, I argue, cash in their gains after each nested game and begin anew. The logic
of cashing in mirrors that of Tsebelis, who proposed the notion of nested games as an an-
swer to the riddle that political actors often fail to make the apparently correct choice.12 A
key contention of this dissertation is that, by shifting our attention to elite splits and con-
flict, we can detect many of the nested games that lead to counterintuitive outcomes such as
democratization.
11Barone and Mocetti 2016.
12Tsebelis 1990.
6
With this broader idea in mind, I argue that when elite wealth is low, different factions
seek to control the state to appropriate income. This leads to greater authoritarian replace-
ment, as the state and its resources provide sufficiently large rents to increase wealth rapidly.
Enclave economies and resource-rich countries with low levels of overall wealth are good
examples of this process: elites with relatively low levels of wealth have a substantial incent-
ive to appropriate these resources and become richer. Once a majority of elites are wealthy,
incentives for conflict decrease and stability is preferred.
A common counterargument here is that, in states where replacement is common, insti-
tutions are weak and easily captured. With stronger institutions, low wealth would not be
the driving force behind authoritarian replacement. In reality, this argument treats institu-
tions as exogenous to the process of wealth accumulation, and they are not. Institutions are
a function of the people that support them, and hold sway over elite and citizen relations
only insofar as they are viewed to have legitimacy. The counterfactual that strong institu-
tions could exist in countries with low levels of elite wealth is illogical, precisely because
strong institutions emerge to arbitrate complex interactions among political subjects who
have an incentive to compromise with other, politically equal subjects. When wealth is
low among the elite, incentives to cooperate are minimal. Only as wealth increases there
is a chance for better institutions. Indeed, the natural resource curse is only such in places
where elite wealth is overall low, and where intra-elite inequality is high –that is, some elite
groups are ultra-wealthy and many others and only above a certain income to qualify them
as elite. Where wealth is high across the board, natural resources abundance is rarely a
7
curse. Moreover, as Chapter 4 shows, even strong institutions fray when inequality among
economic elites increases sharply.
The question then becomes: what happens as wealth increases? Elites have greater in-
centives to secure their wealth, which requires political stability and strong property rights
guarantees. The argument in Chapter 3, therefore, is that democratization transpires when
elites have accumulated high levels of wealth and come to regard democracy as an insurance
mechanism to secure their wealth through the rule of law. In this view of democracy as in-
surance, elites prefer to pay a risk premium in democracy through higher taxes in exchange
for security against expropriation by a competing elite and revolution. In Chapters 2 and 3
I develop further on two key points: first, why elites stop replacing each other in power as
their wealth grows and, second, why democratization ensues as wealth increases instead of
stable authoritarianism.
Lastly, in Chapter 4 I develop a theory of democratic erosion that maintains the logic of
elite cleavages and wealth from the previous two chapters but incorporates a new concept:
the productivity gap. I argue that democratic erosion occurs when a large productivity gap
leads to a wealth cleavage among elites. Elites in low productivity sectors see their current
wealth decline and their future prospects become grim. If the gap between low and high
productivity sectors is large enough, reinvesting wealth into more productive enterprises is
exceedingly costly for stagnant sectors, who have an incentive to seek rents from the state
instead. This gap provides a space for a political entrepreneur to gain the backing of low
productivity sectors and take over the state.
8
Methodologically, this dissertation a diverse array of methodologies to provide credib-
ility and probe its theories. Chapter 1 uses a formal signaling model to show that dictators
expand their coalition through formal institutions in times of weakness to prevent a coup.
For Chapters 2 and 3, I have created a cross-national measure of elite wealth accumulation
per country-year using available data from multiple sources. In Chapter 2 I use multi-state
survival analysis to probe the hypotheses that the likelihood of authoritarian replacement
increases at low levels of wealth and high levels of inequality. In Chapter 3, I use dynamic
probit and linear probability models to show that the probability of a transition to demo-
cracy increases with higher levels of wealth and income inequality. Chapter 4 uses similar
statistical techniques. Lastly, case study evidence is important in this mixed-methods dis-
sertation. Chapters 1 and 2 use the Dominican Republic as a case study, delving into the
country’s history before and during the rule of Rafael Trujillo. Chapter 4 uses the cases of
the United States and Spain to illustrate the current trend toward democratic erosion that
plagues advanced democracies.
My work contributes to the field’s understanding of the dynamics of authoritarian re-
placement from a political economy lens. While work abounds on predatory states, resource-
rich dictatorships, and coups, little work has attempted to systematize the structural causes
of authoritarian replacement. The theory I develop furthers our understanding of these struc-
tural factors. My work also contributes to the shift in the democratization literature toward a
vision of democracy as a positive outcome for authoritarian elites, which they often actively
pursue. I add a political economy angle to this debate by describing how high wealth in-
equality and rapid wealth accumulation promote democracy. Moreover, the research shows
9
that democratic erosion in developed economies is the result of higher inequality and rapid
wealth accumulation. This argument contributes a political economy logic to the nascent
debate around democratic erosion and breakdown. A methodological contribution of my
work is to incorporate multi-state survival analysis into the study of regime transitions.
10
Chapter 2
Institutions as Signals: How Dictators
Consolidate Power In Times of Crisis
Formal institutions in dictatorship, such as political parties and legislatures, are
known to improve authoritarian governance, place constraints on the leader, and
enhance survival by promoting power-sharing. Much less is known about how dic-
tators use institutions to signal strength to the opposition. In this chapter, I advance
a theory of power consolidation in dictatorship that conceives of formal institutions
purely as signaling devices. I argue that, at times of weakness, dictators often follow
an expand-and-signal strategy, whereby they expand the ruling coalition to decrease
the power of potential plotters within it and then create visible formal institutions
to signal strong support to the broader coalition and the public. Doing so directly
decreases (1) the probability that a coup is launched against them and (2) that one
succeeds if launched. The formal model I propose shows that strong leaders who face
a crisis have an incentive to expand the coalition and create institutions, even if they
know they hold the support of a majority of the coalition before a coup is launched.
I use the case of the Dominican Republic before and during Rafael Trujillo’s rule to
illustrate my theory.
11
2.1 Introduction
In 1930, after a perfectly orchestrated coup, Rafael Trujillo began his thirty-year long ten-
ure as self-proclaimed benefactor of the Dominican people. He moved quickly to consolid-
ate power, creating a political party with an expansive regional and national apparatus, the
Partido Dominicano, and strengthening the role of the country’s Senate. Rural supporters
were given important posts in the party, and loyal elites became Senators. In essence, what
Trujillo did with celerity and skill during the first six months of his rule was to (1) expand
his support coalition and (2) use formal institutions to accommodate the expansion. At a
time when he apparently faced no credible challengers, why did Trujillo chose to develop
these formal institutions rather than focus on establishing a strong army, secret police, and
use the repressive tactics that became common in the latter part of his rule?
Formal institutions such as political parties and legislatures are known to improve author-
itarian governance and place at least some constraints on authoritarian leaders.13 They also
enhance survival by promoting power-sharing.14 By incorporating and co-opting different
elites into the regime’s policy-making process, autocrats can enjoy more stable tenures and
better policy outcomes. Indeed, I do not intend to rebuke these assessments. Rather, my
argument is that they are incomplete and that we have not yet theorized about the full range
of reasons why formal institutions in dictatorship are so common and useful to autocrats.
The important intuition from Trujillo’s example, and which constitutes the central claim of
this chapter, is that the institutions he created served as highly effective signaling devices.
13Gandhi 2008; Wright 2008.
14Boix and Svolik 2013.
12
In times of particular weakness, dictators expand their coalition to reduce the relative
power of potential plotters within it, and use institutions to communicate strength to the
rank-and-file. Since the spread of information and coordination among the coup plotters
are crucial for coup success,15 dictators can reduce both the threat that a coup occurs and
the probability that one succeeds if staged by employing the ‘extend and signal’ strategy
that I outline in this chapter.
The reasoning behind expanding the ruling coalition is especially counterintuitive. In
order to reduce the relative power of the potential coup-plotters within the coalition, leaders
incorporate new members with whom plotters will have difficulty coordinating, thus redu-
cing the probability that a coup occurs. An example illustrates my logic best: If, say, a
support coalition consists of ten people, and five of them are plotting against the dictator
without his knowledge, the dictator is one defection away from being deposed. However, if
he expands his coalition to fifteen members by adding five new members, he weakens the
position of the plotters, who no longer have the required numbers to win. The leader pays
a steep cost if he expands: current members of the coalition have to share the pie with new
members, and therefore loyalists may be more likely to defect.16
Leaders face two critical problems when they oversize the coalition. First, they cannot
credibly expect new members to remain loyal indefinitely. The dictator needs these new
members to send a credible signal of commitment to him, become invested in the regime,
15Little 2017; Singh 2014.
16Note that, in this scenario, three of the five old loyalists need to defect for the leader to lose. Provided the
leader does not reduce their rents by a large amount, they are likely to remain loyal. What is important here
is that strong leaders will be willing to pay this cost, and the stronger they are, the higher the cost they can
afford. This will have important implications in the game.
13
and tie their fortunes to the fortunes of the leader. Second, he needs to communicate this
gain in power to the broader winning coalition. Oversizing the coalition is much less useful
if the rank-and-file are unaware of the dictator’s newfound support. This is particularly
important, as we know that information plays a key role in determining which actors join
a coup and which do not.17 The dictator must prevent key players, such as ranking officers
and soldiers, from joining the coup if one is staged. Informational asymmetries may lead
officers to join a coup that may, unbeknownst to them, not have been initially viable. As
Little shows, a coup attempt is likely forthcoming at the beginning of a leader’s tenure.18
It is important for the leader to try to affect the officer’s belief about his strength and make
them less likely to join a coup if one takes place.
Key power-sharing institutions, political parties and legislatures, are useful for the dic-
tator to solve these two problems. First, accepting a public position within an institution,
such as vice-president of the senate or member of parliament, sends a signal of commitment
and loyalty to the leader and carries a cost for the new member. The more closely she is
associated with the regime, the more likely she will be removed or vanished should a coup
occur and succeed. In this way, formal institutions reduce the probability that a coup takes
place. Second, by expanding the coalition through public formal institutions, the leader
communicates strength to the broader coalition, who perceive a signal that the leader is
stronger than they thought he was. This makes officers less likely to defect and thus reduces
the probability that a coup succeeds if staged.
17Singh 2014.
18Little 2017.
14
The expand and signal strategy can be superior to other traditional strategies to garner
support, such as distributing rents or purges. The former suffer from a credibility problem,
since the leader can withdraw funds at any time and the supporters can end their support
whenever they can obtain more elsewhere. The latter, purges, carry a high cost in terms of
popularity and support, since more insecurity may reduce other members’ payoffs of staying
in the status quo and thus induce a coup rather than stifle it. A particularly insidious cost
emerges when purging is used in spurts, as described, for instance, in Svolik: since the
leader cannot be certain about the identity of the plotters, he may actually purge loyalists
who have come under suspicion.19 This may strengthen the chances that a potential coup
succeeds. Expanding the coalition, even though it forces current loyalists to share some of
the pie of resources with new entrants, is a less costly and potentially more effective strategy.
This chapter contributes to the literature on authoritarian governance and power-sharing.
First, by identifying the role of institutions as signaling devices, we can better understand
how they originate and why they vary at different points in time. For instance, it provides an
alternative explanation for co-optation: a leader may co-opt a rival faction in order to almost
instantly reduce the relative power of another rival faction. If these factions are at odds, their
capacity to coordinate is low and, as a consequence, the dictator consolidates power. Institu-
tions then signal this to the broader coalition. Second, modeling how dictators consolidate
power in times of evident weakness, for instance, at the outset of tenure, is an important con-
ceptual shift. While extant work does consider the creation of formal institutions in times
of crises, conclusions are often related to long-run governance – such as policy outcomes
19Svolik 2009.
15
or stability.20 Instead, I consider short-term solutions to short-term problems, and some of
the conclusions I draw differ from those of previous authors. Lastly, a recent trend in the
literature on coups is to model the coordination capacity of the opposition.21 In this chapter,
I complement these works and show how leaders may try to anticipate such coordination by
disrupting elite communication and the information that reaches the lower ranks.
This chapter proceeds in three stages. After a brief introduction to the relevant debates,
I introduce a standard signaling model that shows precisely that for a certain range of relat-
ively strong types of leaders, expanding the coalition and creating institutions is attractive.
I then present evidence from the Dominican Republic that illustrates the logic put forward
in this chapter.
2.2 Extant Work
The question of how authoritarian leaders consolidate power has received renewed attention
in the past decade. Scholars have tended to focus on two different approaches to the issue.
One looks particularly closely at the power dynamics within the regime itself. Conflict can
emerge when an autocratic leader and a set of elites within the regime fight over control
of executive action. Svolik has argued that a the leader has an incentive to accumulate
power in his hands and reduce the influence of the ruling coalition, whereas the latter seeks
precisely the opposite.22 The resulting tension can lead to a coup or a more entrenched and
consolidated leader. Others in this line of research have focused on the relationship between
20Gandhi 2008.
21Little 2017; Singh 2014.
22Svolik 2009.
16
the dictator and the military.23 The central dilemma here is that dictators may prefer a strong
military apparatus to fight an external threat, but a strong military in turn may find it easier
and less costly to conduct a coup. Powell, for instance, argues that placing structural barriers
to the coordination capacity of military personnel, such as fractionalizing different branches
of the military, reduces the likelihood of a military coup.24 The coordination problems
among the military have also been the focus of work by Geddes and Little.25
The second approach to the analysis of dictatorship consolidation has revolved around
the role of formal institutions in improving authoritarian governance, which falls within a
broader debate on authoritarian governance pioneered by Linz and O’Donnell, among oth-
ers, and reenergized more recently by seminal works by Levitsky and Way and Gandhi.26
One formal institution that has received much attention are legislatures. A key argument
in this line of research has been that power-sharing institutions such as legislatures help
dictators co-opt opposition groups into the regime’s policy-making process, which reduces
conflict, solidifies the position of the leader, and helps him obtain his preferred policy out-
comes.27 Boix and Svolik argue that authoritarian institutions increase regime survival by
facilitating power-sharing.28 Gandhi and Przeworski argue that dictators who need greater
cooperation make more policy concessions.29 Other work has helped deepen our under-
standing about the relationship between dictators and their legislative bodies. Wright ar-
gued that regimes dependent on domestic investment rather than natural resource revenue
23Acemoglu, Ticchi and Vindigni 2010; Beasley and Robinson 2010; Powell 2012.
24Powell 2012.
25Geddes 1999; Little 2017.
26see Linz 1975; O’Donnell 1973; Levitsky and Way 2002, 2010; Gandhi 2008.
27Gandhi 2008.
28Boix and Svolik 2013.
29Gandhi and Przeworski 2007.
17
create more binding legislatures, which serve as credible constraints on the leader’s abil-
ity to expropriate wealth.30 Wright and Escribà-Folch also show evidence that legislatures
make dictatorships more stable.31
Another relevant institution are political parties. The debate around the role of political
parties in dictatorship has reached less of a consensus than the debate around legislatures.
Brownlee, for instance, argues that “[r]uling parties ... bridle elite ambitions and bind to-
gether otherwise fractious coalitions.”32 Echoing Geddes, he finds that single-party regimes
tend to be the most long-lasting.33 Slater also ties the survival of authoritarian regimes to
the robustness of political parties, which are valuable mechanisms for elites to act collect-
ively.34 Opposed to this generally positive view of political parties in dictatorship, Wright
and Escribà-Folch argue that political parties, in fact, may have a deleterious effect on re-
gime survival in the long run.35 The authors claim that a strong political party can generate
incentives to democratize for authoritarian elites, who can find protection in a subsequent
democracy. Other work has explored the role of hegemonic parties, which serve to pre-
vent personal dictatorship and ensure long-lasting rewards for ruling elites through a vast
clientelism network.36
A contribution of this chapter is to provide a theory of leadership consolidation in au-
thoritarian regimes that brings together key insights from these two literatures. On the one
hand, I place strong emphasis on conflict between the ruler and his immediate elite circle,
30Wright 2008.
31Wright and Escribà-Folch 2012.
32Brownlee 2007, 33.
33See Geddes 1999.
34Slater 2010.
35Wright and Escribà-Folch 2012.
36Magaloni 2006.
18
rather than a more classic power struggle between a leader and an opposition group. Most
coups come from within the regime,37 hence the focus on this aspect of authoritarian polit-
ics. On the other hand, most of the literature has focused so far on the substantive value of
power-sharing institutions in authoritarian regimes, that is, the constraints that they actually
place on dictators to carry out their policy and the power it affords the opposition to shape
policy. Here I point to the instrumental value of institutions in authoritarian regimes. These
institutions can serve as powerful signaling devices, since they are visible and well-known
in society. They can be used to communicate a signal of strength by the leader to both his
most immediate elite circle as well as to the broader set of actors in the winning coalition.
Thus, I do not mean to directly challenge substantive arguments. I consider that they are,
in fact, complementary to the argument offered in this chapter and that, together, they can
give us a more comprehensive understanding of the relationship between dictators and the
power-sharing institutions under their control.
Indeed, it is important to point out that my argument in this chapter is not in support
of institutions as some form of façade for authoritarian leaders to display a false sense of
power-sharing. In fact, for the signal to be credible, institutions need to be given a raison
d’être, a political purpose. In some cases, legislative powers are weak and wholly depend-
ent on the dictator’s wishes. In others, legislative freedom gives members of parliament
and senators the prerogative to push a certain agenda and obtain actual concessions from
the leader. Either way, institutions require a purpose to serve as credible signals. Institutions
are designed to entrench new elites in their post, so that plotters have a more difficult time
37Svolik 2009; Little 2017.
19
convincing them to switch allegiance. That may mean, sometimes, as in Trujillo’s Domin-
ican Republic, that supporters sycophantically promote legislation to score points with the
dictator, such as the time when Mario Fermı́n Cabral proposed that the capital city of Santo
Domingo be renamed to Ciudad Trujillo, even if the benefactor himself never requested
the change.38 Other times, legislative freedom may lead to reform that forces the leader to
compromise.
My argument also fits well with recent literature on how coups develop from the point of
view of the opposition,39 which echoes earlier arguments by Geddes.40 Little explores the
coordination capacity of officers and the factors that will lead them to join a coup as global
game, and argues that a coup attempt is always forthcoming at the outset of a dictator’s
tenure.41 Singh shows that coup success depends on officers joining it once it’s already
in full swing, and information exchange is key to determine allegiances.42 The better the
plotters communicate their likelihood of success, the more likely they are to seduce potential
converts and succeed. My chapter analyzes how leaders try to thwart the two channels by
which the opposition may succeed in a coup.
First, by expanding the coalition, the leader prevents key top-level actors from launching
a coup in the first place, since their relative strength is lower and the capacity to coordinate
with new members is limited. Second, by using formal institutions as a public signal of
increased strength, the dictator manipulates the information mid- and low- level officers use
38Crassweller 1966.
39Little 2017; Singh 2014.
40Geddes 1999.
41Little 2017.
42Singh 2014.
20
to make the decision whether to join a coup once it starts. We know from Singh that coup
success is largely determined by how many actors joins the coup in the final stages, and the
leader has a substantial incentive to prevent such a snowball effect once he coup is in full
swing.43 If officers believe that the leader is weak, they are more likely to join. Similarly,
the more high-level officers join, the more likely are low level officers willing to join. An
expanded coalition dilutes this belief and makes it more likely that the rank-and-file believe
the leader is strong and the coup will fail.
2.3 Theory
My central claim is that dictators in times of crises consolidate power by using an expand
and signal strategy. It consists of (1) expanding the ruling coalition and adding new mem-
bers to reduce elite coordination capacity; and (2) signaling their strength to the broader
coalition via formal institutions to reduce a snowball effect if a coup is launched. Thus, I
conceive of formal institutions as arenas for signaling, an idea that has not yet been con-
sidered in the literature on authoritarian regimes. In this chapter, I focus exclusively on the
role of a limited set of formal institutions, parties and legislatures.44 There are two reas-
ons for a narrow approach. First is that parties and legislatures have tended to dominate
the debate on institutions in dictatorship;45 and, second, because these are indeed highly
43Singh 2014.
44Gehlbach and Keefer 2011; Gandhi 2008; Wright 2008.
45Brownlee 2007; Gandhi 2008; Gelbach and Keefer 2011, 2012; Wright 2008; Wright and Escribà-Folch
2012.
21
visible institutions that lend themselves easily to the expand-and-signal approach.46 Before
introducing the formal model, I address relevant theoretical issues related to the expansion
of the coalition as well as signaling through institutions.
Expanding the coalition in the midst of a crisis may be counterintuitive, since it may
appear to weaken the position of the leader rather than strengthen it. Yet, introducing a
set of newcomers to the coalition has the advantage of reducing the coordination power
of the plotters within the coalition. New entrants need to be screened and are more likely
to support the leader, at least early on. Trying to persuade them to join a coup may give
the entire plot away, making coordination difficult for the plotters. Careful selection by the
leader is required, but they can select members that have relatively few linkages between
them, with current members of the coalition, or that are part of different opposition groups.47
Making the coordination capacity of the plotters more difficult gains the dictator sufficient
time to weather a crisis and potentially close the window of opportunity for a potential coup
against him.
Formal institutions fulfill two key roles within the expand-and-signal strategy. First, they
force new members of the coalition to be publicly associated with the regime. This is not
an angle that has been much studied in the vast literature on co-optation examined above,
or on other related literatures such as studies on co-optation and repression.48 By extracting
a public commitment to him, the leader ensures that members of the coalition are tied to
his rule, which binds them in future dealings with him or with any potential successor. The
46Extending the logic to other institutions, such as electoral processes (Gandhi and Lust-Okar 2009) or the
judiciary (Ginsburg and Moustafa 2008), is beyond the scope of this chapter.
47see Lust-Okar 2005.
48Escribà-Folch 2013; Frantz and Kendall-Taylor 2014.
22
more associated a member becomes to a given ruler, the more difficult it will be to remain
in power or avoid repression when a new leader comes to power.
Second, institutions signal the leader’s true type to a set of members in the coalition with
whom he may not be able to credibly communicate. Support from mid- and low-ranking
officers and soldiers is key to the success of any coup, and they may be more likely to join one
if they believe the leader is weak and the coup will succeed.49 If the leader receives public
support from new elites, the rank-and-file is more likely to believe that the leader is stronger
than the plotters consider him to be. This reduces the likelihood that a coup succeeds if
launched. Therefore, we should observe authoritarian leaders making fairly extensive use
of seemingly power-sharing institutions such as parties and parliaments, with periods of
expansion coinciding with those moments in which a leader is in a relatively weak position.
Leaders tend to be particularly weak at the beginning of tenure and after growth shocks.
2.4 The Model
The formal model I introduce in this chapter seeks to reflect precisely how expanding the
ruling coalition through power-sharing institutions both reduces the probability that a coup
succeeds and, under some circumstances, persuades plotters to not carry out a coup in the
first place.
The game takes the form of a standard signaling game with two types and a dichotomous
signaling choice. The model consists of two actors, an authoritarian leader (L) and a set of
49See Singh 2014.
23
plotters (P ) from within the dictator’s ruling coalition that seek to depose him. The leader is
uncertain about who the actual plotters are, but he knows whether or not he enjoys majority
support within the coalition. The game begins with Nature drawing a type of leader, who
is strong with probability θ and weak with probability 1 − θ. While the game resembles
Svolik, it differs precisely on this point, i.e. nature moves first by introducing a shock that
weakens the leader.50 The leader’s strength is determined by the level of support the leader
has within the entirety of the winning coalition. If at least half of the members of the winning
coalition back the leader, he always fights and defeats the coup is one is staged with a certain
probability. The fact that he does not win outright accounts for the possibility that the rank-
and-file join the coup as it is unfolding, as Singh shows.51 The strong leader thus knows
that at least half of the coalition supports him before the coup starts, but the figure may vary
as the coup evolves. If, at outset, only half of the coalition minus one (or less) supports
the leader, he quits and loses power if a coup is staged.52 The solution concept is Perfect
Bayesian Equilibrium.
Costs and Commitments
Central to the model is the idea that expanding the coalition and establish certain formal
institutions is costly to the leader. The cost stems primarily from the division of resources
within the coalition. When new members are brought in, current loyal members relinquish
50Svolik 2012.
51Singh 2014.
52I make this assumption for simplicity since it does not alter the results of the model. Theoretically, however,
leaders could survive a coup if less than half of the coalition supports him, but some do not join a coup as it
unfolds. What this would do is lead more weak leaders to imitate strong types, which is already part of the
set of equilibria that result from the model.
24
part of the benefits they receive and are forced to share them with the new members. This is
risky, since it may reduce the level of loyalty of the leader’s traditional supporters and make
them more likely to coordinate with the plotters.
At the same time, only strong leaders can afford to expand the coalition and divide the
spoils among more people without causing a rebellion. This is precisely the source of the
commitment mechanism in the model. I assume that strong types, those who enjoy the
support of at least half of the coalition, are always willing to fight, since their payoff is always
higher than losing government and paying the cost of exile or death. On the other hand,
weak types always quit. Apart from carrying a cost, a credible signal must communicate
that a leader is willing to fight if he is defied. Using functioning institutions as signals does
precisely that: only leaders that are in it for the long haul are willing to incur the expense of
creating and maintaining complex structures such as one or multiple parties and legislatures.
Sending a signal that you are willing to anything to remain in power, and that a majority of
elites are willing to make their support for your regime public, communicates resolve.
Assumption 1: Oversizing is costly, since it the coalition reduces the payoff of sup-
porting the leader for current loyalists within the coalition.
Information Structure and Timing
The leader is informed about his type at the outset of the game. Both the leader and the
plotters know the probability distribution around θ, the prior belief about the strength of
the leader, but the plotters do not know precisely which leader they are facing. The game
proceeds as follows: after Nature has determined the type of leader, the leader decides
25
whether to expand the coalition through institutions or not expand at all. In either case, the
plotters then have a choice between launching a coup or acquiescing in the new or current
status quo. If they do not coup, the game ends. If they do, the leader can then decide whether
to fight the coup or quit. The extended form of the game is represented in Figure 2.1.
Leader
(G ◆, 0) ¬Coup ¬Coup (G, 0)
(⇡G C  ◆, Expand ⇠ ExpandL (G CL,
 Fight F ight(1 ⇡)G Cp) L p q L (1 )G Cp)
Coup Coup
Strong ✓
(a, G) Quit Quit (a, G)
Opposition N Opposition
Weak 1 ✓
(G ◆, 0) ¬Coup ¬Coup (G, 0)
 1 p 1 q( CL  ◆, (CL,
G Fight F ightCp) L Expand G C )hwdu ⇠ Expand L p
Coup Leader Coup
(a, G) Quit Quit (a, G)
Figure 2.1: Signaling Model In Extended Form
Payoffs
The payoffs for the leader are as follows. If he decides not to expand the coalition, all leader
types receive G, the value for continuing to govern the country, if no coup is staged. If a
coup does indeed occur, however, strong types receive φG − CL if they fight, and −a if
they do not. φG is the value for the prize multiplied by the probability of winning a coup
if no expansion occurs. Cp is the cost the leader pays for fighting off the coup. If he quits,
he pays a cost −a, which captures the cost of losing control of government plus other costs
associated with resigning, such as restrictions on freedom, audience costs that may preclude
26
1
him from holding office in the future, or exile. Weak types, on the other hand, would only
pay −Cp if they fight, since their probability of winning a coup if it is staged goes to zero.
If they quit, they also pay the cost of quitting −a.
The payoffs differ slightly for the leader if he decides to expand the coalition through
formal institutions. If no coup is staged, both types receive G − ι, the value of governing
minus the signaling cost. If a coup occurs, strong types get πG−CL− ι if they fight. Here,
π is the probability that the coup succeeds after a signal is sent. As before, weak types that
fight have no chance of winning, so their payoffs of −CL − ι (the cost of fighting and the
sunk cost) are strictly negative. Both types pay a cost of −a if they decide to quit after a
coup is staged.
There are two important conditions to note here. First, the costs to fighting for the weak
type are strictly lower than the costs of quitting, i.e. −CL < −a and −CL − ι < −a. This
ensures that a weak type of leader always prefers to quit rather than fight. Conversely, strong
types always fight, as they know they have enough support to defeat the coup. So φG−CL >
−a and πG − CL > −a. In this set up, sending a public signal that the coalition has been
expanded and more elites show their support for the leader affects the probability that a
coup succeeds. This is meant to reflect the channel of indirect communication mentioned
above. Officers further down the chain of command may not receive credible signals that a
leader is weak or strong, and may join the coup if staged. Sending the signal has the effect
of persuading lower-ranking officers to not join the coup.
π
Assumption 2: > 1. If a strong leader expands the coalition, his probability of
φ
winning is greater than if he does not expand the coalition.
27
As for the plotters’ payoffs, they obtain a payoff of 0 in all situations in which they do not
launch a coup. If they launch a coup, their payoffs will vary according to the type of leader
they are facing and whether they observe the signal. If they observe a signal that the ruling
coalition was extended, their payoffs for fighting are (1 − π)G − Cp if the leader is strong
and G − Cp if he is weak. G is the value for controlling government and it is the same for
both actors. Cp is the cost to the plotters of fighting if the leader does not quit. If they do
not observe any signal and they launch a coup, their fighting payoffs are (1 − φ)G − Cp if
the leader is strong and G− Cp if he is weak. They obtain G every time the leader quits.
Equilibria
Before delving into the effects of asymmetric information on the expansion of the ruling
coalition through institutions, I begin by analyzing the complete-information game. Under
full information, strong leaders would always fight off a coup successfully, while weak ones
would fall. Plotters would always coup against weak types and acquiesce against strong
types. Given this, strong leaders would never be willing to pay the signaling costs, and the
game would end with no expansion and no coup. As for weak types, they quit regardless of
whether they send the signal. So, with complete information, a pooling equilibrium exists
in which no type expands the coalition.
Things change when we give the leader private information about his true type. While a
pooling equilibrium still exists in which leaders do not expand the coalition, another pool-
ing equilibrium emerges in which both types prefer to expand the coalition under certain
circumstances. These two pooling equilibria provide the main results of this chapter. Let us
28
start with the pooling equilibria in which all types decide to oversize the coalition and signal
the expansion through formal institutions. By backward induction, and given the conditions
that strong leaders always fight and weak leaders quit, then the plotters will be indifferent
between staging a coup or not when
p[(1− π)G− Cp] + (1− p)(G) = 0,
(2.1)
p∗
G
= .
πG+ Cp
Thus, if p∗ ≥ θ, their prior belief about the leader’s strength, a coup is launched. They do
not rebel if p∗ < θ. This gives us the plotter’s mixing strategy.
How does a strong leader respond in either case? If p∗ < θ, he knows that a coup will
not occur, and he will obtain a payoff of G − ι. This is because when the probability that
the coup fails is high, the payoff to fighting is strictly negative, (1− π)G−Cp < 0, and the
plotters are better off waiting out the period. Is the leader better off defecting and choosing
to send no signal at all? Off the equilibrium path, the plotters will coup after no signal is
sent if
p[(1− φ)G− Cp] + (1− p)(G) = 0,
(2.2)
q∗
G
= .
φG+ Cp
29
Since, by assumption, π > φ, plotters are more likely to coup after not observing a signal,
p∗ < q∗. If the dictator does not send the signal, the plotters launch a coup and he defeats
them with probability φ. But if he expands the coalition and reduces the likelihood the rank-
and-file join the coup, his probability of surviving the coup attempt increases to π. Whether
he defects now depends on whether q∗ ≥ θ. If this condition holds, then the plotters always
coup and the leader will defect only if he is better off fighting off a coup than sending the
signal. This will be true when G− ι < φG−CL, or ι > (1−φ)G+CL. Only exceedingly
strong leaders, for whom the cost of fighting off the coup is really low and the term (1−φG)
is close to zero, will prefer to not send the signal. If, on the other hand, q∗ < θ, the plotters
never launch and the leader defects and prefers to not institutionalize.
Therefore, with asymmetric information, we obtain a situation in which strong types
whose prior strength lies between p∗ and q∗ may still prefer to send a costly signal and
expand the coalition. It is also expected that strong types for whom θ ≥ q∗ > p∗ never
institutionalize. The reason why this constitutes a pooling equilibrium is that, given the
uncertainty that plotters face, weak types will always try to imitate strong types. If they do
not and separate themselves, they will signal that they are weak and attract a coup, which
will succeed. Hence they will prefer to send the signal, benefiting from the randomization
of the plotters and surviving with probability 1− p∗.
Proposition 1: There exists a range of strong leaders that prefer to expand or oversize
the coalition and pay the signaling cost when p∗ < θ < q∗. Strong leaders will send
the costly signal if ι < (1− φ)G+ CL, and will not otherwise.
30
If p∗ ≥ θ, that is, if the plotters always coup after expanding the coalition, slightly different
outcomes emerge. Since, following my previous logic, q∗ ≥ p∗, in this situation the plotters
always coup no matter what the leader does. This echoes Little’s argument that leaders
always face a coup in period 1 after coming to power.53 In these circumstances, a strong
leader will always fight, while a weak one will quit. Since for the weak leader the cost of
quitting is always−a, he is indifferent between expanding and not expanding. Yet, does the
strong leader still prefer to expand the coalition? That will depend on whether πG−CL−ι >
φG − CL. More specifically, he will prefer to institutionalize when πG − φG > ι, that is,
when he has a chance to obtain more of his value of governing by expanding than the simple
cost of creating the expansion.
Proposition 2: If p∗ > q∗ ≥ θ, strong leaders expand the ruling coalition precisely be-
cause they know a coup is forthcoming, and sending the signal improves their chances
of winning the coup provided πG−φG > ι. Weak types also send the signal to avoid
separating themselves, and benefit from the plotter’s mixing strategy.
This game also has a pooling equilibrium in which both types of leader decide not to send
a signal. For this to be the case, leaders need to be sufficiently strong that plotters will not
coup even in a moment of apparent weakness, or θ > q∗. Another option is that, even if a
coup is forthcoming, or q∗ > θ, then the leader is better off not institutionalizing and instead
fights off the coup, or when φG− CL > πG− CL − ι, or πG− φG < ι. This may be the
case if expanding is too costly, but strong leaders should be able to afford a sufficient level
of institutionalization to increase the probability π of surviving a coup just enough. Thus,
53Little 2017.
31
this equilibrium is sustained on the part of the strong leader only if θ > q∗. As regards the
weak type of leader, again, he has the incentive to imitate the strong type and not send any
signal, as it may benefit from the plotters not launching a coup in such case. He is indifferent
between expanding the coalition or not if the plotters launch a coup as his payoff is −a in
both instances.
Proposition 3: If θ > q∗ ≥ p∗, neither strong nor weak leaders will expand the
coalition through institutions. The strong leader is believed to have enough to survive
a coup even if he does not expand the coalition.
It is important to note that this game only has the pooling equilibria described above. No
separating equilibria can exist, since a weak leader never has an incentive to reveal his
true type. Doing so would always lead to a successful coup. Similarly, no semi-separating
equilibrium exists for the same reason. If, for instance, strong leaders use a mixed strategy,
the weak leader would always want to defect and imitate a strong leader when he chooses
a different strategy. Weak leaders would never mix, since their true type would be revealed
at least some of the time.
In this chapter I have presented a fairly standard signaling game with discrete types and
two signaling choices. In the continuous-signal form of the game, a separating equilibrium
exists for a range of strong types that send a stronger signal than a weak type can ever send.
That is, my model as is assumes that weak leaders can imitate strong leaders and benefit
from the plotters’ mixing strategy, which sometimes leads to no coup occurring. So in
that form of model, (1) leaders that can survive any coup never expand the coalition, (2)
leaders whose support is not sufficient to guarantee that they survive every coup expand the
32
coalition above the level at which a weak type could imitate them, and (3) weak types do
not expand the coalition and fall victim to a coup. The most important takeaway from both
models is the second, i.e. that there is a range of strong types for which it is better to send
a costly signal, expand the coalition through formal institutions. The main result, therefore,
is common across models.
2.5 Institutions Under Trujillo In The Dominican Repub-
lic
In this chapter, I contend that dictators expand the coalition in times of weakness to weaken
the relative position of the plotters and use formal institutions to signal strength to the
broader coalition. The case of Rafael Trujillo, which I review now, illustrates this argu-
ment well. I will focus on two specific events: Trujillo’s removal of Vasquez from power in
early 1930 and the steps he took to consolidate power afterward.
When Rafael Trujillo rose to power in the Dominican Republic in 1930, the country
was politically and institutionally underdeveloped. Internecine power struggles between
coludo and bolo caudillos had led to a period of deep political instability starting in 1899.
Eleven presidents ruled the country between that year and 1914, when US occupation began.
Trujillo trained with the American forces and rose through the ranks during Horacio Vasquez’s
six-year tenure after the US left in 1924, becoming the commander in chief of the armed
forces in 1927.54 As the great depression weakened Vasquez, whose decision to seek an
extension on his term limit proved unpopular, factions developed that sought to overthrow
54Turits 2003.
33
him. A nationalist movement led by Rafael Estrella Ureña launched a revolt on February
23, 1930, and Trujillo seized the moment to strike a pact with Estrella and take power for
himself.
Trujillo’s coup in 1930 is a vivid example of why leaders (1) are constantly under threat
from other members in the ruling coalition and (2) they do not know who these plotters
are. I argue that, precisely for these reasons, leaders are better off using an expand and sig-
nal strategy over other potential solutions, i.e. selective purges and co-optation through in-
creased rents. By making Trujillo commander in chief of the armed forces in 1927, Vasquez
intended for him to become his second in command and his most loyal servant. Yet, in early
in 1930, Trujillo began to plot against Vasquez by forming an alliance with Estrella’s move-
ment. He allowed Estrella’s forces to march freely on Santo Domingo, while telling the
President that he was doing everything within his power to stop them. As Crassweller de-
scribes, Vasquez refused to believe, until the very last second, that Trujillo was behind the
plot to overthrow him; he only realized it when Trujillo himself appeared in his office and
asked him to leave.55
While he had the support of Estrella’s coalition and the army, Trujillo’s position was
not consolidated. New rulers have coups launched against them more frequently56 and
the effects of the Great Depression could sow discontent quickly and erode his position.
Moreover, Trujillo had one main competitor, Estrella himself, whose movement he used
to depose Vasquez. Estrella was the only figure who had the prestige and political capital
to pose a difficult challenge to Trujillo’s rule. While Trujillo controlled the armed forces,
55Crassweller 1966.
56Little 2017.
34
Estrella had the charisma, the coalition, a small political party –the Patriotic Coalition of
Citizens– and a position of power. As part of the deal, Estrella was made interim Pres-
ident and Trujillo would run as a candidate in the August 1930 elections, partly because
the US State Department refused to recognize a potential Trujillo administration.57 Estrella
emerged, therefore, as the main threat to Trujillo’s path to the Presidency.
After becoming President in August 1930, through elections that he helped rig,58 Trujillo
set out to consolidate his power. Indeed, what Trujillo did next helps illustrate precisely the
expand-and-signal strategy I describe in this chapter. Trujillo first dissolved the Patriotic
Coalition of Citizens, Estrella’s party and the one which had brought him to power. He then
founded the Partido Dominicano and established himself as its leader. The party would soon
reach a membership of 1,302,751 Dominicans, as most citizens were required to sign up.
The political elite that emerged was a mix of new Trujillistas from Santo Domingo and an
old rural elite that fit uneasily in the new scheme. Trujillo held a deep disdain for the old ‘ar-
istocracy’ and sought to undermine their power by creating a powerful urban elite made up
of people whom, like himself, had risen from the lower classes.59 He organized the party in
one central, 23 provincial, and 69 communal juntas. Further local and district-level groups
were organized. Leadership of these subgroups became equivalent to social standing, and
occasional purges ensured every provincial and national leader remained alert and loyal.
The expansion of Trujillo’s coalition from a small centralized group of support in the Patri-
otic Coalition of Citizens, in which Estrella Ureña competed for leadership with Trujillo, to
57Crassweller 1966.
58Hall 2000.
59Ibid.
35
a wide-ranging complex organization with many more members like the Dominican Party
took place in less than a year.
The relative power of Estrella and his coalition dwindled as the party elite grew in num-
bers and in support for Trujillo. Indeed, Estrella’s downfall was precipitous. After agreeing
to transfer the Presidency to Trujillo in August 1930 after 5 months as acting President, he
became the Vice-President. He resigned from the position in 1932 after fleeing to Cuba in
late 1931. He returned from exile in 1940 and Trujillo made him Supreme Court Justice,
but tensions between them arose again after Estrella reignited his old Nationalist Party and
died in 1944 in suspicious circumstances.60
More generally, while Trujillo is remembered as one of the most ruthless and long-lasting
dictators in Latin America during the twentieth century, his rule was not solely based on re-
pression executed by a small coterie of sycophants. Quite to the contrary, he built a relatively
complex network of regional and national institutions and used it to manage his power. In
fact, in the Democracy and Dictatorship dataset,61 Trujillo’s Dominican Republic is coded
as one of the most institutionalized dictatorships. Elections were held to elect representat-
ives, multiple parties were allowed both de jure and de facto for long stretches of time, and
a functioning legislature existed that could enact its own laws –with the consent, of course,
of the benefactor. This network of institutions, I argue, was useful to Trujillo not to prevent
domestic conflict and improve policy-making generally. For the former, he had the Military
Intelligence Service, a sharp and spineless police force tasked with repressing all forms of
60Crassweller 1966.
61Cheibub, Gandhi and Vreeland 2010
36
dissent. For the latter, he could find ways to strong-arm any lawmaker into submission.62
Rather, institutions were used as signaling devices to members within and without his im-
mediate support circle that any deviation from the most absolute loyalty would be greatly
punished.
The Dominican Party was Trujillo’s main tool for coalition expansion. Another was
the Parliament, which slowly increased from 31 representatives and 12 senators in 1930
to 58 and 23 respectively in 1957, just four years before Trujillo’s assassination. Why the
expansion? While the legislature retained powers to enact law, and even if multiple parties
were allowed at different points in time, all of its members were staunch Trujillo loyalists. In
the only election in which the Partido Dominicano did not obtain one hundred percent of the
vote, in 1947, no party obtained enough votes to enter either chamber of Parliament. Rather,
these formal institutions were in place precisely as signaling devices. In entering them, the
elites that were part of these institutions made their support for Trujillo explicit by becoming
invariably linked with the regime. While many survived the country’s transition to a weak
democracy in the second half of the 1960s, most notably Joaquı́n Balaguer, the outcome
would likely have been the opposite had a rival faction taken power via a coup. Moreover,
the institutions convened on a regular basis and were part of the political debate of the
time. Senators were important people, and scandals involving them fascinated the popular
imagination. Between the elites in the party organization, and the elites in Parliament, few
were left to contest the regime consistently from the outside, which explains why Trujillo
was never under real pressure to co-opt domestic opposition. Plotters from inside the regime
62Turits 2003; Crassweller 1966.
37
had the unenviable task of convincing the rank-and-file that Trujillo was weak enough that
a coup against him would succeed. By creating such strong a visible institutions, Trujillo
made himself look strong in the eyes of a majority of the armed forces and the broader
populace. The three coup attempts he faced in the first decade in power were easily and
quickly defeated.63
A relevant juncture in Trujillo’s dictatorship came in 1948, perhaps when Trujillo was
at his strongest, with world sugar prices booming after World War II. Multiple parties were
legally allowed in the Dominican Republic under Trujillo, but at different points in time,
parties other than the Partido Dominicano were repressed in practice. For the 1947 elec-
tion, Trujillo allowed multiple parties to run. World War II had a relatively negative effect on
the state’s finances, leading to difficulties between 1943 and 1947. Trujillo decided to create
the image of integration and plurality by allowing multiple parties to run. The Democratic
National Party of Rafael Espaillat, known as navajita (little knife), obtained 3 percent of the
vote. Espaillat was a loyalist who became chief of the secret service. Another party, the Na-
tional Labor Party, obtained also 3 percent of the vote. Its ‘leader’, Francisco Prats Ramı́rez,
was also a loyalist who had also been President of the Partido Dominicano. Trujillo won the
election with over 93 percent of the vote, and neither of the other two parties won sufficient
votes to obtain a single seat in the legislature. After 1948, with the country’s economic
fortunes reversed and the election won, any party other than Trujillo’s Dominican Party was
not allowed to operate in practice, even if they remained legal. It was not until 1961, the
year of Trujillo’s death, that multiple parties operated both de jure and de facto again.64
63Crassweller 1966.
64Cheibub, Gandhi and Vreeland 2010; Crassweller 1966; Derby 2009; Turits 2003.
38
2.6 Conclusion
The main theoretical contribution that this chapter wants to make is that autocrats, when
faced with a particularly difficult predicament, have often resorted to a relatively unknown
and counterintuitive tactic to retain power: they expand, rather than shrink, the ruling coali-
tion and use formal institutions purely as signaling devices meant to generate commitment
from new members and communicate credibly to the rank-and-file that the leader is too
strong for a coup to succeed. Many instances of this exist throughout history. When Rafael
Trujillo came to power, for example, in the Dominican Republic in 1930, he moved quickly
in the first six months of his rule to create a large party apparatus and expand the country’s
senate. In the case of Morocco, which Gandhi uses in her book,65 King Mohammed V faced
strong internal opposition upon his return to the country as rightful King in 1957 just after
independence from France. Within 7 days of his arrival, to counter the powerful movement
that sought to impede his reign, he establish the country’s first Parliament and filled it with
75 loyalists. The celerity with which the institution was created, coupled with the fact that
loyalists were placed in positions of influence, lends credence to the idea that the institution
served as a way to expand the coalition, weaken the relative power of the opposition group,
and tell potential plotters that staging a coup was the wrong choice.
The formal model I have introduced in this chapter captures both of these relevant dy-
namics. On the one hand, and most importantly, using institutions as a public signal that
the ruling coalition has expanded directly increases the probability of winning a coup, since
more members of the rank-and-file perceive the signal and decide not to join a coup if it is
65Gandhi 2008.
39
staged. In doing so, the payoffs for staging a coup for the plotters are lower after the coalition
has expanded, which makes them more likely to not stage a coup in the first place.
The main results of this chapter are straightforward and intuitive: very strong leaders,
who enjoy enough support from their coalition to survive virtually any coup, prefer not to
institutionalize. Weak leaders, who know that they will lose power if a coup is staged, try to
imitate strong leaders and benefit from the plotters’ mixing strategy, which is to not launch
a coup some of the time. In the continuous-signal form of the model, weak types do not
imitate since stronger types manage to separate themselves by sending a signal that weak
types cannot afford. What is most relevant, however, is that there exists a range of strong
types that are always better off sending the signal under certain circumstances.
This chapter bridges an important gap between the literature on coups and dictatorial
consolidation and the one on power-sharing institutions in authoritarian regimes. I provide
a new theory for the emergence of a very specific set of formal power-sharing institutions
in dictatorship: formal organizations such as political parties and legislatures created or
completely reconstructed by the leader. Especially important is to consider actions by a
leader in times of crisis, particularly at the outset of rule and after a severe shock. In these
times, fewer leaders are likely to be strong enough to survive any coup, and visible dwindling
support makes more leaders likely to engage in signaling of the type advanced in this chapter.
More leaders, that is, are likely to fall within the range of types that decide to expand the
coalition in times of crisis. Once created, these institutions are likely to persevere at least
during the leader’s tenure.
40
Chapter 3
Et tu Brute? Wealth Inequality and
Authoritarian Replacement
As the recent example of Robert Mugabe in Zimbabwe shows, transitions within dic-
tatorship remain frequent. What motivates elite factions to seek to replace an au-
thoritarian incumbent? In this chapter, I provide a political economy theory of au-
thoritarian replacement. I argue that high wealth inequality fosters authoritarian
replacement, but that the effect is conditional on overall wealth being low. At low
wealth, elite factions have an incentive to control the state to appropriate income. As
wealth grows, elites shift their focus toward securing their wealth and thus prioritize
finding credible commitments and stability within authoritarianism. I test these hy-
potheses using data from 1960 to 2008 and employ multistate survival analysis. A
case study of Trujillo’s rise in the Dominican Republic illustrates the mechanisms of
the theory. The evidence supports the main theoretical expectation that replacement
is more likely when wealth inequality is high and wealth is overall low.
41
3.1 Introduction
What motivates elite factions to seek to replace an authoritarian incumbent? In this chapter,
I provide a theory of wealth inequality to explain authoritarian replacement. When wealth
inequality is high but wealth levels are low, competing elite factions seek to overthrow one
another from government in order to capture new income through the state. As their wealth
increases, they become progressively more concerned with securing their wealth, and au-
thoritarian replacement decreases. I define authoritarian replacement as a transition from
an authoritarian regime to another, which can occur as a result of a coup, a change in lead-
ership accompanied by a substantial change in the rules of the regime, a popular uprising,
a revolutionary movement followed by a civil war and subsequent new leadership, or elect-
oral loss within autocracy.66 To this day, the politics of authoritarian replacement remain
significantly undertheorized.67
Attention has generally focused on transitions from authoritarian rule to democracy and
democratic breakdown68 or on institutional and economic factors that improve dictatorial
stability.69 Recently, a new wave of research has focused on coups within authoritarian re-
gimes, yielding important insights into the role of dictator-elite relations,70 the coordination
capacity of the plotters,71 and how coups unfold and the role of information diffusion in the
days of the coup.72 A lot less is still known about the motivations of different elite factions
66Geddes, Wright and Frantz (GWF) 2014.
67Wright and Bak 2016.
68Acemoglu and Robinson 2001, 2006; Ansell and Samuels 2014; Boix 2003; O’Donnell and Schmitter 1986;
Przeworski et al. 2000.
69Brownlee 2007; Gandhi 2008; Slater 2010; Wright and Escribà-Folch 2012.
70Svolik 2009, 2012.
71Little 2017.
72Singh 2014.
42
to replace each other, and why they appear to be more common and pernicious in certain
contexts and not others. What factors lead different elite groups to overthrow each other?
Consider the Dominican Republic at the turn of the twentieth century: In the fifteen year
period between 1899 and 1914, eleven different presidents were sworn in. Three of them
fell victim to a coup d’état, five of them resigned their office after brief tenures, and one
was assassinated. The instability resulted from quarrels between two groups, Horacistas
and Jimenistas, for government control. Wealth at the time was low, but the sugar trade was
booming. Greater tax revenue increased incentives for corruption in government. Moreover,
a majority of sugar mills were in the hands of US and Canadian investors by the mid-1910s,
generating income streams in the form of government contracts, licenses and concessions.73
Governing, even for a short period of time, was profitable, and the wealth owned by elite
groups was low enough to justify the risk of taking over the state.
Economic theory and empirics predicting regime transition have until now mostly fo-
cused on authoritarian to democratic transitions and the mechanism of income inequality.74
The literature has paid much less attention to transitions from one authoritarian regime to
another and to alternative economic mechanisms that may explain transitions within au-
thoritarianism.75 In political transitions, conflicts for power usually involve a small and
relatively affluent elite, whose factions compete over power and wealth. I provide a novel
political economy theory of authoritarian replacement based on the role of wealth, which is
the main driver, I argue, of elite competition.
73Bulmer-Thomas 2003, Hall 2000.
74Acemoglu and Robinson 2006; Ansell and Samuels 2014; Boix 2003.
75Houle 2016; Przeworski 2005.
43
I make two key distinctions in this chapter. The first one is between wealth inequality
and levels of absolute wealth. Wealth inequality, a concept that has gained importance of
late, refers to the structure of wealth in the economy at a given point in time, that is, the
division of national output between capital and labor; or, in simpler terms, to differences in
the ownership of capital and its profits among different classes.76 Wealth can be more or less
concentrated in the hands of a few. Wealth levels, on the other hand, refer to the total amount
of wealth owned in society at any given point in time. The second distinction is between
wealth and income. Wealth is a stock and income, a flow. Wealth, or capital, represents the
total assets owned at any point in time, which have been saved or accumulated.77 Income
is the net influx of assets within a defined period of time, usually a year. Individuals need
large amounts of income to transform into savings, which then become their wealth stock.
My theory predicts that, at high levels of wealth inequality, social actors have an incentive
to capture the state in order to increase their short-term income, with their goal set at maxim-
izing their long-run wealth. This effect for authoritarian replacement, I argue, is moderated
by the level of wealth in the economy: when wealth is low overall, the incentives to capture
the state and increase the flow of income are much greater, since they have relatively little
wealth. As wealth levels increase, the incentive to capture the state decreases even in the
face of relatively high inequality.78 A paradigmatic example of such conflict occurred on
Peru during the guano era, from the 1840s until the 1870s: different elite factions replaced
each other in power often in order to enrich themselves by capturing guano income through
76Piketty 2014; Saez and Zucman 2016.
77I use these two concepts, wealth and capital, interchangeably.
78Later I discuss how, at high levels of wealth, elites no longer seek to control the state, since their incentives
shift from seeking new income to securing the wealth they already own through guarantees on property
rights.
44
contracts with British companies.79 Governing was not a medium- or long-term stable ex-
ercise but an expedient and fleeting affair intended to benefit the members of each leader’s
coalition. Inequalities in wealth among these groups were high, and the wealth they owned
relatively low, which incentivized them to seek to control the state and redistribute income
toward themselves.
In this chapter, I model and measure authoritarian replacement.80 I define and operation-
alize the concept here as replacements of authoritarian incumbents by an entirely different
set of elites. I contrast this type of transition with leadership replacement from within the
same coalition, which often does not shift resources or generate more elite redistribution.81
If current elites stay in power in their majority, economic relations will remain the same –as
was the case, for instance, under the Partido Revolucionario Institucional in Mexico from
1920 until 2000. Here, the interest is in authoritarian replacement between opposed elite
factions, and that is why I use the GWF transitions dataset.82
This argument makes important contributions to current debates. First, it helps explain
key dynamics of authoritarian replacement, a topic that has been significantly underthe-
orized. Transitions to democracy have long been at the forefront of comparative politics
research, but the recent trend toward democratic erosion and breakdown is forcing scholars
to study authoritarian politics more intimately. Indeed, the share of democracies has plat-
79Bonilla 1984.
80In my tests, democracy and democratic breakdown are included to account for the full scope of transitions
that countries can experience. This conforms with the multistate survival model used in the empirical section
of this chapter. Including all transitions reduces coefficient bias. However, no theoretical contribution is
offered in this chapter for democratization or democratic breakdown. This will be explained in more detail
in the methods section.
81See GWF 2014; Wright and Bak 2016.
82GWF 2014.
45
eaued in the last decade – 108 democracies existed in 2008, and 108 democracies existed in
2016.83 The number of nondemocracies has not risen either – 75 in 2008 and 75 today, but
a notorious breakdown in Turkey in 2016 and a worrisome trend toward reversal in coun-
tries such as Poland or Hungary foreshadow the potential for a first wave of democratic
breakdowns. These events, coupled with recent transitions within authoritarianism, such as
Mugabe’s ouster in Zimbabwe, underscore the importance of understanding authoritarian
politics at a deeper level and from new perspectives.
Second, the political development literature has focused primarily on the effects of in-
come inequality on regime transitions. I argue in favor of shifting the focus away from
income inequality and toward the concept of wealth, which is better suited to explain struc-
tural political outcomes through the lens of intra-elite conflict. Third, this chapter places a
renewed focus on intra-elite bargaining and conflict,84 moving away from the ‘elite-vs-poor’
logic that has dominated the political development debate for well over a decade.85 Lastly,
it expands on extant work showing that new dictators expropriate competing elites to signal
their exclusive reliance on the group that supports them.86
I use multistate survival analysis to model transitions within authoritarianism and interact
wealth inequality and wealth levels. The dataset includes 144 countries from 1961 to 2008.
I also provide a case study of the Dominican Republic between 1900 and the fall of Trujillo
in 1961 to illustrate the mechanisms of my theory.
83Boix, Miller and Rosato 2013. An updated version of the dataset to 2015 is available at Harvard Dataverse.
84Geddes 1999; O’Donnell and Schmitter 1986; Przeworski 1991.
85Acemoglu and Robinson 2006; Boix 2003.
86Albertus and Menaldo 2012.
46
3.2 Theory
High levels of wealth inequality lead to authoritarian replacement. In what follows I provide
an elite-conflict logic for my argument: in dictatorship, conflict over resources increases
when wealth inequality is high. Disadvantaged economic and social groups can reverse
their fortunes by reaching political power and appropriating resources. However, the effect
of this wealth inequality gap on authoritarian replacement is mediated by the level of wealth
these elite groups already own. When elite factions have low levels of wealth, they will seek
to capture the state in order to appropriate income – as happened in Peru during the guano
era or in the Dominican Republic at the turn of the twentieth century amidst a booming
sugar trade. As wealth levels increase, incentives to secure wealth become more pressing
than incentives to increase income. The need for stronger property rights, in turn, reduces
incentives for authoritarian replacement.
In this section, I will first survey the concept of authoritarian replacement and potential
alternative explanations. I will then define wealth, elites and go on to introduce the main
mechanisms of my political economy theory of authoritarian replacement. I will show that
controlling the state is crucial for elites to achieve their economic goals when wealth in-
equality is high and wealth levels are low.
Authoritarian Replacement
Authoritarian replacement refers to elite groups replacing one another in power, and requires
a change in the group that actually holds power and a shift in the distribution of resources
47
between groups.87 I do not attempt to explain regular leadership transitions from within the
same coalition – I reflect this distinction in the operationalization of the concept. Wright and
Bak make this concise distinction and use the examples of Iran and Nicaragua in 1979 – the
advent of the current clerical regime and Somoza respectively – as instances of leadership
change that both the composition of the elite and the distribution of resources among elites.88
They contrast these two cases with Mexico under the Partido Revolucionario Institucional
(PRI), where leaders alternated in power in an effort to aid authoritarian survival but the
coalition remained the same,89 and with the case of the Argentine military juntas of the
early 1980s. In both of these cases, leadership change amounts to a simple reshuffling of
the executive and the group that holds power is the same. In this chapter, my argument
and logic apply to the former set of cases – and none of the transitions under the PRI are
considered as instances of replacement in the data.
The cases of Iran, Nicaragua are obvious examples of authoritarian replacement, since
they shifted resources in a dramatic fashion between clearly differentiated factions. Other
cases may be less clear-cut. Indeed, some replacements in my data set occur as the result
of military coups, in which it may difficult to judge the extent to which a successor and his
coalition differ from his predecessor and the elites that supported him. This was the case, for
instance, in Ecuador in 1972, where Rodrı́guez Lara ousted Velasco Ibarra in a bloodless
coup;90 in the Central African Republic in 1981, where André Kolingba removed David
87Wright and Bak 2016.
88Ibid.
89Magaloni 2006.
90Acosta 2008.
48
Dacko;91 and in the Dominican Republic in 1930, when Rafael Trujillo toppled Horacio
Vasquez, among others.92
Yet all these examples differ from the cases of the PRI in Mexico or the military juntas in
Argentina and Brazil in the 1970s and 80s in one fundamental aspect, namely, that the new
leader, while part of the old coalition, represented interests that differed from the President
he ousted. The case of Trujillo is exemplary: while he was the power player in the coup
against Vasquez by virtue of controlling the military apparatus, he struck a secret pact with
an opposition faction led by Estrella Ureña and the urban, nationalist elite that sought to
remove Vasquez.93 He shifted resources from an old aristocratic elite toward a new olig-
archy deeply dependent on him.94 Thus, Trujillo’s rise as a clear example of authoritarian
replacement as conceptualized in this chapter, even if he was, admittedly, Vasquez’s second
in command and a large majority of the army remained under his control. In the case of
Kolingba, he was also Dacko’s Army Chief of Staff. Once in power, he built his regime
partly along ethnic lines by appointing Yokamas, his coethnics, to key positions. Yokamas,
a small minority in the Central African Republic, also represented over two-thirds of the
army by the end of his rule.95
It is important to note that the boundaries between the leader’s coalition and competing
elite factions are often blurry. This is because, as is well known, dictators try to co-opt
such rival elites to improve their own chances of survival, which can indeed be a successful
91Ghura and Mercereau 2004.
92Turits 2003.
93Crassweller 1966.
94Hall 2000.
95Ghura and Mercereau 2004.
49
tactic.96 Yet problems remain for the dictator. Primarily, rival factions are fluid and con-
stantly forming, which means that he may often not know precisely how to co-opt every
potential faction that develops or becomes discontented. This problem is more acute when
the leader’s resources are limited and he cannot accede to the growing demands of rival
co-opted factions. The literature has tended, I would argue, to see co-optation as a rigid
process, but it is in fact fluid. Indeed, co-optation requires periodic renewals of the terms
of agreement, as well as the right balance in the benefits offered to co-opted factions. For
the dictator, this process is difficult and dangerous, even more so when some rival factions
own less wealth than other factions.
One particularly perverse manifestation of state capture at low wealth and high inequal-
ity are predatory states, where rulers use their position of power to prey on the citizens and
extract resources through taxation or expropriation.97 Others may take the form of rentier
states, where natural resource abundance often leads to the appropriation and misalloca-
tion of profits and anemic industrial, agricultural and service sectors –commonly known as
Dutch Disease.98 Or, as in the case of Peru in the nineteenth century, enclave economies
can become unstable and see high levels of replacement as a result of elites fighting each
other for political power as they try to profit from contracts with foreign companies.99 My
political economy theory of authoritarian replacement complements these existing debates
and helps explain some, if certainly not all, of the common dynamics identified in them.
96Gandhi 2008.
97Evans 1995.
98Karl 1997.
99Cotler 1979.
50
Elites
How do wealth inequality and wealth levels affect how different economic elite factions
decide to replace an authoritarian leader? Economic elites are defined here as individuals
within a given society who, by virtue of their access to wealth, may exercise political power
and have decisive influence on a state’s executive and legislative powers. They are usually in
the top one-thousandth of the wealth distribution and tend to organize themselves in groups
that represent various economic interests, political beliefs, or nepotism networks. These
groups also can be nested –for instance, oil producers may have a small association within
a staunchly conservative block that advocates for a strong currency. Participation in these
groups is often fluid, and the total set of elites expands as the economy grows. If they do not
themselves nominate a member for political office, these economic elites sponsor a group
of political elites to represent them and often strong-arm them into doing their bidding.
I begin with a simple assumption: elites seek to maximize wealth in the long run. In
dictatorship, they face two main threats: another elite group appropriating their wealth and
a revolution from below. The literature has recently echoed the importance of these two
existential threats. Elites may democratize if their position in authoritarianism becomes
weak but would otherwise be strong in democracy100 or if they have the opportunity to
game democracy.101 They can also stay within authoritarianism if they act as a unitary actor
100Slater and Wong 2013.
101Albertus and Menaldo 2014; Albertus 2015.
51
and pact.102 The existential threat of revolution from below, on the other hand, has been
noted in a majority of the social conflict literature.103
Wealth
Wealth can be conceptualized in two distinct ways. One is structural: who owns wealth and
who does not, and what are the relationships that develop over time between the haves and
the have-nots. This is the idea behind wealth inequality in society. The other corresponds
to the absolute level, or stock, of wealth available in an economy, i.e. how much wealth
there is.104 The two concepts may be related but it need not be so: Cuba, for instance, had
low levels wealth inequality and absolute wealth before 1990, while many Central American
nations had exceedingly high levels of inequality with low levels of wealth in the second half
of the past century. Developed European nations, on their part, tend to have lower levels of
inequality and high levels of wealth. China also broadly fits this category.
I define wealth at the individual level as the value of the productive capital owned by a
person at any point in time. It is important to distinguish wealth from income. Income is
earned in the form of salary and bonuses over a given period time, usually a year —it is,
therefore, a flow. Wealth, on the other hand, is a stock that holds intrinsic value and has the
potential to generate capital income. Wealth includes all productive assets that yield a rate
of return but excludes human capital.105
102Slater 2010; see also Magaloni 2006.
103Acemoglu and Robinson 2000, 2001, 2006; Boix 2003.
104see Piketty 2014.
105Following Piketty (2014, 72), I use the terms wealth and capital interchangeably.
52
I contend that the distinction between wealth and income is central to understanding au-
thoritarian replacement and, more generally, regime change. This idea differs from current
work that has emphasized the importance of income inequality in political transitions.106
Income may come in the form or of a salary or a payment from another person or entity,
or it may be generated by capital. In both cases, income inequality may be high, but the
political implications of inequality differ if personal wealth is the main source of income.
I argue that if elites own little wealth, they will seek to control the state and appropriate
new income. As their wealth increases and generates a majority of their income, they will
become more concerned with protecting their wealth and instituting guarantees for property
rights and political stability. Without being more precise about the concepts of income in-
equality and wealth inequality, we may miss important dynamics tied to income and wealth
separately.
Two final considerations regarding wealth. Human capital cannot be owned by a different
person or group of people and cannot be traded in the open market;107 hence it is excluded
from this theory. Also, while land remains an important theoretical mechanism in many
recent works,108 I focus on all forms of capital here without making distinctions by capital
type explicit. While type certainly matters, and I do hope to extend the logic advanced here
106Acemoglu and Robinson, 2006; Ansell and Samuels 2010, 2014; Boix 2003; Haggard and Kaufman 2012,
2016. It is important to note that these works have indeed considered the importance of wealth, but have
not precisely distinguished wealth from income and considered the implications that each may have, sep-
arately, on political transitions. Boix did so more explicitly by separating the effects of income income
inequality and asset specificity, arguing that elites with fixed assets will be less likely to accept a move to
democracy. However, within the concept of income inequality, he still included both wealth and income.
Other dynamics between wealth, income and regime transitions, such as the ones I identify in this chapter,
remained unexplored.
107Piketty 2014.
108Albertus 2015; Ansell and Samuels 2010, 2014.
53
to particular cases and types of economies, I leave these considerations as implicit in the
certainty that my general conclusions hold across types of capital.
Wealth Inequality, Wealth Levels and Authoritarian Replacement
To show how replacement is most likely when wealth inequality is high and wealth levels are
low, I introduce the following simple accounting identity, which reflects the wealth function
of any given elite at any point in time:
Wt+1 = (1 + rt)Wt + Λt − Tt − ct.
Wealth in period t + 1 is a function of all accumulated wealth in the past, Wt, multiplied
by 1+ the rate of return rt. To this, net income during period Λt is added, while taxes Tt
and consumption ct are subtracted. This calculation produces the net level of wealth of a
member of the elite at any given point in time.109
An elite’s total wealth can increase in two ways: by increasing income (Λt), or by ob-
taining a better rate of return on capital that grows the current wealth stock (rt ∗ Wt). I
assume that everyone starts with a positive level of assets Wt. If Wt is low, increasing
wealth by obtaining a return on capital will prove to be especially slow, considering that it
is reasonable to assume a long-run average rate of return on capital of around 5 percent.110
Thus, elites who find themselves with lower levels of wealth can only obtain so much in
109I do not introduce a formal model explicitly; the formula below will be useful only as an analytical tool to
describe how wealth grows in the long run.
110Piketty 2014.
54
terms of returns to that wealth, and their weak position relative to wealthier elites is likely
to endure in time. The persistence of this weakness is even starker when we consider that
capital increases multiplicatively in the long run, provided the rate of return is positive on
average. The greater the level of Wt, the greater the income from capital, and the faster the
gap between the wealthiest and least wealthy elites grows.
Disadvantaged elites are faced with a dilemma: do they prefer to remain in a weaker
position, or strive to regain their dominant position as one of the wealthiest – and thus more
influential – groups? Two central factors go into this calculus. First, if wealth is too low,
elites must increase their discrete income – and fast. Second, they need to calculate how
important it is to them to hold on to the little they already own, which is never safe in
dictatorship.
Why Take the Risk?
The most effective mechanism for elites to conquer new income is by capturing the state.
They can use it to capture capital from other elites through expropriation and obtain natural
resources or other forms of revenue that have not yet been tapped.111 Ruling elites can also
restructure the economy in such a way that their coalition profits directly from domestic
sectors such as utilities or telecommunications. Ruling is, indeed, profitable.
Yet not all groups take power in the same way. In certain instances of replacement, one
elite has been removed and expropriated completely, and then forced to flee the country or
111Karl 1997; see Ross 2001; Smith 2008.
55
stay at risk for their own lives – as in the aforementioned examples of Iran or Nicaragua.112
Others are less violent, as in guano era Peru, where elite factions simply focused on con-
trolling guano contracts and reining in the profits when they took office.113 However, all
these transitions, as Wright and Bak argue, have one thing in common: a substantial redis-
tribution of resources from one group to another.114
This shift in resources explains one important part of our central puzzle, i.e. why elites
seek to replace each other and not accept the status quo and cooperate with the rulers. Given
the formula above, it pays off to seek to overthrow the government if the potential income
from ruling is higher than the value attached to an elite’s current wealth. Since the income
from takeover will always be quite high everywhere, lower levels of wealth will lead to
increases in the probability of authoritarian replacement.
The only case in which this conclusion would not hold is if an individual elite was so
risk averse that no level of promised income in government could be higher than its value
for its current wealth. Such individual may exist, but it cannot be modal because there are
no guarantees on wealth in dictatorship. One individual may find himself at the wrong end
of selective purges, or without key assets should they become particularly interesting for the
dictator to own or redistribute. Thus, even for the most risk averse elite member, it pays off
to join a plot to overthrow the current leadership if the probability of winning is greater than
the probability that he manages to keep his assets.
112Wright and Bak 2016.
113Bonilla 1984.
114Wright and Bak 2016.
56
When Replacement Ends
As elites grow wealthier, their concern shifts from generating new income flows to securing
the wealth they have accumulated. Their income now stems entirely from their wealth in
the form of interest on bonds or deposits, stock market gains, profit from enterprises in
which they invest, and others. Since elites can take full advantage of the rapid multiplication
of a well-managed and diversified portfolio of wealth investments, they no longer need to
generate non-capital income through capturing the state.
The state may no longer be regarded as a tool to generate new income, but an authoritarian
ruler can still decide to expropriate wealth at a moment’s notice. If uncertainty is high, elites
may be better off trying to replace the leader again to reach political power and thus prevent
the expropriation of their wealth. I argue, however, that elites prefer a better and more stable
alternative to conflict when their wealth is high and wealth inequality decreases. Attempting
to take over the state is risky, and it is difficult for them to ascertain whether it is more risky
to replace the leader than to be expropriated by him.
What wealthy elites require is a solution to the leader’s commitment problem: can elites
formalize a pact by which the leader and the ruling elite credibly commit to not expropriate
other wealthy elite’s property? The answer is that they can. Such agreements usually take
the form of constitutional rules that guarantee and respect property rights, and institutions
such as political parties and legislatures help enforce them.115 If the ruler reneges on his
promises, the affected members will know. His failure to fulfill this commitment may thus
115Gehlbach and Keefer 2011; Wright 2008.
57
lead to his downfall, as other elites now sense uncertainty and seek to depose him. Note,
therefore, that authoritarian leaders are bound by certain commitments, and even more so
when they preside over an economy with wealthy elite groups that can credibly challenge
their rule.
We know that, with a dictator who supposedly has the last word on every issue, it may
be difficult for elites to believe he will not renege, no matter how many checks they try to
place on his power.116 Yet, notice that a perfect credible commitment to other elites is not
required. Rather, he can maintain credibility in his commitments if he acts in such a way that
any rival faction becomes indifferent between mounting a challenge or acquiescing. That
is, his credibility is a function of how much liberty he has to renege on commitments. At
higher levels of wealth, elite groups pose a more credible threat to a dictator, and his power
is thus curbed. The lower the wealth inequality and the higher the level of wealth, the more
likely it is that elites force a credible commitment from the dictator to protect their property,
thus reducing authoritarian replacement.
My argument is, in some ways, similar to Przeworski.117 His argument is that democra-
cies are more stable when the stakes are high for a majority of players, who are too invested
to revert to authoritarianism. As he puts it,“the stakes are too high to risk losing the income
guaranteed under democracy”.118 In democracy, therefore, the costs of regime change are
large, but are lower in authoritarianism. The argument can be extrapolated to countries
where elites own relatively little wealth versus others where elites own a lot. When wealth
116Przeworski and Limongi 1993.
117Przeworski 2005.
118Ibid, 265.
58
levels are high, regime change is costly, as actors are too invested in the current regime.
When wealth is low, on the other hand, regime change is cheaper. Not because elites are
less risk averse, but because the promise of a windfall from taking control of government
eclipses the utility of holding on to little wealth.
3.3 Empirics
My theory predicts that the effect of inequality on authoritarian replacement is conditional
on low levels of wealth. The empirical strategy is to model nondemocratic transitions as
events of authoritarian failure and calculate the probability that a regimes falls in any given
year. The alternative is to consider each authoritarian spell as a separate state, that is, as-
sign each spell its own numerical value and compute the likelihood of being on one state or
another. However, this is conceptually problematic, these regimes do not transition out of
dictatorship. Thus, survival analysis fits the event-based logic of this chapter, but its applic-
ation poses a number of challenges. First, survival analysis can yield biased estimates if one
models only one of the transitions within the full set of changes a subject may experience at
some point.119 Here, any authoritarian regime is at some risk of experiencing democratiza-
tion, and democracies are then at risk of backsliding. The challenge is to model the full set
of potential transitions within a survival framework. Another challenge is truncation, since
119Metzger and Jones 2016; Putter et al. 2007.
59
some authoritarian spells enter the sample some time after becoming at risk for a transition
(left truncated) or our data ends before we observe the final outcome (right truncated).120
To address the first challenge, use multistate survival analysis to capture the full process
of political transitions that countries can experience. Figure 3.1 shows the multistate pro-
cess including all possible transitions conceptually relevant to our theory. A authoritarian
state is at risk of transitioning to another authoritarian state or to democracy, while a demo-
cracy is only at risk of reverting to authoritarianism. While democratization or democratic
breakdown may not central to the theory, they indeed matter empirically. Competing risk
models can only capture parts of the process;121 in our case, a multinomial logistic model
would be limited to the competing risk of authoritarian failure and democratization within
each spell.122
The literature has often used the dynamic probit to model political transitions.123 How-
ever, the dynamic probit can only model two transitions between states, not events, and
it is usually limited to two states since the dependent variable is dichotomous. Moreover,
multistate models do not assume, as pooled models do, that the data generating process
(DGP) is equal for all possible transitions. If we made such assumption, the covariate’s
effects would be the same for all transitions. This is why pooled models of transition pro-
duce biased estimates. Lastly, by using multistate survival model, we can better incorporate
the effect of time in transitions, which has been missing from other studies of political
120See Box-Steffensmeier and Jones 2004.
121Metzger and Jones 2016.
122The results obtain if a simple multinomial logistic model is used. Tests are available from the author.
123Przeworski et al. 2000; Boix 2003; Houle 2009.
60
transitions.124 The methodological contribution of this chapter is to incorporate multistate
survival models into the political economy literature on transitions.125
Authoritarian1(A1) Authoritarian2(A2)
Democracy(D1)
Figure 3.1: All Possible and Recursive Transitions
In the Appendix A, I provide a series of robustness checks. To remove unobservable
heterogeneity and, therefore, potential confounders, I run a series of fixed effects models
using the linear probability model and logistic regression. I also attempt to capture the
inevitable time trend resulting from the natural increase of wealth over time and the decrease
of authoritarian replacement in the data. These results are reported in Table A3 in the
Appendix A.
3.4 Data
The unit of analysis in this chapter is the country-year. The dataset has 3,452 unique ob-
servations and covers 137 countries between 1968 and 2008. This is, to my knowledge, the
first test of long-run wealth accumulation on structural political events such as regime trans-
124Boix 2003; Ansell and Samuels 2010, among others.
125The multistate model is estimated using a discrete-time stratified cox procedure. I provide a description of
the particulars of the model and the dataset structure in Appendix A.
61
itions. As required by the multistate model, the data is duplicated for these observations that
are at risk of both democratization and authoritarian replacement. This process leads to a
total sample size of 5,146.
Failure, the dependent variable, is coded as 0 for those country-year observations in
which no transition occurs; 1 for those in which the authoritarian regime is replaced by
another; 2 for those that experience democratization; and 3 for those that saw democracy fall.
Substantively, the regime failure variable is constructed using a combination of the Geddes,
Wright and Frantz (GWF) dataset and the Democracy and Dictatorship (DD) dataset.126
GWF code authoritarian regime failure as 1 for a country in a given year if: 1) the autocratic
regime is unseated via election and the new government is allowed to take office; 2) the
regime is ousted by a coup, invasion, popular uprising, rebellion or other violent means; and
3) the ruling group substantially changes the rules for selecting the leader and for adopting
key policies. The variable is coded 0 otherwise. I use the DD dataset to code democratic
failure. There are a net total of 116 authoritarian failures that end in replacement, 75 in
democracy, and 39 in democratic breakdown.
I measure economic inequality as the share of capital that accrues to labor (wage share
- WS), which captures cross-class differences between capital-holders and labor. This vari-
able is more fitting to theories that focus on inter-group inequality and class cleavages.127
The higher the share of output that accrues to wages, the more equal the society. The meas-
ure is taken from the INDSTAT2 dataset produced by the United Nations International De-
126Cheibub, Gandhi, Vreeland 2010; GWF 2014.
127Boix, 2003; AR, 2006; Houle, 2009.
62
velopment Organization.128 For the results to be intuitive, the WS variable has been inverted.
Now the greater the variable’s values, the higher the level of inequality. It captures the share
of output that accrues to capital holders, a concept similar to the one used by Houle.129 The
data are available for 163 countries from 1963 to 2008. The advantages of our measure are
the large number of observed data points and its sensitivity. Another widely used meas-
ure of inequality is the Gini coefficient, which captures the distance between the largest
and smallest individual incomes in society. However, the Gini coefficient does not measure
differences across groups or classes. Houle provides a good account of this distinction.130
No yearly cross-national variable exists, to the author’s knowledge, that measures elite
wealth accumulation. Ideally, our measure should capture the total amount of asset stock
owned by the economic elite.131 One option is to obtain the total amount of wealth in an
economy and multiply it by the share that belongs to the top centile. The closest approxim-
ation to this variable can be obtained from Piketty’s data (WID) by multiplying his measure
of total private wealth by the share of wealth owned by the top decile or centile.132 Unfor-
tunately, the sample of countries in the WID data is small and dominated by countries that
have not experienced any recent transitions to democracy.
I construct a new measure using widely available data from the World Bank and contrast-
ing a subsample with the WID data. First, I take yearly cross-national Gross Fixed Capital
Formation (GFCF) in current dollars from the World Bank to estimate the total new wealth
128See Knutsen, 2015; The dataset can be found at https://stat.unido.org.
1292009, 2016.
130Houle, 2009.
131The economic elite is broadly defined as top centile of an economy.
132Alvaredo et al., 2018.
63
generated by capital each year.133 This variable captures a country’s yearly total domestic
investment in fixed assets. It includes land improvements, machinery purchases, private in-
frastructure construction (i.e. residential, commercial and industrial buildings), and the net
acquisition of valuables.134
For each country, I obtain the cumulative sum of new wealth for all years in the sample.
This gives us the amount of wealth generated in a given country-year from the time the coun-
try is first observed. It does not capture total wealth accumulation, as we cannot estimate
an initial value for wealth. GDP or some similar proxy of national wealth could be added as
the initial value, but that would introduce noise and generate its own set of issues, such as
lumping in population. To avoid these identification issues, I simply capture the pattern of
elite wealth accumulation within countries since the moment they are observed, rather than
estimating some initial value of wealth. If elite wealth grows rapidly, elites are more likely
to cash in and transition to democracy. If it does not, elites have an incentive to remain in
the status quo.
I make three substantive improvements to the cumulative sum of wealth described above.
First, countries generate increasing yet similar amounts of wealth year-to-year. The cumu-
lative sum of GFCF, therefore, experiences rapid growth the first few years and linear growth
133Data accessed on April 2018. It can be found at: https://data.worldbank.org/indicator/NE.GDI.FTOT.CD
134An in-depth description of the variable is included in Appendix A, including the World Bank’s own detailed
definition.
64
thereafter.135 To make wealth increase linearly across time, as in the WID data, I replace
each country’s first value of accumulated wealth with its fifth.136
Second, rather than dealing with gross numbers of wealth that are abstract and do not
truly represent the total amount of wealth available, we create a ratio: I divide each year’s
cumulative wealth by the amount of wealth the country produced the first year. Generating
a ratio of wealth allows us to capture precisely within-country elite wealth accumulation
and minimize the effects that cross-national variation may have in generating the variable.
The reason for this is simple: elites in one country are unlikely to compare themselves with
elites in other countries when deciding whether to transition to democracy. Wealth is not
the same in every country, given differences in prices and in pressures on consumption and
savings. It is more useful to net out these differences for the logic we advance in this chapter.
Lastly, I adjust the cumulative sum of wealth for depreciation and state investment. I
subtract 10 percent from each year’s wealth to account for depreciation137 and another 15
percent to account for the share of total investment by the government. The share of wealth
that belongs to the non-elites cannot be netted out, but this is a small part of the measure.
New wealth is largely generated by capital holding elites. To make sure these assumptions
are substantiated, I run a series of stress tests on our final variable. I compare WID’s data on
135Consider a country that creates 100, 110, 120, and 130 units of wealth in four years. The sum will be: 100,
210, 330, and 460 units. Growth from the first period to the second is 110 percent; from the second to the
third, 55%; and from the third to the fourth, 35%. After a few years, growth is linear on average both in our
data and in the WID measure. I refer to the WID measure sometimes as Piketty’s variable for reference.
136I chose the fifth value as growth becomes linear in the data on average after the fifth value of wealth.
137Piketty, 2014
65
total private wealth and elite wealth with a subsample from our dataset.138 A high correlation
with these known measures should provide some external validity for our proxy.
The WID’s net private wealth variable captures the total level of wealth owned by private
citizens in a country-year. The correlation between this measure and our proxy is r = 0.949
on a subsample of 770 observations from 21 countries.139 I then multiply the WID’s net
private wealth variable by the share of wealth owned by the top 5, 1 and 0.1 percent. These
measures directly capture the wealth owned by different sets of elites. The correlation is
high for all three measures. We obtain r = 0.913 for the total wealth of the top 5 percent,
r = 0.925 for the 1 percent, and r = 0.946 for the 0.1 percent. These correlations include
389 observations for 11 countries.140 These high r values are indicative of the strength of
our variable in measuring our theoretical concept of elite wealth accumulation.
3.5 Results
I begin by showing the relationship between the two main variables of interest and author-
itarian replacement descriptively. Figure 3.2 plots the observed probability of replacement
138Our variable cannot capture all forms of wealth that elites may have, such as financial assets, stocks and
savings. Neither does it account for foreign ownership of assets. However, I contend that these omissions
do not harm the validity of our measure. The level and growth of these assets tend to be highly correlated
with the ones in the measure, and foreign ownership, while important, appears to have a small impact on
our measure overall. I provide as evidence of our logic the correlations with Piketty’s private wealth data
for a subsample of countries. These are all above 0.9 and are detailed below.
139These are the US, Canada, China, Russia, the United Kingdom, Australia, Czech Republic, Japan, Mexico,
South Africa, Netherlands, South Korea, Denmark, Germany, France, Greece, Italy, Spain, Norway, Finland
and Sweden. Within-correlation scores are stable across groups, with the exception of South Africa and,
to a lesser extent, Greece. The correlation stays high if each or both countries are excluded from the test.
140The US, Canada, Australia, Japan, Netherlands, South Korea, France, Germany, Italy, Spain and Sweden.
Within-correlation scores are stable across all groups.
66
at different levels of wealth and wealth inequality. This is obtained by dividing the number
of observed instances of replacement by the total number of observations that fall within 20
different quantiles of both independent variables. Authoritarian replacement is more fre-
quent at very low levels of wealth. It then experiences a sharp decline before reaching more
or less the median, with the relationship largely flat at high levels. The effect of wealth
inequality is more ambiguous. Replacement seems, on the one hand, more likely at low
levels of wealth inequality, but the variability is also much higher as shown in the scatter-
plot. After a low point at middle levels of inequality, the relationship appears to be again
stronger at higher levels. The difficulty in assessing the potential effect of inequality on
replacement further suggests an interactive effect. Moreover, the descriptives for both vari-
ables show complex and non-linear relationships. This may complicate the interpretation
of the interaction coefficients in the survival models, but plotting the survival curves and
the joint effect of the two covariates should reveal how and when they are statistically and
substantively significant.141
Table 1 reports the main results of this chapter. Models 1 through 3 estimate the effect
of the independent variables on each possible type of transition: authoritarian replacement
(A1 → A2), democratization (A1 → D), and democratic breakdown (D → A1) without
controls, whereas Models 4 through 6 include all controls and decade dummies. Each set
of three models is estimated jointly in a single Cox stratified model, in which each strata
represents a transition. The model displays the coefficients and not the hazard ratios. Our
theoretical interest lies in authoritarian replacement (Models 1 and 4).
141I do not include multiple interactions with square or cubic splines, which overly constrain the data.
67
0.05 Levels of Wealth
Wealth Inequality
0.04
0.03
0.02
0.01
0.00
5 10 15
Levels of Wealth and Inequality
Figure 3.2: Observed probability in the data of replacement at different levels of wealth
In all models, capital share (inequality) is interacted with wealth, which is inverted here.
The reason for inverting the variable is that it is easier to interpret the sign of the interac-
tion in an intuitive way that fits the main hypothesis if wealth is inverted, i.e. the effect
on authoritarian replacement as wealth levels become lower or decrease, while inequality
increases. As shown in Models 1 and 4, the sign of the interaction is positive, as predicted.
As wealth decreases, increases in inequality lead to a higher probability of replacement; or,
conversely, as inequality increases, lower levels of wealth make authoritarian replacement
more likely –increasing the log-odds by 1.067 for every one-unit change, a relatively large
effect in the log-odds scale. For complex continuous interactions, statistical significance
can only be ascertained by plotting the joint effect.
Figure 3.3a shows precisely this. Since wealth is inverted, it displays the marginal effect
of decreases in absolute wealth on authoritarian replacement at various levels of wealth
inequality. As expected, at higher levels of inequality, a decrease in wealth will have a
68
Observed Probability of Replacement
Multistate Survival Model (MSSM) MSSM (With Controls)
1 2 3 4 5 6
A1 → A2 A1 → D D → A1 A1 → A2 A1 → D D → A1
Wealth (Inv.) t−1 −0.355 −0.034 0.566 −0.225 −0.218 0.828
(0.336) (0.300) (0.952) (0.436) (0.448) (0.773)
Inequality (CS) t−1 −5.624 −0.876 13.860 −5.017 −2.408 17.375
(3.578) (2.618) (8.768) (4.585) (4.456) (6.341)
Wealth (Inv.) t−1 1.024 0.251 −1.305 1.067 0.634 −2.272
x Inequality (CS) t−1 (0.671) (0.525) (1.668) (0.903) (0.892) (1.350)
GDPpc t−1 0.008 0.948 −1.277
(0.351) (0.348) (0.461)
Growth t−1 −0.033 −0.042 −0.097
(0.025) (0.032) (0.121)
Oil t−1 −0.082 −0.150 0.440
(0.082) (0.094) (0.181)
Ethnic Frac. t−1 0.001 −0.001 −0.026
(0.007) (0.008) (0.016)
Trade Open. t−1 −0.002 −0.015 −0.007
(0.004) (0.005) (0.015)
Previous Fail. t−1 0.658 0.327 0.220
(0.134) (0.153) (0.665)
Polity t−1 −0.070 0.002 −0.190
(0.084) (0.046) (0.253)
Polity Sq. t−1 −0.003 −0.016 −0.056
(0.011) (0.010) (0.023)
Time Trend 0.001 0.033 0.159
(0.052) (0.084) (0.114)
1980s −0.674 0.050 −4.827
(0.904) (0.999) (2.099)
1990s −0.088 0.113 −5.208
(1.140) (2.017) (2.551)
2000s 0.126 −1.301 −6.559
(1.359) (2.610) (2.666)
Observations 5592 5146
Pseudo R2 0.028 0.207
Log-Lik. -350.30 -274.281
A1 → A2 - Authoritarian Replacement; A1 → D - Democratization; D → A2 - Democratic Breakdown
Model: Stratified Cox with 3 strata, one for each of the transitions above. SEs in parentheses
Table 3.1: Effects of Wealth Inequality on All Transitions Conditional on Levels of Wealth
69
(a) (b)
Low Wealth
High Wealth
0.2 0.3 0.4 0.5 0.6 0.7 0 10 20 30 40 50
Wealth Inequality Time (Years)
Figure 3.3: Joint effect of wealth and inequality on authoritarian replacement (a). Survival
curves at high levels of inequality (b).
statistically significant positive effect on the probability of replacement. The lower bound of
the 95 percent confidence interval crosses zero when inequality is at 0.43 in the capital share
scare, which is close to its mean of 0.47. For instance, at high levels of inequality (0.69),
a change of 0.516 in the log-odds of replacement means that countries with 14 percent
less wealth will be 70 percent more likely to experience replacement.142 Or, for instance,
countries with 28 percent less wealth are almost two and a half times more likely to see an
authoritarian incumbent replaced by a competing faction at high levels of inequality. This
is because the wealth variable ranges between 0 and 7, so a 1 unit change represents around
a 14 percent difference.
The results are strongly significant in substantive terms. To show this with more clarity,
Figure 3.3b displays the survival curve for regimes with low wealth and high inequality
142The results of the Cox proportional hazards model are presented in log-odds, not hazard ratios.
70
Marginal Effect of Wealth on
Authoritarian Replacement
0.0 0.5 1.0
Survival Rate (High Inequality)
0.2 0.4 0.6 0.8 1.0
versus all others.143 An authoritarian incumbent presiding over a country with low levels of
wealth and high inequality can expect his reign to be relatively short-lived. Within five years,
one in five regimes have already been replaced – the survival rate drops to 80 percent. In
seven years, the survival rate tumbles further to 75 percent, and within fifteen years regimes
have reached toss-up levels, with only 54 percent probability of survival. Conversely, all
other regime types take a full 20 years to drop to a survival rate of 75 percent, and never
reach toss-up levels.
3.6 Robustness Checks
The following robustness checks have been performed and are included in Appendix A
unless otherwise stated. First, note that the main model includes a time trend, which is
important because wealth tends to grow naturally in time. Second, the reason why I report
the naı̈ve model with no controls in Table 3.1 is to show that adding the controls and the
time trend does not generate spurious significance. I include the equivalent of Figure 3.3a
using the model without controls in Appendix A (Figure A1), and the results are unchanged.
Third, to ensure that the results are not a function of the statistical model used in this chapter,
I perform tests with the following alternative models and include the results in Appendix
A: (1) OLS and logistic regression with random effects and (2) a more traditional logistic
regression with clustered standard errors and without random or fixed effects. I develop
143The curve for all other regimes is a weighted average. No CIs are shown as statistical significance has been
defined in Figure 3.3a; survival curves are sufficient to show substantive significance.
71
on the reasoning for these models further in Appendix A. The results obtain with these
alternative specifications.
3.7 The Rise of Trujillo in the Dominican Republic
The Dominican Republic before and after the rise of Rafael Trujillo illustrates the relevant
mechanisms and intuitions underpinning my theory. Before the benefactor, internecine con-
flict ravaged political stability in the Caribbean nation. Factions often replaced each other
in power, with eleven presidents sworn into office in a period of fifteen years. Trujillo’s
coup in 1930 ushered in an era of authoritarian stability, which meant high levels of polit-
ical repression but large increases in private wealth. I now describe these two periods of
Dominican history to illustrate the political economy theory of authoritarian replacement.
Wealth inequality was high in the Dominican Republic between the turn of the twentieth
century and 1930, the year of Trujillo’s coup. The country ranked within the top quart-
ile or tercile in all of Boix’s historical measures on family farms, knowledge distribution
and diversification.144 Data on levels of wealth at the time is not available, but given the
country’s lack of development after independence in 1844, its low GDP, and its incipient
institutions, it is fair to assume that wealth among the elite was low.145 Quarreling elite
factions around this time were primarily rural, so cleavages did not form around factors of
production. Rather, they reflected different regional interests within the landowning class,
144Boix 2003; from Vanhanen 1997.
145Turits 2003.
72
divisions which resulted in caudillo power struggles between the Horacistas and the Ji-
menistas, each named after their respective leader. In the fifteen year period between 1899
and 1914, eleven different presidents were sworn in. Three of them fell victim to a coup
d’état, five of them ‘resigned’ the office after brief tenures, and one of them was assassin-
ated – Ramón Cáceres in 1911. American intervention between 1914 and 1924 provided a
hiatus, and Horacio Vasquez then ruled from 1924 until Trujillo deposed him in 1930.146
One important aspect of my theory is that elite groups seek to control the state to gain
short term income when their wealth is low. This is precisely what we see in the Dominican
Republic during this time. The Jimenistas represented the heritage of Ulises Heureaux,
President until his assassination in 1899. The Horacistas, on the other hand, were the more
liberal wing of the elite and initially emerged as opposition to Heureaux. Important figures
within these two groups replaced each other un power during the tumultuous 1899-1914
period, even if they were all rural elites with common economic interests. Their vehicle
to greater wealth was the sugar trade and government corruption. Sugar exports expanded
rapidly between 1900 and the start of the World War II, reaching 5 percent annual growth
on average between 1913 and 1928.147 Domestic wealth was low and highly concentrated
among rural landowners, with the peasantry increasingly destitute. Rural elites benefitted
from the windfall and, in this context, controlling the state ensured direct profits from a
booming sugar trade.148
146Ibid.
147Bulmer-Thomas 2003.
148Turits 2003.
73
The sugar economy also became increasingly foreign dominated. Indeed, ingenios were
largely owned by US, Canadian, Cuban and Spanish interests in the first decade of the twen-
tieth century. These companies quickly controlled most plantations in the Eastern provinces.
By the mid-1910s, most sugar mills were in the hands of US and Canadian investors after
Cuban and Spanish companies sold their stakes.149 The expansion of foreign-owned plant-
ations and the booming trade of the commodity created an income stream particular to en-
clave economies. Governing was not only profitable because revenue from the sugar trade
could find its way to the pockets of elites in power, but also because the management of
foreign contracts, licenses, and concessions became a large source of income for the ruling
class. This is similar to what happened in Peru during the guano era, where foreign con-
tracts to extract the resource generated large amounts of revenue for elites in power. In turn,
these profits create incentives for different elite groups to take over government.150
The rise of Trujillo after direct American intervention (1914-24) and Vasquez’s relatively
peaceful tenure (1924-30) also illustrates precisely the type of elite struggles that I describe
in my theory and how they unfold. Trujillo was, since 1927, the chief of the Dominican
armed forces, the highest military position, and effectively Vasquez’s second in command.
His meteoric rise forebode his ambition to take the presidency for himself, but he need a
coalition for a coup to stand on firm ground. Coincidentally, a rift began to develop again
among the rural elite. Vasquez was an old politician – he was first President in 1899 –
and became associated with foreign interests and US control. New nationalist elites, led by
Rafael Estrella Ureña and his Republican Party, disapproved of foreign domination in the
149Hall 2000.
150Bonilla 1984; Cotler 1979.
74
economy and the President’s attempt to extend his term limit.151 They launched a revolt that,
according to Crassweller, would have gone nowhere had it not been for Trujillo’s cunning
move to allow Estrella’s forces to march into the palace unopposed.152 Trujillo, who as
Chief of the Armed Forces controlled the country’s scant military defenses, struck a pact
with Estrella beforehand, allowing Estrella to become President first while Trujillo retained
de facto power.
Whomever took power eventually (Trujillo) matters less than the fact that elite factions
struggled to control the state and increase their income, and that such conflict bred a stag-
gering number of authoritarian replacements up until that point. Estrella Ureña’s faction
wanted to take control of the state to wrestle power away from foreign interests and an old
and entrenched ruling elite who appropriated the income generated from the sugar trade.
They sought to capture the existing corruption streams linked to the government, and redir-
ect policy to benefit Dominican landowners such as Estrella himself. Trujillo, initially his
ally, consolidated power on himself and replaced Estrella after one year – a presidency that
was symbolic at most.153
A potential explanation for Trujillo’s abnormal longevity in the Dominican context is
that he was a better dictator. He quickly built a political party to support him, propped up
the army, and created an efficient and ruthless secret police. He was subject to at least three
known coup attempts in the first ten years of his rule, which he swiftly repelled. Yet his
survival was not solely due to skill. Challenges to his rule were much less frequent after
151Turits 2003.
152Crassweller 1966.
153Turits 2003.
75
this initial period, and an important reason is that overall wealth had skyrocketed among a
majority of sectors within the elite.154 No one was richer or owned more businesses than
Trujillo himself, but the dictator was deft at distributing riches among elites. Rural elites
continued to reap large profits from the sugar trade, even if now Trujillo owned a majority
of the assets used in its production. High level urban supporters either obtain large sums for
their political support in public office, such as Senators, or for helping run the businesses
he expropriated. Everyone depended on the Benefactor, and everyone profited from the
dependence. To get an idea of the levels of wealth involved for a relatively underdeveloped
Caribbean island, Trujillo’s own fortune is indicative. By the time of his assassination,
Trujillo amassed the seventh largest fortune in the world by some calculations.155 His son
Ramfis was the ostentatious sidekick to many a Hollywood star, and drove – and ruined –
multiple Ferraris in Madrid, Paris, and the United States. By co-opting a large majority
of elites, he diminished resentment and increased his overall level of support. Other elites
were not better off replacing Trujillo, and the small minority that may have benefitted were
powerless to attempt any uprising due to lack of sufficient support.
3.8 Implications and Concluding Remarks
In this chapter, I have advanced a new theory of wealth inequality and authoritarian replace-
ment. I find that the probability of authoritarian replacement increases at high levels of
wealth inequality, but the effect is conditional on overall wealth being low. When inequality
154Crassweller 1966.
155Crassweller 1966; Turits 2003.
76
is high and wealth is low, elites have an incentive to capture the state to appropriate income.
Consequently, authoritarian replacement is more likely. As the level of wealth increases,
elites become more concerned with securing their wealth, seeking greater guarantees for
property rights. In turn, replacement decreases.
This chapter makes an important contribution to the literature on authoritarian transitions
by providing a new theory of authoritarian replacement based on wealth inequality and elite
conflict. Transitions within authoritarianism that involve groups replacing each other from
power and significantly altering the distribution of resources remain understudied. With the
recent trend toward democratic backsliding, a deeper understanding dynamics within au-
thoritarian regimes is increasingly important. It also complements recent efforts to explain
authoritarian transitions156 and the economic relationship between the leader and the elite
that supports him.157
Methodologically, this chapter is the first one, to my knowledge, to use multistate survival
analysis in long-term structural political outcomes such as regime transitions. This method
can help us model the entire set of transitions that countries may experience and, as has
been shown, obtain more accurate estimates.158 Also, I find a new proxy for private wealth
accumulation that correlates highly with Piketty’s measure and is built on theoretically con-
sistent grounds using World Bank data on gross capital formation per country-year.159
Lastly, I also contribute to the political development literature by making an explicit
distinction between wealth, which is the total accumulated capital stock at any given point
156Wright and Bak 2016.
157Albertus and Menaldo 2012.
158Metzger and Jones 2016; Putter et al. 2007.
159Piketty 2014.
77
in time, and income, which is the flow of new assets every fiscal period. So far, the broader
debate has mostly revolved around the effects of income inequality on regime transitions.160
Wealth and income have been used synonymously, which has made it difficult to unveil
certain important dynamics during political transitions. A shared argument, for instance,
in Boix and Acemoglu and Robinson is that transitions to democracy are less likely at high
levels of income inequality, since the redistribution cost of democracy for elites will be high
if the poor get to set the tax rate.161 However, if we consider this situation in terms of wealth
maximization rather than income maximization, we realize that elite preferences related to
inequality and transitions will differ according to how much wealth they own. Indeed, if
they own a lot of wealth, individuals are likely to be at least as concerned with securing
this wealth as with expanding it. Thus, democracy could be a very attractive –if expensive–
system, since it places emphasis on property rights and separation of powers. I contend
that wealth is a more expansive concept that helps us model elite preferences and intra-elite
conflict in transitions better.162
Understanding the dynamics of authoritarian regimes is more pressing today than at any
point since the end of the Cold War. Some authoritarian regimes are surviving modern-
ization pressures better than many scholars expected, and replacement within dictatorship
remains common, as the recent case of Mugabe in Zimbabwe shows. Most importantly,
however, recent democratic backsliding and breakdowns in Poland, Turkey and Venezuela,
among others, urge us to demand better knowledge of authoritarian politics that expands
160Przeworski et al. 2000; Boix 2003; Acemoglu and Robinson 2006; Houle 2009; Ansell and Samuels 2010,
2014; Haggard and Kaufman 2012, 2016.
161Acemoglu and Robinson 2006; Boix 2003.
162O’Donnell and Schmitter 1986; Przeworski 1991.
78
on and goes beyond issues such as formal institutions, power, economic growth, and trade.
This chapter aims to provide some fresh answers to the problem of authoritarian replace-
ment, but new work is urgently required to explain the sharp downturn in the expansion of
democracy and the revival of dictatorship.
79
Chapter 4
Cashing In: Elite Cleavages, Wealth
Inequality, and Democratization
with Sebastián Vallejo
Recent evidence suggests that transitions to democracy are more likely at high levels
of inequality. To explain this pattern, we develop a theory of democratization in which
elites transition to democracy to secure their accumulated wealth through the rule of
law. When inequality is high, the threat of expropriation by a rival elite or revolution
is greater. The rule of law in democracy works as an insurance mechanism, providing
security for elite wealth in exchange for redistribution. Thus, elites see democracy as
insurance and decide to ‘cash in’ their wealth to secure it by extending the franchise.
We illustrate the theory through a formal model and evaluate it using data from 112
countries from 1966 to 2008. We find that transitions are more likely at high inequal-
ity conditional on high levels of accumulated elite wealth. We add a new dimension
to the debate on inequality and democratic transitions by exploring the role of wealth
in the decision making of elites in the process of democratization.
80
4.1 Introduction
In the 1980s and 90s, countries such as Brazil and Chile in Latin America and Thailand and
the Philippines in Asia expunged dictatorial rule in favor of democracy after decades of in-
dustrialization. Especially striking about these transitions is that they occurred in countries
with high levels of inequality, a fact that runs counter to prominent theories in the polit-
ical development literature.163 Indeed, recent cross-national empirical studies have shown
that high levels of inequality are linked to democratization,164 but our understanding of this
phenomenon remains limited. Why is high inequality associated with transitions to demo-
cracy?
We address this puzzle by identifying a mechanism that explains why transitions occur
more often at high inequality.165 Since high inequality increases the threat of expropriation
by a competing autocrat or revolution, risk-averse elites prefer to transition to democracy
and secure their wealth through the rule of law. Thus, elites ‘cash in’ their accumulated
wealth and use democracy as insurance. They prefer to pay a premium in terms of redistri-
bution in exchange for guarantees on property rights. The Latin American democratization
experience during the third wave informs our logic. Before the 1982 debt crisis, the re-
gion had experienced decades of economic growth through state-led industrialization and
commodity booms. With high levels of economic inequality, profits accumulated to a small
set of elites who grew exceedingly rich by the 1980s and 90s after benefiting directly from
authoritarian rule. Instead of risking their wealth to a competing autocrat or to revolution,
163AR 2006; Boix 2003.
164Ansell and Samuels 2014.
165Ansell and Samuels 2014.
81
elites preferred to democratize and safeguard their wealth through the rule of law in demo-
cracy.
Our research addresses two key questions in the field of political development: (1) why
greater levels of wealth lead to stable democracy166 and (2) how economic inequality affects
democratic transitions.167 Key questions remain in these debates. In particular, even though
some consider it a stylized fact,168 no consensus exists around a mechanism that explains
why countries tend to democratize as they grow wealthier. Neither have we achieved clarity
regarding the precise role of economic inequality in democratic transitions.169
In this chapter, we answer these questions by introducing a theory that connects the
roles of wealth and economic inequality in the process of democratization. We argue that
as wealth levels increase, economic elites become more risk averse to expropriation by an
autocrat or by revolution and see democracy as insurance to protect their stock of capital
through the rule of law. They prefer to pay a risk premium in democracy through higher
taxes in exchange for securing their wealth. Economic inequality, for its part, acts as a
destabilization trigger:170 the higher the inequality, the higher the threat of authoritarian
replacement or revolution. Since established elites risk losing nearly everything after a
successful revolution or autocratic takeover,171 they prefer democracy when both their levels
of wealth and economic inequality are high.
166Acemoglu et al. 2008; Geddes 1999; Kennedy 2010; Lipset 1959; Przeworski et al. 2000; Rueschemeyer,
Stephens and Stephens 1992.
167AR 2006; Ansell and Samuels 2010, 2014; Boix 2003; Houle 2009; Haggard and Kaufman 2012, 2016.
168Geddes 1999; Kennedy 2010.
169Haggard and Kaufman 2016.
170AR 2000.
171Albertus and Menaldo 2012.
82
An important contribution of our argument is to conceptualize democracy as insurance
for elite wealth. When wealth is high, democracy is less costly for elites than previously
theorized. What democracy does, in essence, is provide a complex formal standard for ex-
propriation via taxation and security through the rule of law. Since the level of risk aversion
for elites is increasing in wealth, they will be more willing to accept democracy when they
have a higher level of accumulated wealth stock than when they do not. The novel intu-
ition here is that the value of democracy for economic elites changes as their preferences
move from a narrow focus on generating income toward protecting the wealth they have
accumulated.
A parallel contribution of this chapter is to problematize the costs of transitioning to
democracy for elites. Elites are assumed to pay a high price of redistribution when they ac-
cept democracy.172 Yet, the influence they maintain over the political process allows elites to
reduce their total tax outlays by gaming democracy173 or by transitioning from a position of
strength.174 They pay more than in dictatorship and thus satisfy the redistributive demands
of the democratic society, but they still manage to preserve a majority of their wealth. We in-
troduce a term in our formal model to define precisely this capacity to ‘capture’ democracy
once a transition has occurred. This term allows us to analyze how a lower expectation of
future taxation creates additional incentives for elites to democratize through our insurance
mechanism.
172AR 2006; Boix 2003.
173Albertus and Menaldo 2014
174Slater and Wong 2013.
83
4.2 Democracy as insurance
The core idea of our theory is that elites transition to democracy in order to secure their
wealth and pay a cost of redistribution as insurance. The assumption is that democracy
provides greater security for elite wealth through the rule of law than dictatorship. We
identify two reasons for this: a lack of long-run credible commitments to property rights by
autocrats and the benefits of the rule of law for elites in democracy.175
The first problem that elites face is the inability to reach a credible commitment on prop-
erty rights with the political elite in dictatorship, who have different incentives.176 Economic
elites might have a certain preferred policy agenda, but the dictator may favor a subset of
the elite that launched him into power at the expense of others.177 He can also keep his pref-
erences secret, which exacerbates the informational asymmetries among elite groups. Both
the favoring of a subset of the elite as well as information asymmetries create incentives for
conflict over power and uncertainty over the future of the regime. This argument is related
to Olson’s classic claim that autocratic leaders cannot be forced to keep their promises,178
and that uncertainty over succession implies that commitments cannot last more than one
generation.179 Elites currently benefiting from the favor of the autocrat are not guaranteed
the same position of power under a new dictator. In line with these arguments, we contend
that long-term credible commitments for elites in dictatorship are unattainable, and thus
when faced with a potential for an insurrection from below, democratization solves both
175Albertus and Menaldo 2014; Olson 1991.
176Albertus and Menaldo 2014.
177Albertus and Menaldo 2012.
178Olson 1991.
179Olson, 1993.
84
sides of the commitment issue: the commitment to redistribute resources to the poor and
the commitment among elites to secure property rights.180
The literature has shown that institutions can solve commitment problems in dictator-
ships. Autocracies that rely on domestic investment over natural resources protect property
rights through binding legislatures181 and those with strong political parties tend to offer
more credible commitments to investors.182 Yet, these mechanisms tend to be limited in
scope and cannot fix the long-run credibility problem that autocrats face.183 We also have
relatively little idea of how they apply after an exogenous shock and in contexts of high
inequality. Our argument is that democratic institutions, primarily when transitioning from
a position of strength,184 can be designed to protect the wealthy elite in a way that ensures a
long-term credible commitment to a set of established rules protected by the rule of law.185
The second argument is that democracy provides a long-term credible commitment to
property rights through the rule of law. As argued by Przeworski, certain issues are agreed
upon and specified in formal constitutions, which lay down the rules that everyone will fol-
low.186 Electoral competition offers different political factions the possibility to alternate
in power, and with this, a random probability of holding control over residual issues. Ran-
domness over holding office represents a repeated game in which actors accept the outcome
of an election with the expectation of triumphing some time in the future. Thus, democracy
180AR 2006.
181Wright 2008.
182Gehlbach and Keefer 2011.
183Albertus and Menaldo 2014.
184Slater and Wong 2013.
185The literature has tended to identify the poor as the main beneficiaries of the rule of law in democracy, but
we argue that elites benefit as well.
186Przeworski 2005.
85
reduces conflict through alternation in power. In Przeworski’s terms, democracy allows for
conflict “to be regulated, processed according to rules, and thus limited.”187 This argument
ties together the idea that elites transition from a position of strength and the use of the rule
of law in democracy to protect property rights, which is the overarching interest of elites
and a central part of the specified rules of the game.188
Moreover, recent literature has identified key mechanisms that reduce the cost of demo-
cracy for elites; or, within our theory, the cost of insurance for elite wealth in democracy.
Following Albertus and Menaldo, elites can game democracy by setting roadblocks to re-
distribution ex ante, which makes democracy less expensive.189 Similarly, Slater and Wong
have claimed that elites transition from positions of strength and continue to thrive in demo-
cracy.190 This idea echoes other claims that conservative elites gradually accept and pro-
mote democracy.191 Indeed, an important contribution of our argument is the idea that at
high levels of inequality elites will be more adept at gaming democracy and thriving in it.
Inequality allows a compact and small elite that owns vast amounts of resources to dom-
inate democracy and ensure that the poor do not exceed themselves in their redistributive
demands. In purely rational terms, the poor should accept any offer which improves their
situation in autocracy, and thus the likelihood that wealthy elites dictate the terms of the
transition in high inequality is high. Moreover, high inequality gives an advantage to elites
187Przeworski 2005, p. 270.
188Democracies can certainly expropriate through land reform, nationalization and high estate or corporate
taxes (Albertus and Menaldo 2014; AR 2006; Boix 2003). However, such actions are a product of the
political context and can be reverted in the future through the alternation of power. These forms of expro-
priation test the limits of property rights, but their essence is respected. This is not the same in autocracy,
where property rights can be fully violated without a formal procedure and there is no expectation that the
actions can be reverted by contesting power through elections.
189Albertus and Menaldo 2014.
190Slater and Wong 2013.
191Ziblatt 2017; see Cox 2016.
86
in the democratic process through greater resources and impedes the poor’s ability to com-
pete on an even playing field. This new finding from the literature is incorporated into our
formal model through a ‘capture’ term that reduces the effective tax rate that elites pay in
democracy.
Lastly, an important insight from the democratization literature is the role of external
shocks in the early stages of democratic transitions. O’Donnell and Schmitter contend that
transitions to democracy usually begin with an exogenous shock to the authoritarian regime,
which splits the elite between Hardliners, who seek to preserve the status quo or increase
repression, and Softliners, who prefer to liberalize.192 External shocks, in their view, create
the opportunity for a transition to democracy. Without directly addressing inter-elite com-
petition in our model, we do incorporate the importance of exogenous shocks. We model
these shocks as a stochastic term that changes the cost of opportunity of transitioning for
the elite.193
4.3 Inequality, Repression and Democratization
Why would elites adopt democracy as insurance for their wealth? We argue that high levels
of inequality threaten elite wealth in autocracy through two channels. First, high inequality
increases mobilization by the poor and, consequently, the risk of revolution from below.
Second, it allows for the rise of a populist or caesarian leader that capitalizes on popular
192O’Donnell and Schmitter 1986.
193AR 2001.
87
discontent and challenges established elites. These threats prompt the elite to extend the
franchise and secure their wealth.
Inequality increases the demand for redistribution and thus rises the incentives for the
poor to mobilize. Venieris and Gupta, as well as Alesina and Perotti, have empirically
shown that there is a positive correlation between inequality and political instability and
civil unrest.194 AR also argue that inequality is a mobilizing factor for the poor, which may
lead to revolution if they collectively organize.195 In response to unrest, elites may choose to
increase repression-196 However, elites must also take into account the risk that repression
may fail, leading to democracy or, potentially, to revolution —as AR suggest.197 A failure
to repress effectively sends a strong signal that the ruling elite and their supporters are weak.
If democracy ensues, authoritarian elites may be shut out of the transition process and lose
the capacity to impose, or at least negotiate, their terms.198 This is related to an argument
by Houle, namely, that repression costs need not increase monotonically with inequality.199
Repression has a high probability of success at low and medium levels of inequality, but
may be riskier and more costly at high levels.
Elites are also threatened by the rise of a new autocrat that capitalizes on popular dis-
content and expropriates their wealth once in power. Albertus and Menaldo have shown
that new dictators obtain political stability by relying exclusively on their supporters.200 A
strong signal of this reliance is to expropriate the established elite. New autocrats emerge
194Venieris and Gupta 1986; Alesina and Perotti 1996.
195AR 2000, 2006.
196AR 2006; Boix 2003.
197AR 2006.
198see Albertus and Menaldo 2014; Slater and Wong 2013.
199Houle 2009.
200Albertus and Menaldo 2012.
88
by capitalizing on the discontent that economic inequality creates. Repression may increase
discontent, providing further support for a rising autocrat rather than stabilizing the regime.
Thus, wealthy elites in contexts of high inequality prefer a transition to democracy from a
position of power over a costly challenge to a rising populist autocrat.
Chile illustrates well how elites process the risks and rewards of repression versus trans-
itioning from a position of strength. When Pinochet lost his 1988 referendum, the military
junta was faced with a conundrum: repress or democratize. Inequality was high201 and
the referendum had mobilized the disaffected, making repression uncertain and costly. The
solution was simple: transition from a position of strength and ensure that elements within
Pinochet’s regime could continue to thrive in democracy.202
4.4 The Model
We introduce a formal model to capture the central dynamic of our argument: Elites trans-
ition to democracy to secure their wealth through the rule of law when inequality is high.
We define inequality as the differences in the distribution of resources across classes within
a given economy.203 Inequality is low when wealth is evenly distributed between capital
holders and labor. In this conceptualization, we fall in line with recent literature.204 Elite
wealth, on the other hand, refers to the total amount of resources held by the elite.
201The wage share in Chile in 1990 was only .17, placing it well with the lowest 10 percent of values in our
dataset. That is, it was more unequal than 90 percent of the country-years in our data.
202see Munck and Leff 1997.
203AR 2006; Houle 2009.
204AR 2006; Boix 2003; Houle 2009; Haggard and Kaufman 2012.
89
Our model consists of one agent, the economic elite (E), who decides whether to trans-
ition to democracy or remain in autocracy. The initial state is always an authoritarian regime.
The elite decide the tax rate in autocracy and have an expectation regarding the tax rate in
democracy based on the demands of the median voter if they extend the franchise. To reflect
their capacity to capture democracy, we allow the elite to have a discount on their expected
tax rate in democracy.205 Since we are only interested in democratic transitions, we assume
that democracy is a terminal or absorbing state. The levels of income and wealth in the
economy are stochastic, capturing the importance of exogenous shocks in democratization
processes.206 These exogenous shocks act as triggers to possible revolutions, changing the
opportunity cost for the elite to remain in autocracy.
The tax rate τ in democracy is increasing in inequality. There is a final good y and
a unique asset with total stock α. We begin our analysis of the economy at time t = 0,
where the elite’s assets are derived through income from capital and the capital stock:207
α = i(w) + w. The final good y of the elite will be affected by exogenous shocks to the
economy, which we denote with the term t, and thus yt = tα. We assume t to be a
stochastic term that can take two values
 
g = 1 with probability 1− s
t =  b = a with probability s
205The results of the model hold without this discount parameter, but transitions require higher levels of wealth.
We discuss this in Appendix B.
206O’Donnell and Schmitter 1986; Przeworski 1991.
207For wealthy elites, labor income is low in comparison to the rest of their assets. For simplicity, we assume
it to be zero in our model.
90
where b = a < 1 is a period of recession after an exogenous shock, superscripted b for ‘bad’.
When no exogenous shock occurs, the economy is not in recession and everyone knows
times are ‘good’, or g. Periods of recession change the opportunity cost of a revolution by
the poor.208 We capture this logic in our model by assuming that the risk of revolution will
only exist after an external shock, following the classic democratization literature.209 We
also assume that s < 1/2, so that shocks can be considered rare occurrences.
Post-tax income for the elite is given by ŷt ≡ (1−φτt)[i(wt)+wt], where φ is a discount
factor on taxes to wealth, a term that reflects the capacity of the elite to ‘game’ democracy.
For simplicity, we assume that the tax rate, τt, is set to zero by the elite in dictatorship. The
expected tax rate in democracy for the elite is the median voter’s preferred rate. The post-tax
income of the elite can be re-written as a function of wealth, w, changing the income from
wealth to the rate of return of wealth, r. We also add a term, c, to the income of the elite that
captures consumption. After rearranging the terms, the final equation for post-tax income
for the elite is: ŷt ≡ wt((1− φτt) ∗ (1 + rt)− c).
At any point in time, the expected utility for the elite is represented byE Σ∞ jt j=0δ U(αt+j),
where δ is the discount factor, andEt is the expectations operator conditional on all inform-
ation available. Furthermore, U(α) = ln(α) such that elites are risk averse, and are willing
to pay a higher risk premium (in the form of taxes), as wealth increases. The society starts
in an authoritarian regime (A), where the probability of revolution γ(θ) in any period t ≥ 1
is conditional on the level of inequality, θ, and differentiable with γ(0) = 0 and γ′(θ) > 0
for all θ > 0.
208see AR 2001.
209O’Donnell and Schmitter 1986; Przeworski 1991.
91
The timing of events within a period can be summarized as follows: 1) The state t is
revealed; 2) the elite decide whether to transition to democracy or stay in dictatorship as a
function of the probability that a revolution succeeds; 3) consumption takes place and the
period ends.
The state S is one of (, A) or (,D), where  is the stochastic term mentioned above
and can take the values of  = g or  = b. For each period, the elite decide whether to
stay in dictatorship, π = 0, or transition to democracy, π = 1. The state variables that
condition this decision are wealth, w, and inequality, θ. Following AR,210 there is no threat
of revolution in S = (g, A). In period t, t is revealed and the elite decide whether to extend
the franchise π = 1, or not, π = 0. We start the analysis by introducing an external shock.
Thus, in bad times, the value function for the elite is defined by
V (b, A, w, θ) = max {(1− π)Ṽ (b, A|w, θ) + (π) ṽ (,A,w, θ)}. (4.1)
π∈{1,0}
Here, Ṽ(b, A, w, θ) is the continuation value to the elite of remaining in autocracy, and
ṽ(,A,w, θ) is the continuation value of extending the franchise. If the latter term is greater
than the former, elites decide to democratize, so π = 1. The Bellman equations are defined
as follows:
210AR 2000.
92
Ṽ (b, A, w, θ) = (1− γ(θ)) [ ln ( bwt(1 + rt − c)) ] + δ W (A, w, θ) (4.2)
ṽ (b, D, w, θ) = ln [ bwt (1− φτt)(1 + rt − c) ] + δ W (D, w, θ), (4.3)
In dictatorship, elites maintain their current level of wealth with probability 1 − γ(θ), and
take into account the discounted value of future periods in authoritarianism if they decide
to remain in the current state. For simplicity, we assume that elites lose everything after a
successful revolution. The expected continuation function of remaining in an authoritarian
regime is
W (A ,w, θ) = (1− s) (1−γ(θ)) [ln ( bwt (1 + rt− c))] + s [ln(wt (1 + rt − c))]. (4.4)
In good times we assume that no revolution occurs. Also, as before, the value for the elite
of a successful revolution is simplified to zero. Note that in authoritarian regimes, the elite
set the tax rate to zero.
93
W (D , w, θ) = (1− s) [ ln( bwt (1− φτt) (1 + rt − c)] + s [ln (wt (1− φτt) (1 + rt − c)].
(4.5)
Equation (5) is the continuation function of remaining in democracy. The probability of
good and bad times is defined by s and (1 − s), respectively. In bad times, b = a, and in
good times, g = 1.
The elite will prefer to transition to democracy in state (b, A), maximizing π to 1, if
Ṽ(b, A, w, θ) < ṽ(b, D,w, θ). Solving this, elites in bad times will transition to democracy
when
[ ]
δ s
(γ(θ)) ln(w (1 + r − c)) ≥ − 1 + − ln(1 − φ τ), (4.6)1 + δ(1 s)
that is, when the cost of remaining in an authoritarian regime is greater than the cost of
transitioning to democracy. On the one hand, as the risk of revolution γ(θ) increases, greater
wealth w makes the status quo more costly for elites. On the other hand, a higher capture
term φ reduces the cost of democracy by lowering the effective tax rate τ .211 Note that when
211The right hand of the inequality will always be positive. The ln(1 − φ τ) is negative, as 0 < (1 − φ τ) < 1,
and the term δ s1+ δ(1− s) is positive.
94
there is no threat of revolution, i.e. γ(θ) = 0, the cost of staying in autocracy is zero and
the cost of democracy is always positive,212 so a transition never occurs.
Proposition 1: After an external shock, elites opt for democracy when wealth and the
risk from inequality γ(θ) in autocracy are greater than the cost of discounted taxation
in democracy φτ .
We now solve the formal model when no external shock occurs. In this case, the value
function is given by
V (g, A, w, θ) = max {(1− π) V’ (g, A|w, θ) + (π) v’ (D|w, θ)}, (4.7)
π∈{1,0}
where Ṽe(g, A, w, θ) is the continuation value to the elite after remaining in autocracy and
v’e (g, D,w, θ) the continuation value of democracy. These are given by
V’ (g, A, w, θ) = ln [wt (1 + rt − c)] + δW (A,w, θ), (4.8)
v’ (g, D, w, θ) = ln [wt (1− φτt)(1 + rt − c)] + δW (D,w, θ). (4.9)
212Even if the cost of revolution were zero, inequality can never be zero and thus the taxation cost is assumes
to always take a positive value.
95
The expected continuation functions of remaining in authoritarian regime or transitioning
to democracy are the same as in bad times: W (D,w, θ) and W (A,w, θ). Elites in good
times will transition to democracy when
δ (1 − s) (γ(θ)) ln(w (1 + r − c)) ≥ − (1 + δ) ln(1 − φ τ). (4.10)
In (10), as was the case in bad times, a higher probability of revolution increases the costs
of staying in autocracy as wealth rises. However, as the threat of revolution only emerges in
future periods of bad times, the term δ(1− s) reduces the overall cost of autocratic rule in
good times.213 As with bad times, taxation costs in democracy are reduced by the capture
term φ, but the term 1 + δ makes the overall cost of democracy higher.214 Thus, transitions
in good times are less likely.
Proposition 2: If no external shock occurs, elites opt for democracy when wealth
and the future risk from inequality δ(1− s)γ(θ) in autocracy are greater than the cost
of discounted taxation in democracy φτ .
Comparative Statics
To solve this Markov decision process, we assign values to the parameters, as informed by
the literature.215 We set the future discount parameter to δ = 0.9; a consumption parameter
213δ is the future discount and (1− s) is the probability that there will be bad times in the future.
214This is because δ > δ s1+ δ(1− s) .
215AR 2001; Przworski and Limongi 1993.
96
to c = 0.02; the net rate of return to r = 0.05;216 the penalization term after an external
shock b = 0.7; the probability s of remaining or transitioning to good times to 0.707;217
and the discount factor to taxes in democracy to φ = 0.75, which means that elites manage
to avoid paying 25 percent of their fair share.218 We have a total of four states, and all states
except for regime type can take multiple values. That is, the economy can be in good or bad
times, inequality can be low, medium, or high,219 and wealth is a continuous set of positive
values. We produce value functions for each combination of states, and a decision rule for
elites in all possible states.
Figure 4.1 reports the decision rules for the elite after an external shock. The wealth
parameter w in the model is on the x-axis and the expected cost in each state on the y-
axis. The straight lines report the cost of remaining in authoritarianism at different levels
of inequality and wealth (left side of equation 6.) The dashed lines represent the cost of
transitioning to democracy, which is not a function of wealth (right side of equation 6.)
When the cost of remaining in autocracy surpasses the cost of democracy, elites choose to
transition to democracy. We represent this with a shaded gray area.
The curves show that, provided an external shock, elites are more likely to choose to
transition to democracy at high levels of inequality as wealth increases. The elite decide to
extend the franchise when the risk to their wealth is high, but only after they have already
216Both terms, c and r are net of depreciation. In the model, we adjust these terms to account for a depreciation
of around 10 percent (Piketty 2014).
217We estimate the probability of having a recession from the data.
218Changing the values given to each parameter, within reasonable theoretical expectations, does not alter the
results of the model.
219For the probability of revolution, we set γ(θ) at 0.025, 0.075, 0.2, for low, medium, and high inequality,
respectively. For the tax rate in democracy, we set tau at 0.4, 0.2, and 0.1 for high, medium, and low
inequality, respectively.
97
 Transitions to Democracy at Levels of Inequality and Wealth
Cost of Democracy
Cost of Autocracy High Inequality
Democracy
Med. Inequality
Low Inequality
0 5 10 15 20 25 30
Wealth
Figure 4.1: Democratization as a function of wealth and inequality
accumulated a certain amount of wealth. The cost of remaining in dictatorship is lower than
the cost of democracy in both medium and low inequality, and elites do not transition.
These results mirror the democratization processes we have observed, for example, in
Ecuador and Brazil. In Ecuador, the commodity booms during the military dictatorship led
to urbanization and the “deepening of social and regional inequalities.”220 Elites accumu-
lated large amounts of wealth, leaving the social and economic conditions of the poor mostly
unchanged and giving rise to opposition from unions and indigenous organizations.221 The
strengthened unions and the growing indigenous movement made elites more than willing to
220Larrea 1994.
221Acosta 1995.
98
0.0 0.1 0.2 0.3 0.4 0.5 0.6
avoid turmoil by supporting a democratic transition, an elegant way to preserve their wealth
and establish a democracy on which they would attempt to exert control.
In Brazil, economic growth averaged 7 percent a year from 1960 until 1980 during ISI
and under military dictatorship. The development model stimulated demand through a “de-
liberate effort to skew income distribution in favor of upper income groups,”222 which led
to a fragmented multi-class state, supported by the economic elite (and by international act-
ors). The economic crisis of 1982-85 provided the necessary destabilization to the military
regime, and labor strikes led by Lula da Silva sent a clear signal that the working class
demanded profound political and institutional change. This spurred Figueiredo’s adminis-
tration to try to “carefully control Brazil’s transition from dictatorship to democracy and,
in the process, distribute political capital among the sectors of society that it deemed fit to
govern.”223 At that point there were few downsides to democracy: it would appease worker
demands, lessen conflict and, more importantly for elites, protect the huge wealth they had
accumulated under ISI and military rule at a modest redistribution cost.
Two additional results from the model merit mention. First, democratization is unlikely
without an external shock, which confirms the hypotheses of O’Donnell and Schmitter and
Przeworski.224 Second, the model lends credence to the intuition that capturing democracy
is relevant to democratization processes.225 Increasing the ‘capture’ term φ in our model
makes democracy more attractive to elites, as it reduces the cost of democracy (it shifts the
dashed lines downward.) However, the effect is more pronounced at high levels of inequality,
222Kohli 2004.
223Skidmore and Silva 1988.
224O’Donnell and Schmitter 1986; Przeworski 1991.
225Albertus and Menaldo 2014.
99
where transitions become more likely with less accumulated wealth, especially when 1− φ
is high.
4.5 Empirical Analysis
4.6 Data
Our primary hypothesis is that high levels of inequality and rapidly increasing elite wealth
foster democratization. To test it, we require valid measures of economic inequality, elite
wealth, and democracy. For the latter, we follow the literature and use a dichotomous meas-
ure of democracy for each country-year observation in our dataset.226
We measure economic inequality as the share of capital that accrues to labor (wage share
- WS), which captures cross-class differences between capital-holders and labor. This vari-
able is more fitting to theories that focus on inter-group inequality and class cleavages.227
The higher the share of output that accrues to wages, the more equal the society. The meas-
ure is taken from the INDSTAT2 dataset produced by the United Nations International De-
velopment Organization.228 For our results to be intuitive, we have inverted the WS variable.
Now the greater the variable’s values, the higher the level of inequality. It captures the share
of output that accrues to capital holders, a concept similar to the one used by Houle.229 The
226Boix, Miller and Rosato 2012. We use these authors’ expanded measure to 2015, which can be found
at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/FJLMKT. The results hold
using alternative datasets such as Democracy and Dictatorship by Cheibub, Gandhi and Vreeland (2010).
227Boix 2003; AR 2006; Houle 2009.
228see Knutsen 2015. The dataset can be found at https://stat.unido.org.
229Houle 2009, 2016.
100
data are available for 163 countries from 1963 to 2008. The advantages of our measure are
the large number of observed data points and its sensitivity.
Another widely used measure of inequality is the Gini coefficient, which captures the
distance between the largest and smallest individual incomes in society. However, the Gini
coefficient does not measure differences across groups or classes. Houle provides a good ac-
count of this distinction.230 In Appendix C, we include alternative models using (1) Houle’s
capital share variable and (2) Solt’s disposable income GINI index.231 More information
regarding these measures is provided there.
No yearly cross-national variable exists, to the authors’ knowledge, that measures elite
wealth accumulation. Ideally, our measure should capture the total amount of asset stock
owned by the economic elite.232 One option is to obtain the total amount of wealth in an
economy and multiply it by the share that belongs to the top centile. The closest approxim-
ation to this variable can be obtained from Piketty’s data.233 We can multiply his measure
of total private wealth by the share of wealth owned by the top decile or centile.234 Unfor-
tunately, the sample of countries in the WID data is small and dominated by countries that
have not experienced any recent transitions to democracy.
We construct a new measure using widely available data from the World Bank and con-
trasting a subsample with the WID data. First, we take yearly cross-national Gross Fixed
230Houle 2009. In Appendix C, we provide an an alternative test using Solt’s (2016) variable of inequality,
constructed using an adjusted Gini index and multiple imputation. More detail on this variable is provided
there.
231Houle 2009, 2016; Solt 2016.
232We broadly define the economic elite as top centile of an economy.
233WID from Alvaredo et al. 2018.
234Alvaredo et al. 2018.
101
Capital Formation (GFCF) in current dollars from the World Bank to estimate the total new
wealth generated by capital each year.235 This variable captures a country’s yearly total
domestic investment in fixed assets. It includes land improvements, machinery purchases,
private infrastructure construction (i.e. residential, commercial and industrial buildings),
and the net acquisition of valuables.236
For each country, we obtain the cumulative sum of new wealth for all years in the sample.
This gives us the amount of wealth generated in a given country-year from the time the coun-
try is first observed. It does not capture total wealth accumulation, as we cannot estimate
an initial value for wealth. We could add GDP or some similar proxy of national wealth as
the initial value, but that would introduce noise and generate its own set of issues, such as
lumping in population. To avoid these identification issues, we simply capture the pattern of
elite wealth accumulation within countries since the moment they are observed, rather than
estimating some initial value of wealth. If elite wealth grows rapidly, elites are more likely
to cash in and transition to democracy. If it does not, elites have an incentive to remain in
the status quo.
We make three substantive improvements to the cumulative sum of wealth described
above. First, countries generate increasing yet similar amounts of wealth year-to-year. The
cumulative sum of GFCF, therefore, experiences rapid growth the first few years and linear
235Data accessed on April 2018. It can be found at: https://data.worldbank.org/indicator/NE.GDI.FTOT.CD
236An in-depth description of the variable is included in Appendix C, including the World Bank’s own detailed
definition.
102
growth thereafter.237 To make wealth increase linearly across time, as in the WID data, we
replace each country’s first value of accumulated wealth with its fifth.238
Second, rather than dealing with gross numbers of wealth that are abstract and do not
truly represent the total amount of wealth available, we create a ratio: We divide each year’s
cumulative wealth by the amount of wealth the country produced the first year. Generating
a ratio of wealth allows us to capture precisely within-country elite wealth accumulation
and minimize the effects that cross-national variation may have in generating the variable.
The reason for this is simple: we consider that elites in one country are unlikely to compare
themselves with elites in other countries when deciding whether to transition to democracy.
Wealth is not the same in every country, given differences in prices and different pressures
on consumption, savings and so on. It is more useful to net out these differences for the
logic we advance in this chapter.
Lastly, we adjust the cumulative sum of wealth for depreciation and state investment. We
subtract 10 percent from each year’s wealth to account for depreciation239 and another 15
percent to account for the share of total investment by the government. We cannot remove
the share of wealth that belongs to the non-elites, but we contend that this is a small part
of the measure. New wealth is largely generated by capital holding elites. To make sure
these assumptions are substantiated, we run a series of stress tests on our final variable. We
compare WID’s data on total private wealth and elite wealth with a subsample from our
237Consider a country that creates 100, 110, 120, and 130 units of wealth in four years. The sum will be: 100,
210, 330, and 460 units. Growth from the first period to the second is 110 percent; from the second to the
third, 55%; and from the third to the fourth, 35%. After a few years, growth is linear on average both in our
data and in the WID measure. We refer to the WID measure sometimes as Piketty’s variable for reference.
238We chose the fifth value as growth becomes linear in the data on average after the fifth value of wealth.
239Piketty 2014
103
dataset.240 A high correlation with these known measures should provide some external
validity for our proxy.
The WID’s net private wealth variable captures the total level of wealth owned by private
citizens in a country-year. The correlation between this measure and our proxy is r = 0.949
on a subsample of 770 observations from 21 countries.241 We then multiply the WID’s net
private wealth variable by the share of wealth owned by the top 5, 1 and 0.1 percent. These
measures directly capture the wealth owned by different sets of elites. The correlation is
high for all three measures. We obtain r = 0.913 for the total wealth of the top 5 percent,
r = 0.925 for the 1 percent, and r = 0.946 for the 0.1 percent. These correlations include
389 observations for 11 countries.242 These high r values are indicative of the strength of our
variable in measuring our theoretical concept of elite wealth accumulation. We provide fur-
ther information about the variable and a detailed country-by-country comparison between
these variables in Appendix C. A description of the control variables is also included in
Appendix C.
240Our variable cannot capture all forms of wealth that elites may have, such as financial assets, stocks and
savings. Neither does it account for foreign ownership of assets. However, we contend that these omissions
do not harm the validity of our measure. The level and growth of these assets tend to be highly correlated
with the ones we use in our measure, and foreign ownership, while important, appears to have a small
impact on our measure overall. We provide as evidence of our logic the correlations with Piketty’s private
wealth data for a subsample of countries. These are all above 0.9 and are detailed below.
241These are the US, Canada, China, Russia, the United Kingdom, Australia, Czech Republic, Japan, Mexico,
South Africa, Netherlands, South Korea, Denmark, Germany, France, Greece, Italy, Spain, Norway, Finland
and Sweden. Within-correlation scores are stable across groups, with the exception of South Africa and,
to a lesser extent, Greece. The correlation stays high if each or both countries are excluded from the test.
242The US, Canada, Australia, Japan, Netherlands, South Korea, France, Germany, Italy, Spain and Sweden.
Within-correlation scores are stable across all groups.
104
4.7 Methods
We test our theory by interacting economic inequality and wealth accumulation and estim-
ating their joint effect on democratization. The expectation is that the interaction will be
positive and the marginal effect of wealth accumulation will be statistically significant at
high levels of economic inequality. These tests are reported in the results section.
Two modeling techniques are employed. First is the dynamic probit,243 which is com-
monly used to explore the relationship between inequality and democratization.244 We inter-
act each explanatory variable with the lag of the response variable and produce two coeffi-
cients, one for democracies and one for nondemocracies, for each independent variable. We
are not interested in democratic breakdowns, and therefore we only report the coefficients for
transitions to democracy in the tables. We use the dynamic probit over other logistic regres-
sion specifications for two reasons: including both types of transitions in the model reduces
bias.245 and including the lag of the dependent variable accounts for state dependence. The
second model we use is the linear probability model with country fixed effects (LPMFE).
Even though the dynamic probit’s autoregressive nature does away with much unobserv-
able heterogeneity, it still exploits between-variation. The LPMFE uses within-variation
and helps us be certain that time-invariant unobservables are not affecting the relationship
of interest. We present models with year fixed effects to net out any heterogeneity related
to specific years. We also include a time trend to account for the fact that wealth and events
of democratization increase naturally with time.
243Przeworski et al. 2000.
244Boix 2003; Houle 2009.
245see Metzger and Jones 2016.
105
We use six models in each test. The first four are dynamic probit specifications and
the last two are LPMFE models. First, we provide a naı̈ve model, which describes the
basic relationship between the main variables of interest. It also discards the option that the
relationship is spuriously created by adding controls, which inevitably remove variation.
Second is a full dynamic probit model with the relevant controls. The third model adds year
dummies and the fourth model includes both year and region dummies.246 The fifth model
is the fully specified LPMFE. The sixth adds year fixed effects to the LPMFE.
4.8 Results
Wealth
Inequality
0.02
0.01
0.00
1 2 3 4 5
Quintiles of Wealth and Inequality
Figure 4.2: Frequency of Transitions to Democracy at Different Quintiles of Wealth and Inequality.
We begin by showing the relationship between the two main variables of interest and demo-
cratization descriptively. Figure 4.2 plots the observed probability of a transition to demo-
cracy at different levels of wealth and inequality. This is obtained by dividing the number of
246Houle 2009; Miller 2012.
106
Observed Probability of a 
 Transition to Democracy
observed instances of democratization by the total number of observations that fall within
5 different quantiles of both independent variables. Democratization seems more likely at
high levels of inequality, although the probability begins to increase at medium levels. This
suggests a complex nonlinear effect. The effect of wealth accumulation is less conclusive
on its own, which matches our theory well. The relationship is nonlinear but differences are
inconclusive throughout. The difficulty in assessing the potential effect of both variables on
democratization suggests an interactive effect could be at play.
Our central hypothesis is that high inequality leads to democratization conditional on
rapid wealth accumulation among the elite, who seek to secure their wealth against ex-
propriation through the rule of law. We test it by interacting wealth accumulation with
economic inequality, using the models described above. Table 4.1 shows the results. All
variables have been centered at the minimum for easier interpretation. The two constitutive
terms of the interaction represent the effect of the variable on democracy when the other is
at the lowest observed value in our sample.247
The first two coefficients represent the effect of wealth and economic inequality on demo-
cracy when the other variables are at their minimum values. The statistical and substantive
significance of the two terms is trivial. The test of our theory lies in the interaction terms,
which are positive and statistically significant in all models, as our theory predicts. The
coefficient shows that a marginal increase in wealth has a statistically significant and posit-
ive effect on democratization conditional on higher levels of inequality, and viceversa. The
247Whether these terms are significant is not relevant for our theory, as we are not interest solely on the effect
of each variable at the lowest value of the other. The interaction term, coupled with the marginal effects
plot that follows the table, strongly support our theory.
107
DV: Democracy
Dynamic Probit LPMFE
1 2 3 4 5 6
Elite Wealth -0.829∗∗ -0.967∗∗∗ -1.239∗∗∗ -1.148∗∗∗ -0.018 -0.037
(0.414) (0.355) (0.376) (0.423) (0.030) (0.032)
Inequality -0.194 -0.327∗∗∗ -0.413∗∗∗ -0.437∗∗∗ -0.022∗ -0.027∗∗
(0.122) (0.110) (0.110) (0.119) (0.012) (0.012)
Elite Wealth × Inequality 0.112∗∗ 0.128∗∗∗ 0.165∗∗∗ 0.156∗∗∗ 0.008∗∗ 0.009∗∗
(0.055) (0.047) (0.050) (0.055) (0.004) (0.004)
GDPpc -0.082 -0.091 -0.128 -0.049∗∗∗ -0.054∗∗∗
(0.082) (0.091) (0.095) (0.012) (0.014)
Growth -0.026∗ -0.022 -0.021 0.000 0.000
(0.013) (0.014) (0.015) (0.001) (0.001)
Oil -0.102∗∗∗ -0.113∗∗∗ -0.095∗∗ -0.007∗ -0.007∗
(0.037) (0.042) (0.038) (0.004) (0.004)
Previous Trans. 0.276∗∗ 0.240 0.203 0.031∗∗ 0.034∗∗
(0.140) (0.146) (0.161) (0.014) (0.014)
Time Trend 0.019∗∗ 0.008 0.013 0.001 0.002
(0.009) (0.014) (0.014) (0.001) (0.004)
Ethnic fract. 0.042 -0.014 -0.039
(0.369) (0.381) (0.340)
Religious fract. 0.207 0.074 -0.163
(0.414) (0.443) (0.450)
British Colony -0.109 -0.066 0.149
(0.206) (0.219) (0.217)
New Country -0.656∗∗∗ -0.727∗∗∗ -0.484∗∗
(0.179) (0.188) (0.211)
Constant -0.363 1.080 2.005∗ 1.941∗ 0.518∗∗∗ 0.572∗∗∗
(0.901) (0.983) (1.086) (1.041) (0.114) (0.216)
Country FE - - - - Yes Yes
Year FE - - Yes Yes - Yes
Region FE - - - Yes - -
Observations 3336 3037 3037 3037 3050 3050
Pseudo R2 0.853 0.869 0.880 0.884
Overall R2 0.874 0.873
Log-Lik. -329.472 -266.877 -243.458 -236.403
* p < 0.1, ** p < 0.05, ***p < 0.01. Standard errors in parentheses.
All time-varying IVs have been lagged. GDP, elite wealth, and oil are logged.
Table 4.1: Effect of Wealth Accumulation and Income Inequality on Democratization
coefficients are consistent across dynamic probit specifications (Models 1-4), which means
that the interaction is robust to the addition of controls and year and region dummies. They
are also consistent in the LPMFE (5 and 6).
108
The coefficients for the control variables also display interesting findings and lend cre-
dence to the model specification. Growth is negative and close to statistical significance in
all dynamic probit models, and oil is negative and significant in all models.248 Both vari-
ables are known to affect democratization negatively.249 GDP per capita has a weak negative
effect in the dynamic probit models, as shown in Houle,250 and a stronger negative effect in
the LPMFE.251 Previous transitions, as expected, have a positive effect on democratization,
as countries that have democratized in the past are more likely to experience democratiz-
ation again. New countries that attained independence after de-colonization are also less
likely to democratize, which is consistent with previous research.252 We now assess the
substantive significance of these findings.
Figure 4.3 shows the marginal effects using Models 3 and 6 from Table 4.1. In Model 3,
the marginal effect of increases in income inequality on the probability of democratization
is significant at high levels of wealth accumulation. More specifically, a one unit increase
in inequality will generate a statistically significant change on the probability of democracy
when wealth accumulation surpasses a value of 4.1 in our data, which corresponds to the
83rd percentile. For a country with a value of wealth accumulation of 4.5, such as South
Korea in the mid-1980s, any one-unit increase in income inequality will lead to a seven
248The effect of growth is not significant in the fixed effects models because the controls are not interacted
with the lag of the dependent variable, only the interaction terms are. The effect cancels out for growth
in democratic survival (positive) and democratization (negative). The variable plays the same role as a
control.
249Wright 2008; Houle 2009; Ross 2001.
250Houle 2009, 2016.
251This is explained by the strong negative effect for nondemocracies after adding the wealth variable, which
has never been done in a previous analyses. Indeed, we are extending our work to understand whether the
endogenous effect of greater development is associated to democratization only insofar as elites become
wealthier and seek to democratize for safety. GDP may be best described as having an exogenous effect
(Przeworski et al. 2000), but further research on this issue is needed.
252Houle 2009.
109
(a) Model 3, Table 1 (b) Model 6, Table 1
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Wealth Wealth
Figure 4.3: Marginal effects of inequality on democratization at different levels of wealth accumu-
lation.
percent jump in the probability of democracy. If a country were to go from medium (50th
percentile) to high inequality (80th percentile), its probability of democratizing would in-
crease by 10 percent, a remarkably high probability for a rare event such as democratization.
The prediction in Model 6 is also substantively significant: for countries that reach exceed-
ingly high levels of wealth, such as 5 in our scale, the probability of democratizing climbs
20 percent with every one unit increase in inequality. These results serve as strong con-
firmation of our theory. At higher levels of wealth accumulation, increases in the level of
income inequality have a significant positive effect on the probability of democracy.253
253We show the marginal effects of inequality on democratization at different levels of wealth to model pre-
cisely what the formal model shows: as the risk of expropriation increases, elites decide to transition as a
function of their accumulated wealth. We could also show the marginal effects of increases in wealth at
different levels of inequality, and we do so in Appendix C.
110
 Marginal Effect of Inequality 
 on Democratization
-.1 0 .1 .2
 Marginal Effect of Inequality 
 on Democratization
-.05 0 .05 .1
4.8.1 Alternative Hypotheses and Robustness Checks
Perhaps the empirical relationship we identify in this chapter is driven by a particular form
of inter-elite conflict, namely, a new bourgeoisie that pushes for democracy against an es-
tablished elite that dominates political power. The politically disenfranchised elite obtain a
credible commitment against expropriation through democracy. This is in line with a the-
ory proposed by Ansell and Samuels.254 However, this theory only partially explains the
empirical findings in this chapter. It cannot fully explain the strong association between
wealth accumulation and democracy at high levels of inequality. This alternative explana-
tion would hold only if we assume that wealth accumulates primarily in the new rich, which
is unlikely. Rather, all elites continue to accumulate, and autocracy cannot be arranged in
such a way that it secures everyone’s riches from expropriation. The existence of multiple
wealthy elite groups creates uncertainty over who may take power. Once in power, new dic-
tators rely on their own coalition and expropriate the rest.255 If the established elite does not
transition to democracy, they risk losing everything to a new autocrat or revolution. Indeed,
many transitions are led by conservative elites in power.256 The risk of expropriation and
high wealth lead these elites to democratize and safeguard their wealth.
Our relationship could also be a function of time. The argument could be made that accu-
mulated wealth increases over time and, eventually, international pressures force countries
with greater wealth to democratize. We would still observe the effect of high wealth, but it
would in fact be trivial. Theoretically, this is difficult to justify given that, first, countries
254Ansell and Samuels 2010, 2014.
255Albertus and Menaldo 2012.
256Loxton 2015; Ziblatt 2017.
111
do not naturally democratize as their wealth increases. As we saw in Figure 4.1, there is no
apparent linear relationship between wealth and democracy. A threat to expropriation needs
to exist to persuade elites to cash in and democratize. We argue that inequality provides that
threat after an external shock. Empirically, we include a time trend in our model to rule
out this hypothesis. As expected, the time trend is positive in all models and statistically
significant in the fully specified model without time dummies.
Third, good institutions help elites accumulate wealth faster and also make it easier for
countries to transition to democracy, as they establish democratic qualities to a country ex
ante. While this argument makes intuitive sense, it does not appear to hold empirically.
British colonies would be expected to have better democratization outcomes, as the British
Empire generally replicated its institutional structure in their colonies. However, in our
models the association between British colonies and democratization is generally negative
or weakly positive in some models.257 It never reaches statistical significance. The same
occurs in Houle and Miller.258 To lend greater credence to the idea that institutions do
not drive democratization on their own and could explain the relationship we observe in
this chapter, we replicate Acemoglu, Johnson and Robinson’s seminal piece on the effect
of institutions on development,259 but using democracy in 2015 as the dependent variable.
The sign is positive but not significant in any of the models except for the model with no
controls. We show this in Tables C4 and C5 in Appendix C.
257AR 2013.
258Houle 2009; Miller 2012.
259Acemoglu, Johnson and Robinson 2001.
112
A final argument is that our relationship is driven by GDP per capita and the underlying
correlation between GDP and our wealth variable. It is true that countries with greater
GDP will generate more wealth, and that countries with greater GDP are more often than
not democracies. We control for GDP to rule out this hypothesis, but a high correlation
between GDP and elite wealth could lead to multicollinearity. Here, the correlation is quite
low at r = 0.207. This low r value rules out multicollinearity and makes it unlikely that the
relationship is driven by the similarity between these two variables. In the regression, the
variance inflation factor for the GDP control is 3.2, below any value that would be cause for
concern.
4.9 Conclusion
In this chapter, we propose a theory of democratization based on the role of democracy as in-
surance for elite wealth. We argue that transitions to democracy occur when economic elites
seek to secure their wealth against expropriation in authoritarianism, and pay a premium in
terms of higher taxes in exchange for the security of the rule of law. Democracy is more
likely at high levels of economic inequality because there exists an increased threat to prop-
erty. These threats are revolution from below and expropriation by a competing autocrat
who capitalizes on popular discontent.
The implications of our theory are substantial. First, this is the first works in the field to
address the interrelated role of wealth accumulation and economic inequality. Empirically,
we provide an alternative approach to testing these relationships, modeling the effect of
113
time more precisely and incorporating fixed effects specifications into the debate. We also
provide a novel proxy measurement for wealth accumulation.
Democracy is often understood as an equilibrium between competing forces that limits
conflict. Distinguishing the impact of modernization as a function of the benefit it brings to
the elite better elucidates the motivations of the actors vying to reach that equilibrium. In
doing so, we not only connect the bourgeoisie’s role in the process of democratization that
we find in Moore,260 but also explain why these groups remain powerful in new democracies.
In a 1919 letter to Hungarian workers, Lenin wrote that the ”bourgeois democracy is just
a specific form of bourgeois dictatorship,” in which groups with greater resources impose
their will on those who own less. We contend that elites are interested in democracy insofar
as it guarantees their economic interests better than any alternative system. Understanding
how democracy may benefit all elites rather than harm them helps us see democratization
in a new light.
260Moore 1966.
114
Chapter 5
The Productivity Paradox: How Sharp
Increases in Productivity Motivate
Rent-seeking and Erode Democracy
Why are developed democracies, which appeared impregnable just a few years ago,
backsliding? This chapter provides a theory of democratic erosion from a polit-
ical economy perspective that explains this puzzle. The central argument is that a
large productivity gap between economic groups motivates those with low productiv-
ity to use the state for rent-seeking. Expectations of lower wealth in the future force
these groups to seek control of the state, in an attempt to compensate for their losses
and prevent high productivity groups from continuing to increase the wealth gap. A
Caesarian leader emerges from within the party system or as a political outsider and
provides a political base and promises of rents to these elites in exchange for polit-
ical power. I use both quantitative and case study evidence from the US and Spain to
support my main hypotheses.
115
5.1 Introduction
Democracy’s expansion is over. No new democracies have emerged since 2008.261 Recent
autocratic reversals in Venezuela and Turkey, coupled with the election of leaders like Don-
ald Trump in the US and Andrzej Duda in Poland, have sent political scientists and the public
in general into a frenzy. Why are developed democracies, which appeared impregnable just
a few years ago, backsliding?
This chapter provides a theory of democratic erosion from a political economy perspect-
ive that explains this puzzle. The logic I advance is as follows. First, a political opportunist
capitalizes on the discontent among the middle class created by increasing wealth inequal-
ity. He then requires a certain level of elite support to win power, but elites are alienated
by sharp anti-establishment rhetoric. However, when the productivity gap among economic
sectors is large, groups whose wealth comes primarily from low productivity capital have an
incentive to capture the state and appropriate rents. A Caesarian leader exploits the situation
by providing a political base and the rents to these elites in exchange for political power.
Consider the cases of the Spain and the United States, two countries that have exper-
ienced some level of democratic erosion in the past few years. In Spain, leftist political
party Podemos emerged in 2014 to channel the political discontent caused by the domestic
housing crisis of 2007 and the global financial crash of 2008. Government spending cuts
and labor market reform in the private sector targeted low productivity sectors and workers,
who bore the brunt of the adjustment. Pablo Iglesias, Podemos’ charismatic leader, shot up
261See Boix, Miller and Rosato, 2012. Expanded to 2015 by the authors of the original dataset. Expanded to
2018 by the author of this manuscript.
116
in the polls with promises of a large increase in minimum salaries, the nationalization of
utility companies, and guaranteed monthly income for all citizens. His anti-elite discourse
crystallized in popular catchphrases such as la casta, a slight against individuals with ‘ped-
igree’ and influential family lineage. Given the historical roots of the conservative Popular
Party (PP) party in Franco’s Spain, most of the casta has direct or indirect ties to the PP. In
Spain’s case, the struggle for state control was between low productivity workers and a set
of elites that remained fairly united –no sectors were particularly productive in comparison
to others. These united elites gave the conservative Popular Party greater leeway to crack
down on opposition, increasingly restricting the right to protest through the ley mordaza
and free speech.262
In the United States, on the other hand, the productivity gap took the form of a pro-
nounced elite cleavage, brought on by globalization and technological progress: high pro-
ductivity sectors in technology and finance versus low productivity sectors in natural re-
source extraction, industry and, more specifically, manufacturing. The latter sectors were
against the reelection of the Obama-Clinton coalition, and eventually threw their support
behind Donald Trump rather than form a protective wall against a candidate with non-
democratic traits. Thus, in the US, polarization was most pronouced in the economic elite,
as opposed to among voters263 or the political elite.264
I use the term ‘democratic erosion’ to reflect the gradual process whereby democracy
drifts toward autocracy.265 It must be distinguished from democratic breakdown, which
262Gomez-Reino and Llamazares 2015, Orriols and Cordero 2016, Sola and Rendueles 2018.
263Svolik 2018.
264Bermeo 2003; Linz and Stepan 1972.
265Bermeo 2016; Lust and Weldner 2015.
117
reflects the return of an authoritarian regime where a country had been fully democratic. For
instance, Venezuela suffered a breakdown in 2006 and Turkey in 2016. Certain democratic
practices still exist in these countries, such as multi-party elections, but freedoms have been
curtailed to such an extent that competition is patently unfair and all parties know ex-ante
the result of the election. Conversely, while democratic quality is decreasing in advanced
democracies such as the US, Poland, France or Spain, these countries remain democratic. In
these countries, we are not discussing democratic breakdown, but the potential breakdown
of the democratic equilibrium. Nor are we observing the full dismantling of democratic
procedures, but the reshaping of institutions and systematic attacks on democratic norms.
Thus, to study democratic erosion or ‘backsliding’,266 a more expansive vision of democracy
is required.267
This chapter makes a few important contributions to the emerging debate on democratic
erosion and informs the debate on democratic breakdown.268 It is the first one, to the author’s
knowledge, to address the political economy angle of the recent trend toward democratic
erosion. It ties many important processes that have developed since the end of the Cold
War, such as greater globalization, factor mobility and technological innovation, into the
simple concept of the productivity gap. Indeed, one can think of the productivity gap as the
result of a global economic shift brought on by globalization, in which old industrial sectors
in developed economies are losing out to new and innovative sectors that yield high profits in
the global market. It also provides an explanation that links wealth inequality to both voter
266I use these two concepts synonymously.
267Lust and Weldner 2015; Svolik 2018.
268Bermeo 2016; Diamond 2015; Lust and Weldner 2015; Miller 2018; Svolik 2015, 2018.
118
mobilization by political entrepreneurs and the role of the elites as democratic gatekeepers.
In doing so, it places special emphasis on the ‘choices and choosers’ involved in backsliding
while accounting for important economic factors that influence institutional change.269 I
proceed as follows. After describing and defining the recent process of democratic erosion,
I develop my wealth inequality argument and support it using both quantitative and case
study evidence from the US and Continental Europe.
5.2 Democratic erosion in developed economies
Widespread democratic erosion in advanced economies is a recent phenomenon.270 Various
measures show that democracy expanded in the developed world until around 2012, when
USA
France
Spain
Poland
Brazil
1995 2000 2005 2010 2015
Year
Figure 5.1: Evolution of V-Dem scores by country since 1995.
269Bermeo 2016.
270Bermeo 2016, Lust and Waldner 2015.
119
VDem
0.60 0.65 0.70 0.75 0.80 0.85 0.90
a process of deterioration began.271 Democracies such as Poland or Hungary marched de-
cisively toward autocracy, while liberties were curtailed in other established democratic
polities. In France, the fight against terrorism by French nationals who joined ISIS promp-
ted the state to surveil its own citizens, often with weak evidence of wrongdoing. EU states
have responded to the new influx of immigrants fleeing conflict in Africa, the Middle East
and Central America with solutions of questionable democratic quality that contravene vari-
ous international treaties. Free speech is becoming less free. In Spain, in a rather comical
twist, two puppeteers were jailed for invoking the name of defunct Basque terrorist group
ETA in a 2015 show. A rapper was recently given a 3 year prison term for inflammatory
lyrics against the Crown. Poland’s Duda has been systematically attacking the judiciary in
an effort to wean its powers and act with greater impunity, a similar strategy to Orban’s in
Hungary. Figures 1 and 2 confirm this pattern. Figure 5.1 shows the evolution of V-Dem’s
continuous measure of liberal democracy for five developed countries. Democracy kept on
its slow but steady progress until around 2012, when a decline begins.272 Most countries’
scores have since decreased to unprecedented levels, and some to historical lows, as in Spain
and Poland. The US score is the lowest since Nixon’s presidency.
However, while democracies are losing their luster, the number of democracies and auto-
cracies in the world has remained stable in the last decade, as Figure 5.2 illustrates. Plot (a)
shows the evolution of democracy since 1950 and the sharp uptick in the number of demo-
cratic states during the third wave from the 1970s until the 1990s. Plot (b) then displays
271See the Varieties of Democracy (V-Dem) and Polity datasets.
272There may be a certain degree of coder bias in the V-Dem measures, which are scored retrospectively, but
broadly accurate patterns can be gleaned from these scores.
120
two flat lines that depict the worldwide democratic halt of the past decade. No democracies
are coming into existence, and no autocratic regimes are failing. Part of the explanation
for this puzzle is that declines in democratic quality are not sufficient for countries to lose
their democratic status. Democracies are becoming more like autocracies in weakening the
judiciary and curtailing civil and political rights, but they remain democracies. Thus, the
evidence shows that this is a problem of democratic erosion, not breakdown –at least not for
time being. While we may be on the way to notorious breakdowns, we are not observing
them yet.
(a) Since 1950 (a) Since 2005
1950 1960 1970 1980 1990 2000 2010 2006 2008 2010 2012 2014
Year Year
Figure 5.2: Evolution in the number of democracies from 1950 and 2005.
What is democratic erosion? Part of the problem is that, at least among developed eco-
nomies, backsliding is a novel phenomenon. As such, its theoretical boundaries are fluid.
Lust and Waldner define it as “a change in a combination of competitive electoral proced-
ures, civil and political liberties, and accountability.”273 Adopting a minimalist approach,
focusing only on elections, cannot help describe the myriad ways in which incumbents at-
273Lust and Waldner 2015.
121
Frequency
40 60 80 100 120
Frequency
80 90 100 110 120
tack the liberal aspects of democracy.274 Adopting too expansive a view of democracy would
lead to the conclusion that no democracies are such in the first place. I share Lust and Wald-
ner’s ‘middle ground’ approach, as it helps us analyze democratic erosion in terms of both
procedures as well as civil and political liberties.275
One thing is common across backsliding democracies: elections remain mostly free.
Cases of voter intimidation exist but are few and far between, and the vote is secret. Votes
are tallied up truthfully and electoral fraud is virtually nonexistent. Indeed, it has become
rare to steal elections on election day.276 However, elections are increasingly less fair. In-
cumbents are using their power to capture the process around elections, or what Bermeo
calls ‘strategic manipulation’.277 Incumbents seek to tilt the playing field in their favor,
using the media, government funds, or intimidation of the opposition as tactics. In the con-
tentious Catalan elections of December 2017, which came in the wake of a controversial
independence referendum in October, the Spanish government used its executive powers
over the region278 to tilt the election in favor of unionist parties. Pro-independence parties
had little or no access to advertisements on media outside regional outlets, which were in
turn banned from broadcasting any event, organized by political parties or civil organiza-
tions, that had a pro-independence message. Prime political advertisements in Catalonia
Square in Barcelona’s center were reserved exclusively for unionist parties, which in turn
had 1.5 times the budget of their rivals despite only having 36 seats in the Catalan Parlia-
274Zakaria 2011.
275Lust and Waldner 2015.
276Bermeo 2016.
277Bermeo 2016.
278As a result of the application of Article 155 of the Spanish constitution, the conservative government of
Mariano Rajoy took full control of Catalan institutions on October 27, 2017. The region had until then
been ruled by a pro-independence coalition that won the September 2015 elections.
122
ment —to the pro-independence coalition’s 72. Moreover, Catalonia’s two most influential
political figures, Oriol Junqueras and Carles Puigdemont, were unable to campaign: the
former was jailed without trial on November 2, while the latter avoided jail and fled to Bel-
gium and then Germany. Both countries refused to extradite him back to Spain on armed
rebellion charges, which carry a prison sentence of 30 years.
An important aspect of defining the recent downtrend in democracy as ‘erosion’ instead
of breakdown is that we can judge what is happening irrespective of its outcome for demo-
cratic survival. This is especially important in advanced democracies with strong institu-
tions, as incumbents require time and effort to subvert democracy. Incumbent takeovers as
a form of democratic breakdown exist but are rare, precisely because it is difficult to de-
termine when the process is complete.279 In Poland, for instance, Duda’s ruling right wing
Law and Justice Party (PiS) has sought to reorganize the country’s court system, reducing
judicial independence and increasing the party’s influence over key institutions such as the
constitutional court. In July 2017, Parliament passed a law that effectively dismissed all
justices on the constitutional court not appointed by Duda. Met with fierce public protests,
Duda vetoed the bill, but a new one was introduced in December 2017 lowering the justices
retirement age to 65 from 70, forcing 21 members, including its Chief Justice, to step down.
It also created a disciplinary committee for the constitutional court, which could be used by
the PiS to intimidate judges. Thus, defining attempts at a takeover is also complicated, as we
do not know the real motivations of suspicious incumbents such as Trump or Duterte. What
279Svolik 2018.
123
we can indeed observe is the fact that democratic quality is decreasing, and that is usually
accompanied by the rise of a political opportunist that concentrates power in his hands.
5.3 Shocks, economic inequality and the rise of a political
entrepreneur
The economic shock of 2008 had broad repercussions for developed democracies. Inequal-
ity rose to unprecedented levels, primarily because wealth accumulated in the elite at the
fastest rate since the late 1800s.280 Economic growth was slow in Europe during the re-
covery and faster in the US. In general, however, growth slowed in the period 2008-2018
as compared to the previous decade. New wealth accumulated to the richest individuals in
society and the middle class felt the economic squeeze through higher unemployment and
stagnant salaries. They are worse off in purchasing power today than before the crisis.
Moreover, the 2008 recession came at a time of deep transformation in the global eco-
nomy. Emerging markets have concentrated production for industries that used to be based
in developed economies, offering higher productivity levels and cheaper costs. Since the
turn of the new millennium, globalization and technological progress have shifted the pro-
ductivity spectrum, with high value added in services, technology and advanced financial
management, and low productivity in industry, manufacturing, and mining. The shift had
visible consequences for communities that, for years, had relied extensively on the latter.
Industrial and manufacturing jobs disappeared as industries settled in emerging markets;
280Piketty 2014.
124
retirement plans tied to stock prices, such as General Electric’s, lost half or more of their
value. The 2008 crisis compounded on these issues, accelerating the deterioration of these
communities through the loss of property values and widespread foreclosures. The sense
of loss that follows a crisis or a period of deep transformation fostered political discontent,
primarily in communities where low productivity sectors are located and the consequences
are more visible. Indeed, the rise of unemployment was marked in areas where populist
movements, from the left or the right, gained the most ground. It was also where traditional
parties or candidates were hit the hardest.
The discontent brought on by economic transformation and loss creates an opportunity
for the rise of a political entrepreneur. Traditional parties have difficulty capturing these
voters because they are accustomed to building broad coalitions that include a large portion
of moderate voters. With the belief that losing moderate voters is the death knell of any
political campaign, these parties are slow to react to a rapidly changing political landscape.
On the other hand, new, alternative candidates have an opening to capture these voters’
support by way of a new political brand that promises credible change. Not all succeed,
but the ones that do establish a base of support through on which they can build for future
election cycles.
Why do some political entrepreneurs develop an intent to subvert the system and take
power for themselves, i.e. become Caesarian candidates? Indeed, the 2008 crisis and its
aftermath has led to a proliferation of Caesarian leaders in advanced democracies, with
Trump in the US and Duda in Poland as standard-bearers. These candidates or leaders
125
use personal charisma or prestige as outsiders to gain the popular favor of the masses and
accumulate power in their hands once they take office.
I argue that deep crises like the Great Recession merge economic malaise with a pro-
found lack of belief in the system by voters, who blame it for bringing on the crisis. Tra-
ditional political candidates and elites lose the faith of many voters, who begin to blame
them for their travails. Outsider candidates who champion democracy as is are less likely
to woo these supporters. Fiery anti-establishment candidates, on the other hand, hit this
nerve forcefully and are able to build solid coalitions around this message. When economic
troubles are the result of regular economic cycles or external shocks for which foreign coun-
tries can be blamed, democracies (and political systems in general) are better able to cope
with discontent. For instance, one of the reasons Cuba survived the special period in the
early 1990s was that the leadership was hardly at fault for the fall of the Soviet Union. It
was surely to blame for creating the alliance in the first place, but not for the fall. It was
easier for the leadership to distance itself from the cause of the crisis and sell a message of
unity and perseverance against adversity.
The 2008 crash is, in many ways, the great failure of regulation and, to a lesser extent, of
democracy. It was perceived as a domestic issue in many countries, even though the crisis
originated in the US. Banks suffered globally because of their lax lending practices, and
bailouts were required in a majority of developed democracies. Bailouts, and the perception
that the rich got off unscathed and became even richer, were particularly harmful to people’s
view of democracy. In Spain, Bankia and other smaller institutions were rescued through a
private-to-public debt conversion mechanism orchestrated by the European Central Bank.
126
Public debt ballooned from 50 percent to 100 percent of GDP within one year. Bankia’s
case was particularly harmful, as it cost taxpayers 23 billion euros and the perception was
widespread that the bank paid few costs for its malpractice.281
This lack of belief in the system generates an opening for a new leader to propose strong
leadership to replace the weak and greedy political class; in essence, to offer voters less
traditional democracy in exchange for a potentially better future. Promises to rid the state of
corruption, replace the political elite, improve economic prospects with large-scale reform,
are common threads in these emergent leaders. Explicit voter support for these positions
paves the way for them to take office and accumulate power in their hands in a way that
previous leaders could not.
5.4 Economic elites and why Caesarian leaders triumph
Once Caesarian candidates have built a coalition of voters that can challenge in an election
credibly, they require elite support to win. Elites, however, are usually loathe to give it
precisely because these leaders are independent, unpredictable, and often espouse an anti-
establishment message. As the case of Podemos in Spain shows, it is difficult for alternative
candidates to win office and subvert the system if elites present a united front against them.
In other cases, candidates have been victorious partly because a set of favorable elites has
281Adam Przeworski points out that “something is wrong when a plurality of citizens in a democracy, asked
which institutions have most power in their country, answer ‘banks.’ Access of money to politics is the
scourge of democracy” (Przeworski 2016). He was referring precisely to the case of Spain, where a sur-
vey found that 32 percent of citizens considered banks to be the most powerful institutions, followed by
government at 26 percent.
127
supported them or, at least, allowed them to win. Why would elite groups support such a
candidate?
I argue that support from the elite is driven by the degree of polarization in the economic
elite. Classic works by Linz and Stepan and Bermeo identified elite polarization as the cent-
ral mechanism for twentieth century democratic breakdowns in Europe in Latin America,
but their accounts in fact reflected divisions among the political elite.282 Recent work by
Svolik focuses on the role of voter polarization in democratic backsliding.283 He argues that
highly polarized voters will opt for a candidate that fulfills their policy preferences over a
more democratic candidate that does not. However, considering the growing importance of
money in politics and the crucial role of democratic institutions in managing the diverging
interests of economic elite groups, a renewed emphasis on the motivations of the economic
elite may yield important insights into democratic erosion.
Economic elites are highly polarized when important subgroups cannot accept the elect-
oral victory of another group. In normal times, they may rue another group’s triumph over
theirs, which occurs frequently in democracy. But when polarization is high, they cannot
tolerate it. This is precisely what has occurred in the US, where a subset of the conservat-
ive elite decided it could not accept the victory of the Obama-Clinton coalition in 2016,
whatever the cost. Whether these groups’ beliefs are substantively justified is beyond the
point; to them, the Obama coalition and its debt-inducing policies, with universal health care
at the forefront, was an existential threat. Trump capitalized on this polarization to obtain
282Linz and Stepan 1978; Bermeo 2003.
283Svolik 2018.
128
the necessary backing of an elite group that saw him as the redemptor of the conservative
cause after Obama.284
In Spain, the rise of Podemos in late 2014 and early 2015 led to a delicate balancing act
by established economic elites. Podemos at the time were on the far left, inspired by the
popular protests of the 15th of May, 2011 and the movement they spawned. The party shot
up in the polls, becoming the favorite to win the national elections in most surveys around
November 2014. An El Paı́s poll had them ahead of the socialists at 27.7 and 26.2 percent
of the vote respectively. The right-wing Popular Party’s share decreased from 44.6 in the
2011 elections to 20.7 in that same poll. Another survey in January 2015 revealed that they
would beat the Popular Party in the general election by 3 points. At the time, Podemos spoke
of nationalizing industries and drastic increases in taxes on the rich to revert the post-crisis
fortunes of a battered middle class. The rise of an alternative, populist party put all elite
sectors on alert, and very few if any supported Podemos. Rather, a majority of elites saw the
need for a new conservative party to compete against Podemos and replace the traditional
Popular Party, whose corruption scandals were only adding fuel to the populists’ fire. As
a consequence, many established elites threw their support behind Ciudadanos and Albert
Rivera, a party built around two core concepts: rid the state of corruption and defeat the
nationalists movements in the Basque Country and, more importantly, Catalonia. As of
mid-2018, Rivera is leading in the polls.285
Thus, both in the US and in Spain, conservative elites at one point in time faced the pro-
spect of an intolerable opposition triumphing in an election. The rhetoric in conservative
284See Inglehart and Norris 2016.
285Gomez-Reino and Llamazares 2015, Orriols and Cordero 2016, Sola and Rendueles 2018.
129
circles, misguided or not, was that the Obama-Clinton coalition or the Podemos coalition
would wreak havoc on taxes, jobs, the national debt, and investment. It would lead to greater
centralization and bigger government, through nationalization (Spain) or greater debt, cli-
mate change policy, and universal health care (US).
5.5 Wealth, the productivity gap, and elite polarization
What makes elites polarized to the extent that they are willing to accept an alternative can-
didate who defends certain anti-democratic and anti-establishment positions? Following
Przeworski’s idea that money in politics tends to undermine democracy, I contend that a
particular type of cleavage in the economic elite fosters the rise of Caesarian leaders that
undermine democracy: the productivity gap.286
The basic idea around the productivity gap is that certain groups whose capital is tied to
low productivity sectors turn to the state for rent-seeking. Conversely, groups in high pro-
ductivity sectors continue to accumulate new wealth through innovation, and require only
favorable trade and monetary policy to continue expanding. The productivity gap among
economic sectors, therefore, generates divergent trajectories for their long-run wealth ac-
cumulation. High productivity sectors can expect high returns in the future, and thus their
expectation is that they will be much wealthier down the line. Low productivity sectors
are generating little new wealth from their capital investments. Low returns now and in the
future translate into an expectation for decreasing wealth and influence on policy.
286Przeworski 2016.
130
It is useful to think of this elite choice in game theoretic terms. Elite groups must consider
two broad states: the present state and some future state. Two main considerations exist in
these two states. First are returns to wealth and the second is policy influence. For low
productivity sectors, the present state is comprised of low returns on wealth, as defined
precisely by their low productivity. However, their political clout is still high, given the
important position they occupied in the economy for so long, their high current wealth, the
number of people they employ, and the lobbying connections they developed over time. The
future state, on the other hand, promises to be lower on both fronts if productivity stays low.
For high productivity sectors, the picture is reversed. They have high levels of productiv-
ity but exercise less political influence in the present state; these groups tend to be the result
of innovation and technology shocks and are thus newer, with a less well-developed lobby-
ing infrastructure. However, they expect their wealth and political clout to be higher in the
future provided their productivity stays high.
The diverging productivity paths for both groups affect their political calculus in the
present time. Low productivity sectors have to choose between the status quo or seeking al-
ternative ways to generate income and compensate for the low returns on their capital. High
productivity sectors face a similar choice, i.e. status quo or find alternative ways to generate
capital and maintain power, but their large gains in the present and high expectations for the
future eliminate any incentive they may have to depart from the status quo.
Low productivity sectors, on the other hand, are less convinced about the status quo.
All else equal, they would rather remain in a democratic system in which they have full
guarantees on their property against the whims of an autocrat. But they see their social
131
and economic position so threatened in the future state that they need to seek alternatives.
Reinvesting capital into more productive enterprises is one option. If the productivity gap
is large, new investments will require a long time to produce sufficient returns to close the
gap, and there is no guarantee that the gap will close. Another alternative is to take over the
state and rent seek.
Economic elites in democracy prefer to seek rents from the state when their payoff from
doing so in the future state is higher than investing in more productive assets. If they consider
that they will benefit more from rent-seeking than productive investments in the future, low
productivity groups seek to capture the state in the present period. This is contingent on
their ability to do so, but their greater political influence in the present time enables them.
Indeed, an added incentive to capture the state early is to have a first-mover advantage in
reallocating money and taking control of the coercive apparatus. Thus, one can think of
a continuum along which low productivity elites decide whether to support a democratic
candidate or one that helps them rent seek, as a function of their current ability to influence
the system. As low productivity makes a dent in these groups’ wealth, political influence
decreases and they no longer possess the ability to influence the system. Their optimal
strategy is to capture the state and rent seek before they lose the ability to do so if they
expect to gain more from rent-seeking than productive enterprises in the future.
5.6 Why Caesarian leaders subvert democracy
A high productivity gap provides incentives for some economic elites to take over the state
and obtain rents. However, these elite groups continue to face two issues: a collective
132
action problem, which stems from the doubts that certain elements may harbor about directly
subverting an established democratic system; and uncertainty over the outcome. I contend
that Caesarian leaders emerge to solve these two problems and provide low productivity
elites with a political platform to obtain state rents.
Caesarian leaders gain power by exploiting the situation created by greater elite polar-
ization and offering one group of elites certain policy concessions. Their rhetoric is often
pro-democracy as much as Caesar’s was pro-Republic. The latter famously stated that he
only established his Emperorship in order to fix the Republic’s ills, but that it was his in-
tention all along to restore the Republic as soon as he could. Modern strongmen follow a
similar tactic: they claim to be the only leaders suited to save the system from itself, but
find themselves concentrating greater power and undermining democracy once they win of-
fice. Established elites, faced with the prospect of an unacceptable opposition triumph or a
Caesar, choose the latter.
So far I have argued that polarization in the economic elite through a large productivity
gap facilitates the emergence of a Caesarian leader, who capitalizes on popular discontent.
But we need more clarity as to why these leaders, once in power, subvert democracy. Char-
acter and personality certainly play some role, and it is no coincidence that candidates with
weaker democratic convictions emerge in times of crisis, as I argued before. However, I
contend that there are two structural factors that drive incumbents to take over democracy:
(1) the platform that allowed the leader to rise is built precisely in such a way that the leader
accumulates power and enacts vast reform and (2) worries about post-tenure legal troubles.
133
First, political opportunists that emerge in a post-crisis scenario tend to espouse anti-
establishment rhetoric because rallying against the establishment wins votes. These chal-
lenges, however, require that these candidates make grandiose promises of reform. The
system must be changed wholesale, not piecemeal. Candidates raise the stakes to such an
extent that, should they win office, they must follow through on at least some of the more
notorious promises. Quick, wholesale reform is not possible within the bounds that strong
democratic institutions place on elected officials. Shortcuts are needed, and these shortcuts
are precisely what erodes democracy. I will go into these shortcuts in detail in my case
studies.
Another important and often overlooked reason why leaders have a necessity to fully
subvert the system once in office are preoccupations regarding their legal status after they
leave office. These leaders are likely to have committed certain illegalities either to reach
power or to achieve meaningful reform. While in office, they are faced with a choice to
subvert the system to the point that they can control reelection or, at least, maintain a faithful
following in the top echelons of the state apparatus, or place themselves at the mercy of the
opposition if the latter triumph. In this situation, Caesarian leaders stay the course and
continue to subvert democracy as much as they can to ensure their freedom and the safety
of their wealth once they leave office.
134
5.7 Empirical Analysis
Quantitative evidence is presented first, with tests for democratic erosion and breakdown.
Two case studies, on the United States and Spain, follow.
5.8 Data
The primary hypothesis is that, in democracy, high levels of inequality and a growing pro-
ductivity gap among economic sectors lead to democratic erosion. To test it, valid measures
of economic inequality, elite wealth, and democracy are required. For the latter, I use V-
Dem’s continuous measure of liberal democracy to probe whether wealth inequality and
accumulation interact to produce democratic backsliding.
I use two measures of economic inequality. First, I use the share of capital that accrues
to labor (wage share - WS), which captures cross-class differences between capital-holders
and labor. This variable is more fitting to theories that focus on inter-group inequality and
class cleavages.287 The higher the share of output that accrues to wages, the more equal the
society. The measure is taken from the INDSTAT2 dataset produced by the United Nations
International Development Organization.288 For the results to be intuitive, I have inverted
the WS variable. Now the greater the variable’s values, the higher the level of inequality –it
captures, in essence, the share of output that accrues to capital holders, a concept also by
287Boix 2003; AR 2006; Houle 2009.
288See Knutsen 2015. The dataset can be found at https://stat.unido.org.
135
Houle.289 I use data for advanced democracies from the end of the cold war until 2008 –the
last year for which data are available.
Another widely used measure of inequality is the Gini coefficient, which captures the
distance between the largest and smallest individual incomes in society. I use Solt’s adjus-
ted Gini index, which is one of the few measures to capture inequality levels in advanced
democracies until 2017.290 This compensates for the fact that WS data is available only
until 2008, and thus misses the period in which we have observed the greatest amount of
democratic backsliding. More information about this variable is included in Appendix A1.
I use productivity data from the OECD to estimate the productivity gap among sectors.
Data are available for the following sectors: finance, information and communication, pro-
fessional services, and retail are classified as high productivity sectors that fall under the
broader ‘business excluding agriculture’ umbrella; and industry, manufacturing and mining
which are group under industrial sectors. I compare the aggregate productivity of these two
broad sectors, business services and industry. I use data available from the end of the cold
war until 2017.291 To create the final productivity gap measure, I calculate the difference
in productivity between business services and industry for each observation (the unit of
observation is the country year).
289Houle 2009, 2016.
290Solt 2016.
291A majority of countries have data only up to 2015.
136
5.9 Methods
Tests for democratic erosion use V-Dem’s continuous democracy measure. I use random
effects OLS models to leverage both within- and between-variation across different coun-
tries. The main models are country random effects with clustered standard errors,292 with
year and region fixed effects. I also present the analysis using only OLS models with fixed
effects.
(1) (2) (3) (4) (5) (6)
Prod. Gap 0.174 -0.029 0.084 0.808∗∗∗ 0.311∗∗∗ 0.433∗∗∗
(0.628) (0.160) (0.200) (0.154) (0.106) (0.113)
Inequality (WS) -0.001 0.001 0.000 -0.001 -0.001 -0.003
(0.006) (0.001) (0.002) (0.003) (0.001) (0.002)
Prod. Gap × Inequality (WS) -0.030 -0.001 -0.017 -0.138∗∗∗ -0.056∗∗∗ -0.079∗∗∗
(0.143) (0.031) (0.044) (0.031) (0.020) (0.023)
GDPpc 0.002∗ 0.004∗ 0.003∗∗ 0.006∗∗
(0.001) (0.002) (0.001) (0.002)
Growth -0.000 -0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000)
Oil Prod. 0.001 -0.001 0.002∗∗∗ 0.001
(0.000) (0.001) (0.001) (0.001)
VDem Lag 0.953∗∗∗ 0.828∗∗∗ 0.728∗∗∗ 0.581∗∗∗
(0.012) (0.042) (0.057) (0.043)
Constant 0.760∗∗∗ 0.021 0.096∗∗ 0.836∗∗∗ 0.205∗∗∗ 0.309∗∗∗
(0.025) (0.016) (0.039) (0.015) (0.046) (0.045)
Observations 499 474 474 249 243 243
Ovr. R2 0.051 0.983 0.980 0.247 0.906 0.854
Table 5.1: Effect of the productivity gap and inequality on democratization
292Clustering is important in both random and fixed effects because nonrandom selection and treatment con-
tinue to affect the standard errors substantially (Abadie et al., 2017).
137
5.10 Results
Table 5.1 presents the results for the effect of the productivity gap on democratic erosion,
conditional on levels of inequality as measured by the WS. Models 1 through 3 show the res-
ults for the full sample, which includes all democracies irrespective of their VDem score.
Models 4-6 include advanced democracies only. The subsample in the latter models in-
cludes countries with a V-Dem score greater than 0.8: Australia, Austria, Belgium, Canada,
Chile, Costa Rica, the Czech Republic, Denmark, Estonia, Finland, France, Greece, Ireland,
(a) All Democracies, Mod.2 (b) Advanced Democracies, Mod.5
2.5 3.5 4.5 5.5 6.5 7.5 8.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5
Inequality (WS) Inequality (WS)
(c) All Democracies, Mod.3 (a) Advanced Democracies, Mod.6
2.5 3.5 4.5 5.5 6.5 7.5 8.5 2.5 3.5 4.5 5.5 6.5 7.5 8.5
Inequality (WS) Inequality (WS)
Figure 5.3: Effects of the productivity gap on democracy conditional on inequality.
138
 Predicted Change on V-Dem Score  Predicted Change on V-Dem Score
-.04 -.02 0 .02 .04 -.03 -.02 -.01 0 .01 .02
-.04 -.02 0 .02 .04 -.04 -.02 0 .02 .04
the Netherlands, New Zealand, Norway, Poland, Portugal, Sweden, the UK and the US.293
The differences are stark, primarily in the interaction term between the productivity gap and
inequality. The effect of greater differences in productivity as inequality increases is slightly
negative and not statistically significant for the full sample, but it becomes steeply negative
and statistically significant when we include only the subset of advanced democracies. In
Models 4-6, greater elite wealth as inequality increases significantly decreases the predicted
V-Dem score for a given country.
Figure 5.3 displays the results from Table 5.1 substantively. Plots (a) through (d) show
the marginal effects of increases in the productivity on the V-Dem score at different levels
of inequality. For all democracies, the effect is weakly negative and is not statistically signi-
ficant at any level of inequality, both the the random effects –plot (a)– and fixed effects –plot
(c)– specifications. In the subsample for advanced democracies only, however, the effect is
sharply negative and statistically significant for countries with high inequality, consistent
with Table 5.1. In plot (b), as inequality increases in an advanced democracy, any one unit
increase in the productivity will produce a substantive negative change in the V-Dem score.
For instance, an advanced democracy with inequality levels at 7.5 will see its V-Dem score
decrease by 0.012 units for each unit increase in the productivity gap. In the sample for ad-
vanced democracies, the V-Dem score ranges from 0.8001 to 0.8922, so a decrease of 0.012
represents a 13 percent change in the dependent variable. If the productivity gap increases
from 1 to 3, the loss in the V-Dem score grows to 0.024, or 26 percent.
293A score of 0.8 represents the median score for the sample.
139
In plot (d), the model with country fixed effects, the interactive effect of differences in
productivity and inequality is even starker. A one unit change in the productivity gap at
high levels of inequality (7.5) will produce a decrease of 0.016 in the predicted V-Dem
score. A two unit will lead to a loss of 0.032 in the value of democracy. These figures
represent, respectively, a 17.4 and 34.7 percent change within the range of the V-Dem score
in the sample. The results are consistent with country-specific evidence: for example, the
US peaked at 0.88 in its V-Dem score in 2012, and decreased to 0.83 in 2016, a substantive
decrease that is similar to the one predicted by the model.294
Results obtain if we use Solt’s Gini index measure of inequality, which extends until
2017, as shown in Table 5.2. For the full sample of democracies, the effect of the productiv-
ity gap on democracy is positive as inequality increases, as the interaction term shows. It is
statistically significant at the 0.1 level for the two fully specified models (1 and 2). However,
as with the results in Table 5.1, the sign of the relationship reverses when only advanced
democracies are included in the sample. For these countries, as inequality increases, a lar-
ger productivity gap leads has a negative and statistically significant effect on democracy
–models 4-6.
5.11 Wealth Inequality and Democratic Erosion in Spain
Democratic declines in Europe are apparent: since 2012, a majority of countries have
suffered a regression in their democratic quality. Figure 5.4 displays the evolution of V-
294Again, remember that the inequality measure using WS is only available until 2008, but the predictive power
of the model is consistent with events that occurred after the crisis. I use Solt’s measure to produce results
that are time-consistent with the events of the past decade.
140
(1) (2) (3) (4) (5) (6)
Prod. Gap -0.078 -0.063∗∗ -0.049∗ 0.166∗∗∗ 0.096∗∗ 0.110∗
(0.103) (0.032) (0.026) (0.052) (0.047) (0.054)
Gini Market Income 0.182 -0.034 -0.013 0.198∗∗∗ 0.011 0.050
(0.227) (0.023) (0.064) (0.053) (0.020) (0.057)
Prod. Gap × Gini Market Income 0.175 0.126∗ 0.098∗ -0.348∗∗∗ -0.205∗∗ -0.234∗
(0.218) (0.068) (0.056) (0.110) (0.104) (0.119)
GDPpc (log) -0.001 -0.004 0.002∗∗∗ 0.003
(0.001) (0.003) (0.001) (0.002)
growth1 -0.000 -0.000 -0.000 -0.000
(0.000) (0.000) (0.000) (0.000)
vdemlib1 0.972∗∗∗ 0.872∗∗∗ 0.676∗∗∗ 0.485∗∗∗
(0.022) (0.037) (0.055) (0.053)
Constant 0.669∗∗∗ 0.054∗∗∗ 0.147∗∗∗ 0.760∗∗∗ 0.257∗∗∗ 0.392∗∗∗
(0.110) (0.017) (0.035) (0.025) (0.047) (0.039)
Observations 899 867 867 238 235 235
Table 5.2: Effect of the productivity gap and inequality on democratization using Solt’s Gini.
Dem scores for the UK, France, Denmark and Spain. All countries decline after 2012, two
of them to the lowest levels since 1995 – Denmark and Spain. France is close to its 1995
score after dropping by almost 7 percent since the election of François Hollande. The UK
has decreased 11 percent since 2012, and Spain’s V-Dem score of 0.7 puts it on par with
France, the UK or Ireland in the 1960s.
My theory predicts that an inequality shock provides an opportunity for a political oppor-
tunist to obtain popular support and attempt to take power, but that elite support is required
to win a general election. Elites support an outsider when levels of elite wealth are so high
that groups believe they can obtain their preferred policy from government and no longer
need to compromise. This increases polarization among the economic elite, who prefer a
less democratic candidate over the victory of the opposition. A Caesarian leader exploits
141
UK
France
Spain
Denmark
1995 2000 2005 2010 2015
Year
Figure 5.4: Evolution of V-Dem scores by country since 1995.
this situation to obtain both popular and elite support, win an election with a certain prob-
ability, and subvert democracy once in power.
The case of Spain is a good example of when the lack of elite support prevents a popu-
list, potentially Caesarian leader from emerging. This shows precisely why the second stage
of my argument, elite polarization through high levels of wealth, is so important in foster-
ing or hindering democracy. Spain suffered a double crisis between 2008 and 2010. First,
the housing bubble began to burst in late 2007 after years of rapid growth. The ECB had
maintained interest rates low at 2 percent from 2003 to 2006 to help Germany recover from
its recession.295 With growth at over 4 percent for the period, Spain’s economy overheated
through the housing market, as banks borrowed cheap money from the ECB and packaged
295Reisenbichler and Morgan 2012.
142
VDem
0.70 0.75 0.80 0.85 0.90
it into easy to obtain loans for consumers. This left banks over-leveraged and consumers
unable to pay back their mortgages after the bubble burst and the effects of the Great Re-
cession began to be felt. The recipe for recovery by the ECB were cuts in public spending,
labor market reform, and the nationalization of private bank debt.296
These reforms cut deep into the middle class’ pockets and clawed back some of the
welfare gains that had been made since the advent of democracy in 1978. The discontent
erupted into large scale protest on May 15, 2011. Millions of citizens marched onto the
streets in the country’s main cities and occupied the main squares for months. The move-
ment coalesced around a new political party, Podemos, founded in January 2014 under the
charismatic leadership of Pablo Iglesias. At inception, the party espoused radical left wing
views, some of which were hotly debated. Among these were the nationalization of large
corporations in the energy sector, a large increase in the minimum wage, and a guaranteed
minimum income. Its discourse was overtly anti-establishment, popularizing the concept
of la casta297 to describe established elites who, in their view, had ran the country to the
ground.298
The party first burst onto the scene in the May 2014 elections to the EU parliament,
just three months after it was created. It won 5 seats and 8 percent of the vote, finishing
fourth. General elections loomed in late 2015, and it appeared that Podemos would sweep
into power. Polls in late 2014 and early 2015 had them comfortably ahead of established
parties like the conservative Popular Party (PP) and the socialist PSOE. Pablo Iglesias was
296See Sinn 2014.
297Casta refers to pedigree, to aristocracy, to the elite, and also to chastity and purity. Podemos popularized it
as a pejorative label for established elites.
298Gomez-Reino and Llamazares 2015, Orriols and Cordero 2016, Sola and Rendueles 2018.
143
the most popular leader in the country, had constant media presence, and the sense was that
he would become, in all likelihood, the next president.299
However, Podemos’ project attracted few elites. Promises to nationalize large corpor-
ations were particularly harmful to build an elite coalition, as they increased uncertainty
around which companies would be placed on the state’s books should the party govern.
Large-scale wage redistribution became one of Podemos’ core messages, but few elites
warmed up to the idea of a substantial increase in costs. The party’s elite connections were
also initially weak. Pablo Iglesias was a university professor before his switch to politics,
and most of the executive committee came from similar non-elite backgrounds. Most were
university professors, researchers, and activists. Among the first representatives Podemos
sent to the European Parliament were a high school teacher, a small business owner, and a
researcher, all of whom had no previous experience in politics.
In essence, Podemos represented the low productivity working class who suffered the
brunt of the economic readjustment imposed by the European Union after the financial
crisis. Thus, its appeal was based on a logic of class struggle, in which low productivity
workers from the middle and working classes saw their wages slashed and social benefits
cut as elites sought to create a leaner and more productive economy. Elites, however, were
united in how they approached their future wealth generation and the role of democracy
in that process. No industries were obtaining the ultra high productivity margins of US
technology corporations, and all were more or less affected by the double set of crises that
affect the Spanish economy – the housing crash of 2007 and the global financial crisis of
299Sola and Rendueles 2018.
144
2008. Banks were particularly weakened by large amounts of toxic mortgages handed out
during the 2003-07 boom. Realignment was inevitable, and an entire network of local and
regional savings banks were dismantled or absorbed by national banks after large bailouts.
A majority of the country’s top banks required assitance, with Bankia’s 23 billion bailout
package the most conflictive of all. To a lesser extent, La Caixa (now CaixaBank) and Banc
Sabadell also required assistance, and only Banco Santander navigated the crisis relatively
unscathed.
Large Spanish construction corporations were heavily affected by the housing crisis and
the temporary slowdown in public projects in the years that followed. National utility and
telecom corporations, such as Telefonica or Gas Natural, continued to enjoy dominant mar-
ket positions that would continue no matter the party in power. Clothing giants like Inditex
(known for their Zara brand) and Mango had their sights in the global market and looked at
Spanish politics with relative disinterest. Thus, perhaps in part because the housing crisis
had ripple effects on almost every part of the economy, there was no obvious elite cleavage
that Podemos and its Caesarian candidate, Pablo Iglesias, could exploit.
Elites, however, did see the danger that he posed should he obtain sufficient electoral sup-
port to govern –or at least wield influence within a multiparty coalition government. To pre-
vent this, elites presented a united front and used the tools in their disposal to stop Podemos
from rising further. Established political elites contributed to this. Socialists were particu-
larly weary of Podemos, who were agitating their core base of supporters and threatening
to win them over. Conservative elites saw Podemos as an outside movement that threatened
to break up the constitutional order, which had long been upheld by the traditional parties,
145
PP and PSOE. It certainly did not help matters that Podemos was unapologetically repub-
lican, and in multiple occasions intimated that the royal family should be removed from
their ceremonial office. National newspapers from the left and the right equally attacked or
ignored Podemos, as did all television outlets except, to some extent, La Sexta. The media
were particularly keen on the romantic and practical attachment that some members of Po-
demos had with Venezuela. Iglesias’ personal admiration of Hugo Chavez and Fidel Castro
was duly reported and debated ad nauseam. The beginning of the end for Podemos were
multiple allegations agains Juan Carlos Monedero, who at the time was the party’s most
influential member after Iglesias and his number two, Iñigo Errejón. On January 27, 2015,
with Podemos leading the polls, El Paı́s alleged that Monedero’s academic background and
degrees were false, and even reported that professors he cited as references claimed not to
know who he was. The accusation was recanted by the newspaper just three days later.
Then, on January 28, came the most damning report: Monedero had received payments
from Chavez’s government and other ALBA countries, for consulting services. Monedero
resigned in March, 2015, from his post within the party.
5.12 Wealth Inequality and Democratic Erosion in the United
States
According to V-Dem, democratic erosion in the United States began during Barack Obama’s
second term –see Figure 5.1. Democratic quality declined steadily between 2012 and 2016
before a sharper downturn in 2017. In popular discourse, democratic erosion in the US is
often associated with the rise of Donald Trump, but its roots run much deeper. The fast
146
rise of the Tea Party in the early 2010s increased political polarization in congress, forcing
President Obama’s hand into signing a slew of executive orders that would previously have
been resolved within the legislative branch.
Increased polarization within the legislative seemed to have played a key role in the early
phase of democratic erosion in the US. This is reminiscent of arguments by Linz and Stepan
and Bermeo.300 However, greater use of executive action, while constitutive of democratic
erosion and thus reflected in the data, is not a systematic attack on democratic institutions
and values. There is more evidence that this is occurring under Donald Trump301, and the
rise of the current President is intimately tied to polarization within the economic elite rather
than the political elite.
Indeed, the steepest decline in the United States’ V-Dem score takes place in 2017, the
first year of Trump’s presidency. Trump has attempted to reshape institutions in a way that
nullifies referees, such as courts, push out actors that could threaten him, and slowly un-
balance the playing field. Indeed, the tactics and discourse employed by Trump closely
resemble the classic authoritarian script.302 More importantly for our theory, these traits
already were in full display during the 2015/6 republican primary and presidential cam-
paigns. A populist, nationalist message resonated with a portion of the electorate that iden-
tified their economic travails with perceptions of increased immigration, cheap imports,
and eight years of rule by Democrats in Washington, DC.303 Furthermore, as happened
with Spain’s Podemos at the earliest stages of its emergence, Trump promised to ‘drain
300Linz and Stepan, 1972; Bermeo, 2003.
301Levitsky and Ziblatt 2018.
302Levitsky and Ziblatt 2018.
303Inglehart and Norris, 2016.
147
the swamp’, a catchphrase akin to la casta. Both describe the intention to change the elite
make-up of the upper echelons of political power in Madrid and Washington. Thus, as the
theory predicts, Trump exhibited all the traits of a Caesarian leader during the presidential
campaign, in a way not dissimilar to Pablo Iglesias –even if both leaders were, at least in
principle, at different points in the left-right ideological spectrum.
In Spain, as we saw, economic elites were fairly compact and blocked Podemos’ rise with
campaigns aimed at decreasing its popularity among voters. They then backed Ciudadanos
extensively in an attempt to create the right-wing equivalent of Podemos, using a similar
populist discourse around calls for cleaning up corruption and democratic ‘regeneration’.
However, in the United States, not only did Donald Trump capture the Republican nomina-
tion, but he also won the Presidency in October 2016.
I contend that cleavages among the elite were at the core of Trump’s emergence and
eventual victory in the United States. In particular, an increasingly large productivity gap
between technological/financial elites and industrial/natural resource elites produced an
opening for Trump to obtain sufficient elite support to mount a credible challenge to es-
tablishment Republican (and Democratic) elites. Among Republicans, the push by Charles
and David Koch to make the Republican party embrace libertarian ideology took a practical
turn in the mid-2000s, when the brothers set their minds to capturing Republican majorities
in state legislatures. They observed that a majority of regulation affecting their direct busi-
ness interests was produced in the states, not in Washington. Dominating state lawmaking
bodies would translate into fewer regulations and lesser government meddling into business
affairs. The focus on the states, however, left the national party leadership hollow, and it
148
became too late to pivot toward greater national presence late in Barack Obama’s second
term.304
While the ideological elite of the Republican party abhorred the rise of political outsiders
–Trump and Ted Cruz– in the primaries, and the Koch brothers were no exception, two
subsets of the Republican elite favored the nationalist rhetoric that the candidates offered.
The first were primarily natural resource extracting elites in states such as West Virginia
and Texas, as well as others who wanted to see projects such as the Keystone Pipeline not
sidelined by environmental regulations. The second subset of elites amenable to Trump
were US manufacturers who had lost out with globalization and who saw potential tariff
increases and trade disruptions as a positive for their future standing.305These elites stood in
stark contrast with a high productivity technology sector that is largely behind Democratic
candidates and who espouse openness both in terms of trade and social values. To gain an
idea of the differences among sectors, the five most valuable companies in 1990 were all
industrial, manufacturing and natural resource giants –companies such as General Electric,
General Motors, Ford, Exxon and Mobil. In 2015, the five most valuable companies in the
United States were all located in Silicon Valley or were part of the new post-2000 technology
wave –Apple, Google, Amazon, Microsoft and Facebook.306
For elites that were early Trump supporters, the calculus was not only about maximizing
wealth in the present time. Indeed, the productivity gap across sectors makes differences in
304See Skocpol and Hertel-Fernandez, 2016, for a detailed account of the Koch brothers’ effect on Republican
ideology and strategy in the last two decades.
305See https : //www.nytimes.com/2016/03/28/us/politics/donald − trump − republican −
voters.html
306According to Fortune 500: http : //archive.fortune.com/magazines/fortune/fortune500 archive/full/1990/;
and Forbes Magazine: https : //www.forbes.com/powerful − brands/list/.
149
wealth larger across time, given that wealth grows exponentially. Supporting a Caesarian
candidate that espouses certain undemocratic views may be accepted by low productivity
elites in exchange for state rents that help cover the productivity gap with other sectors. In
sum, the case of the United States illustrates how large productivity differences among the
economic elite provide an opening for a Caesarian candidate to obtain sufficient elite sup-
port to mount a credible political challenge to established political elites. In both Spain and
the United States, high levels of inequality fueled voter discontent, which in turn gave new
candidates like Pablo Iglesias and Donald Trump a political base.307 However, differences
in elite support for these candidates determined their different political fates. A large pro-
ductivity gap created an opening for Trump, whose nationalist rhetoric translated into early
support from elites in manufacturing and natural resource sectors. Other Republican elites
later joined his cause rather than shut him out of the process precisely because it was too
late; the early support was sufficient to push him through. In the case of Pablo Iglesias, a
compact economic elite suffering from a small productivity gap made a concerted effort to
prevent Podemos from reaching power and succeeded, even if at some point Iglesias had a
larger share of voter support in Spain than Trump did in the United States for most of his
campaign.
307Inglehart and Norris, 2016.
150
Chapter 6
Appendices
151
6.1 Appendix A
Table A1: Snippet of the Data with Examples from The USA and Cuba (after expanding)
Country Name Year t t0 Capital Share 1->2 Capital Share 1->3 Capital Share 3->1 Strata
United States 2000 42 41 0 0 65.935 3
United States 2001 43 42 0 0 68.011 3
United States 2002 44 43 0 0 69.585 3
...... ...... ...... ...... ...... ...... ...... ......
Cuba 2000 42 41 65.603 0 0 1
Cuba 2000 42 41 0 65.603 0 2
Cuba 2001 43 42 64.966 0 0 1
Cuba 2001 43 42 0 64.966 0 2
Cuba 2002 44 43 65.260 0 0 1
Cuba 2002 44 43 0 65.260 0 2
Building the dataset:
We set our time counter, t, is set to increase by 1 every year within each spell. Then, a
variable t0 is generated to represent the time at which each observation enters a new spell
of one year –it is akin to, quite simply, t − 1. This way, instead of collapsing each spell
into one single observation and setting t to be the entire duration of that spell, we give t a
starting value of 1, and t0 a starting value of 0. They both then increase by 1 each year within
the spell and are reset once a new spell begins. Each independent variable is then forked
into three separate variables, each representing every possible transition. These variables
are set to zero for the irrelevant ‘strata’ (for instance, the value for the United States for the
authoritarian replacement capital share variable is 0 because the country is not at risk of
experiencing a transition to democracy). The actual values of the independent variables
are kept for the transition for which they are relevant (again for the United States, its capital
share values are kept only in the variable that represents transitions from democracy to
authoritarianism, as this is the only transition the country is at risk of experiencing). The full
model, therefore, produces coefficients for three different variables, one for each transition,
for each independent variable in the sample. Since the process is modeled in its entirety, and
each type of transition (the strata) receives its own baseline, we can produce more accurate
estimates for each coefficient.
Two additional variables are created: a dichotomous status variable, coded 1 for country-
years in which an event occurs and 0 otherwise. A second variable, for the strata, is created
grouping all those observations that are at risk of suffering each type of transition and is
coded 1 for country-years at risk of transitioning to authoritarianism, 2 for those at risk of
democratization, and 3 for those at risk of democratic breakdown. These ‘strata’ will be
used to construct our independent variables as well.
Summary Statistics:
Variable Mean Std. Dev. N
Wealth 8.043 2.55 6912
Capital Share 65.157 10.896 7362
GDPpc 2.739 1.091 6817
Growth 1.776 6.152 6809
Oil 0.204 0.403 8306
Ethnic Frac. 30.05 24.436 7944
Trade Open. 74.925 46.567 7301
Previous Repl. 0.363 0.888 10764
Polity 0.165 7.484 6617
Polity Sq. 56.032 31.953 6617
Table A2: Summary statistics (before expanding)
As Figure A1 shows, the results are unchanged if all controls are removed. Significance was
not induced by any one specific variable or set of variables into the model.
Alternative models
153
0.2 0.3 0.4 0.5 0.6 0.7
Wealth Inequality
Figure A1: Marginal effect of decreasing levels of wealth on replacement at different levels
of wealth inequality, no controls.
Figures A2 and A3 present the results of random effects models, which are my main ro-
bustness check. Here I run a linear probability model with random effects by country (A2)
and a logistic regression model also with random effects by country (A3). The dataset for
these models includes only transitions from one authoritarian regime to another —that is,
only those observations in the first strata of the sample used in the main results section.
Consequently, the sample is reduced to 2,455 observations. I use a random effects model
over a fixed effects specification. An F-test of the null hypothesis that the fixed effects are
not zero indicates that it cannot be rejected. This means that the primary assumption of the
fixed effects model is violated, and a random effects model is preferred. This is confirmed
by a Hausman test, which shows that the random effects and fixed effects estimates are not
significantly different from each other. The null hypothesis that the individual effects are
uncorrelated with the variables of interest cannot be rejected and, therefore, the random
effects model is more efficient. The results are consistent with Figure 3 (a) in the main pa-
154
Marginal Effect of Wealth on
Authoritarian Replacement
−0.4 −0.2 0.0 0.2 0.4 0.6 0.8
per. At high levels of wealth inequality, a lower level of wealth will increase the probability
of authoritarian replacement significantly (remember that the variable that captures wealth
levels is flipped, and the line is thus positive).
0.2 0.3 0.4 0.5 0.6 0.7
Wealth Inequality
Figure A2: Joint effect of wealth and inequality on authoritarian replacement. Displays the
predicted probability of replacement using an LPM with random effects and clustered stand-
ard errors.
155
Marginal Effect of Wealth on
Authoritarian Replacement
−0.010 −0.005 0.000 0.005 0.010 0.015
0.2 0.3 0.4 0.5 0.6 0.7
Wealth Inequality
Figure A3: Joint effect of wealth and inequality on authoritarian replacement. Displays the
predicted probability of replacement using logistic regression with random effects and
clustered standard errors.
Below, in Figure A4, I report the predicted probabilities using a simple logit specification
without random effects. The reason for showing the results of a simpler model is to reject
the possibility that the findings are generated solely as a function of the more complicated
models used. The results are also consistent with what we found using the other models.
156
Marginal Effect of Wealth on
Authoritarian Replacement
−0.4 −0.2 0.0 0.2 0.4 0.6 0.8
0.2 0.3 0.4 0.5 0.6 0.7
Wealth Inequality
Figure A4: Joint effect of wealth and inequality on authoritarian replacement. Displays the
predicted probability of replacement using logistic regression.
6.2 Appendix B
Model with no ability to capture democracy
To further test our formal model, we assume that elites are not able to capture democracy
by giving a value of 1 to the capture term, φ. In this scenario, elites pay the full tax rate
in democracy. Thus, similar to equation 6, elites will prefer to transition to democracy in
state (b, A), maximizing π to 1, if Ṽ(b, A, w, θ) < ṽ(b, D,w, θ). Solving this, elites in
bad times will transition to democracy when
[ ]
δ s
(γ(θ)) ln(w (1 + r − c)) ≥ − 1 + − ln(1 − φ), (6.1)1 + δ(1 s)
157
Marginal Effect of Wealth on
Authoritarian Replacement
−0.4 −0.2 0.0 0.2 0.4 0.6 0.8
that is, when the cost of remaining in an authoritarian regime is greater than the cost of
transitioning to democracy. Figure A1 reports the decision rules for the elite after an external
shock, assuming there is no capture term. Following Figure 1 in the mainuscript, the wealth
parameter w in the model is on the x-axis and the expected cost in each state on the y-
axis. The straight lines report the cost of remaining in authoritarianism at different levels
of inequality and wealth (left side of equation 1.) When the cost of remaining in autocracy
surpasses the cost of democracy, elites choose to transition to democracy. We represent this
with a shaded gray area.
The curves show that the relationship between levels of inequality and transitions at dif-
ferent levels of wealth is maintained. Note, however, that it takes more wealth for democracy
 Transitions to Democracy at Levels of Inequality and Wealth
Cost of Democracy
Cost of Autocracy
High Inequality
Med. Inequality
Low Inequality
0 20 40 60 80
Wealth
Figure B1: Democratization as a function of wealth and inequality
158
0.0 0.2 0.4 0.6 0.8
to be attractive enough for elites to decide to transition, which makes intuitive sense. The
less beneficial democracy is to elites, the less likely they are to transition. Thus, our theory
holds for this specification of our model, but we still believe that a more realistic assumption
takes into consideration the limits of nascent democracies at providing a leveling playing
field for all participants in society. It is more likely that elites manage to create a system
from which they benefit in some way, however small. If assuming a cost 1−φ of 95 percent,
they transition when our theoretical value for w is 50.4. This is in contrast with w = 64.55
if 1−φ = 1, meaning no capture at all. If elites manage to save 10 percent, the threshold for
wealth decreases to 39.74. Even with small levels of capture, the likelihood of democratiz-
ation increases sharply. However, what it most relevant about this extension of the model is
that countries still only transition to democracy at high levels of inequality and wealth.
159
6.3 Appendix C
We first describe our set of control variables and provide descriptive statistics for all meas-
ures used in the analysis. Further tests and robustness checks are then provided. We con-
clude with more information on our measure of wealth.
Control variables and descriptive statistics
In this chapter we follow Boix (2003), Houle (2009) and Miller (2012) in the selection of
controls. First, we incorporate the lag of GDP per capita and economic growth. Income
levels or economic trends could impact the relationship between inequality and democratiz-
ation, as these variables are known to be related theoretically to our main independent and
dependent variables. We control for ethnic and religious fractionalization, following Boix
(2003) and Houle (2009). Religious and ethnic cleavages could be driving the relationship
between inequality and democratization, affecting both variables systematically. We also
incorporate a continuous variable for oil production taken from the World Bank. Oil should
have a negative relationship with democratization (Ross 2001). From Houle (2009) and
Miller (2012), we control for whether a country was a British colony; whether the country
was formed after World War II (new country); and previous transitions to democracy. A
descriptive summary of the variables used in the analysis follows in Table C1.
160
Table C1: Summary statistics
Variable Mean Std. Dev. N
Democracy 0.465 0.499 8915
Elite Wealth 2.581 1.419 7153
Inequality 6.427 1.399 3857
GDPpc 7173.877 14704.487 8831
GDPpc (log) 7.504 1.703 8831
Growth 3.906 6.416 8506
Oil Production (log) 0.26 2.469 7836
Ethnic fract. 0.434 0.254 7656
Religious fract. 0.428 0.234 7598
New Country 0.604 0.489 11426
British Colony 0.244 0.429 11426
Time Trend 28.5 16.741 11426
Previous Trans. 0.205 0.468 11229
Decade 1984.138 16.717 11426
Robustness checks 1: alternative measures of inequality
We corroborate our results using three measures of inequality from two sources: Houle
(2009, 2016) and Solt’s SWIID data (2017). Table C2 shows the results of our models using
Houle’s Capital Shares variable. This measure represents the share of output that accrues to
capital holders, and captures inter-class inequality well. The interaction term is significant
at the 0.5 level in 3 of the models and at the 0.1 in the other 3. However, even in those models
where significance is lower, the size of the probit and LPMFE coefficients indicates that the
relationship is likely significant at the 0.5 somewhere in the marginal effects curve. This is
shown in Figure C1. Both plots (a) and (b) report the marginal effect of increases in income
inequality at different levels of wealth. The conclusion that democratization becomes more
likely at high levels of wealth as inequality increases is common to both, even though only
Model 3 is significant at the 0.05 level in the table. The results obtained in the chapter hold
using Houle’s variable.
161
DV: Democracy
1 2 3 4 5 6
Elite Wealth -0.635∗ -0.636∗∗ -0.672∗∗ -0.527 -0.022 -0.040∗
(0.335) (0.317) (0.311) (0.339) (0.023) (0.024)
Inequality (Houle) -0.707 -1.261 -1.359 -1.284 -0.161∗ -0.202∗∗
(1.119) (1.170) (1.147) (1.260) (0.096) (0.097)
Elite Wealth × Inequality (Houle) 0.896∗ 0.911∗∗ 0.961∗∗ 0.826∗ 0.059∗ 0.072∗∗
(0.461) (0.421) (0.411) (0.440) (0.030) (0.030)
GDPpc -0.050 -0.048 -0.107 -0.032∗∗∗ -0.034∗∗∗
(0.061) (0.063) (0.069) (0.009) (0.010)
Growth -0.016∗ -0.014 -0.013 0.000 0.000
(0.008) (0.009) (0.010) (0.000) (0.000)
Oil -0.069∗∗ -0.068∗∗ -0.066∗∗ -0.005∗ -0.005
(0.033) (0.033) (0.034) (0.003) (0.003)
Previous Trans. 0.253∗∗ 0.241∗∗ 0.214∗ 0.035∗∗∗ 0.036∗∗∗
(0.104) (0.107) (0.114) (0.011) (0.011)
Time Trend 0.007 -0.016 -0.011 0.001∗∗∗ 0.002
(0.006) (0.019) (0.020) (0.001) (0.004)
Ethnic fract. 0.026 0.023 0.024
(0.283) (0.285) (0.261)
Religious fract. 0.157 0.143 -0.037
(0.298) (0.299) (0.311)
British Colony -0.059 -0.049 0.100
(0.173) (0.174) (0.178)
New Country -0.536∗∗∗ -0.541∗∗∗ -0.334∗
(0.155) (0.157) (0.196)
Constant -1.417∗ -0.534 -0.200 -0.317 0.351∗∗∗ 0.375∗
(0.792) (0.928) (0.905) (0.943) (0.082) (0.193)
Country FE - - - - Yes Yes
Year FE - - - - - Yes
Region FE - - - Yes - -
Decade FE - - Yes Yes - -
Observations 5192 3899 3899 3899 4608 4608
Pseudo R2 0.857 0.870 0.870 0.874
Overall R2 0.901 0.901
Log-Lik. -513.678 -347.227 -345.779 -335.293 2540.811 2589.560
Table C2: Dynamic Probit results using Houle’s Capital Shares variable
* p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors in parentheses.
Table C3 reports the results of our models using Solt’s adjusted disposable income Gini
index. The measure “incorporates data from the OECD Income Distribution Database, the
Socio-Economic Database for Latin America and the Caribbean generated by CEDLAS
162
(a) Model 3 Table A2 (b) Model 4 Table A2
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6
Wealth Wealth
Figure C1: Marginal effect of inequality on democratization at different levels of wealth,
using Houle’s Capital Shares data.
and the World Bank, Eurostat, the World Bank’s PovcalNet, the UN Economic Commis-
sion for Latin America and the Caribbean, national statistical offices around the world, and
academic studies while minimizing reliance on problematic assumptions by using as much
information as possible from proximate years within the same country.”308 The data are the
result of a multiple imputation procedure, and our models are run using the ‘mi estimate:”
command in Stata 13. The dynamic probit models (1-3) show no statistical significance for
the interaction term between inequality and wealth. However, the relationship is strong and
in the predicted direction in all fixed effects specifications (Models 4-6). Figure C2 plots
the marginal effects using Model 5 from Table C3, and the results are similar to the fixed ef-
fects specifications using UNIDO’s variable from the main chapter or Houle’s capital shares
measure.
308https://fsolt.org/swiid/. See Solt (2016) for more information.
163
 Marginal Effects of Inequality 
 on the Pr(Democratization) 
-.5 0 .5 1 1.5 2
-1 -.5 0 .5 1 1.5
(1) (2) (3) (4) (5) (6)
Elite Wealth -0.525 -0.279 -0.199 -0.082∗∗∗ -0.050∗ -0.054∗
(0.338) (0.433) (0.426) (0.024) (0.028) (0.028)
Gini Disp. Income -2.848 -2.378 -2.066 -0.887∗∗∗ -0.792∗∗∗ -0.780∗∗∗
(2.196) (3.113) (3.108) (0.195) (0.210) (0.211)
Elite Wealth × Gini Disp. Income 1.244 0.876 0.685 0.234∗∗∗ 0.202∗∗∗ 0.197∗∗∗
(0.810) (1.019) (1.001) (0.058) (0.062) (0.062)
GDPpc -0.158∗ -0.139 -0.029∗∗∗ -0.023∗∗
(0.089) (0.092) (0.010) (0.012)
Growth -0.015 -0.013 0.000 0.000
(0.010) (0.011) (0.000) (0.001)
Oil -0.098∗∗∗ -0.095∗∗∗ -0.007∗∗ -0.006
(0.035) (0.036) (0.003) (0.004)
Previous Trans. 0.285∗∗ 0.286∗∗ 0.019∗ 0.018
(0.116) (0.118) (0.011) (0.011)
Time Trend 0.004 -0.041∗∗ 0.000 0.000
(0.007) (0.016) (0.001) (0.001)
Ethnic fract. 0.054 0.053
(0.356) (0.364)
Religious fract. 0.045 0.020
(0.361) (0.365)
British Colony -0.221 -0.217
(0.217) (0.220)
New Country -0.530∗∗∗ -0.547∗∗∗
(0.204) (0.209)
Constant -0.641 0.281 0.486 0.434∗∗∗ 0.565∗∗∗ 0.511∗∗∗
(0.906) (1.279) (1.383) (0.079) (0.098) (0.107)
Observations 4237 3598 3598 4237 4095 4095
Table C3: MI Estimate results using Solt’s disposable income Gini.
* p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors in parentheses.
Robustness checks 2: alternative hypotheses
We replicate the analysis by Acemoglu, Johnson and Robinson (2001) in which they instru-
ment ‘institutions’ through settler mortality. We have taken their publicly available replica-
tion materials from their website (https://economics.mit.edu/faculty/acemoglu/data/ajr2001),
and added a variable for whether a country was a democracy in 2015 from the Boix, Miller
164
(a) Model 5 Table A3
0 1 2 3 4 5 6 7
Wealth
Figure C2: Marginal effect of inequality on democratization at different levels of wealth,
using Solt’s Disposable Income Gini Index.
and Rosato dataset. We use this variable in the second stage instead of log GDP per capita,
which the authors use in the original. We then do the same but with VDem’s continuous
measure of democracy (v2x polyarchy). Thus, we do not estimate the effect of institutions
(1) (2) (3) (4) (5)
Old Institutions (Instrument: Settler Mortality) 0.191∗∗ 0.153 0.185 -0.105 -0.037
(0.090) (0.126) (0.186) (0.086) (0.193)
Latitude 0.487 0.849 1.179∗∗ 0.584
(0.721) (0.633) (0.493) (0.592)
Africa dummy -0.617∗∗∗
(0.171)
Asia dummy -0.539∗∗
(0.224)
Other Continent dummy 0.053
(0.318)
Constant -0.670 -0.506 -0.753 1.274∗∗ 1.045
(0.592) (0.717) (1.124) (0.552) (1.197)
Observations 64 64 60 37 64
Table C4: Replication of Acemoglu and Robinson’s (2001) Table 4 using BMR’s dichotom-
ous measure of democracy as DV.
* p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors in parentheses.
165
 Marginal Effects of Inequality 
 on the Pr(Democratization) 
-1 -.5 0 .5 1
on economic development today, but their effect on democracy today. Results are in Tables
C4 and C5. We do not report the first stage as it is the same as in the original chapter.
The models used are the same as in the original chapter, as well. Model 1 is the naı̈ve
model with no controls, Model 2 adds latitude and Model 3 removes the ‘Neo-Europes’, i.e.
the United States, Canada, Australia, and New Zealand, from the sample. Model 4 runs the
model without African countries and Model 5 includes the full sample and uses continent
dummies instead. As we can see from both results, only the first model is positive and
statistically significant. As controls are added, the relationship loses statistical significance
and reverses sign in Models 4 and 5. This provides evidence that strong institutions are
important for development but they do not necessarily lead to democratization.
(1) (2) (3) (4) (5)
Old Institutions (Instrument: Settler Mortality) 0.087∗∗ 0.062 0.048 0.018 -0.043
(0.036) (0.045) (0.066) (0.034) (0.082)
Latitude 0.330 0.333 0.487∗∗ 0.416∗
(0.262) (0.247) (0.223) (0.233)
Africa dummy -0.241∗∗∗
(0.084)
Asia dummy -0.193∗
(0.098)
Other Continent dummy 0.241
(0.211)
Constant 0.001 0.105 0.184 0.426∗ 0.889∗
(0.236) (0.257) (0.409) (0.216) (0.520)
Observations 60 60 56 34 60
R2 0.120 0.213 0.075 0.234 0.261
Table C5: Replication of Acemoglu, Johnson and Robinson’s (2001) Table 4 using VDem’s
continuous liberal democracy index as DV.
* p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors in parentheses.
166
More on our wealth proxy
This is the definition of the raw Gross Fixed Capital Formation (GFCF) data from the World
Bank, according to the organization:
“Gross fixed capital formation (formerly gross domestic fixed investment) includes
land improvements (fences, ditches, drains, and so on); plant, machinery, and equip-
ment purchases; and the construction of roads, railways, and the like, including schools,
offices, hospitals, private residential dwellings, and commercial and industrial build-
ings. According to the 1993 SNA, net acquisitions of valuables are also considered
capital formation. Data are in current U.S. dollars.”
Our variable cannot fully net out non-elite wealth accumulation from the variable, but we
suspect it does not affect the measure much. The high correlation with Piketty’s measure
of total elite wealth provides some external validity for this assumption. To emphasize this
point, we plot the evolution of both variables in countries that match between samples in
Figure C3. Time in years is in the x-axis. The blue line represents our wealth proxy, whose
values are in the left y-axis, and the red line plots Piketty’s variable, with its values in the
right y-axis. In most countries, the variables evolve close to each other. Only in Canada,
and to a much lesser extent Australia, the line becomes sharper for Piketty’s variable earlier
than for ours. For the rest, the trend for both lines is similar and only vary in their response
to some growth shocks, such as Japan in 1990. The correlation between both variables is
above r = 0.88 for all these countries, and the within correlation averages r = 0.92. The
between correlation figure for these two variables is r = 0.946.
167
US France
Piketty Our Proxy
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
Year Year
Canada Germany
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
Year Year
Spain Italy
1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
Year Year
Australia Japan
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
Year Year
Sweden South Korea
1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010
Year Year
Figure C3: Evolution of Piketty’s elite wealth variable and our proxy by country.
168
Elite Wealth (Piketty) Elite Wealth (Piketty) Elite Wealth (Piketty) Elite Wealth (Piketty) Elite Wealth (Piketty)
0 10 20 30 40 0 20 40 60 80 0 5 10 15 20 25 30 0 10 20 30 40 50 60 0 10 20 30 40
2220.417 12220.417 22220.417 1097.746 7097.746 15097.746 3009.726 9009.726 15009.726 21009.726 2352.655 8352.655 14352.655 22352.655 21356.21 221356.21 421356.21
Wealth_ch Wealth_ch Wealth_ch Wealth_ch Wealth_ch
Elite Wealth (Piketty) Elite Wealth (Piketty) Elite Wealth (Piketty) Elite Wealth (Piketty) Elite Wealth (Piketty)
0 200 600 1000 1400 0 20 40 60 80 100 0 10 20 30 40 2 4 6 8 10 12 5 10 15
4758048 12758048 20758048 832158 2432158 4032158 5632158 3746.593 11746.593 21746.593 5675.936 21675.936 37675.936 2582.315 10582.315 18582.315
Wealth_ch Wealth_ch Wealth_ch Wealth_ch Wealth_ch
Marginal effects of wealth at different levels of inequality
Lastly, as we mention in footnote 26 in the main text, we include the marginal effects of
wealth at different levels of inequality in Figure C4. Both models show a positive slope that
is statistically significant at high levels of inequality. The effect of wealth on democratization
when inequality is high is statistically and substantively significant. We refer to reader to the
main text for our argument regarding substantive significance, which applies here as well.
(a) Model 3, Table 1 (b) Model 6, Table 1
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Inequality Inequality
Figure C4: Marginal effect of wealth on democratization at different levels of inequality,
using UNIDO inequality data as in the main chapter. Table 1 is from the main text.
169
 Marginal Effect of Wealth 
 on Democratization
-.3 -.2 -.1 0 .1
 Marginal Effect of Wealth 
 on Democratization
-.1 -.05 0 .05 .1
References
170
Acemoglu, Daron, Davide Ticchi and Andrea Vindigni. 2010. “A theory of military dictat-
orships.” American Economic Journal: Macroeconomics 2(1):1–42.
Acemoglu, Daron and James A Robinson. 2000a. “Democratization or repression?”
European Economic Review 44(4):683–693.
Acemoglu, Daron and James A Robinson. 2000b. “Why did the West extend the franchise?
Democracy, inequality, and growth in historical perspective.” The Quarterly Journal of
Economics 115(4):1167–1199.
Acemoglu, Daron and James A Robinson. 2001. “A theory of political transitions.” Amer-
ican Economic Review pp. 938–963.
Acemoglu, Daron and James A. Robinson. 2006. Economic Origins of Dictatorship and
Democracy. Cambridge University Press.
Acemoglu, Daron and James A Robinson. 2013. Why nations fail: The origins of power,
prosperity, and poverty. Broadway Business.
Acemoglu, Daron, Simon Johnson and James A. Robinson. 2001. “The Colonial Origins of
Comparative Development: An Empirical Investigation.” The American Economic Re-
view 91(5):1369–1401.
Acemoglu, Daron, Simon Johnson, James A Robinson and Pierre Yared. 2008. “Income
and Democracy.” American Economic Review 98(3):808–842.
171
Acosta, Alberto. 1995. Breve historia económica del Ecuador. Vol. 2. Corporación Editora
Nacional.
Albertus, Michael. 2015. Autocracy and redistribution. Cambridge University Press.
Albertus, Michael and Victor Menaldo. 2012. “If you’re against them you’re with us: The
effect of expropriation on autocratic survival.” Comparative Political Studies 45(8):973–
1003.
Albertus, Michael and Victor Menaldo. 2014. “Gaming democracy: elite dominance dur-
ing transition and the prospects for redistribution.” British Journal of Political Science
44(3):575–603.
Alesina, Alberto and Roberto Perotti. 1996. “Income distribution, political instability, and
investment.” European economic review 40(6):1203–1228.
Alvaredo, Facundo, Lucas Chancel, Thomas Piketty, Emmanuel Saez and Gabriel Zucman.
2017. “World inequality report 2018.” The World Inequality Lab, http://wir2018. wid.
world .
Ansell, Ben and David Samuels. 2010. “Inequality and democratization: A contractarian
approach.” Comparative Political Studies 43(12):1543–74.
Ansell, Ben and David Samuels. 2014. Inequality and Democratization. Cambridge Uni-
versity Press.
Barany, Zoltan. 2009. Democratic breakdown and the decline of the Russian military. Prin-
ceton University Press.
172
Besley, Timothy and James A Robinson. 2010. “Quis custodiet ipsos custodes? Civilian
control over the military.” Journal of the European Economic Association 8(2-3):655–
663.
Boix, Carles. 2003. Democracy and redistribution. Cambridge University Press.
Boix, Carles, Michael Miller and Sebastian Rosato. 2013. “A complete data set of political
regimes, 1800–2007.” Comparative Political Studies 46(12):1523–1554.
Boix, Carles and Milan Svolik. 2013. “The foundations of limited authoritarian government:
Institutions, commitment, and power-sharing in dictatorships.” The Journal of Politics
75(02):300–316.
Bonilla, Heraclio. 1984. Guano y burguesı́a en el Perú. Lima: Instituto de Estudios Peru-
anos.
Box-Steffensmeier, Janet and Bradford Jones. 2004. Event history modeling: A guide for
social scientists. Cambridge University Press.
Brownlee, Jason. 2007. Authoritarianism in an Age of Democratization. Cambridge Uni-
versity Press.
Bulmer-Thomas, Victor. 2003. The economic history of Latin America since independence.
Cambridge University Press.
Carothers, Thomas. 2002. “The end of the transition paradigm.” Journal of democracy
13(1):5–21.
173
Cheibub, José Antonio, Jennifer Gandhi and James Raymond Vreeland. 2010. “Democracy
and dictatorship revisited.” Public Choice 143(1-2):67–101.
Cotler, Julio. 1979. State and regime: comparative notes on the Southern Cone and the
Enclave Societies. In The new authoritarianism in Latin America, ed. David Collier and
Fernando Henrique Cardozo. Princeton University Press pp. 255–282.
Cox, Gary W. 2016. Marketing sovereign promises: Monopoly brokerage and the growth
of the English state. Cambridge University Press.
Crassweller, Robert D. 1966. Trujillo: The life and times of a Caribbean dictator. Macmil-
lan New York.
Derby, Lauren H. 2009. The Dictator’s Seduction: Politics and the Popular Imagination in
the Era of Trujillo. Duke University Press.
Escribà-Folch, Abel. 2013. “Repression, political threats, and survival under autocracy.”
International Political Science Review 34(5):543–560.
Evans, Peter B. 1995. Embedded autonomy: States and industrial transformation. Princeton
University Press.
Frantz, Erica and Andrea Kendall-Taylor. 2014. “A dictator?s toolkit: Understanding how
co-optation affects repression in autocracies.” Journal of Peace Research 51(3):332–346.
Gandhi, Jennifer. 2008. Political institutions under dictatorship. Cambridge University
Press.
174
Gandhi, Jennifer and Adam Przeworski. 2006. “Cooperation, cooptation, and rebellion
under dictatorships.” Economics & Politics 18(1):1–26.
Gandhi, Jennifer and Adam Przeworski. 2007. “Authoritarian institutions and the survival
of autocrats.” Comparative Political Studies 40(1):1279–1301.
Gandhi, Jennifer and Ellen Lust-Okar. 2009. “Elections under authoritarianism.” Annual
review of political science 12:403–422.
Geddes, Barbara. 1999. “What do we know about democratization after twenty years?”
Annual Review of Political Science 2(1):115–144.
Geddes, Barbara, Joseph Wright and Erica Frantz. 2014. “Autocratic breakdown and regime
transitions: A new data set.” Perspectives on Politics 12(02):313–331.
Gehlbach, Scott and Philip Keefer. 2011. “Investment without democracy: Ruling-party
institutionalization and credible commitment in autocracies.” Journal of Comparative
Economics 39(2):123–139.
Gehlbach, Scott and Philip Keefer. 2012. “Private investment and the institutionalization of
collective action in autocracies: ruling parties and legislatures.” The Journal of Politics
74(2):621–635.
Ghura, Mr Dhaneshwar and Benoı̂t Mercereau. 2004. Political Instability and Growth: The
Central African Republic. Number 4-80 International Monetary Fund.
Ginsburg, Tom and Tamir Moustafa. 2008. Rule by law: the politics of courts in authorit-
arian regimes. Cambridge University Press.
175
Gomez-Reino, Margarita and Iván Llamazares. 2015. “New left populism in contemporary
Spain? the upsurge of Podemos.” Council for European Studies .
Haggard, Stephan and Robert R Kaufman. 2012. “Inequality and regime change: Demo-
cratic transitions and the stability of democratic rule.” American Political Science Review
106(03):495–516.
Haggard, Stephan and Robert R Kaufman. 2016. Dictators and democrats: masses, elites,
and regime change. Princeton University Press.
Hagopian, Frances and Scott P Mainwaring. 2005. The third wave of democratization in
Latin America: advances and setbacks. Cambridge University Press.
Hall, Michael R. 2000. Sugar and Power in the Dominican Republic: Eisenhower, Kennedy,
and the Trujillos. Greenwood Publishing Group.
Houle, Christian. 2009. “Inequality and democracy: Why inequality harms consolidation
but does not affect democratization.” World Politics 61(04):589–622.
Houle, Christian. 2016. “Why class inequality breeds coups but not civil wars.” Journal of
Peace Research 53(5):680–695.
Inglehart, Ronald and Pippa Norris. 2016. “Trump, Brexit, and the rise of populism: Eco-
nomic have-nots and cultural backlash.”.
Karl, Terry L. 1997. The paradox of plenty: Oil booms and petro-states. University of
California Press.
176
Kennedy, Ryan. 2010. “The contradiction of modernization: A conditional model of endo-
genous democratization.” The Journal of Politics 72(03):785–798.
Knutsen, Carl Henrik. 2015. “Reinvestigating the Reciprocal Relationship between Demo-
cracy and Income Inequality.” Review of Economics & Institutions 6(2).
Kohli, Atul. 2004. State-directed development: political power and industrialization in the
global periphery. Cambridge University Press.
Larrea, Carlos. 1994. “The mirage of development: Oil, employment, and poverty in
Ecuador (1972-1990).”.
Levitsky, Steven and Daniel Ziblatt. 2018. How democracies die. Crown.
Levitsky, Steven and Lucan A Way. 2010. Competitive authoritarianism: Hybrid regimes
after the cold war. Cambridge University Press.
Levitsky, Steven and Lucan Way. 2002. “The rise of competitive authoritarianism.” Journal
of democracy 13(2):51–65.
Linz, Juan J. 1975. “Totalitarian and authoritarian regimes.” Handbook of political science
3:175–411.
Lipset, Seymour Martin. 1959. “Some social requisites of democracy: Economic develop-
ment and political legitimacy.” American political science review 53(1):69–105.
Little, Andrew T. 2017. “Coordination, learning, and coups.” Journal of Conflict Resolution
61(1):204–234.
177
Loxton, James. 2015. “Authoritarian successor parties.” Journal of Democracy 26(3):157–
170.
Lust-Okar, Ellen. 2005. Structuring conflict in the Arab world: Incumbents, opponents, and
institutions. Cambridge University Press.
Magaloni, Beatriz. 2006. Voting for autocracy: Hegemonic party survival and its demise in
Mexico. Cambridge University Press.
Metzger, Shawna K and Benjamin T Jones. 2016. “Surviving Phases: Introducing
Multistate Survival Models.” Political analysis 24(4):457–477.
Miller, Michael K. 2012. “Economic development, violent leader removal, and democrat-
ization.” American Journal of Political Science 56(4):1002–1020.
Moore, Barrington. 1966. Social origins of dictatorship and democracy: Lord and peasant
in the making of the modern world. Boston: Beacon Press.
Munck, Gerardo L and Carol Skalnik Leff. 1997. “Modes of transition and democratization:
South America and Eastern Europe in comparative perspective.” Comparative Politics
pp. 343–362.
O’Donnell, Guillermo. 1973. Modernization and bureaucratic-authoritarianism: Studies
in South American politics. University of California Press.
O’Donnell, Guillermo and Philippe Schmitter. 1986. Transitions from Authoritarian Rule:
Tentative conclusions about uncertain democracies. Johns Hopkins University Press.
178
Olson, Mancur. 1991. “Autocracy, democracy, and prosperity.” Strategy and choice
131(157):131–157.
Olson, Mancur. 1993. “Dictatorship, democracy, and development.” American political
science review 87(3):567–576.
Orriols, Lluı́s and Guillermo Cordero. 2016. “The breakdown of the Spanish two-party
system: the upsurge of Podemos and Ciudadanos in the 2015 general election.” South
European Society and Politics 21(4):469–492.
Piketty, Thomas. 2014. Capital in the twenty-first century. Harvard University Press.
Powell, Jonathan. 2012. “Determinants of the Attempting and Outcome of Coups d’état.”
Journal of Conflict Resolution 56(6):1017–1040.
Przeworski, Adam. 1991. Democracy and the market: Political and economic reforms in
Eastern Europe and Latin America. Cambridge University Press.
Przeworski, Adam. 2005. “Democracy as an Equilibrium.” Public Choice 123(3):253–273.
Przeworski, Adam. 2016. “Democracy: A never-ending quest.” Annual Review of Political
Science 19:1–12.
Przeworski, Adam and Fernando Limongi. 1993. “Political regimes and economic growth.”
Journal of Economic Perspectives 7(3):51–69.
Przeworski, Adam, Michael Alvarez, Jose Cheibub and Fernando Limongi. 2000. Demo-
cracy and Development: Political Institutions and Well-Being in the World, 1950-1990.
Cambridge University Press.
179
Putter, Hein, M Fiocco and RB Geskus. 2007. “Tutorial in biostatistics: competing risks
and multi-state models.” Statistics in Medicine 26(11):2389–2430.
Reisenbichler, Alexander and Kimberly J Morgan. 2012. “From ?Sick Man? to ?Miracle?
Explaining the Robustness of the German Labor Market During and After the Financial
Crisis 2008-09.” Politics & Society 40(4):549–579.
Ross, Michael. 2001. “Does oil hinder democracy?” World Politics 53(3):325–361.
Rueschemeyer, Dietrich, Evelyne Huber Stephens and John D Stephens. 1992. “Capitalist
development and democracy.” Cambridge, UK .
Saez, Emmanuel and Gabriel Zucman. 2016. “Wealth inequality in the United States since
1913: Evidence from capitalized income tax data.” The Quarterly Journal of Economics
131(2):519–578.
Singh, Naunihal. 2014. Seizing power: The strategic logic of military coups. JHU Press.
Sinn, Hans-Werner. 2014. “Austerity, growth and inflation: remarks on the Eurozone’s
unresolved competitiveness problem.” The World Economy 37(1):1–13.
Skidmore, Thomas E and Mário Salviano Silva. 1988. Brasil: de castelo a Tancredo, 1964-
1985. Paz e Terra Rio de Janeiro.
Skocpol, Theda and Alexander Hertel-Fernandez. 2016. “The Koch network and republican
party extremism.” Perspectives on Politics 14(3):681–699.
Slater, Dan. 2010. Ordering power: Contentious politics and authoritarian leviathans in
Southeast Asia. Cambridge University Press.
180
Slater, Dan and Joseph Wong. 2013. “The strength to concede: ruling parties and demo-
cratization in developmental Asia.” Perspectives on Politics 11(03):717–733.
Smith, Alastair. 2008. “The perils of unearned income.” The Journal of Politics 70(3):780–
793.
Sola, Jorge and César Rendueles. 2018. “Podemos, the upheaval of Spanish politics and the
challenge of populism.” Journal of Contemporary European Studies 26(1):99–116.
Solt, Frederick. 2016. “The standardized world income inequality database.” Social science
quarterly 97(5):1267–1281.
Svolik, Milan W. 2009. “Power sharing and leadership dynamics in authoritarian regimes.”
American Journal of Political Science 53(2):477–494.
Svolik, Milan W. 2012. The politics of authoritarian rule. Cambridge University Press.
Turits, Richard Lee. 2003. Foundations of despotism: peasants, the Trujillo regime, and
modernity in Dominican history. Stanford University Press.
Vanhanen, Tatu. 1997. Prospects of democracy: A study of 172 countries. Psychology
Press.
Venieris, Yiannis P and Dipak K Gupta. 1986. “Income distribution and sociopolitical in-
stability as determinants of savings: a cross-sectional model.” Journal of Political Eco-
nomy 94(4):873–883.
Wright, Joseph. 2008. “Do authoritarian institutions constrain? How legislatures affect
economic growth and investment.” American Journal of Political Science 52(2):322–343.
181
Wright, Joseph and Abel Escriba-Folch. 2012. “Authoritarian institutions and regime sur-
vival: Transitions to democracy and subsequent autocracy.” British Journal of Political
Science 42(02):283–309.
Wright, Joseph and Daehee Bak. 2016. “Measuring autocratic regime stability.” Research
& Politics 3(1).
Ziblatt, Daniel. 2017. Conservative Political Parties and the Birth of Modern Democracy
in Europe. Cambridge University Press.
182