POLITICAL LEGITIMACY, CITIZENS’ MOBILIZATION, AND LEADERS’ SURVIVAL
by
Changkuk Jung

A DISSERTATION
Submitted
to Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
POLITICAL SCIENCE
2011

ABSTRACT
POLITICAL LEGITIMACY, CITIZENS’ MOBILIZATION, AND LEADERS’ SURVIVAL
by
Changkuk Jung
Leader survival has been addressed mainly by international relations students, while regime
survival and government survival have been dealt with mainly by comparativists. Most studies that examine leaders’ survival neglect how their constituents (both voters and nonvoters)
view the leaders’ competence, parties’ representation and accountability of political systems,
and how the electorates’ behaviors, both conventional and unconventional, instigate political instability that may substantially increase the risk of deposition. I adopt a theoretical
structure of the selectorate theory proposed by Bueno de Mesquita et al. (2003), but adapt
it as well by focusing on the role of the masses. By using individual-level data sets such
as the World Values Survey, Afro-Barometer, and Asian-Barometer (Chapter 3), and the
Korea Social Science Data (Chapter 4), the effect of electoral politics on political trust and
a leader’s survival is examined; and the leaders’ data of Archigos and other country-level
data (Chapter 5) allow me to examine leaders’ survival.
Chapter 2 reviews literature about the potential causes of leaders’ survival by focusing on
elections, perceived legitimacy, and political movements. First, political legitimacy is viewed
as both a consequence of voters’ perception of the selectorate institutions and as a cause
for leaders to secure the support of citizens. Finally, leaders’ survival is understood in the
context of mass political movement and voters’ electoral behavior.
Chapter 3 argues that the loyalty norm for each leader provides an institutional context
for voters to perceive how legitimate their system is. For the mass public, a multilevel analysis
shows that the institutional legacy of the loyalty norm matters for electoral winners and nonpartisans as much as the short-term loyalty norm for electoral losers. This is because the
mass public’s attitudes rely on their memory of how often leadership turnover has occurred

in the polity while the leader’s behavior is more based upon the short-term loyalty norm.
In Chapter 4, South Korea’s impeachment experience illustrates how electoral politics
and partisan status affect the citizen’s perception of a president’s impeachment at the individual level. Admitting a limitation of the validity of variables measured at the aggregate
level, especially for the loyalty norm measure, I take advantage of the unique experimentlike impeachment process in South Korea in 2004. South Korea’s recent experience of the
impeachment provides differing responses by partisans and non-partisans. Loyalty to the
Uri Party and partisan status influence voters’ impeachment support. As partisans are more
likely to have any preferences for the issue of impeachment, they are labeled ranging from
“loyal”, to “dissenting”, to “defecting”. In contrast to partisans’ wide range of impeachment
support, non-partisans’ moderate level change of the impeachment support illustrates two
types of non-partisans: sympathetic and ideological non-partisan. This means that ideology matters when a fledgling incumbent party tries to hold non-partisans’ interests in the
conservative party system.
Chapter 5 supposes that stability in leader survival is the key interest to a leader herself
although it may not provide an accountable leader. The hypothesis to test in this chapter
is whether leader longevity is influenced by mass movement that is indicated as electoral
behavior, or depends on a leader’s strategy of survival determined by the loyalty norm
institution. Based on a composite data set that provides information on electoral politics
and the winning coalition institution, I employ three measures of electoral and social stability
– the level of mass threats, electoral non-participation, and electoral competition – to test
these hypotheses. The empirical findings indicate that electoral non-participation and mass
threats are key determinants to the risk of leader deposition, while the leader herself has
room to manipulate the risk by constraining the level of mass media.

To my beloved wife, Ki-ae Jang, infinitely supportive and encouraging,
To my amazing son, Jae-ung, generous and wise beyond his years,
and
To my parents, Kum-suk Kim and Chung-hyun Jung, for their unending trust and moral
and financial support.

iv

ACKNOWLEDGEMENTS

I am grateful to all of those whose guidance and support have allowed me to have faith in
this research project. Each of the members of my Dissertation Committee, Dr. Brian Silver,
Dr. Michael Bratton, Dr. Eric Chang, Dr. Michael Colaresi, and Dr. Sandra MarquartPyatt, has provided me personal and professional guidance during the course of the graduate
program. I would especially like to thank Dr. Brian Silver, the chairman of my committee.
As my advisor and mentor, he has taught me more than I could ever give him credit for
here. He has shown me, by his example, what a good scholar (and a father) should be,
and among many of his thoughtful encouragements, the “Goldstar-to-LG" analogy to my
academic career has truly made a lasting impact on my life.
I would like to express my gratitude to two current and past directors of the graduate
program, Dr. Steve Kautz and Dr. Sandy Schneider, both of whom provided a great deal
of opportunities for me to focus on my research with abundant resources, including the
opportunities to work for the OLC of the ICPSR at the University of Michigan over two
years, fieldwork trips to South Korea and Taiwan, and participation in the SIPP in Stanford
University. The project was supported by the Department of Political Science and the
graduate school at Michigan State University.
This work would not have been possible without endless support of my family and friends.
First of all, I would like to thank my parents, Kum-suk Kim and Chung-hyun Jung, whose
support and love have been with me in whatever I try to accomplish and whenever I experience failure. I wish to thank my loving and supportive wife, Ki-ae Jang, who has sustained
me and our family even when we faced difficult times. Without her care, love, and encouragements, my life path would have been so different, and I could not have finished this
work. My wonderful son, Jae-ung, has provided me with unending inspiration and joy. His
enthusiasm and love for saxophone, piano, (sometimes, cello, tae-kwon-do, soccer,) and ice-

v

hockey have amused me, and all of them gave me a good diversion from my work. I am also
thankful to my parents-in-law, Ok-ja Park and Gil-sung Jang, for their support, care, and
unchanging trust in me. My sincere thanks go to our dearest friends who are like family,
Susy and Larry Karnes, for their love and care. Numerous movie nights, concerts, and many
cherishable moments with them have always provided me and my family with warm and
cozy feelings by which we have been able to do away with any homesickness. And last, but
most importantly, I am thankful for God’s grace and faith in Him that I could learn from
my late loving grandmother and her kneeling prayers.

vi

Contents

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

x

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
CHAPTER 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Research Question and Argument . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Electorate’s Beliefs and Behaviors in the Selectorate Theory . . . . . . . . .
CHAPTER 2 THEORY AND LITERATURE REVIEW . . . . .
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Political Trust Revisited: Election and Selectorate Politics . . . .
2.2.1 What is Political Trust? . . . . . . . . . . . . . . . . . . .
2.2.2 Political Trust, Policy, and Participation . . . . . . . . . .
2.2.2.1 Trust and Policy Attitudes . . . . . . . . . . . .
2.2.2.2 Trust and Voting . . . . . . . . . . . . . . . . . .
2.2.2.3 Trust and Mass Political Movements . . . . . . .
2.3 Non-voting and Mass Mobilization: Risky for Leader? . . . . . . .
2.3.1 Who is an Anti-system Non-voter? . . . . . . . . . . . . .
2.3.2 Why are Non-voters Mobilized? . . . . . . . . . . . . . . .
2.4 Unpacking Leader Survival: The Electorate in Selectorate Politics
2.4.1 The Electorate and Mass Threats to a Leader . . . . . . .
2.4.2 The Electorate in Selectorate Politics . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

1
3
6

. . . 10
. . . 10
. . . 10
. . . 11
. . . 13
. . . 13
. . . 16
. . . 18
. . . 20
. . . 20
. . . 22
. . . 25
. . . 26
. . . 28

CHAPTER 3 THE PARTISAN-W/S NEXUS OF POLITICAL TRUST
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Selectorate Theory of Electoral Politics: Explaining Political Trust . . . . .
3.3 Partisans, Loyalty Norm, and Political Trust . . . . . . . . . . . . . . . . .
3.4 Connecting (Non-)Partisans and the Loyalty Norm . . . . . . . . . . . . .
3.4.1 Partisans with the short-term loyalty norm . . . . . . . . . . . . .
3.4.2 Non-partisans with Historical Loyalty Norms . . . . . . . . . . . .
3.5 Data and Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5.1 Dependent variable . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5.2 Main independent variables . . . . . . . . . . . . . . . . . . . . . .
3.5.3 Covariates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.6 Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.7 Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. 30
. 30
. 32
. 35
. 41
. 41
. 44
. 49
. 49
. 51
. 52
. 53
. 56
. 60

CHAPTER 4 SOURCES OF PRESIDENT’S SURVIVAL IN SOUTH KOREA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

vii

4.2
4.3
4.4

.
.
.
.
.
.
.
.
.

80
83
86
86
87
89
90
92
95

CHAPTER 5 CITIZEN’S MOBILIZATION AND LEADER’S SURVIVAL
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Selectorate Theory of Electoral Politics Revisited . . . . . . . . . . . . . . .
5.2.1 From Electoral Politics to Mass Threats to Leader Survival . . . . . .
5.2.2 Alienated Nonvoters . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.2.3 From Votes to Violence . . . . . . . . . . . . . . . . . . . . . . . . . .
5.3 Hypotheses on Nonvoter, Protester, and Revolutionary . . . . . . . . . . . .
5.3.1 Vanishing Voters and Mass Movements . . . . . . . . . . . . . . . . .
5.3.2 De-mobilized Nonvoters . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.3 Free Press as a ‘Mobilizer’ . . . . . . . . . . . . . . . . . . . . . . . .
5.3.4 Leader Survival, Mass Threats and Nonvoters . . . . . . . . . . . . .
5.4 Methodology and Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.1 Dependent Variables: Mass Political Movements and Leader Survival
5.4.1.1 Mass Political Movements . . . . . . . . . . . . . . . . . . .
5.4.1.2 Leader Survival . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.2 Main Independent Variables . . . . . . . . . . . . . . . . . . . . . . .
5.4.2.1 Election Loser . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.2.2 Nonvoters . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.2.3 Winning Coalition (W) . . . . . . . . . . . . . . . . . . . .
5.4.2.4 Mass Threats . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.2.5 Free Press . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.3 Covariates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.3.1 Mass Political Movements . . . . . . . . . . . . . . . . . . .
5.4.3.2 Leader Survival . . . . . . . . . . . . . . . . . . . . . . . . .
5.5 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.5.1 Mass Political Movements, Nonvoters and Free Press . . . . . . . . .
5.5.2 Leader Survival, Revolutionary Threats and Nonvoters . . . . . . . .
5.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

120
120
122
122
126
129
130
134
137
139
141
144
144
144
145
145
145
147
149
150
151
151
151
152
153
153
158
161

4.5
4.6
4.7

Elite Politics and Regionalism in South Korea
The Leader-Party Nexus and Partisan Support
Data and Measures . . . . . . . . . . . . . . .
4.4.1 Dependent Variable . . . . . . . . . . .
4.4.2 Main Independent Variables . . . . . .
4.4.3 Covariates . . . . . . . . . . . . . . . .
Method . . . . . . . . . . . . . . . . . . . . .
Estimation Results . . . . . . . . . . . . . . .
Conclusions . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.

CHAPTER 6 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
6.1 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
6.2 Limitations and Extensions . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
APPENDIX A CHAPTER 3: CODING . . . . . . . . . . . . . . . . . . . . . 182

viii

APPENDIX B CHAPTER 3: HISTORICAL W/S FORMULA . . . . . . . 185
APPENDIX C CHAPTER 3: ROBUST CHECK FOR INFLUENCE CASE186
APPENDIX D CHAPTER 3: COUNTRY CASES . . . . . . . . . . . . . . 191
APPENDIX E CHAPTER 4: BRANT TEST . . . . . . . . . . . . . . . . . 192
APPENDIX F CHAPTER 4: COMPREHENSIVE COMPARATIVE LOYALTY MODEL . . . . . . . . . . . . . . . . . . . . . . . . . . 193
APPENDIX G CHAPTER 4: MULTINOMIAL LOGIT MODEL WITH
IDEOLOGY VARIABLE . . . . . . . . . . . . . . . . . . . . 198
APPENDIX H CHAPTER 5: MODELS FOR NONVOTER’S SUBSTANTIAL CHANGES (OVER 5% INCREASE) . . . . . . . . . 201
APPENDIX I

CHAPTER 5: VARIANCE INFLATION FACTOR . . . . 204

APPENDIX J

CHAPTER 5: GOODNESS OF FIT

. . . . . . . . . . . . . 205

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212

ix

LIST OF TABLES

3.1

ANOVA of Variance in Political Trust Across Levels (Model 1 - Model 4) . . . .

62

3.2

Descriptive Statistics of Dependent Variable and Main Independent Variables
(Model 5 - Model 8) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

63

Multilevel Linear Model of Partisanship and W/S: Dependent Variable: Political
Trust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

64

Multilevel Linear Model of Partisanship and W/S: Dependent Variable: Political
Trust (Perceived Personal Economy included) . . . . . . . . . . . . . . . . . . .

67

Multilevel Linear Model of Partisanship and W/S: Dependent Variable: Political
Trust (Democratizing countries only) . . . . . . . . . . . . . . . . . . . . . . . .

70

4.1

Partisanship by Region (%) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

97

4.2

The Leader-Party Nexus/Contrast & (Non-)Partisan Support: Expected Response to Impeachment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

98

4.3

Support for Impeachment by partisanship (%) . . . . . . . . . . . . . . . . . . .

99

4.4

Partial Proportional Odds Models of Impeachment Support . . . . . . . . . . . 100

4.5

Partial Proportional Odds Models of Impeachment Support (Comp. Loyalty) . . 104

5.1

Party Identification and Voting Behavior in Elections . . . . . . . . . . . . . . . 163

5.2

Hypothetical Effect of Voters/Nonvoters on Mass Political Movements with the
Low or High Press Freedom Interaction . . . . . . . . . . . . . . . . . . . . . . 164

5.3

Hypothetical Relationship of Leader Survival: The Effect of Nonvoters and Mass
Threat on Leader Deposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

5.4

The Effect of Nonvoter Size on Mass Political Movements . . . . . . . . . . . . . 166

5.5

The Effect of Nonvoter Size and Mass Threat on the Risk of Leader’s Deposition 168

3.3

3.4

3.5

C.1 Chapter 3: Multilevel Linear Model of Partisanship and W/S Controlling for
Influence Cases (Nigeria and Mali), Dependent Variable: Political Trust . . . . . 186
D.1 Chapter 3: Countries (59 country-year samples) . . . . . . . . . . . . . . . . . . 191
E.1 Chapter 4: Brant Test for Loyalty & Comparative Loyalty . . . . . . . . . . . . 192

x

F.1 Chapter 4: Partial Proportional Odds Models of Impeachment Support (Comparative Loyalty) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
G.1 Chapter 4: Multinomial Logit Models of Impeachment Support (Loyalty) . . . . 198
H.1 Chapter 5: Models for Nonvoter’s Substantial Changes . . . . . . . . . . . . . . 201
I.1

Chapter 5: VIF from Model 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204

xi

LIST OF FIGURES

3.1

Hypothesized Citizen’s Trust (Short-term W/S) . . . . . . . . . . . . . . . . . .

73

3.2

Hypothesized Citizen’s Trust (Historical W/S) . . . . . . . . . . . . . . . . . . .

74

3.3

(a) Predicted Probability of Political Trust by the Loyalty Norm and Partisanship
(Short-term W/S) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

75

(b) Predicted Probability of Political Trust by the Loyalty Norm and Partisanship
(Historical W/S) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

76

3.4

4.1

Hypothetical Relationship between Loyalty and Impeachment . . . . . . . . . . 108

4.2

Distribution of the Loyalty by Election Winner (For interpretation of the references to color in this and all other figures, the reader is referred to the electronic
version of this dissertation) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

4.3

Distribution of the Loyalty by Election Loser . . . . . . . . . . . . . . . . . . . . 110

4.4

Distribution of the Loyalty by Non-partisan . . . . . . . . . . . . . . . . . . . . 111

4.5

Distribution of the Comparative Loyalty by Election Winner

4.6

Distribution of the Comparative Loyalty by Election Loser . . . . . . . . . . . . 113

4.7

Distribution of the Comparative Loyalty by Non-partisan

4.8

Discontinuity of Impeachment Support by Cohort in 2004 . . . . . . . . . . . . 115

4.9

Predicted Probability of Impeachment Support (Strongly Opposed vs. Opposed/(Strongly)
Supportive): Loyalty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

. . . . . . . . . . 112

. . . . . . . . . . . . 114

4.10 Predicted Probability of Impeachment Support (Strongly Opposed vs. Opposed/(Strongly)
Supportive): Comparative Loyalty . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.11 Predicted Probability of Impeachment Support (Strongly Opposed\Opposed vs.
(Strongly) Supportive): Loyalty . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.12 Predicted Probability of Impeachment Support (Strongly Opposed\Opposed vs.
(Strongly) Supportive): Comparative Loyalty . . . . . . . . . . . . . . . . . . . 119
5.1

Predicted Effects of Electoral Mobilization on Revolutionary Movements (2-way
interaction) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

xii

5.2

Estimated Baseline Cumulative Hazard (Stratified by W in Model 7) . . . . . . 171

C.1 Influence Country Robust Check for Model 4 . . . . . . . . . . . . . . . . . . . . 189
C.2 Influence Country Robust Check for Model 8 . . . . . . . . . . . . . . . . . . . . 190
J.1 Predicted Effects of Electoral Mobilization on Revolutionary Movements with the
Low or High Press Freedom Interaction (3-way interaction) . . . . . . . . . . . . 205
J.2 Goodness of Fit (Model 4)

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

J.3 Goodness of Fit (Model 5) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
J.4 Goodness of Fit (Model 6) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
J.5 Goodness of Fit (Model 7) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
J.6 Goodness of Fit (Model 8) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

xiii

Chapter 1
INTRODUCTION

On March 12, 2004, South Korea witnessed an unprecedented political event when the
opposition-controlled National Assembly impeached President Roh Moo-hun by a vote of
193 to 2, suspending all his presidential powers. As he was elected as the 16th President of
South Korea with a razor-thin margin in December 19th, 2002, the liberal president was afflicted by the lack of legitimacy that had often been a challenge to himself and his supporting
coalition. Until Kim Dae Jung became his predecessor from 1998 to 2002, conservative parties had dominated Korean politics through either three conservative party merge or through
inveterate interregional rivalry since the authoritarian Chun Doo Hwan regime had agreed
to step down from the people’s uprising in June, 1987.
Against the dominant conservative rule of South Korea (hereafter Korea), “Roh’s personal background, political style, and electoral constituents amount to something radically
new" to Korean politics (Hahm and Kim, 2005: 28). Roh and his reform-minded followers
defected from the incumbent Millennium Democratic Party (MDP) that was founded by
Kim Dae Jung, and the defection clearly represented their awareness of the citizens’ yearning for political reform and innovation. Roh’s presidential campaign also relied on class and
generational politics that had a basis of political reform. His defection culminated in his
followers’ inauguration of a new political party, the Open Uri Party (i.e., Open Our Party).
Although the legislative impeachment was initiated when Roh publicly supported the
new Open Uri Party (Uri Party, hereafter) to gain more legitimacy and support for his
reform drive as well as the victory in the upcoming general election, Roh’s tendency of appealing directly or populistically to the citizens, rather than pursuing existing institutional
channels, had really troubled the opposition parties including now the MDP. As the Korean

1

election law prohibits public officials from openly supporting a particular political party, 1
his audacious support for the Uri Party is believed as his calculated strategy to provoke
the opposition parties intentionally. Within a week, a constitutional agency of the National
Elections Committee (NEC) formally ruled that President Roh violated his duty to stay in
electoral neutrality. However, Roh defied the state agency and its rule on his pre-election
“influences”. Roh’s refusal to apologize for the violation of election law provoked the opposition parties that then seriously initiated impeachment proceedings. The opposition parties
that dominated the Assembly with over two-thirds of the seats wielded their constitutional
power of impeachment only about a month before the National Assembly election.
On March 9, President Roh’s opponents in the National Assembly introduced an impeachment motion. By allowing the president to have 72 hours to defuse the situation, his
foes might have been placated by his simple apology and a promise of maintaining neutrality
until the 17th general election. On March 11, however, Roh characteristically refused to back
down, giving a nationally televised speech.
Why were there candle vigils in South Korea to support the quite unpopular president’s
reinstatement? In a counterfactual statement, would the Court’s decision be different if the
opposition parties could win the 17th general election and get a majority in the National
Assembly?2 Beyond the Korea case, how is the risk of deposition understood by looking
at the relationship between electorates’ perception of the political system and their protest
potential? The determinants to explain leader survival may include presidencies being interrupted, prime ministers losing a vote of confidence, and even parties transferring power
to the opposition.
1 Until May 2004, President
2 In the counterfactuals, we

Roh was not yet formally a member.
may pose a similar question of why Roh was willing to take
the risk of impeachment by refusing to accept the opposition’s demand for an apology. Many
political leaders, we assume, try to keep power rather than risking it.

2

1.1

Research Question and Argument

Scholars explaining leaders’ survival frequently deal with the effects of political institutions
on whether political leaders and the parties in power can maintain their time in office. Based
on the predominant institution-based explanation of leaders’ survival, domestic institutions
ascribe different goals or interests to incumbent leaders and challenging candidates. Both
competing sides should be aware of the constraints given by institutions and the immediate
institutional reforms.
Leaders’ maximization of political survival is argued to be largely constrained and influenced by institutional characteristics that affect a leader’s and opponents’ electoral strategies
and outcomes. These include the regime types, partisan resources in the assembly, president’s
share of first-round vote, and different party goals between presidential and parliamentary
systems (Bueno de Mesquita et al., 2003; Maeda and Nishikawa, 2006; Kim and Bahry, 2008).
Regarding postwar leadership turnover studies in international relations, highly stressed is
the impact of domestic regimes, significantly interacting with conflict outcomes such as war
costs, and also in the context of rivalry relationships (Bueno de Mesquita and Siverson, 1995;
Goemans, 2000a; Colaresi, 2004b; Skocpol, 1979).
Most studies that examine survival of political leaders tend to neglect how their constituents (both voters and nonvoters) view the leaders’ competence, parties’ representation
and accountability of political systems, and how electorates’ behaviors, both conventional
and unconventional, instigate political instability that may substantially increase the risk of
deposition. Although the effect of electoral and mass politics on leaders’ tenure seems to be
an inherently important concern, the relatively recent development of comprehensive leader
survival data sets has paid more attention to the institutional and structural determinants,
rather than electorates’ behavior and attitude.
Unlike leader survival literature, however, regime transition literature highlighted the
role of mass protests that were preceded by the stages of liberalization and further democratization during the Third Wave – which started from Southern Europe, to Latin America, to
3

Asia, to post-communist Europe and to sub-Saharan Africa, and thus ended many autocrats’
rule (Huntington, 1993; Bratton and Van de Walle, 1997; McFaul, 2002).
While what is focused in the ‘transitology’ literature is political elites’ response to mass
politics at a critical juncture of the democratic transition (Colomer, 2000; Przeworski, 1991,
1992; Karl, 1990; O’Donnell, Schmitter and Whitehead, 1986; Rustow, 1970; on democratic
breakdown, Cohen, 1994; Linz, 1978a), the effects of mass politics on leader survival have
been regarded as a by-product of the macro-political phenomena of regime transition. The
literature largely has focused on regime survival but not on leader survival (Bernhard, Nordstrom and Reenock, 2001; Bernhard, Reenock and Nordstrom, 2003; Svolik, 2008). Therefore,
it is puzzling why the literature on the survival of political leaders has paid little attention
to the citizens’ perceived political legitimacy and the electorates’ voting behavior that could
stabilize political system, thereby extending the leaders’ survival.
In this sense, this dissertation is an effort to link the disjointed literature of mass politics
and leaders’ survival, especially in chapter 5. Before focusing on this, chapter 3 addresses
what causes citizens’ perceived legitimacy that can be a key indicator of protest politics, and
focuses on how the selectorate institutions determining leaders’ survival can also contribute
to the understanding of citizens’ perception. As citizens’ perception/attitude toward political
institutions in chapter 3 is not well translated into their electoral behavior of non-voting in
chapter 5, these two chapters of the cross-national large-N studies do not entail an intervening
variable of political legitimacy between the effect of selectorate institutions and leaders’
survival. This is because there exists a difficulty to make a proxy for perceived legitimacy as
a determinant for leaders’ survival because of limited public opinion data. Instead, chapter 5
centers on the electorates’ non-voting behavior to incorporate mass politics in an explanation
of the survival theory.

4

Some Implications from Regime Change Theory
Legitimacy is dependent upon the leaders’ competence, parties’ representation and accountability of political systems. It has been witnessed that people’s voice can induce many
processes of liberalization and further democratization that shortened many autocratic leaders’ survival since the Carnation Revolution in 1974 of Portugal. However, there exists the
literature on democratic breakdown (i.e., democratic survival, not leaders’ survival) that
deals with dissatisfaction with democracy and lack of confidence in institutions (Lagos 2003,
Valenzuela 2004).
By re-examining the actor-centric, pact-making, and top-down explanation of the thirdwave democracy, other key democracy scholars center on the role of mass politics and ‘democracy from below’ by focusing on political participation and competition and the balance of
power between old elites and oppositional masses during the transitional period (Bratton and
Van de Walle, 1994; McFaul, 2002). Similarly, during democratization and regime breakdown, there may have been a higher correlation between discontent with existing regimes
and lack of confidence in political institutions, and there might have been resultant outcomes
of mass protests and even military coups (Lagos, 2001; Valenzuela, 2004; Carlin, 2006).
Borrowing from the aforementioned regime transition arguments centering on the interaction between mass politics and electoral outcomes, my theory centers on the finding that
the leader’s survival literature lacks consideration of mass threats either from electoral or
non-electoral politics, while focusing heavily on the institutional and external constraints
given to leaders. A highlight on electoral and mass politics may be also appropriate for the
study of leader survival in a sense that a leader’s response to mass protests may include a
democratic transition to secure her survival. An old leader’s deposition could be catalyzed by
democratically committed masses under favorable institutions that allow for the opposition’s
mobilization against the ancien régime elites. On the other hand, with unfavorable institutions for the masses and (maybe the resultant) lack of mass mobilization for demonstrations
and protests against the ancien régime, some postcommunist countries such as Turkmenistan

5

and Uzbekistan kept their presidents from within the old communist elites, i.e., the former
first secretaries of the Communist Party.
While regime change need not result in automatic leadership change, it inherently bears
high probabilities of leadership turnover, more often with revolutionary changes and decolonization processes. Having revolutionary threats from a popular uprising, leaders consider
some useful measures such as previous electoral outcomes and protest potentials that could
indicate that leaders are at risk of deposition. With no experience of free and fair elections
and mass demonstrations, in contrast, old regime elites could mistakenly accept liberalizing
processes and the founding elections in which they could not win. Therefore, the risk of
deposition of leaders is argued to be a function of citizen’s electoral behavior in chapter 5,
especially voters’ non-voting behavior.

1.2

Electorate’s Beliefs and Behaviors in the Selectorate Theory

This dissertation analyzes the associations between political trust and the selectorate institutions (chapter 3), and between mass political movements and leaders’ survival (chapter 5)
by revisiting “selectorate theory”. Bueno de Mesquita et al. (2003) propose a simple and
elegant theory of the winning coalition and the selectorate to explain key political outcomes
such as corruption, war and peace, and leaders’ survival. Even though the new concept and
measurement of the main variable of winning coalition has raised many controversies over
their conceptual understanding (Clarke and Stone, 2008; Kennedy, 2009), the leader’s coalition of the winning or ruling system seems to provide useful measures for both the leader
herself and citizens to perceive whether their political playground is favorable for the former
or leveled out for the latter.
Bueno de Mesquita et al. (2003) define the selectorate as “the set of people whose endowments include the qualities or characteristics institutionally required to choose the government’s leadership and necessary for gaining access to private benefits doled out by the
government’s leadership” (p. 42): They defines the winning coalition as “a subset of the se6

lectorate of sufficient size such that the subset’s support endows the leadership with political
power over the remainder of the selectorate as well as over the disenfranchised members of the
society” (p. 51). If measured appropriately, the size of the winning (or in autocracy, ruling)
coalition is supposed to be translated into a mechanical evaluation of the leader’s available
resources, which in turn could determine whether the leader wages wars or maintains peace,
whether she may be liable to be corrupt to secure her limited resources or accountable for
the electorate, and thus whether she could secure the winning coalition’s support for her
survival or not.3
Consequently, the smaller ratio of the winning coalition size (W) to the selectorate size
(S), which implies greater “loyalty norms”, gives the leader more chance to secure her incumbency.4 This is because the small size of winning coalition would not consume all the
resources from her coffer that the leader uses for the coalition’s support for her survival
while maintaining the loyalty norms. Therefore, as it is reasonable to assume that the leader
always has limited resources to meet her coalition’s demands, the smaller ratio of W to S facilitates a greater loyalty norm, thereby reducing the leader’s burden of resource allocations
and extending her survival.
This top-down, mechanical and institution-focused selectorate theory of leaders’ survival,
however, neglects the important role of the citizens’ perceived political legitimacy that can
be expressed in public opinion data and can be captured in electoral behavior. Borrowing the
concept of the loyalty norm, not in a mechanical and structural sense but in a psychological
and behavioralist context, in Chapter 3, I first focus on whether and how two categories of the
electorate, such as electoral losers and non-partisans differently, perceive the political system
from electoral winners. Similar to the findings of the political trust literature (Anderson et al.,
2005), I find that both segments of this non-winning electorate have lower trust levels than
3 Leader

is referred to as female pronoun here while challenger/opposition is referred to
as male pronoun, as Bueno de Mesquita et al. (2003) did.
4 While the ratio of W to S is defined as the loyalty norm, increased size of W means less
loyalty to leaders. That is, the small group of winning coalition usually carries strong and
focused support for their leaders.

7

winners. More interestingly, both non-winners are sensitive to the levels of the loyalty norm
in a different manner. While political institutions are negatively perceived by non-winners,
it is argued that the increased chances of being winners, i.e., the greater size of the winning
coalition and equivalently the smaller loyalty norm to a specific leader, may translate the
negative perception of electoral losers into more positive belief in government. Specifically,
political institutions characteristically defined by the loyalty norm provide an interactive
relation with non-partisan’s perception of the system since non-partisan’s psychological traits
of distrust, or more accurately alienation, can be rather influenced by long-term accumulation
of the loyalty norm, instead of short-term changes.
In Chapter 4, I discuss the aforementioned South Korean impeachment case in more
detail. Korea’s impeachment case in the year 2004 provides a supporting example for the
previous chapter’s key finding that electoral losers’ consent with political institutions complies with the short-term changes of the loyalty norm while non-partisans’ political trust
appear to be relatively insensitive to this kind of the loyalty variation.5 This chapter deals
with the relationship of the support for the party in power (President Roh’s URI party)
as the independent variable and the support for President Roh’s impeachment trial as the
dependent variable according to different partisan status, and finds that electoral losers’
support for impeachment, like trust level of political institutions, appears to change more
extremely than non-partisans’ when the support for the Uri Party changes in the shortterm. Furthermore, an election survey that was done between Roh’s step-down following
the impeachment vote and the final decision of the Constitutional Court allows for a rare
opportunity for me to provide an illustrative example for more extended examination of the
relationship between electoral politics and leader survival in the next chapter.
My final theoretical results and contribution center on the findings in Chapter 5. The
findings suggests that political alienation measured by an increased number of non-voters
is more likely to bring about frequent mass political movements, which in turn destabilizes
5 The

long-term loyalty can not be examined in a single survey data.

8

the incumbent’s tenure and thus risks her deposition. A measure of the non-voter’s nonvoting behavior as a proxy for political alienation in aggregate-level data is unavoidable and
requires me to shift my focus from non-partisans to non-voters. While a discussion of the
appropriateness of this shift is found in Chapter 5, other than for the purpose of analysis
the non-voter’s measure is more appropriate to examine the behavioral outcomes of mass
political events.
In the concluding chapter, I review the main findings and implications from each chapter,
and discuss future projects that can be derived from this dissertation.

9

Chapter 2
THEORY AND LITERATURE REVIEW

2.1

Introduction

In this chapter, I discuss how selectorate theory has been developed from and developed
into studies on beliefs and behaviors. I highlight both the literature on social values and
perceptions, considering how citizens’ perception and behavior affect political survival in
certain political systems, and whether perceived legitimacy can be also shaped by institutional conditions and then affect leaders’ survival. Therefore, I review both the literature
on political trust and on leaders’ survival, as these are the variables of interest. I review the
related theoretical arguments and discussions on these topics.
Trust in political institutions can be viewed as a dependent variable and considered as
one of key indicators of political legitimacy. Political trust has been a key concept for many
scholars in political science to explain democratic development. After defining political trust,
the consequences of political trust on voting and mass political movements is surveyed in light
of system stability and leader survival. Then, electoral de-participation and its aftermath
of mass movements is discussed in terms of the proxy of political distrust. Finally, the
consequences of electoral de-participation and mass political movements on leader survival
are discussed as a final causal explanation.

2.2

Political Trust Revisited: Election and Selectorate Politics

For an understanding of how the non-winning electorate, i.e., electoral losers and non-voters,
in a certain institutional context evaluate and threaten a leader, the general perception of
the political system by the electorate must be carefully examined. Trust or confidence
that citizens place in political institutions is an important indicator for this kind of general

10

perception and a key to political development that can be based on a mature citizenship.
Whether in advanced industrial countries or in democratizing countries, political trust allows
citizens to be involved in their own governing systems, while government that represents
citizens establishes the stability and legitimacy of its democratic representation and policies
(Bianco, 1994; Levi and Stoker, 2000).

2.2.1

What is Political Trust?

Since Easton’s seminal work on support for the political system in the late 1960s, political
support and trust have been key concepts in understanding political development. Easton
(1965; 1975) and Gamson (1968) proposed, and the concept was empirically tested and
specified by Miller (1974a; 1974b) and Citrin (1974).
Easton (1965) provided a relatively clear definition of political trust and differentiated the
types of the support for the political system. By differentiating two types of input - demands
and support - into the political system, the political system produces policy outputs, which
again affect subsequent demand and support. He defines support as “an attitude by which a
person orients himself to an object either favorably or unfavorably, positively or negatively
[...] support refers to the way in which a person evaluatively orients himself to some object
through either his attitudes or his behavior” (436). Therefore, support is a key attitudinal
variable that can be easily translated into behavioral consequences in the working of the
political system.
Two types of support were distinguished by Easton (1975): specific support and diffuse
support. Specific support entails citizen’s relatively fleeting satisfaction with performance
of the political authorities. He noted that specific support entails the satisfaction obtained
by members of a system resulting from the perceived policy outputs and performance of the
political authorities.
Diffuse support is defined as “the reservoir of favorable attitudes and good will” for the
objects of the political system. With that, citizens can accept and tolerate undesirable

11

outputs produced by their political system. Unlike specific support, it is closely associated
with generalized attitudes and perceptions toward three political objects, the community,
regime, and authorities.
The main idea about political trust as an intervening variable between electoral politics
(election winner/loser/non-partisan) and job security of a leader inherently faces a limitation
of research design. Macro-level leader survival analysis does not allow for direct measurement
of the individual-level psychological effect. In this situation, borrowing from existing theory
and research from political psychology, I employ the construct of accessibility that plays an
important role in sorting out institutional effects on an individual’s attitude to the political
system.
Here, accessibility can be defined as “the extent to which information or attitudes are
retrieved from memory and used in making judgments or decisions” (Fazio, 1986; Chanley,
2002: 480). In this vein, political trust toward government (parliament, president, judiciary,
army, police, etc) is argued to be more accessible in memory than an individual’s less strongly
held feelings (Fazio, 1986). As I will discuss in the first part of my theoretical causality
model in chapter 3, citizens’ perception of a system’s institutions and their attitudes to
the trust level of the institutions could be sensitive to the institutional development since
its history. When political trust as an attitude is accessed from the citizen’s memory of
the institutional development (i.e., historical loyalty norm (W/S)), it can produce different
individuals’ perception of the political institutions than the current institutional context
(i.e., short-term loyalty norm (W/S)). Thus, citizens may consider political institutions in
evaluative terms relying on their own memory.
Specifically, research in social and political psychology contends that attitudes that are
to affect decisions, judgements, or behavior must be both available and accessible from
memory (Aldrich, Sullivan and Borgida, 1989; Higgins and King, 1981). Fazio’s (1986)
model of attitude accessibility posited that more accessible attitudes are more prone to
guide the processing of relevant behavior and information. Individuals’ attitudes to the

12

political institutions accessible from their memory of institutional development are in this
context hypothesized to be of substantial importance in understanding the attitudinal effect
of political alienation and its subsequent effect of nonvoting behavior on leader survival. I
will discuss the second part of this causal chain in chapter 5.

2.2.2
2.2.2.1

Political Trust, Policy, and Participation
Trust and Policy Attitudes

Trust in government has long been considered to be connected to citizens’ satisfaction with
government’s policies. In this sense, if citizens’ expectations of government policies are
achieved, trust in government can be expanded (Miller 1974a; Hetherington 1998). However,
very few have focused on the other direction; i.e., political trust can have an impact on the
formation of policy attitudes. Hetherington (2005) posits “why trust matters” in light of
the consequences of political trust/distrust, while noting that many have already tried to
identify the causes of this political attitude since the 1970s (Citrin, 1974; Miller, 1974a,b).
Likewise, exceptions include Chanley, Rudolph, and Rahn (2000), Hetherington and Nugent
(2001), and Hetherington and Globetti (2002).
An individual’s ideological beliefs have long received attention in the study of welfare policy attitudes, despite well-known reservations regarding the individual’s capability to engage
in ideological thinking (Converse 1964). The ideology argument presumes that attitudes
toward the welfare state are rooted in more general value systems regarding the proper relationship between the individual, the state and other institutions such as labor markets
and voluntary organizations (Feldman and Zaller 1992). These contradictory values and
beliefs, which have been labeled achievement and equality by Lipset (1963) and economic
individualism and social equality by McClosky and Zaller (1984), can provide the ideological
justification for either supporting or opposing welfare programs.
In support of the welfare state is the belief that all citizens have basic social rights

13

including the right to an acceptable level of economic welfare and security and the right
to live according to prevailing social standards. In opposition to the welfare state is the
concept of economic individualism, which assumes that each person is responsible for his
or her own welfare, and which understands individual well-being to be an outcome of hard
work. According to this view, citizens are expected to do what they can to be economically
self-sufficient. The welfare state undermines this principle by excusing some citizens from
their economic responsibilities and by fostering inappropriate behavior among recipients of
benefits.
Several studies support the thesis that attitudes toward the welfare state are connected
to more general ideological dispositions. Jacoby (1994) found a more coherent structure
in public attitudes toward social welfare expenditure than other government expenditures,
which he hypothesized reflected stronger ideological conflict over this issue. Both Sears et
al. (1980) and Hasenfeld and Rafferty (1989) found that endorsement of the concept of
social rights was a key predictor of welfare state support. Similar findings were reported in
open-ended interviews conducted by McClosky and Zaller (1984). And analysts find that
ideology shapes spending attitudes across a wide range of issues (Jacoby 1994, 2000) and
is particularly influential among those whose ideological orientations are more accessible
(Huckfeldt et al. 1999).
Hetherington (2005) develops a theory designed to explain the conditions under which
political trust will influence citizens’ policy attitudes. Citizens are more likely to support
an expansion of services when the government that will deliver those services is perceived
as trustworthy. While this theory seems to assume a simple “feed back” or reciprocal effect
between political trust and policy attitude, his key argument lies in the activation of the
trust heuristic that is tied to perceived sacrifice or risk associated with a particular policy;
specifically, political trust is expected to be activated when individuals are asked to sacrifice
their own material interests for the advancement of political minorities. He argues that
increasing political distrust, not increasing conservatism, explains why policy agendas have

14

become more conservative in the United States. His understanding of political trust/cynicism
in terms of a cost-benefit perspective regarding the policy agenda is well described in the
following statements: “people do not need to trust the government much when they benefit
from it [...] [P]eople need to trust the government when they pay the costs but do not receive
the benefit” (4).
In this context, Rudolph and Evans (2005) successfully pick up where the previous
ideology-government-spending literature left off. They provide significant support to Hetherington’s “sacrifice-based theory of political trust” by looking at the moderating effect of
ideology on political trust. In doing so, while Hetherington shows that political trust figures
more prominently in shaping public support for redistributive spending than distributive
spending, their analysis demonstrates that that ideological stance of citizens conditions the
effects of political trust on attitudes toward distributive as well as redistributive spending.
The revised ‘cost-benefit analysis’ in political attitudes is linked with the ideological stance
of citizens, which implicitly signifies how political trust can be strongly associated with sacrifice, not only from economic benefits, but also from political benefits, the latter of which
include ideological costs (i.e., sacrifice). Conservatives tend to have more pronounced effects
of political trust on support for government spending than liberals (Rudolph and Evans,
2005).
In the following sub-section, I review the literature connecting between trust and vote,
and between trust and mass political movements, mainly focusing on the consequences of
political trust/distrust. Specifically, if one is alienated from the political system, the strength
of alienation should be greater: “perceptions of the polity are far more powerful factors in
producing alienation than social background variables” (Schwartz, 1973: 11). The consequences of political trust, as represented by the link of distrust-alienation, can be citizens’
behavioral outcomes such as political participation, either conventional or unconventional
(Barnes et al., 1979). Borrowing from Inglehart’s distinction, these participatory acts can be
divided into two distinguishing ones: the first one is elite-directed and expressive participa-

15

tion such as voting; and second one is citizen-directed and instrumental participation such
as mass political events (Inglehart, 1997). Below, political trust has different consequences
according to these two participatory acts, and in addition, empirical analyses focus on these
two effects on leader survival in chapter 5.

2.2.2.2

Trust and Voting

Early theorizing about political alienation and disaffection suggests that distrust might bring
about a lack of political engagement (Almond and Verba, 1963; Finifter, 1970), especially
seeming to be the case when decline of voting turnout in the United States coincided with
the decline in political trust. Contrasting to these early theories, many scholars find no
relationship between trust and vote turnout (Miller, 1980; Rosenstone and Hansen, 1993).
The weakening of partisanship and declining belief of “external” efficacy, not political trust,
are largely considered as two major reasons for the decline of electoral participation in
the U.S. (Abramson and Aldrich, 1982). Similarly, Miller, Goldenberg and Erbring (1979)
suggests that political confidence affects participation in presidential elections, only with an
indirect effect through external efficacy.
Many scholars have discussed whether citizens’ perception of trust in government is a
prerequisite for democratic governance (Stokes, 1962; Citrin, 1974; Miller, 1974a; Citrin
et al., 1975; Abramson and Finifter, 1981; Nye, Zelikow and King, 1997; Warren, 1999; Pharr
and Putnam, 2000; McDonald Michael and Popkin Samuel, 2001; Mackenzie and Labiner,
2002; Dalton, 2004). Miller (1974a) argued that in the United States low levels of trust in
government represented a prevailing and enduring dissent with government, which might
be “increasing the potential for radical change” (951). And, he interpreted the persistent
drop of political trust and political efficacy as an indication of “a situation of widespread,
basic discontent and political alienation” (915). Political trust to Miller was centered on
the difference between citizens’ utilitarian motives and their expressive expectations; as the
discrepancy between motives and expectations grows, more public discontent prevails.

16

With a rejoinder, Citrin (1974) accepted Miller’s main conclusion that “policy-related
discontent is a source of political cynicism” and proceeded to focus on the following three
issues: “(1) The meaning of political trust as measured by the Trust in Government Scale;
(2) the independent impact, if any, of attitudes of political cynicism on political actions at
the individual level; and (3) A re-examination of the claim that a continuation of “centrist”
policies will inhibit the restoration of public confidence in the political process” (974, original
italic). Yet, “political cynics” who focus their dissatisfaction on incumbent authorities may
support political regime, thereby bearing negative evaluations of the incumbent national
administration, rather than general political regime.
Regarding the link between cynicism and uncustomary/unconventional political action,
Citrin found that increased cynicism was not strongly correlated with any “oppositionist”
political behavior of protest and with withdrawal from electoral politics, although admitting
that “political mistrust or alienation intervene between a sense of discontent and “oppositionist” actions requires [...] that cynical and trusting respondents differ in their behavior,
particularly at high levels of policy dissatisfaction” (980) (cf. Citrin and Green, 1986; Chanley, Rudolph and Rahn, 2000, 2001).
Yet, the null finding of a relationship between trust/distrust and participation using the
National Election Studies (NES) does not correctly suggest the effects of the distrustful and
even alienated voter’s attitudes on voting patterns with a different trust index (Seligson,
1983; Muller, Jukam and Seligson, 1982). Seligson’s use of both the NES trust index and
Muller’s (1977) “Political Support-Alienation” (PSA) index in survey data gathered in Mexico
allows scholars to re-examine the relationship between trust and vote turnout. This finding
suggests not only strong evidence against the validity of the widely-used NES trust index,
but also shows a strong relationship between political distrust and low vote turnout.
Regarding political trust and vote choice, Hetherington (1999) argues that politically
distrustful voters cast their votes more for candidates from the nonincumbent major party in
two-candidate races, while in the situation of three viable candidates, third-party candidates

17

get most benefits from declining political trust at the expense of two major parties. Similarly,
Miller and Listhaug (1990) suggest that voters channel distrust through support for protest
parties in multiparty systems.
In contrast, Koch (2003) argues that “political cynicism is not causally prior to third party
candidate preferences” (57). When using an instrumental variable of partisan independence
“as measured 2 years prior to the election” for third party support, he finds a significant
effect of third party support on political cynicism. The use of the instrumental variable of
non-partisan status is confirmed by finding in the ANES data that “it is the third party
candidacy that joins independence and political cynicism, two sets of political attitudes
that are otherwise unrelated to each other” (58). Citizens less likely to be partisan appear
to better embrace what a third party candidate criticizes about policies and social issues,
thereby deepening political cynicism. For example, the citizens who supported and were
involved in Perot’s 1992 presidential campaign were found to be far more distrusting than
the average citizens (Atkeson et al., 1996).
The aforementioned endogeneity of political trust/cynicism implies two things in this
dissertation: first, non-partisan status may tend to lead to political cynicism with the help
of viable alternatives that may entail opposition actors or supporting institutions; second,
how viable alternative actors/institutions are matters in linking partisan status and political
trust. These are discussed in chapter 3 in the form of the winning coalition and the selectorate
that may engender some viable alternatives to the non-winning electorate.

2.2.2.3

Trust and Mass Political Movements

While Seligson’s finding may be more applicable for non-US populations or only in Mexico,
the null finding from the NES trust index may cover another important consequence of political trust. More trusting voters in the United States may not be likely to go to the ballot box,
but distrusting voters may be likely to engage in unconventional political participation such
as protests and anti-government demonstrations. Gamson (1968) noted that “a combination

18

of high political efficacy and low political trust is the optimum combination for mobilization
– a belief that influence is both possible and necessary” (48).
Borrowing from Gamson’s (1968) argument of the relationship between discontent and
system-challenging behavior, many scholars focused on the possibility of increased unconventional participation caused by distrust and alienation. Most research on this hypothesis
lies in survey analysis during the 1960s’ and 1970s’ turmoil that found mixed results for
the link of distrust and protest activity/potential (Aberbach, 1969; Jackson, 1973; Citrin,
1977; Pierce and Philip, 1989; Muller, Jukam and Seligson, 1982). Earlier than Gamson’s
work, Easton (1965) specified two types of political discontent; diffuse and specific discontent, like diffuse and specific support. Diffuse political discontent is also considered to afford
a strong incentive for aggressive political disobedience or violence. This is so because diffuse
alienation from the political system “provides a normative incentive for aggressive political
participation” (Muller and Jukam, 1983: 172).
More specifically, Muller, Jukam and Seligson (1982) contended that only regime-oriented
diffuse discontent, and not incumbent-focused fleeting dissatisfaction, were significantly associated with aggressive anti-system behavior (e.g. Citrin, 1974, 1977). The use of diffuse
support/discontent implies that more general measures of political trust should be considered in examining the system stability that eventually affects leadership stability. Alienated
citizens thus seem to be no good for a leader’s security if they are accumulated.
In a narrower range of countries’ examples, Tarrow (2000) finds that loss of confidence
in government due to the government’s provision of misinformation is often a key cause
of political activism, and thus distrusting citizens are more prone to involve themselves
in contentious politics. However, distrusting but engaged citizens build up some kind of
“working trust” in government officials. In a somewhat similar context, Norris, Walgrave
and Van Aelst (2005) complements Tarrow’s argument of translation of “discontent into
constructive ... political engagement” in a similar context of postindustrial society’s protest
politics (Hall, 1999: 455). In the 1999 Belgian-Flanders general election study with a series

19

of surveys of protestors engaged in seven different demonstrations in Belgium, while the
authors disconfirm the Crisis of Democracy argument concerned about anti-state protests
characterized by unconventional political participation (Crozier, Huntington and Watanuki,
1975), they regard demonstrations as conventional in Belgian politics, and demonstrators
are not considered as “anti-state radicals who belong to socially marginal groups or who
despise conventional forms of political participation. By contrast, they are more similar to
the Belgian population as a whole than civic joiners and party members” (203).

2.3

Non-voting and Mass Mobilization: Risky for Leader?

Alienated citizens involved in abstention from elections are reasonably assumed to be prone
to find some alternative ways to express their voices if they can be mobilized. Hence, scholars
raised the following question: How is electoral abstention and de-participation derived from a
lack of political trust translated into mass political events? In the next sub-section, I review
the literature to attempt to answer this question. Conceivably, answering this question
raises another important question of how electoral de-participation poses the risk of leaders’
deposition.

2.3.1

Who is an Anti-system Non-voter?

Nonvoting is conventionally viewed as an individual-level phenomenon, as discussed in the
previous literature. Verba, Schlozman, and Brady (1995) focus on the three key individuallevel explanations for nonvoting: (1) because they do not want to, (2) because people are
not able to, and (3) because they are not mobilized to vote (see also Verba, Nie and Kim,
1987; Oppenhuis, 1995; Dalton, 2006). Thus, nonvoting results from ignorance, indifference,
dissatisfaction, or inactivity (Ragsdale and Rusk, 1993). This dissertation examines how
nonvoting behavior can be translated into anti-system “nonbehavior” in an aggregate form
in terms of its impacts on mass political movements and eventually the risk of leader’s

20

deposition. Thus, how can nonvoting be understood, and how does it have an aggregated
form of expression such as mass political movements?
Non-voters can be understood in two ways: using a the cultural/sociological approach or
an economic approach (cf. Barry, 1970).1 While the economic approach to voting behavior
emphasizes voters’ calculation of costs and benefits of voting (Downs, 1957; Olson, 1974;
Riker and Ordeshook, 1968), the sociological approach highlights the social context that
confers normative and habitual voting behavior on voters (Almond and Verba, 1963; Lipset,
1960). Similarly, the sociological approach appears to bear a strong relationship with “elitist”
theories of democracy (Dahl, 1954; Almond and Verba, 1963), and the economic approach
has much in common with the “institutional/constitutional” approach to democracy (Downs,
1957; Riker and Ordeshook, 1968; Ferejohn and Fiorina, 1974; Shepsle, 1979).
The economic approach to nonvoting is more centered on the political system, rather
than beliefs and attitudes, and entails institutional/constitutional reforms. The nonvoting
paradox was first introduced by Downs (1957) and developed by Riker and Ordeshook (1968),
who proposed innovative ideas of the nonvoting dilemma at the individual-level parameters
of the utility function. They focused on the concept of the “D term”:
E(U ) = (P ∗ B) − C + D
where E(U ) represents the expected utility of voting; P represents the probability that
the individual’s vote will actually affect the outcome; B is the expected benefit from the
preferred candidate’s winning the election; C is the cost of voting including being informed
about the campaign and going to the ballot box; and D represents the intrinsic benefit of
the act of voting including psychological satisfaction resulted from fulfilling one’s civic duty.
As the D term is not influenced by the probability of P , and C can be complemented by a
larger D, the expected utility of voting can be positive. Thus, some institutional conditions
1 Other

than this distinction, the propensity for non-voters or minority groups to overreport can be another important concern in the literature of non-voting behavior survey
(Silver, Anderson and Abramson, 1986; Abramson and Claggett, 1991).

21

were suggested to explain how a certain institution provides increased selective benefits
that increase the size of the D term (Aldrich, 1993), how to reduce the costs C of voting
through registration laws and polling hours (Rosenstone and Wolfinger, 1978), and why more
competitive elections increase the perception of P (Cox and Munger, 1989).
Under the rubric of a sociological approach, Dahl’s elitist theory assumes that the existence of non-voters is a healthy sign of citizen’s satisfaction with the political system and
democracy, and high rates of participation, on the other hand, indicate polarization and
extremism (cf. Berelson, Lazarsfeld and McPhee, 1954; Schattschneider, 1960). Unlike the
optimistic view of nonvoting, pessimists such as Edelman and Walker suggested that nonvoting is an inevitable response to “symbolism” and political culture that are applied by elites
who engender alienated masses and make them “acquiesce” in the elite’s decisions (Edelman, 1964, 1971; Walker, 1966). Thus, for these political scientists, a substantial increase of
abstention in voting would threaten democratic stability.
This dissertation suggests a hybrid approach between the economic and sociological
schools, which admits the importance of institutional conditions and constraints and individual electorate’s attitudes and behaviors toward state institutions of winning coalitions
and the selectorate. Bueno de Mesquita et al.’s (2003) conceptualization of the winning
coalition and the selectorate in terms of leader survival is argued to provide an institutional environment for the electorate’s perception of political alienation. Hypothetically, as
the winning coalition size allows for excluded groups (either non-partisans or non-voters)
to perceive how much their institutions are likely to alienate them, those excluded groups’
alienation can be strengthened (in greater W) or weakened (in smaller W) by the winning
coalition size. This is one of the hypotheses that will be tested in chapter 3.

2.3.2

Why are Non-voters Mobilized?

A classical dilemma between mobilization and political participation for a ruling elite in
developing countries was suggested Samuel Huntington (1968). Huntington argued that

22

when political participation is too great for institutional capabilities to assimilate it, system
instability can result. Indeed, he contended that “the political essence of revolution is the
rapid expansion of political consciousness and the rapid mobilization of new groups into
politics at a speed which makes it impossible for existing political institutions to assimilate
them. Revolution is the extreme case of the explosion of political participation” (266).
While he restricted the term revolution to the rapid and widespread transformations of
whole societies, “a fundamental continuity between revolution and lesser forms of conflict”
exists in his formulation of revolution (Tilly, 1978: 434). Therefore, to Huntington (1968),
any imbalance resulting from violence or “other forms of disorder” is argued to account for
revolution (358).2
Charles Tilly criticized Huntington, arguing that while Huntington did not make any
distinction between pre-revolutionary participation and participation in revolutions themselves, “we must disaggregate revolutions into its components instead of treating it as a
unitary phenomenon ... and we must specify and trace the relations of each major segment
to the changing structure of power” (436-7). Tilly’s centering on disaggregation and tracing
major segments suggests an alternative mode of specifying revolution to Huntington’s emphasis on mobilized groups and their preceding structural changes (cf.Moore, 1966; Skocpol,
1979). A prominent explanation of an aggregate psychological hypotheses could provide a
useful frameworks to understand revolution and political violence (Davies, 1962; Gurr, 1970;
Feierabend and Feierabend, 1966).
Thus, why should we focus on nonvoters? As described in the optimistic views in the
“elitism” approach (Dahl, 1954; Berelson, Lazarsfeld and McPhee, 1954; Schattschneider,
1960; Schumpeter, 1952), political quiescence is regarded as an important benefit to elites
as well as to leaders. With political quiescence, the leader extracts taxes without serious
resistance, maintains no opposing “voices” to the party in power, implements her policies that
may contribute the state’s economy, and thus protects the political legitimacy of her rule
2 Huntington

coupled violence with this kind of “other forms of disorder”.

23

(cf.Davenport, 2007). Even so, the pessimist view of the “elitism” approach raises questions
of the benefits, since widespread abstention from voting indicates a lack of regime support
that could be extended to threats to leaders (Edelman, 1964, 1971; Walker, 1966).
The aforementioned study by Seligson (1983) also noted the problem of widespread abstention from voting, in part due to the lack of political support. He observed that “Mexican
political leaders repeatedly make reference to the ‘problem of abstentionism’ and, in recent
years have waged vigorous campaigns against it. There is little question that elites view
growing abstention from the vote as a clear sign that support for the Mexican system of
government is declining” (13). Earlier, Milbrath and Goel (1977) suggested that political
support is strongly related to institutionalized political participation such as voting. They
also pointed out that when socio-economic status was controlled for, the relationship between political affect and participation tended to disappear. This is so because low SES and
political alienation seemed to be highly correlated.
Economically alienated citizens are also often categorized as politically alienated ones.
This notion is related to one of two main models of political violence: the relative deprivation
model and the rational action model. In the relative deprivation model mainly suggested
by Gurr (1970), many scholars found a strong relationship between inequality and mass violence (Muller, 1985; Muller and Seligson, 1987; Gurr, 1968a, 1970) and between potential
separatism and rebellious movements (Muller and Weede, 1990). Minority groups in terms
of either socio-economic or ethnic status tend to be liable to become involved in mass political movements, either at the national level or state/province level (cf.Wilkinson, 2004).
According to rational action models, costs of protest/rebellion are also dependent on the
structure of the political system and the leader’s behavior in response to protest (Muller and
Weede, 1990; Lichbach, 1998).
In a somewhat similar vein, Green (1984), while confirming Karl Deutsch and Myron
Weiner’s notion of a “crisis of participation”, argues that political non-participation should
be considered as an important factor explaining revolutions and other kinds of mass political

24

movements. “Popular involvement in the Iranian upheaval was inextricably linked to the
absence of participation before it began. It can be argued that a greater awareness on the
part of the Shah to the politicization accompanying Iranian development might have helped
him to preserve his throne.”
From both models, non-voters may be alienated economically and ethnically, and nonvoters’ alienation may be translated into their mobilization for protest, which could be
facilitated by institutional conditions such as the winning coalition size. Hypothetically,
increased size of the winning coalition can be argued to result in less isolation of the alienated
non-voters from the benefits of a leader’s policy provisions, thereby decreasing political
alienation for non-voters. This is also one of the hypotheses that will be tested in chapter 3.

2.4

Unpacking Leader Survival: The Electorate in Selectorate Politics

Leadership turnover has been continuously intriguing scholars in both comparative politics
and international relations to examine in diverse contexts such as regime type, war costs,
election timing, and domestic institutional settings. Bueno de Mesquita and Siverson (1995)
found in a systematic fashion that different regime types and war costs were key determinants
of leadership turnover during post-war periods. Goemans (2000a) similarly showed that “only
leaders of mixed regimes are likely to lose power and suffer additional punishment [in the
form of exile, imprisonment, or death] whether they lose a war moderately or disastrously”
(555).
Focusing on the rivalry context of international relations, Colaresi (2004b; 2004a) adroitly
suggests that after war, job security of a leader can be more secured in the context of
international rivalry than in one outside of rivalry, thereby a ‘hawkish’ leader in the rivalry
context being less likely to be deselected from power. These findings claiming that “wars in
many cases lead to violent revolutions and leadership turnover” were largely derived from

25

previous work of Skocpol (1979) and Tilly (1996) (Colaresi, 2004b: 713). 3 Therefore, war
outcomes were theorized and tested to breed political violence and thus leadership turnover.
Other possible progenitors of political violence, other than war outcomes and costs, should
center on delving into leadership replacement. As Norpoth (1987) argued that “war and
economics have few rivals when it comes to making or breaking governments” (949), the
provision of ‘butter’ should thus be considered as another key progenitor of political violence
that could threaten the job security of a leader. Similar to economic growth, inequality was
suggested as a predictor of political violence and leadership replacement (Midlarsky, 1988;
Muller, 1985; Muller and Seligson, 1987; Muller et al., 1989).
The theory and analysis developed from the relationship between inequality and violence
put forward a specific referent of those who may feel deprived relative to other citizens in
the society. Land’s and income’s patterned inequality do not seem to change at a fast rate,
and thus relatively deprived citizens seem to be liable to be economically alienated and to
be the sources of political violence, which may turn into political alienation. Slow-changing
patterns and perpetuated socio-economic conditions are reasonably assumed to produce the
alienated who may initiate every-day forms of resistance such as peasants’ ‘foot-dragging’,
and perhaps rebellious against a ruling elite as well as against their system (Scott, 1985,
1990;Midlarsky, 1988; Muller and Seligson, 1987).
In the following sub-sections, I survey the components that influence job security of a
leader, first votes and mass threats to leadership tenure, and selectorate politics for the
electorate.

2.4.1

The Electorate and Mass Threats to a Leader

How should the alienated or the powerless be examined based on more regular and dynamic
circumstances than unchanged socio-economic conditions, both in terms of their likelihood
3 The

reverse causal explanation from revolution to war is also suggested by Stephan Walt

(1996).

26

of becoming involved in political violence and in their support for the leader? Elections
and the electorate’s behavior can provide scholars with rich data sources. Recently, evident
is the identification of the relationship between fighting and voting, that is, linking “two
research programs, one on the sources of democratic consolidation and the other on the
causes, consequences, and dynamics of internal conflict” (Dunning, 2011: 2).
Dunning (2011) points to work in Jack Snyder’s From Voting to Violence (2000), which
links fighting and voting and also connects comparative politics and international relations.
To Dunning and other scholars contributing to the Journal of Conflict Resolution, elections
can be understood as either substitutes for fighting, complementing fighting, or suppressing
fighting. In the recent special issue of the journal, Chacon, Robinson and Torvik (2011) imply
that increased electoral competition engenders more fighting between parties and groups that
have an equal chance of winning elections. This is based on their assumption of higher costs of
being involved in fighting than participating in elections. If parties lack widespread popular
support, they also have a low probability of winning a civil war. Therefore, less electoral
competition may result in more peace.
In contrast, Machado, Scartascini and Tommasi (2011) in the same issue of the journal
argue that “actors who see little or no chance of having their interests taken into account
in the formal decision-making process are more likely to take to their plights to the streets”
(347). Arguably confirming the earlier work of Milbrath and Goel (1977), they show that
the aggregate number of street protests at the national level is negatively associated with
measures of “institutional quality”, which is indicated by the effectiveness of lawmaking bodies, the degree of judicial independence, and the stability of the party system. Their finding
coincides with the argument advanced by Przeworski (1991) who convincingly suggested that
“compliance depends on the probability of winning within the democratic institutions. A
particular actor i will comply if the probability it attaches to being victorious in democratic
competition, p(i), is greater than some minimum; call it p ∗ (i) . . . the more confident
the actor is that the relationship of political force will not take a adverse turn within the

27

democratic institutions, the more likely is this actor to comply; the less risky the subversion,
the less likely are the potential antidemocratic forces to comply” (30).
In a similar vein, Vreeland (2008) argues that civil war onset is assumed to be a function of the political participation components of Polity IV, rather than the chief executive
components of it. His admission of the importance of political participation by the citizens
highlights that an unstable political system characterized by civil conflicts and wars can be
a function of the instability of political participation by the citizens.

2.4.2

The Electorate in Selectorate Politics

Regime change and leadership turnover have been carefully examined by scholars of democracy. Specifically in sub-Saharan Africa, Bratton and Van de Walle (1994) highlight democratic transitions in Africa to occur more commonly from below. “Of twenty-one cases of
transition in sub-Saharan Africa between November 1989 and May 1991, the initiative to
undertake political reform was taken by opposition protesters in sixteen cases and by incumbent state leaders in only five cases” (461). With this notion, they argue that “when rule is
built on personal loyalty, supreme leaders often lose touch with popular legitimacy. They
lack institutional ties to corporate groups in society that could alert them to the strength of
their popular support” (462). The founding elections of sub-Saharan African new democracies suggest that cohesive opposition movements in Zambia and Niger were strong enough to
trounce the ruling parties and deprive them of control of the legislatures (Bratton, 1997).4
For the sake of leadership survival, which is not uncorrelated with regime change, fear of
rivals and opposition protesters may drive leaders to emasculate the very state institutions
that secure their loyal support (Migdal, 1988); or conditional on the free resource availability
and the extant size of winning coalition, the leader may expand or suppress the size of the
winning coalition (Bueno de Mesquita and Smith, 2010). Bueno de Mesquita and Smith
4 Regarding

the review and the role of election boycotters will be discussed in chapter 5’s
theoretical part.

28

(2009) show that increasing the winning coalition size can be in the interests of political
outsiders as well as leaders and coalition members when mass political movements especially characterized by revolutionary threats are threatening to a leader. By increasing the
size of the winning coalition, it can satiate potential revolutionaries by incorporating them
into political domains, and the leader’s policies can be rationalized in terms of rewarding
supporters.
Regarding the electorate’s response to the leader’s policy and institutional reforms derived
from mass movements from below, Przeworski (1991) formally modeled the self-enforcing
democratic equilibrium. He concludes that “[d]emocracy will evoke generalized compliance,
it will be self-enforcing, when all the relevant political forces have some specific minimum
probability of doing well under the particular system of institutions” (30-31). Further, Weingast (1997) highlights the role of the electorate by incorporating it into Przeworski’s model.
In his extension of Przeworski’s model, democratic stability that is a function of restrictions
on governing elites is only possible “if there exists a citizen consensus to react against tomorrow’s incumbents if they attempt to rig elections” (255). His notion of the citizens’ or the
electorate’s consensus signifies the important roles of electorate politics.

29

Chapter 3
THE PARTISAN-W/S NEXUS OF POLITICAL TRUST

3.1

Introduction

A recent study on political survival or leader longevity by Bueno de Mesquita et al. (2003)
proposes a simple and elegant institutional condition to explain political leaders’ survival
in a variety of political contexts such as presidentialism, parliamentarism and mixed (semipresidential) systems. The key institution is the loyalty norms, measured by the ratio of
the size of winning coalition (W ) to the size of selectorate (S ). Conceptually, the loyalty
norms have a negative relationship with the ratio of W to S, however: The higher the W/S
ratio, the less private goods distributed to each winner in the coalition; and equivalently, the
smaller the loyalty norms, the less they are loyal to the incumbent leaders.
Measuring a winning coalition is heavily dependent on the Polity indicators. As described in Appendix A, Bueno de Mesquita and colleagues use regime type (military/civilianmilitary/civilian), competitiveness of executive recruitment (XRCOMP), openness of executive recruitment (XROPEN), and the competitiveness of participation (PARCOMP).1 These
indicators are appropriate measures for a winning coalition in a sense that the indicators are
mainly based on the ways of executive recruitment and electoral participation for that recruitment. Using these indicators, their model entails that both the leader and challengers
offering policy benefits intend to form a coalition large enough to maintain or take power.
These elites should appeal to both groups of the winning coalition and the selectorate: the
selectorate is defined as those who “have a government-granted say in the selection of leaders"
(Bueno de Mesquita et al., 2003: 42); and the leader should form a supporting coalition of
enough citizens to maintain and stabilize her tenure though a winning coalition (W), “a subset of the selectorate of sufficient size such that the subset’s support endows the leadership
1

My revised measure of the winning coalition is described in Appendix A.
30

with political power over the remainder of the selectorate as well as over the disenfranchised
members of the society" (Buneo de Mesquita et al., 2003: 51).
What matters most to winners of a winning coalition is the degree to which a leader can
secure her power and tenure by spending enough resources for the supporting the fcoalition
against challengers who would propose better policies in pursuit of getting more defectors
from the leaders’ winning coalition. Bueno de Mesquita et al. focus on how a leader can
avoid defection of the winners to challengers by taking into account the affinity between
leaders and the selectorates, focusing on the size of W.
While some researchers contend that selectorate theory cannot capture the multifaceted
concept of democracy, others argue that the size of winning coalitions may tap only into
the presence/absence of competitive elections (Clarke and Stone, 2008; Kennedy, 2009).
Moreover, Bueno de Mesquita et al. pay little attention to the effect of protests on political
survival. Contrary to the limited institutional approach to explain leader’s survival, Kathryn
Hochstetler (2006) highlights the importance of public opinion and argues that the effect of
public opinion on the probability of street protest should be included in the political survival
literature.
Selectorate theory looks mostly (if not only) at the effects of political elites’ behavior and
citizens’ behavior, which seems independent of the effect of mass public attitudes. So the
disjuncture of selectorate theory and mass public attitudes is inevitably severe, and it ends up
missing another key part of the causal relation that includes mass public attitudes toward
the system and political elites. Furthermore, a leader and challengers interact with each
other according to their respective utility maximization, which this chapter argues should be
based upon both the size of the winning coalitions and the selectorate and on the citizens’
basic attitudes toward the system such as the partisan status of the selectorate (Anderson
et al., 2005).
In a similar vein, citizens’ attitudes toward the system are considered key factors in explaining feelings about the incumbent leaders and thus whether leaders can succeed or not

31

(Hetherington, 1998). While Bueno de Mesquita et al. highlight the elites’ behavior mechanically determined by the loyalty norm, they ignore the importance of culturally defined
functions of the loyalty norm; that is, they do not attend to how both electoral winners and
electoral losers in the selectorate view the institutions of the loyalty norms as well as elites’
behavior and how non-partisans support the system. Moreover, the culturally defined loyalty norms can interact with the effects of partisan status on political legitimacy. It entails
how selectorate theory explains the variance of political trust among winners and losers in
selectorate.
Therefore, this chapter develops an important link between two disjointed literatures
by using macro-/micro-level data to confirm that both literatures will contribute to each
other.2 The next section examines the literature concerned with political trust and relates it
to research on leadership stability. I introduce a theoretical connection between the loyalty
norm and partisan’s preference in terms of political legitimacy. Next, issues of measurement
and data analysis are discussed. After presenting the results from the multilevel analysis with
robust check, I discuss the importance of the loyalty norm in partisan support for regime
and spell out avenues for potential research in leader survival literature.

3.2

Selectorate Theory of Electoral Politics: Explaining Political
Trust

Political support has been a key concept for cultural theory of democracy, along with social
capital (Almond and Verba, 1963;Putnam, Leonardi and Nanetti, 1994). High institutional
confidence in a civic culture often emphasizes diffuse political support and is positively
associated with current regime legitimacy and the effective function of institutions (Easton,
1965, 1975;Fuchs, Guidorossi and Svensson, 1995;Gibson, Caldeira and Spence, 2003). Unlike
2 Scholars

contributing to political consent literature argue that satisfaction with democracy and confidence in institutions vary between winners’ and losers’ status (Anderson et al.,
2005; Anderson and Guillory, 1997). Their main institutional variables are the effective number of parliamentary parties, the disproportionality between seats and votes, and ‘alternation
in power’.
32

institution-oriented scholars, culturalists give more stress to identity, values, and norms,
thereby individuals’ rational evaluation of government performance being less important
(see, e.g., North, 1990; Akerlof and Kranton, 2000).
Despite much interest in institutional theories of democracy, system support scholars
have reinvigorated cultural arguments and even have incorporated institutional theories in
the study of democratic institutions and citizens’ behavior (Rothstein and Stolle, 2008; Rothstein, 2009; Uslaner, 2003; Hooghe and Stolle, 2003). Socially trusting individuals tend to
have more significant levels of political trust (Almond and Verba, 1963) while Bahry and
Silver (1990; 1987) show that greater social trust was associated with greater citizen activism
in both regime complaint and regime critical political action. Some theorists have had difficulty finding significant partial correlations between political trust and social trust at the
individual level, while aggregate-level analyses show stronger correlations (Mishler and Rose,
2005; Rothstein, 2002; Uslaner, 2002; Newton and Norris, 2000; Wright, 1976).3 Brehm and
Rahn (1997) also find significant effects of social trust on confidence in government by controlling for 13 other variables in the US data. With a more inclusive country sample, Freitag
and Bühlmann (2009) contend that the positive effect of more power-sharing institutions on
social trust is magnified for the individuals having “bridging” social capital and the positive
effect of income equality on social trust is also strengthened for people having more life
satisfaction.
While many students of political support try to disentangle the concepts of socially and
politically engaging attitudes and behavior, most (if not all) admit that contextual variables
matter in generating political support, and thus we need to attend to associated microand macro-level determinants causing public support in a certain context. Regarding microlevel constraints, the political consent literature has emphasized the distinction between
3 By

using rescaled 11-point social trust and political trust variables, recent study on
the relationships between political trust and social trust, and between satisfaction with
democracy and social trust shows a significant correlation at individual-level (Zmerli and
Newton, 2008).

33

winners and losers of elections; losers naturally tend to have lower evaluations of their own
system, depending upon the electoral systems (Anderson and LoTempio, 2002; Anderson
and Guillory, 1997); individuals having left ideology are also less likely to accept governing
authority to the extent that inequality is able to reinforce their discontent (Anderson and
Singer, 2008). In addition, when the political consent literature focuses on the variation
among individuals, compliant and unconventional political activities in the context of a
repressive regime implies that variations of political activism have different impacts on the
citizens’ level of public support (Bahry and Silver, 1987, 1990).
Macro-level constraints usually appear in combination with micro-level characteristics;
for example, the effects of electoral systems and income inequality in a country on political
support depend on individuals’ characteristics. For the latter, the Soviet Union’s repressive
regime itself allowed variations of political activism to occur, although people’s psychology
better explains the variations of political activism (Bahry and Silver, 1987, 1990). This
incorporation of micro- or individual-level variables into the study of formal rules/institutions
has been supported by Douglas North (1990), who states that more study of culturally
derived norms of behavior and their interaction with formal rules/institutions represents
“beginning the serious study of institutions.”
With a surge of research that integrates both individual- and country-level characteristics in mass political support, diverse combination has been possible for a limited contextual
country sample; that is, the studies are focused either on the established or on the new
democracies (Anderson et al., 2005; Anderson and Guillory, 1997; Cho and Bratton, 2006;
Cho, 2010, see Anderson and Singer, 2008 for an exception4 ). Moreover, while institutional
or generally contextual variables can be either structural, institutional, or cultural in nature (Kedar and Shively, 2005), most contextual parameters focus mainly on the current
institutional impact on system support. This may end up ignoring how system support
is also influenced by long-term contextual parameters. Therefore, in the study of political
4 Besides

their use of multi-level analysis, they include both Western and Central-Eastern
Europe while they are all European countries.
34

support, students are tempted to have more inclusive countries including established and
new democracies with appropriate institutional history variables in their models. This kind
of temptation is more pronounced when leader’s interests (mainly, her5 survival and stable
governance) do not seem to be unrelated with the mass public’s levels of political support
not only in the established countries but also the new democracies.
Connecting public support with leadership and institutional stability has been tried in a
limited context, mainly in Latin America (Hochstetler, 2006; Helmke, 2010). It is intriguing
to know that while leaders’ survival study across countries attends to the statistical significance of public protest and social mobilization in explaining interrupted presidents (Kim
and Bahry, 2008), what directly instigates citizens to be dissatisfied with leaders is missing.
A more mechanical institution-based explanation of leaders’ tenure stability has been done
by Bueno de Mesquita et al. (2003). To better explain how the leadership and institutional
instability occur across countries, analysis with individual-contextual parameters along with
historical or cultural components is helpful and even necessary.
Political trust is regarded as a good empirical indicator of how much citizens would
legitimize their political institutions and leaders’ tenure. Citizens’ legitimization of the
political system is expected to be critically dependent upon the relation between political
elites and the citizens’ vote choices. Without understanding the variation of individual-level
perceptions, political confidence may still be left out as an explanandum in the mechanical
limitations of institutional “explanans”.

3.3

Partisans, Loyalty Norm, and Political Trust

According to Easton (1965, 1975), political support has been analyzed at three levels of
the political system, i.e., authorities, regimes, and communities (Dalton, 1999, 2004). Especially, in the Miller-Citrin debate, it has been evident that both authorities and regime
5 Here

I follow the usual use of gender pronoun as seen in many survival literature; leaders
are female while challengers are male (Bueno de Mesquita et al., 2003).

35

become important predictors of trust in government. More specifically, while Miller and
colleagues referred to unresponsive political processes as a key predictor of declining trust
in government, Citrin and colleagues argued that incumbent authorities are key sources
of trust (Miller, 1974a;1974b;Miller and Listhaug, 1990;?;Citrin and Green, 1986). More
recently, Keele (2005) found that the effect of authorities (i.e., partisan loyalties) matters
for trust in government relative to evaluations of the political process, while Hetherington
(1998) contends that regime support has stronger impact on authority support rather than
the reverse.
Under the same community level, therefore, system support measured by citizens’ evaluation of their basic public institutions implies two aspects of the regime level, namely trust in
politicians/authorities and institutional performance. The authority level refers to specific
support based on partisanship, while the latter regime level with diffuse support or political
confidence. Scholars contributing to the system support literature explain institutional confidence by using the effects of partisan status as a main predictor of political support and
institutional legitimacy. This is because trustworthy institutions and stable leadership are
two inter-connected pillars of democracy.
As the voluminous literature in the American context has uncovered, political trust is
more associated with political winners and losers, rather than social winners and losers
(Abramson, 1983; Putnam, 1995; Orren, 1997). Przeworski (2005) contends in a similar
vein that citizens’ preference for democracy, being independent of income, affects elites’
calculation of losers’ rebellion. Hence, the losers’ levels of institutional legitimacy matter
more than those of the other sides, i.e., winners and non-partisans. This is because the most
deprived and dissatisfied group is the losers, especially the repeated losers of elections.
Institutional legitimacy ensures that despite the economic performance of the government, state institutions can be viewed by electoral winners as trustful and valued, and are
accepted in principle (Gerber and Huber, 2010, 2009). Significantly high levels of perceived
legitimacy among the mass public seem to come from either repetitive elections (i.e., the

36

experience of democracy) or the quality of elections (i.e., the fairness of elections) (Dahl,
2006; Manin, 1997; Dahl, 1991). While losers with low levels of political trust may tend to
actively engage in attempts to overthrow the existing regime (Przeworski, 1991;2005), partisans’ evaluation of the system does not simply reflect voters’ partisan status resulting from
an election outcome. Instead, the winner-loser gap in institutional legitimacy has been noted
as an important predictor of institutional legitimacy in the advanced industrial democracies
(Anderson et al., 2005; Anderson and Guillory, 1997; but, see also for exceptions Moehler
and Lindberg, 2009; Cho and Bratton, 2006).
As the level of perceived legitimacy varies according to partisan status, the losers’ consent
literature raises appealing arguments to the declining trends of political trust in the advanced
industrial democracies. As long as most elections are not blatantly unfair and stopped,
citizens (including electoral losers) tend to regard the election outcomes as possibly legitimate
in both electoral (i.e., illiberal) and liberal democracies. Electoral losers among politicians
themselves as well as the mass public who supported them would not readily protest against
’unfavorable outcomes’ to them (e.g., Esaiasson and Göteborg, 2009).
Therefore, losers in elections can give at least moderately credible signs for regime legitimacy and political stability in a certain institutional context, regardless of the levels
of democracy. The losers’ consent argument is based upon how partisan status interacts
with institutional traits such as party system, electoral institutions, and ’alternation effect’
(Moehler and Lindberg, 2009; Anderson et al., 2005; Bratton, 2004). These findings confirm
that political trust is strongly associated with citizens’ preference for open and responsive
government.

Representation and Implementation
While many political theorists have emphasized electoral democracy that provides equal
representation and procedural fairness in a democratic regime, they mainly deal with the
representation side of the political system (Dahl, 2006; Manin, 1997; Dahl, 1991). Beetham

37

(1991) noted, however, the output-side legitimating process of the government’s production
of goods and services as one of four important mechanisms of power legitimation (Beetham,
1991). Easton’s much earlier system analysis of political life had also noted the reinforcing
effect of system “output” on system support (Easton, 1965). Recently, some empirical findings
in both advanced industrial democracies and transitional democracies show that political
legitimacy measured by institutional trust can also be achieved by the implementation side
of the political system (Rothstein, 2009; Levi, Sacks and Tyler, 2009).
Regarding the implementation side, Rothstein (2009) presents an alternative legitimacyproviding mechanism through the “quality of government.” Moreover, even to create effective
representation, the quality of government may matter more than the representative capacity
of electoral systems. In a similar context, using three implementation institutions of the
police, courts, and tax departments in sub-Saharan African countries, Levi, Sacks and Tyler
(2009) contend that administrative competence along with procedural justice and government performance plays the most effective roles of generating the mass public’s legitimating
belief in political system.
While losers tend to focus more on the chances of becoming winners, winners care about
the benefits after validating “home team effects” to their candidates (Anderson and LoTempio, 2002). During this process, losers as well as winners would legitimize the political process
such as elections, especially after turnovers of power take place. Losers can readily concede
their defeat by having a prospective win in the next elections, realizing they would get more
private goods from the potential winning in the future, thereby helping democracy to be
sustained (O’Donnell, 2007; Przeworski, 1991). Besides the “home team effects” of winners’
significant system support, the winners’ positive evaluation of system also varies depending
upon how well winning itself can translate into their preferable policies in a certain context,
especially when their historical experience of policy implementation could reflect meager
provision of private and public goods.
Hence, I can expect that there should be some drops of the winners’ levels of perceived

38

legitimacy depending upon whether or how well the electoral outcomes are converted into
preferable policies, not only through their current representation but also through their historical experience of resource allocation institution. While current institutional implementation effects are important, it is also important to conceive long-standing characteristics
of institutional heritage in the analysis of the effects of institutional constraints on citizens
who usually bear institutional history in mind. Scholars are starting to incorporate this
characteristic into measures of governing institutions (Gerring et al., 2005; Gerring, Thacker
and Moreno, 2008).
Despite widely accepted interactive effects of the representation side of political institutions (i.e., electoral rules) and partisanship on legitimating beliefs, non-partisans’ (or independents’) institutional interaction has not been well studied in the losers’ consent literature.
Usually, non-partisans are considered the reference category and easily assumed to have moderate effects between two opposing groups of partisans.6 If the implementation side as well
as the representation side can be incorporated in the analysis, students of political consent
can understand better how non-partisans as well as losers interact with diverse contextual
country-level effects. This is because non-partisans are arguably more sensitive to implementation than representation.
In conjunction with the political consent literature, Colomer’s (2003) comparison between
plurality rule and majority rule for single-winner political systems helps us to understand
how losers under plurality could mute their negative feelings of losing elections. He argues
that “plurality rule is most vulnerable to losers’ strategies" (99). This can be understood
in the context of the ratio of W to S. The plurality rule requires only one-quarter or less of
the selectorate to win so that losers might have a chance to be winners in future coalitions
(Bueno de Mesquita et al., 2003).7
6 Recent focus on non-partisans, however, includes Anderson and Paskeviciute (2009).
7 Bueno de Mesquita et al. (2003) implies that oversizing leader’s coalition can be induced

by higher chance of her winning coalition’s defection to a new challenger. Also, the chance
of defection depends upon "the risk of exclusion in the challenger’s winning coalition" and
"the risk of exclusion from the leader’s winning coalition if the challenger fails to come to

39

For example, as a plurality rule entails a smaller W/S or greater loyalty norms than
majority rule whose W/S is about 1/2, election losers or the minority in a plurality rule
would be less likely to be included in the winning coalition than losers in majority rule,
thereby making leader’s welfare provision more likely to take the form of private goods.
Losers could introduce “relatively few new issues or create a few new alternatives” (99) and
incumbent leaders would need to recruit their supporters from outside of a winning coalition
to suppress supporters of challengers (Bueno de Mesquita et al., 2003: 345). Thus, I propose
that losers in small winning-coalition institutions will have less leeway to be included in the
future winning coalition even in single-winner political systems.
Another good example illustrates regarding the relationship between electoral systems
and the budget spending. More refined distinctions of electoral systems between singlemember district (SMD) and proportional representation (PR) are suggested for their effects
on the budgetary cycle that do not seem to unrelated with political elites’ calculation of
the political system. In the political business cycle (PBC) literature, politicians prefer to
use a specific form of budgetary cycle under certain electoral incentives; Eric C. C. Chang
(2008) finds that under SMD systems budgetary cycles take the form of more district-specific
spending to attract their loyal supporters, while proportional representation systems are
more associated with higher social welfare spending and expanding public goods, rather
than assembling loyal supporters.
Therefore, my argument extends the mechanism of electoral representation to more general institutions involving resource allocation for selectively represented voters. It is also an
extension in a sense that the loyalty norm institution broadly considered as the difference
between SMD (i.e., understood as large W/S with small loyalty) and PR (i.e., understood
as small W/S with greater loyalty) is a more comprehensive contextual variable with greater
explanatory power by considering the interaction of voters’ characteristics and elites’ resource allocation. Thus, the interaction of mass public and political elites accounted for by
power" (p.278-9). W’s defection, therefore, is more likely when W/S is small and the loyalty
norms are great.
40

the loyalty norm institutions and the voters’ consent permits examination of how resource
allocation is contingent on electoral incentives in explaining citizens’ beliefs about political
legitimacy.

3.4

Connecting (Non-)Partisans and the Loyalty Norm

The loyalty norms allow a leader to better understand how to satisfy her winning coalition,
stabilize the polity, extend her tenure and even care for her post-exit political life. As the
leader’s utmost interests lie in keeping her incumbency, current or short-term W/S seems to
be critical to her. With a limited time-span expectation of her tenure, she can calculate how
much resources are needed for her coalition. Yet at the same time, the loyalty norms provide
an institutional context for the selectors and the disenfranchised to perceive how legitimate
their system is. For the mass public, institutional history of the loyalty norms matters as
much as the current level of the loyalty norms. This is because the mass public’s attitudes
rely on their memory of how frequently leadership turnover has occurred in the polity while
the leader’s behavior is more based upon the short-term W/S.

3.4.1

Partisans with the short-term loyalty norm

It is important for selectors to perceive whether or how much welfare services can be available
to them. In particular, partisans are important actors in the short-term loyalty norm context
because partisans would feel either satisfied with welfare gaining by winning representation or
feel deprived of welfare loss by losing representation. Regarding the citizens’ perception, this
implies that most private goods will be distributed to the represented, i.e., to the winners in
elections according to the winning platform, although as shown above the amount of welfare
depends upon the winning coalition size.
From the short-term loyalty norms, partisans’ levels of perceived legitimacy can vary,
referring to political elites’ short-term and biased representation through office-seeking ten-

41

dency, rather than to their consistent and impartial policy implementation through policyseeking tendency (e.g.,Strom, 1990; Erikson and Romero, 1990; Adams, 2001b,a). Instead,
barely attached to parties are non-partisans who are not expected to have significant interests
(at least according to their lack of party affiliation) in getting more welfare via the winning
coalitions in the future. Thus, the short-term W/S is more associated with partisans and political elites’ office-seeking, rather than with non-partisans’ interests in fair implementation
of public goods.8
Winners in a small short-term loyalty norms are expected to have higher levels of political
trust because of increased party loyalty from electoral victory (i.e., “home team effect”). In
contrast, losers in the small short-term W/S have lower levels of political trust as their
weakened party loyalty from electoral defeat translates into lower levels of political trust.
However, the loyalty of winners to their leader is not automatically translated into trust in
political institutions. The winners’ loyalty in the winning coalition perceived as strong in
small size of W/S by leaders can be regarded as no more than temporary "support" for
her survival, and she may readily replace the current winning coalition with a greater size
of future winning coalition if she has a hard time getting support from the current winners
(i.e., if more chances of defection of W to challengers are noticed).
An increase of short-term loyal norms (i.e., a small size of W/S) would have winners
maintain their existing amount of welfare through high loyalty to, and high affinity with
the leader, and thus give more leverage to leaders than to her challengers who are not able
to initiate any institutional reform. The leader may manipulate the size of the winning
coalition and the selectorate to secure her tenure, and it will lead to the winners’ defection
to her opponent (Bueno de Mesquita et al., 2003;278). Bueno de Mesquita et al. (2003;333)
suggest that “as in rigged electoral systems, the selector’s preferences are very close to those
of the disenfranchised. The small value of W/S in these forms of governance means selectors
outside of W have little prospect of entering a future coalition when a new leader takes
8 Being

a non-partisan seems to be more associated with public goods preference, rather
than private goods that is more likely to be represented by being partisan.
42

office.”
Decreasing loyalty to the leader and her party leads to low levels of political trust among
winners (Keele, 2005; Anderson and Paskeviciute, 2009), especially when the leader may
be put into a situation where she has to include more winning coalitions (i.e., oversizing
winning coalition) or even replace repeated winners with repeated losers. This situation is
more likely to occur when old winners’ welfare needs cannot be met with the incumbents’
limited resources, when old winners can find more affinity with the challenger who may
be more competent, or when the size of the selectorate is large enough for the leader to
replace the current W whose demands cannot be met by her resources with losers and nonpartisans. However, the short time frame of current loyalty norms does not provide voters
with full historical memories of winner-loser rotation.
In the large short-term loyalty norms context9 , the leader may not have enough leverage
to manipulate the size of winning coalition. Both winners and losers will calculate the
probabilities of changed affinities with the leader and her opponents, and winners calculate
the risks of being excluded in a future coalition after they defect to one of her challengers.
As the increased chance of being included in the future winning coalition is more visible in
the large short-term W/S, increasing as well as replacing the winning coalition will dilute
the winners’ affinity with the incumbent leaders which may in turn decreases winners’ “home
team effects”. This is because the increased or replaced W will decrease the amount of
winners’ private goods by diverting original winning platforms. Hence, the old winners who
can remain in the winning coalition do not have exclusive welfare provisions while managing
to keep the “home team effect” with weakened party loyalty.
In contrast, losers in small short-term loyalty norms are now able to become winners
with more private goods, thereby gaining party loyalty and obtaining more legitimacy in the
system. Bueno de Mesquita et al. (2003) note that especially when both the winning coalition
and the selectorate are small, losers and non-partisans might have strong attachment to their
9 This

includes the situation when W and S are both small.

43

own institutions. If selectors outside of the winning coalition can become winners, they will
have significant stream of private goods as was done to old winners. Moreover, the high
ratio of W/S (or lower loyalty norms) means frequent leadership turnover, which increases
the probability of becoming new winners. The short-term loyalty norms, however, are not
able to allow voters to realize how frequently the leadership turnover has really taken place.

3.4.2

Non-partisans with Historical Loyalty Norms

Many institutionalists tend to ignore the cumulative effects of institutions except for the
recent works by Gerring and his colleagues (Gerring, Thacker and Moreno, 2008; Gerring
et al., 2005; Gerring, Thacker and Moreno, 2005; Moehler and Lindberg, 2009). The institutional effects should be considered to “cumulate over time as new institutional rules
begin to condition actions and expectations” (Gerring, Thacker and Moreno, 2005). As the
variable of interest in this section entails the causal effect of the long-term historical W/S,
the measurement of W/S centers not only on a country’s current levels of W but also on its
institutional experience and history. Especially, the cumulative effects seem to be more significant for the impacts on citizens as people’s memory matters for their political attitudes.
“[T]he collective memory about the actual operations of the institutions” accounts for trust
in institutions of law and order (Rothstein, 2000, p.493).
In the political consent literature, Moehler and Lindberg (2009) use Huntington’s “twoturnover-test” as a key contextual variable to condition sub-Saharan African citizens’ levels
of perceived legitimacy. Although they do not use a direct cumulative effect of turnover,
turnover experience conditioning citizens’ perceived legitimacy is well connected to citizens’
cumulative memories of contextual changes. Repeated losers and winners tend to narrow
their legitimacy gap, respectively, by being new winners and new losers through the experience of the power alternation (see also, Bratton, 2004).
Historical loyalty norms are a historically accumulated stock of W/S and provides a
long-term institutional experience by which citizens evaluate their own institutions. Hence,

44

historical loyalty norms provide a long-term collective memory of the institution by which
citizens evaluate leadership legitimacy. Unlike short-term loyalty norms, historical loyalty
norms are helpful in distinguishing different levels of legitimacy between partisans and nonpartisans because citizens’ memory derived from historical loyalty norms can tell how much
political elites have been accountable and how much political institutions have been inclusive
or exclusive. Current and prospective levels of perceived legitimacy can be better reflected
through a historical measure of the loyalty norms, rather than the short-term loyalty norms.10
In small historical loyalty norms, partisans are likely to have experienced more frequent
leadership turnover and thus have more chances of being either winners or losers. In large
historical loyalty norms, however, less frequent leadership turnover has taken place, which
in turn allows partisans to be either repeated winners or repeated losers.
Winners in large historical loyalty norms are expected to have the highest levels of political trust among partisans and non-partisans. It would result from high party loyalty and
high affinity with the incumbents because winners’ repeated winning status enabled by large
historical loyalty norms (i.e., small size of historical W/S) is likely to provide the most loyal
coalition to the leader and guarantees a future stream of private goods. In large historical
loyalty norms, repeated winning and losing can be expected unless losers have new issues and
alternatives to make the leader defect to repeated losers or to allow a competent challenger
to take advantage of the new situations.
As with repeated electoral victories in small stock of historical winning coalition (i.e., in
large loyalty norms to winners), repeated defeats make losers lose affinity with the failing
challenger. The oversizing winning coalition that opens a window of opportunity for repeated
losers to become new winners could have been unnecessary to the leader and thus has not
been expected in the small stock of historical winning coalition. The potential rebellion
by repeated losers and the challenger could have forced the leader to provide some welfare
provisions to the selectorate outside of her coalition.11 It could have occurred more often
10 More weights should be given to prospective levels of legitimacy in the leader’s
11 The disenfranchised may be more likely to immigrate, rebel, or even support

45

stability.
civil war

than not when both the winning coalition and the selectorate are small since the small size
of the selectorate guarantees more private goods for new winners.
From repeated electoral defeats in small historical winning coalition, losers are left with
either “exit” or “voice” option (Hirschman, 1970). With the limited resources, co-optation of
loser by oversizing the winning coalition is not costless for leaders (Gandhi and Przeworski,
2007). The “voice” as political dissents and protests by repeated losers, however, can enforce
leaders to provide rewards to the selectors outside of her coalition. Especially, oversizing
winning coalition is likely to happen and even necessary in the process of democratization
with increasing the size of the selectorate, which still continues the context of small stock of
historical winning coalition.
The leader in this context has also incentives to bind the non-winning groups (losers and
non-partisans) to the political system with moderate levels of political trust. This is because
the incumbents’ survival by maneuvering the small stock of historical winning coalition has
been supported by making repeated losers and non-partisans pay taxes, rather than protests
or anti-government demonstrations, for the leader’s resources distributed to her winning
coalition and to appropriate public good spending. This was evident in South Korea during
the democratic transitional period: President Kim Young Sam allocated disproportionately
more pork-barrel benefits to the western provinces, Cholla province (Honam), as well as to
his own turf, Kyongsang province (Youngnam), to assuage the complaints from the repeated
losers in the presidential elections as well as to guarantee his “graceful retirement” (Horiuchi
and Lee, 2008; see also Kwon, 2005).
In the case of large stock of historical winning coalition, the experience of rotation of
winner-loser is expected to increase partisans’ efficacy and the system’s representation.
Policy implementation is more party-oriented, with competition of many political parties.
Specifically, public goods for partisans and non-partisans are available in case of a large
historical stock of the winning coalition and the selectorate. Losers among partisans are
(Bueno de Mesquita et al., 2003).

46

likely to have experienced being winners despite the fact that the good memories of electoral
victory have faded out due to a frequent leadership turnover. If insignificant benefits from
public goods can be compensated for by their experience of political representation, which
is often the case in a large stock of historical winning coalition,12 both partisans of winners
and losers are expected to have no effects of the historical stock of winning coalition and
selectorate on political legitimacy.
New winners in a small historical loyalty norms (i.e., a large stock of historical W/S) lack
high affinity with the new leaders because of their experience of old leaders’ manipulation of
the winning coalition and defection to selectors outside of the winning coalition. Moreover,
oversizing winning coalition has occurred at least once when historical stock of W/S is close
to one13 , which could have disabled old winners from taking advantage of more private
goods. Through partisans’ experience of the leaders’ defection for elites’ own sake and of
losing (the chances of)14 significant private goods, partisans would not have full confidence
in the changed system, as well as having new leaders who would exclude partisans in any
coalition in the future. To partisans, an increased stock of historical W/S provides more
information about how the political elites’ strategies could decrease the partisans’ interests.
Regarding the welfare provision side, more stock of historical W/S means less private goods
and more general but insignificant public goods. Therefore, partisans are hypothesized to
have decreased levels of political trust as the accumulated stock of W/S is increasing, with
the proviso that winners would have higher confidence than losers.
Compared to partisans’ decreasing legitimacy in the small historical loyalty norms, nonpartisans’ legitimacy level is expected to decrease and to be prompted to politicize when
the size of historical winning coalition increases. In a small historical winning coalition,
12 This

is because frequent rotation with a limited time of winning experience still provide
external efficacy to partisans.
13 If the increased historical W/S results from decreasing size of the selectorate, voters in
high historical W/S are expected to have low trust levels of political institutions. This is
because voters have experienced to become the disenfranchised in their institutional (i.e.,
constitutional) history.
14 These chances are for losers.
47

non-partisans currently would have no chances to get the benefits given to winners such as
being represented and having private goods. Moreover, non-partisans’ who lack affinity with
any opposition parties and their leaders would put themselves in a psychologically marginal
status, thereby leaving themselves particularly cynical toward institutions over time (Abramson, 1983). Therefore, the leader would not pay much attention to them when securing her
winning coalition. Through the long history of exclusion and alienation, increased stock of
historical winning coalition instead strengthens a decreasing level of non-partisans’ political
trust (see also Keele, 2005).
Therefore, the hypotheses below will be tested;
• H1a (Winner discontent hypothesis): Electoral winners have decreasing levels of political trust as short-term loyalty norms decrease (i.e., as a short-term W/S increases).
• H1b (Losers/non-partisan consent hypothesis): Electoral loser/non-partisans have increasing levels of political trust as short-term W/S increases.
• H2a (Partisan Hypothesis): Electoral winners/losers have moderate levels of political
trust as historical loyalty norms decrease (i.e., as a historical W/S increases).
• H2b (Non-partisan Hypothesis): Electoral non-partisans have decreasing levels of political trust as historical loyalty norms decrease.
Confidence in institutions is used for key variables in this study. The effects of partisan status
allow us to examine how differently partisans interpret their losing/winning status in different
levels and types of W/S. As the political consent theory would implies (Anderson et al.,
2005), a small short-term winning coalition indicates less inclusive political representation
for losers/non-partisans through more private-good policy implementation to winners, while
a large short-term winning coalition indicates more inclusive representation to losers/nonpartisans through less private-good provision to winners (H1a and H1b; see Figure 3.1).
[Figure 3.1 & Figure 3.2 ]
48

Unlike the theory of political consent, historical stock of winning coalitions and historical perception of the loyalty norms allow citizens to rely on their memories of institutional
changes and thus makes the non-partisans’ system legitimacy distinct. By considering historical memories of the loyalty norms and the interaction with partisan status, I suppose
that large historical winning coalition will aggravate non-partisans’ alienation and exclusion
from the political system, thereby leading to a significant legitimacy drop for non-partisans.
With this long-term perception of a historical winning coalition, partisans and non-partisans
diverge in their legitimization of political institutions (H2a and H2b; see Figure 3.2).

3.5

Data and Measures

Combining country-level contexts and individual-level behavior in the empirical analysis is
the key in the research questions of this chapter. Citizens’ attitudes and behavior data
should be sufficient across a significant number of countries. I use three survey data sets
from the World Values Survey (hereafter, WVS) conducted in 1999-2004 (the 4th wave), the
Afrobarometer (AFB) in 2002-2004 (the 2nd round), and the East Asian Barometer (EAB)
in 2001-2003 (the 1st round). Countries used in the empirical analysis are shown in Table
B.1 in Appendix B.

3.5.1

Dependent variable

I examine attitudes toward the political system by focusing on confidence or trust in public
institutions (i.e., political trust). To measure how much confidence the respondents have in
public institutions, the survey questions directly ask their trust levels accorded to various
institutions. As each data set has political trust indicators on different institutions, I could
come up with six major institutions such as parliament, president, government, the police,
the armed forces/military, and civil services.15 After combining the three data sets, I recoded
15 In

case of the WVS, parliament, civil service, the armed forces, and the police compose the four basic institutions as government question includes many missing observations.
49

political trust variable into the interval scale ranging from 0 to 1.0.
Summary statistics show that there are significant variations of the citizens’ attitudes
toward political institutions across countries. Specifically, while most countries’ levels of
political trust range between 0.45 and 0.57 with about 0.12 Inter-quartile Range and 0.5 of
mean, there are 8 country-year samples that have less than 0.4, and 3 countries more than
0.7, respectively. The lower mean countries includes Mexico, Peru, Argentina, the Czech
Republic, Macedonia, Greece, Lithuania, and Nigeria in 2003; the higher mean countries are
Mali, Senegal, Tanzania in 2001, and Bangladesh.
Regarding more rigorous findings of systematic difference of the citizens across countries,
the variance of the political trust variable can be decomposed by ANOVA model into interand intra-country variations. As many comparative political behavioralists contend, both
levels of analysis should be appropriately executed with multi-level analysis for rigorous
statistical findings about citizens’ attitudes toward government institutions (Raudenbush
and Bryk, 2002; Steenbergen and Jones, 2002; Anderson and Singer, 2008). The ANOVA
model results are shown in Table 3.1.
[Table 3.1]
From the one-way random-effects ANOVA, the incorporation of country-level analysis
is legitimized through the measure of intra-class correlation (ICC). Raudenbush and Bryk
(2002) recommend the ICC for a preliminary step in a multilevel analysis. As ICC can
be calculated by the ratio of the country-level variance to the total variance of political
trust (i.e., the total variance between and within countries), the ICC in the analysis will be
0.0103/(0.0103+0.0432) = 0.193. So the between-country variability takes about 19.3% of
the variance in political trust. The ANOVA model, therefore, indicates that individual-level
However, government was added to the basic four institutions in the EAB while in the AFB,
president question replaces the civil service question that does not exist in this wave of the
survey. (see Appendix A for question wording)

50

variables cover more variance (about 81%) of the dependent variable while the country-level
variables also account for a significant portion of the variance.

3.5.2

Main independent variables

At the individual level, partisan status, i.e., winner-loser in election and non-partisan independents, is considered key variable explaining the variation of political trust. It is measured
by using survey questions on the respondents’ vote intention, reported votes cast, and party
affiliation (see Appendix A for coding of each data set). These three components of partisan
status from each data set could lead to an inconsistent measure, especially for the measure
of reported votes cast, because after election voters would support opposition parties or be
independent while they voted for a incumbent party. However, this potential inconsistency
will be minimal considering that changed partisan status will not be substantial until the
next coming election. Moreover, through the use of the loyalty norm as another key variables of interaction terms in this study, the W/S in either short-term or in historical context
provides more expanded time frame than the partisan status.
One key variable of interest in the country-level variables is short-term W/S in each
country. As Bueno de Mesquita et al. (2003) admit that their selectorate theory “remains
a primitive theory in need of enrichment with more institutional details and improved measurement,” the measurement of the loyalty norm (W/S) in this chapter
Basically, I followed Bueno de Mesquita et al.’s measurement of selectorate and winning
coalition size by using the POLITY data (Jaggers and Gurr, 1995) and Arthur Banks’ data
set. The winning coalition (W) and selectorate (S) are appropriate proxies for institutions
for policy implementation and representation; W is measured by two Polity indicators on
executive recruitment (openness and competitiveness), and two other indicators are for competitiveness of participation from the Polity and regime type from Arthur Banks’ data. But
I revised W by adding 0.5 into the dichotomous components of the measure of W. For example, I coded “factional” and “transitional” as 0.5, rather that 0, in the PARCOMP (i.e.,

51

competitiveness of participation), which is one of four components of W. By having detailed
measure of W, I refined the regime characteristic more clearly. For the short-term measure
of W, I use the recent 5-year mean of W for each country (see Appendix A for coding).
There is considerable variation of the new measure of short-term W across the countries
sampled in this chapter.16 The average short-term W is .83. And the highest level of 1 in the
short-term W was found in all the advanced industrial countries and the Czech Republic,
Lithuania, Slovenia and Taiwan, while the lower levels less than two standard deviations
from the mean (about <.47) include Ghana and Nigeria in 2001.
Another primary variable of interest is historically stocked measure of W, S and W/S,
which I called historical W, S and W/S. Historical measures of W/S have been divided into
two variables; one is for accumulated measure of 100 years up to a year before the surveys
year; the other for accumulated measure of 100 years up to the year that five year the shortterm W/S was calculated. I assume that recent history matters more with higher weights
as suggested by Gerring et al. (2005). The latter historical measure should be used in order
to avoid overlapping period of time in the final models. Descriptive statistics of dependent
variable and main independent variables are listed and summarized in Table 3.2 below
[Table 3.2]

3.5.3

Covariates

I also sought to control for associated variables with political trust found in the previous
analyses. At the individual-level, I controlled for the level of political interest, perceived
measure of personal economy, and standard demographic variables (age, gender, education).
Social capital variables such as social trust and membership in voluntary associations are
16 Regarding

short-term S, there are only four countries that has less than 1; they are
Nigeria in 2000, Pakistan in 2001, Algeria in 2002. In the case of Algeria in 2002, there is
no Arthur Banks’ data for the regime type between 2000 and 2002. So I had to code all
the components of W and S up to 1999 for Algeria. This is not a problem in the use of the
survey data (the WVS) because Algerian presidential election occurred in 1999 and therefore
I could still use the individual data for the partisan status.
52

intentionally deleted from the models. Both variables seems to be correlated with systemic
variables, not with individual-level variables (Pharr and Putnam, 2000; Chapter 5).
At the country-level, both representation and implementations sides of political institutions were controlled for. These include level of economic development, level of democracy,
variation of electoral system (Effective Number of Parliamentary Party and Gallagher’s
Disproportionality Index between seats and votes), and government structure (Federalism/Unitarism and Parliament/Presidentialism/Semi-presidentialism). Coding procedures
for all variables are listed in Appendix A.

3.6

Method

The multilevel analysis entails one nesting higher level of variables (i.e., country-level) and
the other nesting lower level variables (i.e., individual-level). This analysis is based upon
widely-accepted concerns that statistical techniques for nested data structures need to take
into account the strong likelihood that observations within groups or nesting units are more
alike than observations from different groups or nesting units. Even using fixed effect models,
the inference will be limited to the samples by ignoring random variability involved with
group-level context (Luke, 2005). If these are theoretical reasons for using multilevel analysis,
there are some statistical problems arising if one ignores the multilevel nature of the data
such as underestimation of standard errors, clustering, and non-constant variance (Snijders
and Bosker, 1999).
The first step for the multilevel models is to specify the individual-level model. For
example, I model confidence in political institutions at the respondent i in the country j as
follows:

P olT rustij = β0j + β1j (Loser)ij + β2j (N onpar)ij + β3j (Educ)ij + β4j (Age) + β5j (F emale)ij
+β6j (Interest)ij + β7j (Economy)ij + rij
(3.1)

53

The models the political trust levels of partisans and non-partisans explained by country
mean of β0j and individual variations from the country mean specified with the individual
level’s independent variables and the individual error term rij .
Then, the second step entails modelling the country mean as a function of the countrylevel characteristics that are specified in the Eq. (3.2).

β0j = γ00 + γ01 (histW/S)j + γ02 (shortW/S)j + γ03 (EN P P )j
+γ04 (Disproport)j + γ05 (U nitary)j + γ06 (P arliament)j
+γ07 (Democracy)j + γ08 (GDP/cap)j + γ09 (AdvDem)j

(3.2)

+u0j ,
Taking advantage of the multilevel models, I included several cross-level interactive terms
in the combined models as shown in the Model 3, 4, 7, 8, 11 and 12 (see Table 3.3, 3.4,
and 3.5). The cross-level interactive effects can provide more information regarding how
partisans and non-partisans view their respective political institutions in the institutional
context of the short-term loyalty norms levels as well as in the historical loyalty norms.
My expectations are as follows: the positive effect of a large short-term winning coalition
on political trust will be significant for election losers since losers can share benefits of
private goods with winners in the short run; the negative effect of a large historical winning
coalition on citizens’ attitudes toward their own political institutions will be significant for
non-partisans since non-partisans’ alienation can be worsened in the long run.
In the framework of multilevel modeling, these political support hypotheses can be tested
by modeling the effect of individual’s partisan status in country j that has both a current
level of W/S and a institutional history of W/S. More specifically, I can simultaneously
estimate Eq.(3.3) and Eq.(3.4), which has the estimated political trust gap between losers
and winners (β1j ) and between non-partisans and winners (β2j ) in country j:

β1j = γ10 + γ11 (histW/S)j + γ12 (shortW/S)j + u1j

54

(3.3)

β1j = γ10 + γ11 (histW/S)j + γ12 (shortW/S)j + γ13 (EN P P )j + γ14 (Disproport) + u1j
(3.4)
This equation is combined into the Eq. (3.1) to test the hypothesis that the effect of both
kinds of the loyalty norms (W/S) is strengthened for electoral losers. Hence, the combined
model tests the gap between how winners and losers evaluate their country’s political institutions changes with changes in the short-term loyalty norms.
Similarly, I also test how the effect of long-term historical winning coalition is dependent
upon (non-)partisan status by including another set of cross-level interactive terms in the
combined model that estimate cross-country variations in the effect of (non-) partisan status
on political trust as a function of historical winning coalition levels. Essentially, the combined
model also tests the difference between how winners and non-partisans evaluate their political
system will increase with changes in the long-term institutional history of winning coalition.

β2j = γ20 + γ21 (histW/S)j + γ22 (shortW/S)j + u2j,

(3.5)

β2j = γ20 + γ21 (histW/S)j + γ22 (shortW/S)j + γ23 (EN P P )j + γ24 (Disproport) + u2j,
(3.6)
It is assumed that there is a constant effect across countries for all of the other individuallevel variables. Equations below are combined into the individual-level equation, Eq. (3.1),
to be the full model.

β3j = γ30 ,

β4j = γ40 ,

β5j = γ50 ,

β6j = γ60 ,

β7j = γ70 .

(3.7)

The models are estimated with centering variables in both levels. Except for the dummy
variables of winners, losers, and non-partisans, all the other individual-level variables are
55

centered at their country mean (Enders and Tofighi, 2007). This is because centering at
group-mean allows the estimates to be free from the effect of country-level differences in the
mean values. Every country-level variable except the variables of interest, i.e., short-term
W/S and historical W/S, is centered at its grand mean so that the intercepts in the models
can be the overall conditional mean for the sample countries. Two loyalty norms (W/S)
variables are coded as they are for the easy interpretation of interaction terms.

3.7

Estimation Results

The results of the multilevel analysis appear in Table 3.3, 3.4, and 3.5. Four models from
each set of sampled countries estimate the effects of the independent variables on the dependent variable, i.e., citizens’ level of political trust. In Model 3 and Model 4, I include
the interaction terms between two types of W/S and Election Loser and Non-partisan and
measure the effect of W/S on the citizens’ level of political trust for the average electoral
winners in the sample.
The cross-level interaction terms derived from Eq. (3.3) through Eq. (3.6) is to test
the two hypotheses that the election losers’ level of political support is increased in the
higher short-term W/S while the non-partisans’ level is decreased by higher historical W/S.
Moreover, two electoral system variables (ENPP and Disproportionality) appear to be significant with expected signs in the interaction with losers, while only non-partisans’ trust
level appears to increase with more disproportional electoral system.17
17 Election losers and non-partisans are expected to be more sensitive to the levels of dispar-

ity between vote and seat than winners. However, the direction of sign could be ambivalent:
for a negative sign, we suppose that more disparity worsens losers’ dissent against political
institutions (although all models show insignificant positive signs). As the disproportionality, unlike relatively obvious number of parties (ENPP), can be better perceived by losers
who are more sensitive to election outcomes, disproportionality is expected to worsen losers’
dissent. For a positive sign of non-partisans as it is the case in the models, more disparity
ameliorates non-partisans dissent. Given that disproportionality is more noticeable in a more
accountable electoral system such as SMD plurality, non-partisans’ dissent could be alleviated by clarity of responsibility derived in part from the disparity. Besides, non-partisans’
alleviated dissent results from their status itself. Less representative electoral system would

56

[Table 3.3]
I also test the same hypotheses by including an individual-level variable of Perceived personal economy. The variable appears to be significant, meaning more economically satisfied
respondents tend to trust more in their political institutions. Although it reveals all the same
statistically significant signs except for Female, the number of country is decreased down to
41 from 59 due to the missing observations. While Democracy level and Real GDP/cap is
not statistically significant in all the models, the direction of signs are unexpected negative
signs. With more limited country sample, Real GDP/capita appears statistically significant
and negatively associated with political trust in Model 9 in Table 3.5.18
[Table 3.4]
I excluded all the advanced industrial countries to make sure the previous findings are
confirmed although I estimate the models with the dummy variables of region. In Model 912, the results confirm the robustness of the findings in the previous models with the sample
that does not include the advanced industrial democracies. Regarding the regional dummy
variables, African & Middle East variable is used as a reference, so it is deleted in the models
in Table 3.5. Latin America countries and post-communist countries in the sample appear
to be lower trust levels than Africa and Middle East countries.
[Table 3.5]
A robust check was done by controlling for two influential countries by using the Cook’s
distance measure. As a rule of thumb, a cut-off value given for the Cook’s Distance is 4/n
justify why non-partisans remain independent. Thus, the signs of non-partisan’s interactive
term with disproportionality is expected to be positive.
18 As real GDP per capita is based on the purchasing power parity conversion of the GDPs,
“the real per capita GDP of low-income countries relative to that of high-income countries
is greater than is indicated by comparisons based on exchange rate conversions of GDPs
to a common currency”, which composes the nominal GDPs (Kravis, Heston and Summers,
1978). This explains in part why high political trust levels in the low-income countries are
pronounced.
57

in which n is the number of groups in the level-2 factor under evaluation, i.e., 59 countries
(Van der Meer, Te Grotenhuis and Pelzer, 2010). Therefore, the cut-off value is 0.068, which
makes Nigeria in 2003, Mali and the U.S. be influential cases across robust-check analyses.
After controlling for these countries, the results still hold, as shown in Model C.1 and Model
C.2. (see also the Cook’s distance in Appendix C).19
Regarding the individual-level control variables, perceived level of political trust increases
with age, political interest, and personal economy situation. As people become older, they
tend to conform authority and prefer the status quo, thereby having more trust in political
system. People’s interest level in politics and perceived level of personal economic well-being
are also positively associated with support level. However, trust declines with education.
Losers and non-partisans show significantly lower trust levels than winners. This confirms
the arguments of political consent literature. Besides, the cross-level interaction models of
historical W/S for non-partisans (Model 3, 4, 7, 8, 11, and 12) support the hypothesis
of strengthening negative effect on the perceived level of political trust. Another crosslevel interaction of short-term W/S for election losers appears to be statistically significant
across the models with strengthening positive effect on the levels of trust. On average, nonpartisans in countries with lower historical W/S have higher levels of political confidence
in their political institutions, while election losers with higher short-term W/S have higher
levels of political trust. And, the short-term W/S levels for electoral winners do not even
appear to be statistically significant across all models, although the positive signs appear as
expected.
Both figures illustrate the results. Short-term W/S does not even appear to be statistically significant in how electoral winners perceive political institutions’ legitimacy. However,
electoral losers are more sensitive to levels of short-term W/S, i.e., more increasing legitimacy occurs to electoral losers (see Fig. 3.3). Historical W/S levels have a statistically
significant effect on how electoral winners perceive their own country’s political institutions.
19 As

shown in Appendix C.3, Model 3 does not have influential values with the 0.098
(=4/41) of cut-off value.
58

While electoral losers do not have a significant difference from winners’ sensitivity to levels
of historical W/S, non-partisans are more sensitive to the levels than partisans, i.e., more
decreasing legitimacy occurs to non-partisans (see Fig. 3.4).
[Figure 3.3 & Figure 3.4]
Model 8 predicts that electoral losers in the highest short-term W/S country in the
sample, i.e., all the established democracies and 5 new democracies (1.00)20 , will be about
8.1% higher than electoral losers in the lowest short-term W/S country in the sample, Ghana
(0.45).21 Regarding the conditional effects of historical W/S, Model 3 predicts that nonpartisans’ trust level in the lowest historical W/S country in the sample, i.e., Namibia (0.346),
will be about 14.3% higher than the level of trust for non-partisans in the highest historical
W/S country in the sample, i.e., the U.S. and Lesotho (1.00).22
Therefore, historical W/S, on average, has significant effects of decreasing confidence for
non-partisans in the basic political institutions while, on average, increased short-term W/S
increases election losers’ perceived levels of political trust. Specifically, from the models in
the Table 3.4, I expect the negative effect of historical W/S to be greater for those who do
not have party affiliation, and the positive effect of short-term W/S will be more pronounced
among electoral losers. Under the different time frames of W/S context, citizens would have
different evaluations of their political system across partisan or non-partisan status.
20 5

new democracies include Algeria, the Czech Republic, Lithuania, Slovenia, and Taiwan. The mean difference of political trust between the established democracies and 5 new
democracies is 0.08 (8%) by which the established democracies have more.
21 The calculation is derived from the following formula; for losers’ trust gap, ((.687+(.054*1)+(.202*1*1))-(.687+(-.054*.45)+(.202*.45*1))) = .081
22 The predicted values of the differences are calculated as follows: For the gap between
non-partisans, ((.687+(-.110*.346)+(-.109*.346*1))-(.687+(-.110*1)+(-.109*1*1))) = .143

59

3.8

Conclusion

The results reported above suggest following important inferences. First, the loyalty norms,
implying relative size of winning coalitions out of the selectorate, matter in two different ways:
short-term preferences and long-term experiences. With both the limited and extended time
frame in the citizens’ psychological schema, contextual variables of short-term loyalty norms
and historical loyalty norms matter for how citizens perceive their political institutions in
relation with political elites’ interests in those contexts. In particular, election losers evaluate
institutions more positively in countries with a higher short-term winning coalition and low
loyalty norms to a leader. This is so because strong loyalty norms by election winners
can turn into a ‘shared and low loyalty norm’ by both winners and losers in election; and
alienated citizens such as non-partisans view the system more negatively in countries with
a large historical winning coalition. In the shared loyalty norms by partisans (winners and
losers only), non-partisans are excluded and alienated from political system throughout their
experience under multiple institutional changes.
Second, the models confirm that voters do not uniformally perceive the political system.
Electoral losers are significantly more confident than other voters in political institutions if
the size of winning coalition recently increases (that is, small but shared short-term loyalty
norms): As their expectations for winning in future elections increases due to the increased
size of the winning coalition, losers can expect that they will get a significant addition of their
future welfare provision. Unlike the short-term loyalty norms, non-partisans are prompted
to politicize and to have significantly decreasing confidence in the historical loyalty norms
where non-partisans can get access to their memory of being alienated in that institutional
history.
Thus, two different concepts derived from the same institutions of loyalty norms produce
the heterogeneous responses from citizens, contingent upon their partisan status. Short-term
loyalty norms are most sensitive among electoral losers while historical loyalty norms allow
citizens to perceive how political elites maximize their benefits from political system, and
60

constrain non-partisans in terms of political alienation.
The results support the hypotheses that large short-term loyalty norms increase support for the political institutions among electoral losers (H1b for electoral losers only), and
large historical loyalty norms undermine support for the system among partisans (H1b),
especially more among non-partisans (H2b). Although winners show negative sign in the increased short-term loyalty norms, it is not statistically significant (i.e., disconfirming H1a).
These findings are also robust. Sampling the new democracies and controlling for influential
countries confirm the results.

61

Tables & Figures for Chapter 3
Table 3.1: ANOVA of Variance in Political Trust Across Levels (Model 1 - Model 4)
Fixed Effect
Average Country Mean (β00 )
Random Effect
Country Effect (τ00 )
Individual Effect (σ 2 )
Log Likelihood

Coef. Est.
0.506**
Variance Component
0.0103**
0.0432**
9,446

62

Std. Err.
0.013

N

59
63,423

Table 3.2: Descriptive Statistics of Dependent Variable and Main Independent Variables
(Model 5 - Model 8)
Political Trust
Short-term W
Short-term S
Short-term W/S
Historical W
Historical S
Historical W/S

Obs. Mean
46330 0.508
41
0.774
41
0.946
41
0.818
41
21.598
41
29.743
41
0.726

63

Std.Dev.
0.245
0.184
0.172
0.121
11.363
12.715
0.183

Min. Max.
0.000 1.0
0.156 1.0
0.200 1.0
0.450 1.0
3.675 46.49
4.900 46.49
0.346 1.0

Table 3.3: Multilevel Linear Model of Partisanship and W/S: Dependent Variable: Political
Trust
Model 1(A) Model 1(B)

Model 2

Model 3

Model 4

Individual-level Variable
Election Loser

-0.064***
(0.002)

Non-partisan

-0.088
(0.002)

Education

-0.009***
(0.000)

Age

0.001***

-0.064*** -0.064*** -0.264*** -0.211***
(0.002)

(0.002)

(0.057)

(0.057)

-0.088*** -0.088***

-0.083*

-0.054

(0.042)

(0.044)

(0.002)

(0.002)

-0.009*** -0.009*** -0.008*** -0.008***
(0.000)

(0.000)

(0.000)

(0.000)

0.001*** 0.001*** 0.001*** 0.001***

(0.000)

Interest in Politics

(0.000)

(0.000)

(0.000)

0.001

0.001

0.001

0.001

0.001

(0.002)

Female

(0.000)
(0.002)

(0.002)

(0.002)

(0.002)

0.019***
(0.001)

0.019*** 0.019*** 0.018*** 0.018***
(0.001)

(0.001)

(0.001)

(0.001)

-0.181*

-0.186*

-0.166*

-0.134†

-0.116

(0.081)

(0.080)

(0.070)

(0.075)

(0.075)

-0.003

-0.040

-0.137

-0.119

(0.137)

(0.120)

(0.125)

(0.125)

0.004

0.004

0.022†

0.022*

0.008

(0.012)

(0.012)

(0.011)

(0.011)

(0.012)

0.003

0.003

0.005*

0.005*

0.003

(0.002)

(0.002)

(0.002)

(0.002)

(0.002)

Country-level Variable
W

0.032
(0.153)

S

0.043
(0.141)

Historical W/S
Short-term W/S
ENPP
Disproportionality

Continued. . .

64

Table 3.3 (cont’d)

Model 1(A) Model 1(B)
Democracy level

Model 2

Model 3

Model 4

-0.011

-0.003

-0.002

-0.002

(0.010)

(0.009)

(0.007)

(0.007)

(0.007)

0.000

0.000

0.000

0.000

0.000

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

0.029

0.034

0.032

(0.057)

(0.056)

(0.056)

-0.002

0.011

0.013

(0.054)

Real GDP/cap

-0.014

(0.053)

(0.053)

Africa & Middle East
Asia
Latin America

-0.162** -0.156** -0.157**
(0.056)

(0.054)

-0.099*

-0.092†

-0.093†

(0.050)

(0.049)

(0.049)

-0.032

-0.062

(0.052)

(0.050)

-0.065†

-0.086*

(0.039)

(0.038)

0.254***

0.218**

(0.074)

(0.070)

0.053

0.038

(0.054)

Post-Communist

(0.054)

(0.053)

Cross-level Interaction
Loser*Historical W/S
Non-partisan*Historical W/S
Loser*Short-term W/S
Non-partisan*Short-term W/S
Loser*ENPP

0.026**
(0.008)

Non-partisan*ENPP

0.015*
(0.006)

Loser*Disproportionality

0.003†
(0.002)

Non-partisan*Disproportionality

0.003*
Continued. . .

65

Table 3.3 (cont’d)

Model 1(A) Model 1(B)

Model 2

Model 3

Model 4
(0.001)

Constant

0.617***

0.691*** 0.726*** 0.776*** 0.747***

(0.138)

(0.127)

(0.118)

(0.120)

(0.120)

0.010

0.010

0.006

0.007

0.007

Election Loser

n/a

n/a

n/a

.004

.004

Non-partisan

n/a

n/a

n/a

0.002

0.002

0.041

0.041

0.041

0.040

0.040

AIC

-22,196

-22,200

-22,200

-23,200

-23,200

BIC

-22,051

-22,100

-22,000

-23,000

-22,900

Log-Likelihood

11,114

11,114

11,117

11,637

11,624

N. of Respondent

63,423

63,423

63,423

63,423

63,423

59

59

59

59

59

Variance of Random Effect
Country-level

Individual-level

N. of Country

†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

66

Table 3.4: Multilevel Linear Model of Partisanship and W/S: Dependent Variable: Political
Trust (Perceived Personal Economy included)
Model 5

Model 6

Model 7 Model 8(A) Model 8(B)

Individual-level Variable
Election Loser

-0.075*** -0.074*** -0.246**
(0.002)

Non-partisan

(0.079)

(0.087)

(0.077)

-0.086*** -0.086***

-0.059

0.065

-0.030

(0.058)

(0.073)

(0.056)

-0.012*** -0.012*** -0.011***

-0.011***

-0.011***

(0.000)

(0.000)

(0.000)

0.001*** 0.001*** 0.001***

0.001***

0.001***

(0.000)
Age

-0.212**

(0.002)

(0.003)
Education

0.082

(0.003)
(0.000)

(0.000)

Interest in Politics

(0.000)

(0.000)

(0.000)

-0.003

-0.003

-0.004†

-0.004†

-0.004†

(0.002)

Female

(0.000)
(0.002)

(0.002)

(0.002)

(0.002)

0.020*** 0.020*** 0.019***

0.019***

0.019***

(0.001)

(0.001)

(0.001)

0.015*** 0.015*** 0.015***

0.015***

0.015***

(0.001)

(0.001)

(0.001)
Perceived Personal Economy

(0.001)

(0.001)
(0.001)

(0.001)

Country-level Variable
W

0.046
(0.166)

S

0.163
(0.149)

Historical W/S

-0.127

-0.102

-0.110

(0.089)

(0.096)

(0.094)

(0.096)

0.077

-0.008

-0.074

-0.054

(0.165)
ENPP

-0.174†

(0.098)
Short-term W/S

-0.198*

(0.147)

(0.155)

(0.154)

-0.000

0.022

0.027

0.011

0.008

(0.016)

(0.017)

(0.016)

(0.017)

(0.018)

Continued. . .

67

Table 3.4 (cont’d)

Model 5

0.007**

0.004

0.004

(0.003)

(0.003)

(0.003)

(0.003)

-0.009

-0.004

-0.003

-0.009

-0.003

(0.009)

(0.009)

(0.010)

(0.009)

-0.000

-0.000

-0.000

-0.000

-0.000

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

0.010

0.031

0.011

0.019

(0.106)

(0.102)

(0.099)

(0.101)

-0.020

0.001

-0.019

0.001

(0.095)

(0.091)

(0.089)

(0.090)

-0.172†

-0.165†

-0.186*

-0.174*

(0.091)

(0.087)

(0.085)

(0.086)

-0.127

-0.104

-0.140

-0.116

(0.110)

Real GDP/cap

0.006*

(0.010)

Democracy level

0.004
(0.003)

Disproportionality

Model 6

(0.106)

(0.104)

(0.105)

Africa & Middle East
Asia
Latin America
Post-Communist

Model 7 Model 8(A) Model 8(B)

Cross-level Interaction Term
Loser*W

0.200*
(0.089)

Non-partisan*W

0.031
(0.065)

Loser*S

-0.295**
(0.101)

Non-partisan*S

-0.094
(0.078)

Loser*Historical W/S

-0.023

(0.063)

(0.064)

-0.091†

-0.117*

-0.109*

(0.050)
Loser*Short-term W/S

-0.041

(0.067)
Non-partisan*Historical W/S

-0.055

(0.049)

(0.047)

0.221*

0.202*
Continued. . .

68

Table 3.4 (cont’d)

Model 5

Model 6

Model 7 Model 8(A) Model 8(B)
(0.094)
0.048

0.033

(0.070)

Non-partisan*Short-term W/S

(0.089)
(0.065)

Loser*ENPP

0.030**
(0.011)

(0.002)

0.004*

0.004*

(0.001)
0.623*** 0.713*** 0.713***

0.002

(0.002)

Constant

(0.008)

0.002

Non-partisan*Disproportionality

0.019*

(0.008)
Loser*Disproportionality

(0.011)

0.022*

Non-partisan*ENPP

0.027*

(0.001)

0.463**

0.687***

(0.156)

(0.155)

(0.159)

(0.160)

(0.157)

0.011

0.008

0.010

0.010

0.010

Election Loser

n/a

n/a

0.006

0.005

0.005

Non-partisan

n/a

n/a

0.003

0.002

0.002

0.044

0.044

0.043

0.043

0.043

AIC

-12,900

-12,900

-13,700

-13,607

-13,600

BIC

-12,700

-12,700

-13,400

-13,292

-13,300

6,454

6,454

6,855

6,839

6,842

46,330

46,330

46,330

46,330

46,330

41

41

41

41

41

Variance of Random Effect
Country-level

Individual-level

Log-Likelihood
N. of Respondent
N. of Country

†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

69

Table 3.5: Multilevel Linear Model of Partisanship and W/S: Dependent Variable: Political
Trust (Democratizing countries only)
Model 9 Model 10 Model 11 Model 12(A) Model 12(B)

Individual-level Variable
Election Loser

-0.077*** -0.077*** -0.273**
(0.003)

Non-partisan

(0.099)

(0.093)

(0.094)

-0.085*** -0.085***

-0.103

0.048

-0.075

(0.070)

(0.074)

(0.064)

-0.013*** -0.013*** -0.012***

-0.012***

-0.012***

(0.001)

(0.001)

(0.001)

0.000*** 0.000*** 0.000***

0.000***

0.000***

(0.001)
Age

-0.229*

(0.003)

(0.003)
Education

0.091

(0.003)
(0.001)

(0.000)

Interest in Politics

(0.000)

(0.000)

(0.000)

-0.005*

-0.005* -0.006**

-0.006**

-0.006**

(0.002)

Female

(0.000)
(0.002)

(0.002)

(0.002)

(0.002)

0.020*** 0.020*** 0.019***

0.019***

0.019***

(0.001)

(0.001)

(0.001)

0.014*** 0.014*** 0.015***

0.015***

0.015***

(0.001)

(0.001)

(0.001)
Perceived Personal Economy

(0.001)

(0.001)
(0.001)

(0.001)

Country-level Variable
W

0.045
(0.169)

S

0.148
(0.150)

Historical W/S

-0.168†

-0.141

-0.149

(0.087)

(0.101)

(0.096)

(0.099)

0.165

0.062

-0.065

-0.045

(0.158)
ENPP

-0.211*

(0.095)
Short-term W/S

-0.261**

(0.143)

(0.163)

(0.159)

0.011

0.029†

0.036*

0.013

0.011

(0.016)

(0.017)

(0.016)

(0.017)

(0.018)
Continued. . .

70

Table 3.5 (cont’d)

Model 9 Model 10 Model 11 Model 12(A) Model 12(B)
Disproportionality

0.008**

0.004

0.004

(0.003)

(0.003)

(0.003)

(0.003)

-0.006

-0.002

-0.000

-0.005

-0.000

(0.010)

(0.008)

(0.008)

(0.009)

(0.008)

-0.000*

-0.000

-0.000

-0.000

-0.000

(0.000)

(0.000)

(0.000)

(0.000)

(0.000)

-0.030

-0.015

-0.014

-0.003

(0.047)

Real GDP/cap

0.007*

(0.003)
Democracy level

0.004

(0.046)

(0.046)

(0.046)

-0.165** -0.171**

-0.171***

-0.168**

Asia
Latin America

(0.058)

(0.054)

(0.055)

-0.138*

-0.129*

-0.142**

-0.129*

(0.057)

Post-Communist

(0.055)
(0.055)

(0.055)

(0.055)

Cross-level Interaction Term
Loser*W

0.212*
(0.106)

Non-partisan*W

0.075
(0.074)

Loser*S

-0.306**
0.108

Non-partisan*S

-0.118
(0.081)

Loser*Historical W/S

-0.021

(0.069)

(0.071)

-0.080

-0.105*

-0.098*

(0.053)
Loser*Short-term W/S

-0.044

(0.075)
Non-partisan*Historical W/S

-0.066

(0.051)

(0.048)

0.257*
(0.113)

(0.105)

0.100

Non-partisan*Short-term W/S

0.229*
0.086
Continued. . .

71

Table 3.5 (cont’d)

Model 9 Model 10 Model 11 Model 12(A) Model 12(B)
(0.081)
Loser*ENPP

(0.073)
0.033**
(0.011)

(0.002)

0.004**

0.004**

(0.002)
0.534*** 0.664*** 0.727***

0.003

(0.002)

Constant

(0.008)

0.003

Non-partisan*Disproportionality

0.021**

(0.008)
Loser*Disproportionality

(0.011)

0.022**

Non-partisan*ENPP

0.031**

(0.002)

0.515***

0.687***

(0.150)

(0.136)

(0.151)

(0.136)

(0.149)

0.010

0.007

0.010

0.010

0.010

Election Loser

n/a

n/a

0.006

0.005

0.005

Non-partisan

n/a

n/a

0.003

0.002

0.006

0.045

0.045

0.045

0.044

0.045

AIC

-10,200

-10,200

-10,900

-10,905

-10,900

BIC

-10,100

-10,100

-10,700

-10,602

-10,600

5,131

5,131

5,502

5,487

5,490

41,350

41,350

41,350

41,350

41,350

36

36

36

36

36

Variance of Random Effect
Country

Individual

Log-Likelihood
N. of Respondent
Number of Country

†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

72

0.6
0.4
0.2

Winner
Loser
Non-partisan

0.0

Political Trust

0.8

1.0

Figure 3.1: Hypothesized Citizen’s Trust (Short-term W/S)

0.0

0.2

0.4

0.6

Short-term W/S

73

0.8

1.0

1.0

Figure 3.2: Hypothesized Citizen’s Trust (Historical W/S)

0.6
0.4
0.2
0.0

Political Trust

0.8

Winner
Loser
Non-partisan

0.0

0.2

0.4

0.6

Historical W/S

74

0.8

1.0

0.6
0.4
0.2

Winner
Loser
Non-partisan

0.0

Political Trust

0.8

1.0

Figure 3.3: (a) Predicted Probability of Political Trust by the Loyalty Norm and Partisanship
(Short-term W/S)

0.5

0.6

0.7
Short-term W/S

75

0.8

0.9

1.0

1.0

Figure 3.4: (b) Predicted Probability of Political Trust by the Loyalty Norm and Partisanship
(Historical W/S)

0.6
0.4
0.2
0.0

Political Trust

0.8

Winner
Loser
Non-partisan

0.3

0.4

0.5

0.6

0.7

Historical W/S

76

0.8

0.9

1.0

Chapter 4
SOURCES OF PRESIDENT’S SURVIVAL IN SOUTH KOREA

4.1

Introduction

From Chapter 3, political trust is conceived as citizens’ attitude toward an aggregated political institution, that is, a whole political system. By looking at the cause of political trust
in the light of citizens’ partisanship and a winning coalition and selectorate institution (i.e.,
the loyalty norms), I argue that the leader’s institutional change through either expansion or
contraction of the size of winning coalition derives citizens to perceive their political system
generally (cf. Bueno de Mesquita and Smith, 2010). With respect to both the short-term and
the historically accumulated loyalty norms, citizens’ attitude toward political institutions is
not independent of a current and previous leader’s institutional change.
From the findings in Chapter 3, just as the institutional change made by a leader and
party in power can affect citizens’ attitude toward political institutions, I suppose that microlevel variables such as partisan status derived from electoral politics should be directly linked
to the perception of leader survival and stability, especially in a critical time period like an
impeachment process. Thus, in order to understand leader’s survival in micro perspective,
it should be a next logical step to examine leader’s survival in terms of the citizens’ loyalty
to the leader and her party. Basically, this step seems to require us to examine the citizens’
attitude toward leader’s survival in the consideration of citizens’ different partisan status
and their loyalty to the leader. If all of this is possible, we can test and verify more direct
causal relationship between leader survival and the interaction of partisan status and the
loyalty norms at the individual level.
Moreover, with the benefit of using individual level data, we may control for citizens’
different attitudes toward specific political institutions so that the effect of partisan status
and the loyalty norms on leader survival can be better examined. For example, the level of
77

citizens’ trust in president is affected by citizens’ attitude toward an incumbent president
while the approval rating of an incumbent prime minister is more associated with citizens’
trust in parliament than with trust in other institutions. Therefore, in explaining leader’s
survival at an individual level data, I need to include disaggregated political institutions so
that I can sort out some confounding effects in resulting from close association of leader and
a specific institution.
In this context, the South Koreans’ experience of an impeachment process seems to provide an unique opportunity to examine the role of citizens’ perception of a leader and the
party in power. Specifically, the Constitutional Court’s reinstatement of the impeached
president by the National Assembly was believed to be substantially influenced by the 17th
General Election’s outcome, which means that voter’s attitude directed to the impeached
president eventually had a say in the impeachment process. Presidency in South Korea
is characterized as a significantly strong presidential system, similar to the French semipresidential system.1 Although the South Korean Constitution allows constitutional mandate for both the presidency and the legislature, the legislature (i.e., the National Assembly)
tends to be subordinate to the president in the policy-making process (Tavits, 2007; Mo,
2001). From this, relatively greater clarity of responsibility is one of main characteristics in
the strong presidency of South Korea. A presidency with clarity of responsibility requires
more substantial support than less noticeable policy-makers from both political elites and
the mass public.
Under the strong and responsible presidency in South Korea, citizens’ support for the
leader varies depending on their partisan status and their levels of loyalty norms to their
parties. In democratizing countries like South Korea wherein the party system is not stable
but volatile in terms of the duration of party existence, the support for an incumbent party
1 Unlike

definitional semi-presidential system, however, Korean president can avoid collective responsibility of the prime minister and the cabinet by only allowing individual PM
responsibility when the legislature (i.e., the National Assembly) requires accountability of
the executive branch.

78

can be well translated into the loyalty norms to the party’s leader: The president in South
Korea thus can not be dissociated with a figure as the leader of the party in power. Moreover,
although the party system is very much personalized, on average, partisans tend to be more
confident toward their political systems than non-partisans (e.g.,Anderson and Paskeviciute,
2009).
In this chapter, I use a case study of impeachment to evaluate and illustrate both the
strength of the presidency and the roles of key partisan groups in presidency support for
this institution. The contrast between partisans and non-partisans is a focus in a sense
that the loyalty of the selectorate has a conditional influence on the mass perception of the
impeachment process and its legitimacy. Especially, the difference in the perception between
winners and non-partisans implies the importance of a stable party system and its function
during the course of consolidating democracy.
This case study focuses on events surrounding the impeachment of President Roh Moohyun in 2004. In brief, the sequence of events is following; One of the presidential candidate
of Roh Moo-hyun, who was self-educated human rights lawyer, was elected in November,
2002. Relatively weak legitimacy earned by the victory with razor-thin margin provided not
only a great challenge to the “honeymoon period” of his governance, but also a motivation
for him to found his own party, the Uri Party. With President Roh’s defection of his own
then-incumbent Millennium Democratic Party, two major opposition parties could secure
the majority in the National Assembly. As result, both parties could accuse President Roh
of his pre-electoral intervention with the impeachment motion, which had been aggravated
by his continually inappropriate remarks and his aids’ subsequent corruption scandals.
Yet, President Roh survived the legislative attempt of impeachment after obtaining legitimacy from the decision made by the Constitutional Court on May, 2004. Before the
Court’s decision was finalized, the newly founded Uri Party had won majority in the National Assembly after the 17th general election in April, 2004. The election seemed to be a

79

referendum-like election against the impeachment attempt.2
This chapter proceeds as follows. First, leader legitimacy in South Korea will be examined
from the perspective of mass politics. Then, I will discuss how party politics in South
Korea after the 2002 presidential election has changed the existing region-based party system
wherein the party loyalty has been easily translated into the loyalty to the party leader who
has the corresponding regional power. Also, the generational gap in the party loyalty will be
discussed. Thirdly, the abnormal case of the President Roh’s period and his also abnormal
experience of the impeachment process provides a distinct opportunity to test the effect of
partisan status and the loyalty norm to the president on the mass perception of the leader’s
survival. Lastly, the conclusion discusses how unstable political parties’ competition for the
strong and accountable president across the regions of South Korea and the generations has
isolated non-partisans from the political arena.

4.2

Elite Politics and Regionalism in South Korea

Authoritarian rule, with the exception of a more democratic parliamentary system briefly
before the General Park’s military coup, showed a stark contrast with democratization era
after 1988. Heo and Stockton (2005) show the major differences as follows. First, the ruling
party’s dominance in the legislature has ended. Second, the proportional representation
system has expanded in the electoral system. Third, with the PR system, the number of
effective parties has also increased. And lastly, the parties’ life, on average, is shorter in the
democratization era (p.680).
A democratic transition process in Korea triggered by massive pro-democracy demonstra2 In

2003, however, President Roh suggested holding an national referendum asking a vote
of confidence, while the Korea Constitution only limits the use of referendum to “important
policies relating to diplomacy, national defense, unification, and other matters relating to
the national destiny” (Article 72) and to part of procedure for amending the constitution
(Article 130). Later, the Court dismissed the case “with warning that efforts such as Roh’s to
bend the constitution in a plebiscitarian direction could become a cover for authoritarianism”
(Hahm and Kim, 2005; p. 35).

80

tion, especially in June, 1987, resulted in Chun’s concession, named the June 29th Declaration by Roh Tae Woo who was handpicked as would-be successor to then President Chun.
Korea’s democratic transition culminated in a new constitutional reform heralding the Sixth
Republic (Park, 2002). The ruling party of the DLP (Democratic Liberal Party) founded by
the ‘3-party merger’ in 1990 showed a clear example of how critical “leadership compatibility”
is in determining party behavior (Kim, 1997).3 While the merging parties allied in accordance with their conservative ideology, two major figures had the same Youngnam regional
base.
Another example of an elite pact occurred between the two Kims. The 1997 elite pact
initiated by Kim Dae Jung induced coalitions between progressives and hard-core conservatives, which in turn seemingly resulted in a more narrowed ideological gap between them.
Kim Dae Jung (DJ) and Kim Jong Phil (JP)’s coalition (a.k.a. DJP coalition) derived
from an agreement on the constitutional reform, mainly including the shift from presidentialism to a parliamentary system, was an irresistible incentive to JP, as JP’s regional base
of Choongchung consists of relatively small number and less loyal support for its regional
leader of JP.
Kim Young Sam and Kim Dae Jung had been figures representing Honam and Youngnam regions, respectively, “maintaining commanding positions within their own parties and
tended to magnify the rivalries and animosities between the regions” (Lee, 1999:p. 214 ; see
also, Horiuchi and Lee, 2008). Although both Kims could become presidents consecutively,
many in the Honam region felt that they had been discriminated against during the successive authoritarian rules by the Youngnam region. This evinces the complicated regional
and personal-based party system in Korea, which has induced constant disappearance and
reshuffling of parties between elections to maximize their electoral fortune.
Many scholars ascribe the weak Korean party system to inveterate regionalistic cliques
3 More

than a year later, main two opposition parties of the New Democratic Party, “led
by the top opposition figure Kim Dae Jung, and the small Democratic Party, led by Lee Ki
Taek.” (New York Times, Sept. 11th, 1991)

81

and a personalized party system (Lee, 1998; Kang, 2005). Regionalism has been an inveterate cause of the failure of party politics in new democracies. Hahm (2008) points out that
political elites’ pack-making in 1991 “led to a major ideological reconfiguration in Korean
politics - from authoritarianism versus democrats to conservatives versus progressives...also
strengthened another major cleavage in Korean politics - regionalism.” (p. 135) With these
peculiarities, the system of shortened party longevity compared to other similarly democratizing countries seems to be clear in Korea. A stable party system has not been settled upon
as personalized parties have frequently emerged, merged, been renamed, or disappeared. After competitive legislative elections restarted in 1985, no opposition party ran in consecutive
elections except for a minor party, the United Liberal Democrats, which ran in both the 1996
and 2000 elections(Heo and Stockton, 2005).4 Korea’s average party longevity, for example,
is 4.9 years, which is much lower than Taiwan’s party system (32.2 years) that has taken
similar path of democratization and socio-economic development (Lee, 2009).
Lee (2009) traces the origin of personalism in the Korean party system back to the
authoritarian era. With shortened life span of parties through party mergers and political
calculations, the early democratizing period witnessed that regional and personal-based party
system has strengthened party loyalty that is well converted to the voter’s loyalty to the party
leaders. However, this results in a large share of non-partisans in their party affiliation,
although there seems to vary across regions (see Table 4.1), and partisan loyalty can be
easily converted to the loyalty to the party leaders, rather than loyalty to party ideology and
party organization.
[Table 4.1]
4 However,

there were two military backed parties of the Democratic Republican Party
(1963–78) and the Democratic Justice Party (1981–91) that ran in consecutive elections
during the authoritarian and transitional periods.

82

4.3

The Leader-Party Nexus and Partisan Support

Since democratization, the long-standing fault-line of authoritarian (military) and democratic (civilian) has converted into a newer fault-line, i.e., the progressive-conservative ideology in the 17th general election (Kim, Choi and Cho, 2008; see Kim, 1998). Although
ideologically-oriented parties did not develop at the beginning of the democratic transition
in 1987, voters’ diverse demands beyond the developmental state argument and across generations have been substantial. Specifically, the ideological gap was dramatically drawn along
generational differences in the presidential election in December, 2002. With an existing
dimension of a regional cleavage, the new dimension of a generational cleavage between the
old and the young made partisan politics in Korea look more complicated (see Fig.1 below).
This gap, for example, was delineated with the strong contrast of both major parties’ candidates, Roh Moo-hyun from the Millennium Democratic Party (MDP, hereafter) and Lee
Hoi-chang from the Grand National Party (GNP, hereafter).
Owing to the younger generations’ and the progressives’ support, President Roh achieved
victory in the 16th presidential election in 2002. Moreover, in the preliminary election he had
relied on populist strategies and had appealed to his career as a human rights lawyer to be a
big surprise to the major groups of the party as well as to the public. Many Koreans started
to believe that Roh’s electoral victory marked the ’end of three Kims’ era (Im, 2004). Yet,
the razor-thin victory in the presidential election did not give President Roh full leverage
and legitimacy for his progressive policies and innovative pledges.
President Roh had neither leadership skills nor the power for party discipline within the
MDP. President Roh was the first president who defected, rather than renaming, his own
incumbent party to found a new political party, Open Uri Party (Uri Party, hereafter). His
defection was considered by his supporters as an essential move because his campaign slogan
had been heavily based on departure from existing corrupt political powers such as the MDP
and the GNP, the major opposition party.
The Uri Party was a symbol for the liberals and the progressives. So it had a reputation
83

representing the young reforming group of the 386 generation, and Roh was believed to lead
this whole group and its penultimate organization of the Uri party. Therefore, Roh tried to
make the party more salient among the public even before the 17th general election campaign
was officially started.
So how would the attempt at impeachment likely be viewed by different groups of voters?
Supporting the Uri Party could reflect how loyal the voters were to the ideas of liberal reform
and to President Roh as well. This was especially the case for the election winners in the
previous presidential election who felt close to the newly founded Uri Party. They are the
group who still supported the Uri Party, confirming the leader-party nexus. They were likely
to strongly oppose the impeachment attempt motioned by the National Assembly, thereby
voting for the Uri Party in the general election in April, 2004. The majority of election losers
from the GNP may have been frustrated with the temporary “impeachment coalition” with
the MDP, which could have instigated more support for both the leader and the Uri Party
(see Tab. 4.2 & Fig. 4.1.’s lower-right quadrant).
In contrast, among those who supported President Roh in the presidential election, the
founding of the Uri Party could be considered to be a defection to the then-incumbent
party of the MDP. These frustrated winners could be expected to support the impeachment
motion.5 They might feel closer to the previous incumbent MDP that had strong regional
support from the Honam region, rather than to the more progressive-reform-oriented Uri
Party. These winners’ view on the impeachment attempt could be expected to be more
supportive than loyal winners to the Uri Party. To loyal losers to the GNP, the response to
the impeachment could be expected to be more supportive than any other partisan status.
They could believe that the “impeachment coalition” of the major opposition GNP and the
previous ruling MDP was a necessary step for them to regain power. (see Tab. 4.2 & Fig.
4.1’s lower-left quadrant).
[Table 4.2]
5 In

Table 4.1, more than 10% of winners can be regarded as “dissent winners”.
84

Non-partisans are more likely than partisans to see a liberal and progressive young party
such as the Uri Party be a good alternative to the old style politics. As a number of scholars
assume that the nexus of partisanship and system support exists, non-partisan status tend
to bear dissent of the existing party system (Holmberg, 2003; Torcal, Gunther and Montero,
2002; Dalton, 1999; Miller and Shanks, 1996; Budge, Crewe and Farlie, 1976). Specifically,
for those who are not identified with any existing parties, more diverse parties are preferable
within a wider ideological spectrum. In this vein, non-partisans who especially experienced
the conservatives’ ‘3-Party Merger’ and the strategic ‘DJP coalition’ could be expected to
welcome the progressive and liberal newcomer of the Uri Party to the conservatively-skewed
party system (see Tab. 4.2 & Fig. 4.1.’s upper-right quadrant).
Finally, non-partisans are expected to support the impeachment motioned by mainly the
two opposition parties (the GNP and the MDP) that these non-partisan may not have felt
close to.6 While they could be expected to have as little affiliation with the Uri Party as
with any existing parties, they are hesitant to support the impeachment motion against the
legitimately elected president. Non-partisans’ sympathy for President Roh is expected to be
greater than partisans who lack loyalty to the Uri Party. This kind of sympathy may not fit
with current and direct public provisions of the ruling party’s public services. Instead, the
non-partisans reject any existing parties including the Uri Party. Their preferences are more
oriented to the leadership itself, instead of party and ideology (see Tab. 4.2 & Fig. 4.1’s
upper-left quadrant).7
[Figure 4.1]
6 Lee

(1999) pays attention to party non-identifiers’ apathy and their threats to fledgling
democracy in Korea.
7 This category of non-partisans would have contrasting preferences toward the leader and
the party. This is especially because of the ambivalence of non-partisans’ support for the
progressive ruling party and their low confidence in President Roh. Thus, non-partisans’
conflictual attitudes toward the leader and the party is expected to have a positive impact
on the perceived support for the passed impeachment motion.

85

4.4

Data and Measures

I use data from a Korea Social Science Data Center Survey, conducted in April, 2004, about
a month after the impeachment motion was accepted by the National Assembly and right
after the 17th general election. The data, therefore, reflects well how the voters think of
the legitimacy of the impeachment and whether they could make their voice heard to this
unprecedented process.

4.4.1

Dependent Variable

There are two questions regarding the impeachment process: Q.33 (whether impeachment is
agreeable or not) and Q.34 (whether the respondents participated in the anti-impeachment
vigils or not). The four category question of Q.33 is chosen for the dependent variable, and
its distribution can be found in Table 4.1.
I collapsed the “Supportive” and “Strongly Supportive” categories. This is because it is
accepted to collapse categories containing small numbers (about 5% of the samples in this
data) so that collapsing categories improves the asymptotic approximations. It also improve
the fit of the model, especially when the data apparently deviate from the parallel regression
assumption (Murad et al., 2003; p. 155).
Besides, a cultural response effect may be involved here. As Javeline (1999) proves, more
acquiescence (i.e., agreement) bias in a polite culture as in Kazakhstan results in the propensity of individuals to agree with an assertion regardless of its content. Like in Kazakhstan,
Korean respondents in the survey may have been hesitant to reveal their strong opinion on
the impeachment. Since the whole society is aware of the mood of anti-impeachment rallies
through the media, the difficulty in responding lies not so much in respondents’ strong opposition to the impeachment as in their moderate support for the impeachment. For those
who support the impeachment, therefore, acquiescence bias can be found, yet at the same
time the strongly supportive respondents may not be different from moderately supportive

86

respondents. Some significant number of the moderately supportive respondents tend to be
more acquiescent in the balanced question design by assigning themselves to the “supportive”
category.

4.4.2

Main Independent Variables

Support for the impeachment reflects both partisans’ and non-partisans’ perception against
the leader and the party. Based on their partisan status, their expectation of the incumbent
party and leader will vary. This is especially the case for democratizing countries where
election winners who are the beneficiaries from the leader’s private goods. Instead, those of
the selectorate out of winning coalition, i.e., election losers and non-partisans, are expected
to have fewer benefits directly from the leader and the incumbent party, except that they
still enjoy the public goods.
Winning coalition is not equal to the total number of people supporting the leader. It
is the number of supporters who “are essential to maintaining the leader in power and who
receive policy concessions and/or personal benefits in return for their support” (Bueno de
Mesquita et al., 2010). The minimal necessary number of future winners are pursued by
the leader and the incumbent party as well. Therefore, in democratizing countries, a newly
founded party may try hard to get as many supporters as possible from the normal voters.
In the Korea case, two characteristics could be attractive to voters, especially to nonpartisans, during the course of the development of the Uri Party: First, as the catch-all party
has been prevalent in the advanced industrialized countries, which is well explained by the
median voter theorem (Black, 1948; Downs, 1957), the newly elected young president may
as well expand a skewed ideological spectrum of the conservatives to include the more liberal
and progressive ends of it. In a country of the NICs (the Newly Industrial Countries) since
late 1970s, way before the democratization process started in mid-1980s, diverse needs should
be met and various voices are expected to be heard in the new millennium. Therefore, the
newly elected party, the Uri Party, was thought of as a good experiment for the expansion

87

of the ideological spectrum.
Second, President Roh was viewed as a symbolic figure to socio-economic “outsiders”
as well as to election losers. The former could become new winners in their representing
process of democracy while successfully voting out a serious conservative contender of the
region-based GNP. However, this kind of new representation was possible not mainly by the
existing other region-based MDP, but primarily by the self-educated human rights lawyer.
Especially, for non-partisans and the younger generation, the new young symbolic figure
must have been a good inducer to the ballot box in the 2002 Presidential Election.
Therefore, I include two variables, loyalty to the Uri Party and partisan status, as key
independent variables in explaining the variation in the dependent variable, the support for
the impeachment. The loyalty to the Uri Party can be measured by two types of question.
One is a direct measure of the loyalty, asking the feeling thermometer about the Uri Party
in a scale of 0-10 range. Another assesses comparative loyalty by comparing the two major
opposition parties, the MDP and the GNP. I operationalize the concept by subtracting the
sum of the loyalty of two opposition parties from the direct measure of the Uri Party loyalty.
Then, the scale of the range is from -10 to 10, implying that zero is the ideal position of
no loyalty to any of the three parties. Moreover, I expect that the interactive relationship
between loyalty and partisan status will be an important mechanism to explain impeachment
support.
The histogram of the distribution can be shown in Figure 4.2 - 4.7 below. Election winners
in both the measures of loyalty appear to be left-skewed, while election losers appear to be
right-skewed. Non-partisans consistently centers on the middle in the distribution of both
the measures.
[Figure 4.2 - Figure 4.7]

88

4.4.3

Covariates

Political trust in specific institutions can be found in Q.39 (1) through Q39. (4). After
priming the impeachment issue by preceding Q.33 (whether impeachment is okay or not)
and Q.34 (whether the respondents participated in the anti-impeachment vigils or not), the
generalized institutional trust can be perceived as specific institutional support during the
course of the impeachment process. Then, how diffuse support can be measured in this
data is the question. Therefore, I include each institution’s trust level of specific support as
control variables in the models.
Another important control variable set should be respondents’ region. For the expediency
of the analysis, there exist five regions: Capital Seoul and its neighboring Kyunggi region,
Choongchung region, Honam region, Kangwon region, and Youngnam region. By making
the Seoul and Kyunggi region as a reference region, I specify four other regions in the models.
Last but not least, the usual demographic variables of Age, Gender, and Education should
be included. For example, the Age variable can be differently specified in the models. According to the Age-Period-Cohort model (Mason, Winsborough and Poole, 1973), Cohort
(year of birth) is identified by subtracting Age (year since birth) from Period (current year).
Thus, I should exclude the Age variable when I include the Cohort dummy variable to avoid
the identification problem caused by the mentioned additive and linear relations among the
three variables (Winship and Harding, 2008). The democratic transition in 1987 was an
important moment especially to the younger generation that is susceptible to the externally
given critical juncture during the process of political socialization. This political crystallization time period affected by critical events is expected to heavily impact the individual’s
voting patterns (Hyman, 1959; Ventura, 2001). Thus, I coded four dummy variables according to the critical events including the 1950 Korean War cohort, the 1987 democratic
transition cohort, the 1997 financial crisis cohort, and the 2002 Presidential election cohort.
I calculated each cohort with different methods as all four contexts have different implications to the respective cohorts. First, the respondents who reached age 20 in 1987 were

89

assigned to the 1987 democracy cohort as explained by many political socialization literature. Second, the respondents who reached age of 46 through age of 65 are considered
most heavily inflicted by the 1997 financial crisis that plummeted stock prices. According
to the life cycle hypothesis (Ferson and Harvey, 1991; Modigliani and Brumberg, 1954), this
age group had used their substantial amount of saving to invest in share of common stock.
However, I coded the 1997 financial crisis cohort up to age of 60 so that the 1950 War cohort
can be covered from age of 68, i.e., age of 14 in 1950. This is because the Korean War
continued until July, 1953 and is regarded as one of the most catastrophic events.8 Lastly,
the 2002 Presidential Election cohort is coded from age of 18 to age of 36 so as to cover the
Age variable exhaustively.9 Each cohort is well distributed along the Age variable, and a
nonparametric scatter plot appears below (see Figure 4.8). A cross-tabulation on support
for impeachment by partisanship is summarized in Table 4.3.
[Table 4.3]

4.5

Method

Ordered logit or probit models can be used for the ordinal dependent variable (Long, 1997).
As the support for impeachment attempt of the dependent variable has four categories,
coded as 1 = Strongly Opposed, 2 = Opposed, 3 = Agreed, and 4 = Strongly Agreed, either
estimation method can be employed. However, with parallel regression assumption being
frequently violated in the real world, I need to test the assumption by using either a Loglikelihood Ratio test or a Wald test (i.e., Brant test)(Long and Freese, 2006). Both tests
confirm that the parallel lines assumption (or proportional odds assumption) is violated at
the 0.001 level. Specifically, Brant test shows that the violations are for the variables of
Youngnam region and Loyalty (also Comparative Loyalty).
8 More discussion can be found below in the conclusion section.
9 When I recoded this relatively extended age cohort (19 years)

by splitting in half, the
results hold the same in both the direction and the statistical significance.

90

To relax the parallel regression assumption in the ordinal dependent variable case, generalized ordered logit models can be an alternative to the constrained ordered logit/probit models. Two generalized ordered logit models exit. First, Fu’s(1998) program, gologit, was designed to estimate the model, although it only allowed the least constrained model with relaxing the parallel lines assumption for all the independent variables. Second, gologit2 developed by Richard Williams (2006) can estimate Partial Proportional Odds Models (PPOM),
where the parallel regression assumption is only violated by one or a few of the variables
(see also Peterson and Harrell, 1990). 10
Therefore, we can say that the parallel lines model estimated by an ordered logit model
is a special case of the generalized ordered logit model. As the parallel lines model has the
same β coefficients for all values of j, the model can be written as

P (Yi > j) = g(Xβ) =

exp(αj + Xi β)
, j = 1, 2, ..., M − 1
1 + [exp(αj + Xi β)]

(4.1)

Unlike the parallel lines model, the partial proportional odds model has the same β
coefficients in some variables for all values of j, while others can differ. If the model can be
written as
P (Yi > j) = g(Xβ) =

exp(αj + X1i β1 + X2i β2 + X3i β3j )
, j = 1, 2, ..., M − 1 (4.2)
1 + [exp(αj + X1i β1 + X2i β2 + X3i β3j )]

, the β1 and β2, for example, are the same for all values of j but the β3 are different from
the two.
As mentioned earlier, the variables of Youngnam region and Loyalty/Comparative Loyalty appear to be free to differ from the other variables in the results of the Brant test (see
Appendix). This legitimizes the use of the PPOM.
10 Another

possibility for estimation is using a multinomial logit model. By allowing the
variable of interest to be in nominal categories, the violation of parallel line assumption will
not be a concern any more. Also, the advantages of this estimation include that each category
can be estimated according to their respective effects from a base category. The results of the
multinomial logit model appear in Appendix G, and the base category is “strongly opposed"
to impeachment. The estimation results is not different from the PPOM model.

91

4.6

Estimation Results

According to the survey after the election, partisanship and the perceived levels of impeachment appear to be strongly correlated, meaning that winners in the presidential election
in 2002 tend to show strong objection to the impeachment attempt while losers and nonpartisans show some support for it (see Table 4.1). However, the simple cross-tabulation
does not give us a full picture of the relationship between partisan status and support for
impeachment.
Before specifying the models according to the variables mentioned above, a nonparametric scatter plot produced by STATA’s regression discontinuity package of rd provides
a legitimate reason for including the cohort dummy variables, replacing the Age variable.
The jump between the 2002 election cohort and the 1987 transition cohort signifies the
lower support of the impeachment by the transition cohort. Although the older 2002 cohort
appears to support the impeachment, the younger 1987 transition cohort bears relatively
strong opposition to the impeachment. The latter’s political crystallization of the transition
period seems to give legitimacy to the democratically elected president and thus opposes the
old-style politics.
[Figure 4.8]
In Table 4.4, Election Loser and Non-partisan variables appear to be statistically significant with positive sign as expected. As the PPOM model produces two result panes
for the binary dependent variable11 , the variables’ trend across the category can be traced.
For example, compared to the reference category of Seoul and Kyunggi region, the Honam
region variable shows strong opposition to the impeachment with the odds of (relatively)
11 The

first result pane is Strongly Opposed as 0 and Opposed and collapsed Supportive
as 1, and the second one is Strongly Opposed and Opposed as 0 and collapsed Supportive
as 1.

92

supporting impeachment being decreased by a factor of 0.45412 , holding all other variables
constant. While Youngnam region tends to be more supportive with positive but marginally
significant support for the impeachment, it turns to be negative in the second result pane.
This means that when considering Comparative Loyalty that allows the party loyalty to the
existing opposition parties, Youngnam region respondents shows moderate opposition to the
impeachment with statistical significance at the .05 level (see Table 4.5). This reflects public
moods of anti-impeachment rallies, widely spread across the country.
[Table 4.4]
Regarding the interaction terms between partisan status and two kinds of the loyalty to
the Uri Party, Model 3 and Model 6 consistently show positive and significant coefficients
across both result panes in Table 4.4 and Table 4.5. The positive signs across the models
imply the non-partisans’ consent to the ideological orientation of the Uri Party while having
increased support for the impeachment by the factor of 1.23 or 23%, holding all variables
constant (see Table 4.2’s upper-right quadrant).13 At the same time, relative opposition to
the impeachment by non-partisans, compared to the “loyal losers” to the MDP and the GNP
and “dissent winners”, seems to be derived from the non-partisans’ preference for President
Roh (see Table 4.2’s upper-right and upper-left quadrants). This, for the non-partisans,
indicates the conflictual attitude towards the leader and the party.
Both partisans of winners and losers in the 2002 Presidential election appear to strongly
oppose the impeachment. These are “defecting losers” and “loyal winners” to the Uri Party.
These partisans seem to perceive a highly correlated relationship between the leader and the
party. After the momentous 2002 Presidential Election, election winners as well as election
12 As

the moderately Opposed is also coded 1, the supporting impeachment can be said
as relatively surprising impeachment. Besides, this factor decrease is equivalent to the decreasing odds of (relative) supporting impeachment by 54.6%, holding all other variables
constant. And, the factor of .454 is calculated by the exponent of the coefficient of -0.790
according to the logit regression estimation.
13 Non-partisan’s ideological distribution also support this argument.

93

losers were frustrated by the corruption scandals of the president’s relatives and aides and
also by an unlawful election intervention made by President Roh himself.14 These winners
tend to see the nexus of the leader and the Uri Party, that is, president’s close relationship
with the Uri Party, and their loyalty norm was settled down even during the time of the
controversial pre-election intervention before the 17th general election.
Regarding election losers, they tend to show more opposition to the impeachment if they
have more sympathy for President Roh. This sympathy seems to result from election losers’
frustration of the super majority of two opposition parties’ wielding power and from their
perception of a spillover from the sympathy of Roh into preference for the Uri Party, that is
an example the leader-party contrast.
[Table 4.5]
Following Figure 4.9 and Figure 4.10 illustrates clearly more extreme change of partisan’s
levels of impeachment support than non-partisan’s level along the loyalty to the Uri Party.
The gap between the winner’s and loser’s support for the impeachment increases in Fig.
4.9, while the gap between the winner’s and non-partisan’s support for the impeachment
gets wider than one between two partisans. Relatively more supportive non-partisans than
“loyal winners” and “defecting losers” has been argued above as ideological non-partisan.
In contrast, (person-based) sympathetic non-partisans show clear contrasting preferences
between the President Roh and the Uri Party, having as low confidence in the Uri Party as
do “dissent winners” and “loyal losers.”
[Figure 4.9 & Figure 4.10]
14 The

reason for the opposition parties’ impeachment mainly focused on the President’s
support for the Uri Party in the upcoming 17th general election. Under the Korean election
law, it is prohibited for public officials to openly support a particular party. Although the
GNP and the MDP’s majority in the National Assembly posed a threat to President Roh,
he denied the opposition’s requests of apologies, and it facilitated the GNP and the MDP’s
impeachment process.

94

4.7

Conclusions

South Korea’s recent experience of the impeachment provides various responses of partisans and non-partisans. With specific political contexts of regionalism and personalism,
the citizens in South Korea are highly influenced by these “background factors”(Caporaso,
2009). Honam region and Choongchung region appear contrasting supporting levels for
the impeachment against the reference category of Seoul and Kyunggi, while the specific
Partial Proportional Odds Models show changing pattern of Youngnam region in a sense
of comparative loyalty. That is, compared with Seoul and Kyunggi, Youngnam region is
more likely to moderately oppose or to support the impeachment than to strongly oppose
it. In the binary Comparative Loyalty of “Strongly Opposed/Opposed”(0) and “(Strongly)
Supportive”(1), however, the negative sign of Youngnam implies more prevalent moderate
opposition to the impeachment in this region.
Loyalty to the Uri Party and partisan status result in variations of impeachment support.
As partisans are more likely to have any preferences for the issue of impeachment, they are
labeled ranging from “loyal”, to “dissent”, and to “defecting”. Contrasting wider range of
partisans’ supporting levels of the impeachment, non-partisans’ moderate level of the impeachment support change illustrate two types of non-partisans; (person-based) sympathetic
and ideological non-partisan. The survey also implies that there are more increased nonpartisans during the process of the impeachment. Both non-partisans show more liberallyoriented ideological spectrum. This means that ideology matters when a fledgling incumbent
party try to grip non-partisans’ interests in the conservative party system.
Regarding the Cohort dummy variables, there exist some additional findings: If I added
age of 68-72 (i.e., age of 61-65 in 1997) to the 1997 Financial Crisis cohort and excluded
it (i.e., age of 14-18 in 1950) from the 1950 War cohort, the estimation results still hold in
both coding methods, except for the 1950 War cohort dummy variable. According to the
estimation results in Table 6.1. and Table 6.2, this dummy variable appears to be statistically
significant at the 0.10 and at the 0.50 level, respectively.
95

Specifically, in both measures of loyalty to the Uri Party, the 1950 War cohort appears
to be more supportive to the impeachment attempt than the 2002 election cohort. While
the 1997 crisis cohort appears consistently to be more supportive than the election cohort,
however, the shrinked War cohort does not appear to be statistically different from the
election cohort. That is, this cohort, like the loyal election cohort to the Uri Party, seem to
oppose the impeachment attempt.
I suppose this is resulted from another cultural aspect. The aged respondents seem to
bear a paternalistic and Confucian view on President whose image can be reflected as a king
in a country or a father in a family. Then, it would be difficult for them to support the
impeachment attempt. Although the election cohort and the war cohort compose the “loyal
cohorts” to the Uri Party, additional model specification including squared term of the Age
variable does not appear to be significant. Besides, omitting the Age variable and the Cohort
variable does not change the results in terms of the direction and the statistical significance.

96

Tables & Figures for Chapter 4
Table 4.1: Partisanship by Region (%)

Region
Partisanship Soodokwon

Total

237

34

2

20

222

515

(22.67%)

(1.20%)

(41.67%)

(53.75%)

(34.33%)

182

37

26

10

79

334

(24.67%)

(15.57%)

(20.83%)

(19.13%)

(22.27%)

303

79

139

18

112

651

(52.67%)

(83.23%)

(37.50%)

(27.12%)

(43.40%)

722

150

167

48

413

1,500

(100.00%)

Total

Youngnam

(41.97%)

Election Winner

Kangwon

(25.21%)

Non-partisan

Honam

(32.83%)

Election Loser

Choongchung

(100.00%)

(100.00%)

(100.00%)

Source:Korea Social Science Data Center: Interviewed in April, 2004

97

(100.00%) (100.00%)

Table 4.2: The Leader-Party Nexus/Contrast & (Non-)Partisan Support: Expected Response to Impeachment
Impeachment Support

Loyalty to the Uri Party

Change (∆)

(Low)

(High)

Moderate

Non-partisan’s Contrast

Non-partisan’s Consent

(Non-partisan Type)

(Sympathetic Non-partisan)

(Ideological Non-partisan)

Extreme

Partisan’s Dissent

Partisan’s Nexus

(Partisan Type)

(Loyal Loser/Dissent Winner)

(Defecting Loser/Loyal Winner)

98

Table 4.3: Support for Impeachment by partisanship (%)

Partisanship
Support for Impeachment

113

359

577

(37.54%)

(57.44%)

(41.07%)

155

133

194

482

(44.19%)

(31.04%)

(34.31%)

164

45

57

266

(14.95%)

(9.12%)

(18.93%)

55

10

15

80

(3.32%)

(2.40%)

(5.69%)

479

301

625

1,405

(100.00%)

Total

105

(11.48%)

Strongly Agreed

Total

(34.24%)

Agreed

Winner

(32.36%)

Opposed

Non-partisan

(21.92%)

Strongly Opposed

Loser

(100.00%)

(100.00%)

(100.00%)

Source: Korea Social Science Data Center: Interviewed in April, 2004

99

Table 4.4: Partial Proportional Odds Models of Impeachment Support
Model 1

Model 2

Model 3

-0.790***

-0.759**

-0.773**

(0.239)

(0.239)

(0.241)

0.341†

0.330†

0.334†

(0.199)

(0.199)

(0.199)

0.380

0.360

0.309

(0.261)

(0.260)

(0.263)

0.596†

0.567†

0.533

(0.340)

(0.342)

(0.343)

-0.004

-0.003

-0.008

(0.045)

(0.046)

(0.047)

0.000

-0.011

-0.016

(0.143)

(0.143)

(0.143)

-1.122***

-1.114***

-1.123***

(0.132)

(0.132)

(0.133)

0.755***

0.747***

0.752***

(0.144)

(0.144)

(0.145)

0.190

0.199

0.222

(0.142)

(0.143)

(0.145)

0.634***

0.640***

0.068

(0.184)

(0.184)

(0.431)

0.479*

0.396*

-0.776

(0.197)

(0.193)

(0.484)

-0.255***

-0.257***

-0.334***

Strongly Opposed vs.
Opposed/(Strongly) Supportive
Honam Regiona
Youngnam Regiona
Choongchung

Regiona

Kangwon Regiona
Education
Female
Age

0.033***
(0.006)

Trust in President
Trust in Nat’l Assembly
Trust in Party
Election Loser
Non-partisan
Loyalty (Uri Party)

Continued. . .

100

Table 4.4 (cont’d)

Model 1

Model 3

(0.036)

(0.036)

(0.050)

0.685***

0.723***

(0.165)

(0.166)

0.988***

0.995***

(0.235)

(0.237)

1.160*

1.295*

(0.561)

1987 Demo. Cohort (Born in

Model 2

(0.573)

1953-1968)b

1997 Fin. Crisis Cohort (Born in 1938-52)b
1950 War Cohort (Born in & before 1937)b
Loser*Loyalty

0.093
(0.072)

Non-partisan*Loyalty

0.203**
(0.077)

Constant

0.937†

1.827***

2.315***

(0.517)

(0.444)

(0.500)

-0.790***

-0.759**

-0.773**

(0.239)

(0.239)

(0.241)

-0.375

-0.385

-0.356

(0.239)

(0.238)

(0.239)

0.380

0.360

0.309

(0.261)

(0.260)

(0.263)

0.596†

0.567†

0.533

(0.340)

(0.342)

(0.343)

-0.004

-0.003

-0.008

(0.045)

(0.046)

(0.047)

0.000

-0.011

-0.016

(0.143)

(0.143)

(0.143)

Strongly Opposed/Opposed vs.
(Strongly) Supportive
Honam Regiona
Youngnam

Regiona

Choongchung Regiona
Kangwon Regiona
Education
Female

Continued. . .

101

Table 4.4 (cont’d)

Model 1

Model 3

-1.122***

-1.114***

-1.123***

(0.132)

(0.132)

(0.133)

0.755***

0.747***

0.752***

(0.144)

(0.144)

(0.145)

0.190

0.199

0.222

(0.142)

(0.143)

(0.145)

0.634***

0.640***

0.068

(0.184)

(0.184)

(0.431)

0.479*

0.396*

-0.776

(0.197)

(0.193)

(0.484)

-0.366***

-0.370***

-0.461***

(0.044)

(0.044)

(0.063)

0.685***

0.723***

(0.165)

(0.166)

0.988***

0.995***

(0.235)

(0.237)

1.160*

1.295*

(0.561)

Age

Model 2

(0.573)

0.033***
(0.006)

Trust in President
Trust in Nat’l Assembly
Trust in Party
Election Loser
Non-partisan
Loyalty (Uri Party)
1987 Demo. Cohort (Born in 1953-68)b
1997 Fin. Crisis Cohort (Born in 1938-52)b
1950 War Cohort (Born in & before 1937)b
Loser*Loyalty

0.093
(0.072)

Non-partisan*Loyalty

0.203**
(0.077)

Constant

-0.512

0.398

0.945†

(0.529)

(0.446)

(0.520)

Pseudo R-Square

0.256

0.255

0.259

Number of Cases

928

928

928

Continued. . .

102

Table 4.4 (cont’d)

Model 1
a Omitted
b Omitted

Model 2

Model 3

category is Seoul and Kyunggi Region
category is the 2002 Presidential Election Cohort (Born in & after

1969)
†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

103

Table 4.5: Partial Proportional Odds Models of Impeachment Support (Comp. Loyalty)
Model 4

Model 5

Model 6

-0.682**

-0.653**

-0.654**

(0.248)

(0.248)

(0.248)

0.371†

0.363†

0.369†

(0.212)

(0.212)

(0.213)

0.566*

0.552*

0.555*

(0.264)

(0.263)

(0.264)

0.535

0.522

0.514

(0.350)

(0.351)

(0.351)

-0.004

-0.004

-0.005

(0.047)

(0.048)

(0.048)

0.038

0.029

0.036

(0.148)

(0.148)

(0.149)

-0.990***

-0.985***

-0.981***

(0.140)

(0.139)

(0.140)

0.479**

0.474**

0.461**

(0.149)

(0.149)

(0.150)

0.158 0.164 ancial

0.175

Opposed/(Strongly) Supportive
Honam Regiona
Youngnam Regiona
Choongchung Regiona
Kangwon Regiona
Education
Female
Age

0.026***
(0.006)

Trust in President
Trust in Nat’l Assembly
Trust in Party

(0.150)

0.207

(0.196)

(0.218)

0.398†

0.326

0.067

(0.204)

(0.201)

(0.249)

-0.286***

-0.288***

-0.315***

(0.031)

Comparative Loyalty (Uri Party)

0.270

(0.195)
Non-partisan

(0.151)

0.261

Election Loser

(0.150)

(0.031)

(0.041)

Continued. . .

104

Table 4.5 (cont’d)

Model 4

0.659**

0.632*
(0.247)

1.098†

1.111†

(0.570)

1950 War Cohort (Born in & before 1937)b

(0.172)

(0.245)

1997 Fin. Crisis Cohort (Born in

0.524**

(0.171)
1938-52)b

Model 6

0.523**

1987 Demo. Cohort (Born in 1953-68)b

Model 5

(0.572)

Loser*Comp. Loyalty

-0.011
(0.063)

Non-partisan*Comp. Loyalty

0.116†
(0.063)

Constant

0.516

1.212**

1.289**

(0.498)

(0.415)

(0.424)

-0.682**

-0.653**

-0.654**

(0.248)

(0.248)

(0.248)

-0.511*

-0.517*

-0.522*

(0.257)

(0.257)

(0.259)

0.566*

0.552*

0.555*

(0.264)

(0.263)

(0.264)

0.535

0.522

0.514

(0.350)

(0.351)

(0.351)

-0.004

-0.004

-0.005

(0.047)

(0.048)

(0.048)

0.038

0.029

0.036

(0.148)

(0.148)

(0.149)

Strongly Opposed/Opposed vs.
(Strongly) Supportive
Honam regiona
Youngnam Regiona
Choongchung Regiona
Kangwon Regiona
Education
Female
Age

0.026***
Continued. . .

105

Table 4.5 (cont’d)

Model 4

Model 5

Model 6

-0.990***

-0.985***

-0.981***

(0.140)

(0.139)

(0.140)

0.479**

0.474**

0.461**

(0.149)

(0.149)

(0.150)

0.158

0.164

0.175

(0.150)

(0.150)

(0.151)

0.261

0.270

0.207

(0.195)

(0.196)

(0.218)

0.398†

0.326

0.067

(0.204)

(0.201)

(0.249)

-0.440***

-0.443***

-0.464***

(0.044)

(0.044)

(0.056)

0.523**

0.524**

(0.171)

(0.172)

0.659**

0.632*

(0.245)

(0.247)

1.098†

1.111†

(0.570)

(0.572)

(0.006)
Trust in President
Trust in Nat’l Assembly
Trust in Party
Election Loser
Non-partisan
Comparative Loyalty (Uri Party)
1987 Demo. Cohort (Born in 1953-68)b
1997 Fin. Crisis Cohort (Born in

1938-52)b

1950 War Cohort (Born in & before 1937)b
Loser*Comp. Loyalty

-0.011
(0.063)

Non-partisan*Comp. Loyalty

0.116†
(0.063)

Constant

-1.493**

-0.790†

-0.716†

(0.505)

(0.416)

(0.426)

Pseudo R-Square

0.302

0.301

0.303

Number of Cases

905

905

905

Continued. . .

106

Table 4.5 (cont’d)

Model 4
a Omitted
b Omitted

Model 5

Model 6

category is Seoul and Kyunggi Region
category is the 2002 Presidential Election Cohort (Born in & after

1969)
†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

107

1.0

Figure 4.1: Hypothetical Relationship between Loyalty and Impeachment

0.6
0.4
0.2
0.0

Supporting Impeachment

0.8

Winner
Loser
Non-partisan

0.0

0.2

0.4

0.6

Loyalty (Uri Party)

108

0.8

1.0

0

.1

.2

Density

.3

.4

.5

Figure 4.2: Distribution of the Loyalty by Election Winner (For interpretation of the references to color in this and all other figures, the reader is referred to the electronic version of
this dissertation)

0

2

4
6
Loyalty to the Uri Party

109

8

10

0

.1

Density
.2

.3

.4

Figure 4.3: Distribution of the Loyalty by Election Loser

0

2

4
6
Loyalty to the Uri Party

110

8

10

0

.1

Density
.2

.3

.4

Figure 4.4: Distribution of the Loyalty by Non-partisan

0

2

4
6
Loyalty to the Uri Party

111

8

10

0

.05

Density

.1

.15

Figure 4.5: Distribution of the Comparative Loyalty by Election Winner

-10

-5

0
5
Comparative Loyalty (the Uri Party)

112

10

0

.05

Density

.1

.15

Figure 4.6: Distribution of the Comparative Loyalty by Election Loser

-10

-5

0
Comparative Loyalty (the Uri Party)

113

5

10

0

.05

.1

Density
.15

.2

.25

Figure 4.7: Distribution of the Comparative Loyalty by Non-partisan

-5

0
5
Comparative Loyalty (the Uri Party)

114

10

0

1

2

3

Figure 4.8: Discontinuity of Impeachment Support by Cohort in 2004

2002 Elect.Coh

0 Democ.Coh.

Fin.Crisis Coh.

115

War Cohort

.2

Pr(Opposed/(Strongly) Supportive)
.4
.6
.8

1

Figure 4.9: Predicted Probability of Impeachment Support (Strongly Opposed vs. Opposed/(Strongly) Supportive): Loyalty

0

2

4
6
Loyalty (Uri Party)
Winner
Non-partisan

116

8
Loser

10

0

Pr(Opposed/(Strongly) Supportive)
.2
.4
.6
.8

1

Figure 4.10: Predicted Probability of Impeachment Support (Strongly Opposed vs. Opposed/(Strongly) Supportive): Comparative Loyalty

-10

-5

0
Comparative Loyalty (Uri Party)
Winner
Non-partisan

117

5
Loser

10

0

Pr((Strongly) Supportive)
.2
.4
.6

.8

Figure 4.11: Predicted Probability of Impeachment Support (Strongly Opposed\Opposed
vs. (Strongly) Supportive): Loyalty

0

2

4
6
Loyalty (Uri Party)
Winner
Non-partisan

118

8
Loser

10

0

.2

Pr((Strongly) Supportive)
.4
.6
.8

1

Figure 4.12: Predicted Probability of Impeachment Support (Strongly Opposed\Opposed
vs. (Strongly) Supportive): Comparative Loyalty

-10

-5

0
Comparative Loyalty (Uri Party)
Winner
Non-partisan

119

5
Loser

10

Chapter 5
CITIZEN’S MOBILIZATION AND LEADER’S SURVIVAL

5.1

Introduction

On every Monday between September 25, and December 18, 1989, German citizens in Leipzig
in East Germany continued anti-government demonstrations that had been inspired by the
Nikolai Church’s peach prayers on Mondays since 1982. During this time, the East German
Socialist Unity Party (SED)’s general secretary, Erich Honecker, resigned, and on November
9, 1989, the Berlin Wall fell. The mass demonstrators demanded political liberalization,
open borders, and German unification.
The Monday demonstrations in Leipzig triggered a huge wave of mass political movements
across East Germany to express the public’s long-standing discontent, mainly caused by the
lack of political freedom and poor standard of living. Although elections were held regularly,
no opposition party existed, and election results were often manipulated. Moreover, with
limited information access by the mass public, only some high-level authorities could sense
a precipitous drop in public support for the regime throughout 1980s. However, the ignition
of the people’s silent dissent had been searching for a right place and right time, and the
spark was made in Leipzig.
In addition to the mass protest in Leipzig in East Germany, the on-going process of the
“Jasmine Revolution” in Sidi Bouzid in Tunisia and in Cairo in Egypt in 2011 as well as the
other “Colored Revolutions” highlights the roles of mass dissent toward regimes and leaders.
The Tunisian mass protest was also ignited by a young street vendor whose self-immolation
dramatically aroused other silent dissenters in Tunisia and eventually deposed 23-year-long
dictator of Zine el-Abidine Ben Ali. The spillover effect of mass protests went over the border
to oust another long-standing dictator, Hosni Mubarak in Egypt.
Mass threats of demonstrations and protests have been good litmus indicators for lead120

ers to read how well their regimes and governments are doing for their citizens. In either
advanced democracies or hybrid regimes such as electoral democracies and competitive authoritarian countries, mass threats seems to involve a direct risk to the incumbent leaders.
In the previous chapters, I have focused on the differential attitudes of potential discontent dissenters (i.e., election losers and non-partisans) from loyal supporters (i.e., election
winners) toward political institutions (Chapter 3) and toward a leader who experienced a
most risky situation of an impeachment trial in South Korea (Chapter 4). In this chapter,
I will examine whether and how the election losers’ and nonvoters’ changes in their electoral behavior can be translated into mass behavior such as general strikes, anti-government
demonstrations, riots and even revolutions. As the two illustrative cases exemplified at the
beginning of this chapter, mass political movements can reflect citizens’ discontent toward
the existing political systems. The differential behavior among partisans and between voters
and nonvoters signifies substantially different results in mass political movements, thereby
resulting in the risk of deposition of leaders.
This chapter proceeds as follows. First, the electoral and behavioral building blocks
of selectorate theory are introduced. I derive how electoral politics characterize the effect
of a leader’s winning coalitions on mass political movements. By focusing on the changes
in electoral competition (represented as election loser’s size) and participation (represented
as nonvoters’ size), the core electoral and behavioral building blocks complement what the
structure of selectorate theory is missing in its wide application to political phenomena. I
highlight how important it is to incorporate individual-level’s determinants into selectorate
politics. Moreover, the interaction effects of electoral behavior and free press are discussed
in the selectorate politics institution, as well as how mass political movements are affected
by these interaction effects.
The second stage discusses in more detail the direct and indirect effects of electoral
politics on leader survival and the direct effect of mass threats on leader survival. More
interestingly, the increase in the size of population, the non-voting coupled with mobilization

121

in mass political movements, seems to be more likely to increase the risk of deposition.

5.2
5.2.1

Selectorate Theory of Electoral Politics Revisited
From Electoral Politics to Mass Threats to Leader Survival

In chapter 3, we saw how election losers and non-partisans view political systems through
the different institutions of loyalty norms (i.e., W/S) to leaders. While election losers are
sensitive to short-term changes of W/S, non-partisans rely on long-term changes of W/S as
their attitudinal reference point. That is, election losers trust a system that allows more
opportunities of being winners derived from the increased winning coalition within recent
years (i.e., short-term W/S). With this, they can get the benefits of private goods, although
the amount of private goods decreases with expanded winning coalition size.
A leader’s increase in winning coalition size in response to the demand for liberalization
from strong opposition parties may end up appealing to election losers through favorable policies and thus trying to include them in the winning coalition. In this case, non-partisans,
being solely excluded from the winning coalition, tend to be isolated and alienated in the
political community.1 As state resources spent by a leader to obtain support from her winning coalition are now extended to election losers, non-partisans are not beneficiaries for
increased amount of public spending. In this sense, non-partisans tend to be less alienated
and thus trust political institutions when they perceive that a leader’s winning coalition has
been rather constrained and kept small for a long time. Therefore, while having stayed alienated from the system during the course of the increasing winning coalition, non-partisans’
perception of political trust is more likely to be related to the accumulated experiences of
the loyalty norms, i.e., historical or long-term loyalty norms (W/S) than to the short-term
loyalty norms (W/S).
1

Non-partisans can also be identified as they feel apathetic to and satisfied with the system. We will discuss this aspect when the link of nonvoters and non-partisans are discussed
below.

122

In this chapter, I will bring up the causal logic of how perceived system legitimacy
such as political trust can be indicated by electoral participation, and nonvoters’ electoral
behavior stemming from the outcome of alienated non-partisans can be associated with
mass political movements and leader survival. Following Abramson and Aldrich’s (1982)
and Finkel’s (1985) notions that decline of vote turnout in the U.S. largely results from the
combined effect of the weakening partisanship and the lack of “external” political efficacy,
non-partisans’ government trust or efficacy shown in the previous chapters may well be
indicated by the changed size of nonvoters in this chapter.
Based on the aforementioned close association of non-partisans’ attitude and nonvoters’
behavior, we need to make following two assumptions: (1) if non-partisans do not vote,
non-partisans are considered as alienated voters, rather than apathetic and satisfied; (2)
among electoral behaviors derived from non-partisans’ attitudes, non-partisans’ distrust of
the system is largely translated into non-voting. These assumptions sound plausible for the
following reasons. As discussed in Chapter 3, non-partisans’ distrust of political institutions is significant only with accumulated historical winning coalition (W), which suggests
that non-partisans’ distrust can have mainly originated from perceived political alienation,
rather than apathy and satisfaction, through the experience of accumulated history of W’s
expanding.
For the second assumption, non-partisans’ distrust of the system results largely in nonvoting behavior since nonvoters have no supporting parties and no trust in political institutions. As the purpose of this concluding empirical chapter is to see whether and how
non-partisans’ attitudes toward the system affect mass movements and eventually influence
the deposition risk of the leader, the use of nonvoters’ variation in size as an indicator of
non-partisans’ attitudinal change sounds plausible. The origins of non-partisan’s attitudes
and the outcomes of nonvoters’ behavior, therefore, can be characterized as a conceptual
distinction illustrated in Table 5.1. Here I use non-partisan and nonvoters, with a caveat
that non-partisan denotes attitudinal aspects while nonvoter indicates behavioral aspects,

123

both of which imply political alienation when they are combined as shown in lower-right
quadrant of Table 5.1.
[Table 5.1]
From the Table 5.1, I distinguish four different types of voters and nonvoters whose
attitudes can be understood in terms of their party identification. Partisan voters can be
either winners or losers in elections through which party identification will determine which
side voters belong to, and thus they can be labelled as ‘allegiant voters’ to their affiliated
parties. Pragmatic voters are those who may have specific policy goals when they vote.
While they do not have any party identification, strongly policy-oriented voting behavior is
expected. Inefficacious nonvoters are those whose behavior is not committed to voting while
their attitude is affiliated with a certain party.2 From dissatisfaction with the incumbent
performance of their traditional party, inefficacious/boycotting nonvoters would reveal either
allegiance or dissent to the party system and would tend to make their voice by going to
the polls, rather than taking to the street. As political alienation can be better referred to
as an enduring orientation than as transient feelings of discontent, non-partisan nonvoters
who appear to be consistent in terms of electoral behavior and partisan attitude signify
consistent and enduring orientation of citizens’ alienation from political system as well as
from candidates in the elections (cf. Citrin et al., 1975). Thus, they can be labelled as
‘alienated nonvoters’.
More specifically, both winners and losers are not alienated from political systems, and
thus have a chance to be included in the winning coalition (W), especially when W’s size gets
large. In chapter 3, we can see that the short-term W can increase the loser’s political legitimacy significantly in the cross-country multilevel analysis. For non-partisan voters, they
have some potential involving protesting since they are pragmatic in terms of their specific
policy goals. In that sense, the characteristics of their participation in mass political events
2 If

inefficacious nonvoters boycott in the elections, those affiliated parties are mainly
opposition parties.
124

could be highly focused in terms of policy goals, and demonstrations are regarded as another
conventional political participation or civic expression, rather than as unconventional participation that can destabilize the political system and ultimately risk a leader’s deposition
(cf. Norris, Walgrave and Van Aelst, 2005; Dalton, 2000; Inglehart, 1997, 1977).
Regarding partisan nonvoters, they are eventually a part of nonvoters. As they are committed to a certain party identification, their rejection of the electoral process is temporary
or rather in a fickle fashion since they are supporters of a certain party that is likely to boycott elections and then rejoin future elections to get the seats in the parliament. Or, they
are erstwhile supporters for a disappointing party. Thus, through election boycotting either
organized by opposition parties or motivated by partisan individuals, partisan nonvoters do
affect electoral participation.
In this vein, election boycotters, when identified as partisan nonvoters, can be mobilized
not to vote only during a limited time period, thereby dropping turnout sharply and then
boosting it dramatically; For example, in Ghana in 1992, the opposition initiated a major election boycott after declaring that the incumbent party had snatched a victory from
the opposition by electoral unfairness. This resulted in a substantial turnout drop by 26.4
percent of registered votes in the December parliamentary election from the previous presidential election in November, 1992. However, when the opposition came back to the electoral
process in 1996 without boycotting, turnout went up to 78 percent. Similarly, Bangladesh
in 1996 also witnessed the same trend of plummeting and soaring of turnout during the two
elections in 1996 (Beaulieu, 2006). In both cases, partisan boycotters were mobilized, either
in confronting a flawed electoral process or in a strategic “ruse by opposition parties that
concluded that they stand no chance of winning” (Bratton, 1998: 53). A sudden drop of
voting turnout resulting from election boycotting is, therefore, not caused by boycotters’
political alienation from the electoral process.3
3 Another

example of nonvoters involves election suspension resulted from military coups
and civil wars. In either case, it is reasonable to assume that all voters including loyal
winners to a previous incumbent leader become alienated nonvoters.

125

Moreover, the risk of deposition seems to be hardly influenced by easily recovered turnout
after boycotting. Rather, election boycotting provides a leader with useful information
whether to increase or decrease the size of winning coalition, depending upon how much
the leader can utilize free resources that could be available to her to purchase support from
boycotting opposition parties (Bueno de Mesquita and Smith, 2010). In this sense, I center
on the role of non-partisan nonvoters as ‘alienated nonvoters’ in explaining the causes of
mass political movements and the risk of deposition.

5.2.2

Alienated Nonvoters

Many political stability scholars suggest that a combination of efficacy and trust in the political system matter in understanding citizens’ collective dissent (Lipset, 1959: 86; Gamson,
1968; Wrong, 1979; chap.5; Barnes et al., 1979; Gamson, 1990). William A. Gamson argues that “[p]olitical alienation includes both an efficacy (or input) dimension and a trust
(or output) dimension. ... The efficacy dimension of political alienation refers to people’s
perception of their ability to influence; the trust dimension refers to their perception of the
necessity for influence” (1968: 42). Similarly, Finkel (1985) notes in his panel survey study
that participation enhances political trust and external efficacy and vice versa (for a crosssectional study with similar reciprocal results, see Barnes et al. (1979)). His findings imply
that non-participation is a function of the lack of political trust and external efficacy.
From these studies, citizens’ dissent and political alienation seems to be closely associated
with political participation. Electoral participation, i.e., voting itself, may provide policy
benefits that election losers and nonvoters may have missed in past elections.4 In order
to examine the nonvoters’ attitudes, Finifter’s dimensions of political alienation seem to
be useful in understanding the consequences of nonvoters’ either voluntary or involuntary
alienation (1970). She distinguished four dimensions of alienation: political powerlessness,
4 Nonvoters

become either to be apathetic to political outcomes or to be stressful with the
cross-pressure of real conflict that can force voters to lose interest in the election (Lazarsfeld,
Berelson and Gaudet, 1944).
126

political meaninglessness, perceived political normlessness, and political isolation.5 As long
as election losers show their partisan status in the existing party system, they can still
engage in the political process not only through elections, but also through demonstrations
as another mode of civic expression (Norris, Walgrave and Van Aelst, 2005). In contrast,
at the individual-level explanation of electoral behavior, nonvoters’ disengagement from the
election largely results from their lack of resources for participation, the lack of interest to
engage in politics, or their lack of social capital for participation (Verba, Nie and Kim, 1987;
Oppenhuis, 1995; Verba, Schlozman and Brady, 1995; Dalton, 2006). That is, nonvoting
results from ignorance, indifference, dissatisfaction, or inactivity (Ragsdale and Rusk, 1993).
And yet, other scholars of political participation consider vote avoidance as a political
activity in the context of a small-coalition institution such as communist China and the
Soviet Union (Shi, 1999; Roeder, 1989). In a somewhat similar vein, when voters cannot
freely ‘voice’ in elections in a small winning coalition and constrained system, non-partisan
nonvoters may abstain from voting “without revealing that they planned to do so”, which
can be termed as ‘silent boycott’ (Bratton and Lambright, 2001). The election boycott of
Uganda’s Referendum 2000 on the introduction of multiparty competition was mainly organized by the opposition parties. But, nonvoters were mainly “unwilling to side publicly with
a stay-away organized by established political parties”, although the results of Referendum
2000 on the surface suggested that citizens supported for the Museveni’s inclusive, no-party,
‘movement system’. Thus, the multiparty sympathies were mutely voiced by non-partisan
nonvoters .
Related to non-partisan’s perception affected by leader’s manipulation of the key institution of the winning coalition size, their political alienation from the existing party system
may well be expressed as a nonvoters’ ‘exit’ from the system. Political participation such as
5 Finifter’s

four types are developed from Seeman’s (1959) four meanings of alienation. He
originally included one more type of self-estrangement. Political meaninglessness involves “an
individual’s inability to distinguish any meaningful political choices .... consequently, [one
cannot] use [choices’ probable outcomes] to change social conditions” (p. 390). Therefore,
meaninglessness implies more frustrated ’powerlessness’.

127

campaigning, voting, communal activity, and personalized contacts matters as potential precursors of mass political movements (Verba and Nie 1972). Among these activities, as voting
seems to bear its highest stake in terms of leader survival, whether the election is nominal
in competitive authoritarianism or substantial in advanced democracies, we now need to see
how citizens’ non-voting mediates the relations between non-partisans’ alienation and mass
mobilization.
Following his distinction of trust in a specific leader, national government, and country,
Citrin (1974) argues that a sense of alienation from the political community level leads
people to be “oppositionist” and to withdraw from electoral politics. He also argues that
it does “not [automatically] imply that increased cynicism about the incumbent national
administration will reduce turnout at presidential election” (979). Although he notes that
the feeling of alienation and cynicism about the political community may not affect turnout
in a presidential election, the changes of turnout, especially a significant drop of participation
in elections, is a good predictor for political alienation and dissent about the existing political
system.
Electoral politics provides a venue where continuity and volatility of membership in a
certain institution can be understood in a context of partisan membership. As political parties are a quintessential political institution and organizations that mediate the state-society
relation, voters’ and nonvoters’ behavior should be a key variable in explaining political upheaval and decay resulting from party and electoral politics. Elections provide a legitimating
process for both election winners and election losers to perceive whether the institutional
procedures secure a fair selection of leaders or not (Moehler and Lindberg, 2009: 347). Participating in elections itself arouses citizens who legitimize the existing political systems.
Therefore, non-participation or non-voting behaviors should be an variable of interest to
political stability scholars, too.

128

5.2.3

From Votes to Violence

The central argument in this chapter is that the lack of electoral participation and a decline
in partisanship represents high risk of mass political movements and thus poses greater
risk of a leader’s deposition. This kind of electoral behavior is said to be derived from
citizens’ political apathy, and furthermore the lack of perceived legitimacy in their political
system (Kaase and Barnes, 1979; Kaase, 1988). This argument is aligned with Campbell et
al.’s earlier findings on partisanship; at the individual level, partisan status provides more
motivations to vote, to attend to the election campaign, and attend to the election outcomes
(1960).
Voters’ non-participation in the elections may not be a dangerous signal for leaders or
may be considered as even a favorable situation for their survival. With an apathetic and
satisfied citizens’ non-voting, leaders take advantage of the citizens’ political quiescence
(Edelman, 1971). Earlier, E. E. Schattschneider (1960) warns of a form of full participation
that “if everybody got into the act the unique advantages of this form of organization would
be destroyed, for it is possible that if all interests could be mobilized the result would be
a stalemate” (35). However, as threatening citizens’ behavior to the political system has
long been considered as the main target of government repression, political quiescence is
considered a major benefit to leaders. For example, quiescence includes supporting the
extraction of taxes, contributing to the state’s economy, and delivering no opposing ‘voices’
to the party in power, thereby being a major part of the political authorities’ legitimacy as
a protector (Davenport, 2007). Defined as a “law of coercive responsiveness”, the repressive
state’s actions are reasonable responses to the citizens’ potential threat to the political
authorities.6
Related to the political quiescence literature, in unconsolidated democracies where demo6 Reversed

causal relationship from mass threat to repression is also discussed. That
is, the state’s repression may also lead to a form of response of mass threats; however, it
is characterized as highly inconsistent findings by many repression scholars (Hibbs, 1973;
Lichbach and Gurr, 1981; Muller, 1985; Gurr and Moore, 1997).

129

cratic practices are easily violated by the incumbent leaders, “a healthy dose of political
skepticism . . . is associated with greater attitudinal resistance to [democratic] breakdown
in the form of a military coup” (Seligson and Carrión, 2002: 58). Contrary to the conventional belief of a negative relationship between system support and approval of coups, the
V-curve between system support and approval of military coups found in the Peruvians’
attitude proves that uncritical support for a system represents those who might well accept
the violation of democratic practices in unconsolidated democracies such as Peru.
Nonvoters’ attitude toward regime stability can be induced from Seligson and Carrión’s
work. As high levels of support for coups is significantly associated with both extremes
of system support and non-support, those who are expected to support coups are either
alienated nonvoters identified as low system supporters or apathetic nonvoters identified
as satisfied (if not strong) system supporters. Similarly, some scholars note that a decline
in partisanship indicates either increased political apathy or the rising concerns of protest
behavior, both of which may pose a similar potential threat to the incumbent government
as well as to the existing political order, or both (Barnes et al., 1979; Kaase, 1990).
Not only for the characteristics of system support in unconsolidated democracies, but
also the excessive changes of electoral participation have been considered a major threat
to political development and institutionalization in the emerging democracies. Competitive
elections with excessive participation are also considered to contribute to the increase of
protest involving other complicated issues such as resource allocations and ethnic cleavages
and liberalization processes. However, increased electoral ‘departicipation’ by voters will be
a good predictor for increased protest levels, as discussed in more detail below.

5.3

Hypotheses on Nonvoter, Protester, and Revolutionary

What if any political system cannot embrace either extremist parties or democratic transition, that is, it cannot try to keep a small size of the winning coalition? Do existing political
parties of the incumbent and opposition parties facilitate mass political movements while
130

keeping the size of winning coalition as it used to be? This is because as described above
elections’ role of a forum for participation can incorporate potential dissidents including
election losers as well as nonvoters.
According to the literature on contentious politics, there are three so-called master variables of the social movement literature: resource mobilization and organizational forms or
“mobilizing structures”, framing strategies, and political opportunities and context. These
ways of understanding mass mobilization of political movements are useful but developed in
the context of long-standing liberal democracies. While we consider both advanced and new
democracies and even competitive authoritarianism, the fact that we focus on the effects of
elections on mass movements and on leader survival allows us to narrow our perspective to
one of the frameworks. This is because through a ‘political opportunities and constraints’
framework we can see the interaction of masses and elites.
Tarrow (1998) lists five dimensions of opportunity and suggests that rising protest can
be expected according to the following conditions: (1) the opening of access for new actors;
(2) the evidence of political realignment within the system; (3) the appearance of influential
allies; (4) emerging discord among the elite; and (5) a decline in the state’s capacity or will
to repress dissent (Tarrow, 1998: 76). I will focus on conditions (1), (2), (3) and (5) in
this section. The condition (4) and (5) have more implications for the interaction between
masses and leaders will be discussed in the next section.
Voting turnout is regarded as a key and sole mechanism for political participation that
involves a majority of the citizens. Mostly, voters have some or substantial affiliations with
certain political parties. Joseph Schumpeter (1952) suggested that limited citizens’ participation was sufficient to bring stable and accountable government, only when these are free
and fair elections contested by parties and politicians legitimize governments’ representation
at regular intervals. Later, E. E. Schattschneider (1960) defined party’s organization influenced by “some kind of political bias ... is itself a mobilization of bias in preparation for
action” (p. 30: italicized originally).

131

Regarding resource distribution through elections and parties, many analysts view elections and parties as the most important tool to ease diverse forms of distributional conflicts
in intra-elite relations as well as in the state-society relations. This is the case not only in
democracies but also in autocracies where leaders provide electoral markets to incorporate
potential dissenters and thus strengthen the political hold by using these “autocratic survival strategy” (Geddes, 2005; see also Blaydes, 2011; Brownlee, 2007; Gandhi and Lust-Okar,
2009)7 .
By using vote buying and turnout buying, i.e., by rewarding unmobilized voters (supporters), parties in authoritarian regimes are able to mobilize their potential supporters (Stokes,
2005). Electoral authoritarianism can be disaggregated into “hegemonic electoral authoritarian” regime and “competitive authoritarian” regime (Diamond, 2002; Levitsky and Way,
2002). Electoral authoritarianism can develop either into durable authoritarianism or into
opportunities for democratization (Brownlee, 2007). Liberalization through elections can be
seen as either using “a safety valve for regulating social discontent and confining opposition
... [or as] turn[ing] a regime’s pressure valve into a springboard for entering government” (8).
Brownlee (2007) makes an interesting comparison between Egypt and Malaysia for the former cases and Iran and the Philippines for the latter cases. The characteristics of elite unity
and uncompetitive elections led to durable authoritarianism in Egypt and Malaysia, while
the opposite characteristics of elite discord and contested elections resulted in opportunities
for democratization.
Therefore, selectorate theory can be extended in this sense: a small winning coalition
characterized by competitive elections and elite discord in authoritarianism could have more
potential to be expanded, thereby being a good indication for democratic transition. In
contrast, a small winning coalition with uncompetitive elections and elite unity could result
in a more constrained the size of winning coalition, which also perpetuates the existing
authoritarian regime as well as the autocrat.
7 Brownlee

focuses on the effectiveness of parties while the role of election is highlighted

in Blaydes.
132

Similar to liberalizing impacts on leader survival in electoral authoritarian regime, democratic elections are also expected to bring divergent outcomes depending on the elite structure
(i.e., winning coalition structure) and electoral competitiveness. Since mass mobilization is
more freely accepted in a democratic regime than in an authoritarian counterpart, the role of
mass threats through protest, riots, demonstrations, and even revolutions should be included
in the explanation of democratic leader survival.
Furthermore, in many non-advanced democracies, mobilization of voters is highly associated with organizing supporting networks for a party through vote buying and even
turnout buying. Riker and Ordeshook (1968) suggest that voters go to the polls because
they tend to overestimate the tiny chances of being decisive voters or receiving some large
direct benefits from their paradoxical act of voting. Also, party mobilization of nonvoters
can be found more frequently in new democracies, rather than old democracies (Karp and
Banducci, 2007). However, for the voters in developing countries, especially for the poor
voters, clientelistic practices force voters to be susceptible to the benefits of consumption
goods or other substantial private goods more directly than the wealthy voters or the voters in advanced democracies. In a somewhat similar context of welfare spending literature,
Hicks and Swank (1992) contend that increased electoral turnout is usually from among poor
voters, so increased welfare spending is due to increased turnout.
Thus, what are the elements in electoral non-participation and competition that create
problems for regime stability if vote and turnout buying is not successful? To answer this
question, I will survey the existing literature on electoral non-participation and competition
and their consequences on mass political movement. During this survey, the key relationship between electoral non-participation/competition and mass political movement will be
theorized, and specific testable hypotheses will be developed.

133

5.3.1

Vanishing Voters and Mass Movements

Tarrow’s condition (1) implies a political opportunity in which the dissidents are able to gain
partial access to participation. This incentive should be great for them to engage in a mass
political movement since the narrower participation level in the preexisting rules could force
the dissidents to use an alternative unconventional participation of protests and demonstrations. As a mass movement usually represents an extremist view of politics, the political
interest of nonvoters can be awakened by a mass movement (Lipset, 1960). Therefore, a
good indication for this, I argue, is significant changes of nonvoter’s size between election
years.
The most famous decisive role of nonvoters in risking the regimes as well as leaders has
been Reinhard Bendix’s “radicalization of the electorate”, i.e., the mobilization of ‘apolitical’
segments such as nonvoters and the young in the late 1930 and 1932 during Germany’s
Weimar Republic period. Following the divergent notions of Bendix (1952) and Lipset (1960)
on the surge of German National Socialism, both apolitical nonvoters’ radicalization and the
impoverished middle-class’s mobilization seem to matter in understanding the success of
Hitler and the failure of the previous regime (see also Linz, 1978a, 1976; O’Lessker, 1968).
Although nonvoters’ radicalized and thus mobilized elections risk deposition of leaders,
nonvoters’ “losing interest” in elections itself also poses substantial threats to leaders.8 This is
because they can be mobilized with more compatible political parties. Therefore, substantial
increase of electoral non-participation derived from a decline in partisanship would either
indicate rising political apathy, or the increase of protest potential (Kaase and Barnes, 1979;
Kaase, 1988; Craig and Maggiotto, 1981; Muller, 1979; Muller, Jukam and Seligson, 1982;
Cheles, Ferguson and Vaughan, 1995).
By using actual voting data across ten advanced democracies, Przeworski (1975) highlights a significant destabilizing effect of electoral non-participation on the pattern of voting
8 Lazarsfeld,

Berelson and Gaudet (1944) argues that nonvoters would escape from any
real conflict resulted from cross-pressures and thus lose interest in the election.

134

behavior and democratic stability. Borrowing Huntington’s concept of “developed” political
system and political institutionalization, Przeworski (1975) demonstrates that institutionalization of the party system regarding voting patterns along with electoral (de)mobilization is
based on variations in the membership of political institutions. Although Huntington (1968)
argues rapid and excessive increase of mass mobilization is a threat to political development
and thus leads to political decay, he posits that “demobilization of groups has a much sharper
destabilizing effect upon patterns of voting than incorporation of new voters” (65; italicized
added).
Yet, contrary to Huntington’s (1968) negative outcomes of increasing political participation, he does not seem to worry about the potential instability caused by incorporation
of new voters, but in more concerned about the danger of rapid and abrupt increases of
electoral apathy or non-participation. He notes that “when groups cease to play according to
the institutionalized rules, alternative forms of political mobilization may emerge: “parties”
become replaced by “movements” ” (p.67). He suggests that the real threat to a democratic
system as well as political elites is posed by non-participation or withdrawal of voters from
the electoral process, thereby losing the legitimized process of elections as a form of conflict
resolution in a society.
Anderson and Mendes (2006) contend that while individuals who identified themselves
as nonvoters are expected to have more negative attitudes towards a political system that
implies more protest potential9 , their multilevel analysis across old and new democracies
shows that nonvoters tend to show less protest potential, compared to voters.10 The aforementioned Finkel’s (1985) finding suggests that more electoral participation leads to system
support and feeling of trust and external efficacy, and vice versa. Both research designs
seem appropriate in consideration of limited data resources; for the former, their dependent
9 See their footnote 32 in p. 102
10 Moreover, election losers among

those voters have more increased protest potential than
winners, and the increasing effects are strengthened in the new democracies. With specific
focus on ethnic violence in India, Wilkinson (2004) highlights state-level electoral competition
as the main cause of ethnic violence.

135

variable is limited to protest potential, rather than real mass threats to the system.11 For
the latter, the U.S. case with a panel data design provides strong internal validity, while
lacking external validity. At the individual-level survey analysis, by contrast, previous studies suggest inconclusive results in terms of the effects of electoral participation on system
legitimacy and stability.
These inconclusive results about the effect of electoral non-participation on system stability illustrate ambiguous causal relations between election participation and protest potential.
This is so because we do not focus on how much citizens are politically alienated, thereby
understanding less of the attitudes of the alienated and frustrated toward the political system
and its consequences for protest potential.
The conceptual dimensions of political alienation can represent relative deprivation or
‘anti-system radicalism’ in their attitudes toward political system (Gurr, 1968a,b; Crozier,
Huntington and Watanuki, 1975). Latent attitudinal frustration can be translated into
aggression and thus protest potential. Moreover, Rosenstone and Hansen (1993) found that
the fewer the number of participants in political activities, the greater the inequality in
political participation (238). These kinds of frustration and inequality largely can be found
among the more disadvantaged nonvoters who are disengaged from conventional political
activities such as working in election campaigns, contacting authorities, and contributing
money to candidates or parties. Regarding unconventional political participation such as
demonstrations, protests, strikes, and revolutions, alienated and frustrated nonvoters are
more likely to get involved in such anti-system movements than advantaged citizens.12 We
11 This

is legitimate in their study since they use survey-based method, but we may still
be interested in knowing if real mass threats can be a function of nonvoter’s and election
loser’s variation in terms of their respective sizes.
12 The more advantaged citizens are said to be more engaged in different kinds of unconventional political participation than the disadvantaged citizens (Marsh and Kaase, 1979:
112-26). These includes boycott, rent and tax strikes, blocking traffic, and occupying buildings. These activities are not directly related with ‘anti-system radicalism’ or derived from
relative deprivation. More specifically, Norris, Walgrave and Van Aelst (2005) shows that
Belgium’s demonstrations reflecting diverse social issues do not verify the relative deprivation hypothesis. They complement earlier the Political Action Study by identifying strategic

136

can hypothesize that electoral non-participation can facilitate alternative unconventional
participation if nonvoters’ alienation can no longer be tolerable any more.
The first hypothesis follows:
H1: As nonvoters’ size change increases positively, the level of mass political movements
will increase.

5.3.2

De-mobilized Nonvoters

Politically alienated parties can exist as much as nonvoters since political parties can be
organized by the dissenters’ dissatisfaction with the status quo (cf. Schwartz, 2000). For
example, ideologically extremist parties represent a barometer of citizen alienation as voters’
support for the extremist parties reflects their discontent with major parties and the working
of democracy. According to Tarrow’s condition (2), shifting electoral alignments open a new
possibility of contention.
Yet, while the existence of the extreme parties is associated with executive, not legislative,
instability and the extreme party voters are significantly associated with rioting, “none of
the extremist party measures was associated with riots or protests, once account was taken
of citizen alienation ... the presence of extremist parties tended to dampen rather than
exacerbate turmoil in the streets” (Powell, 1981; Powell, 1986b:372). By including more
voices in the democratic system either by incorporating the extremist parties or by increasing
their chances of sitting at the bargaining table with leaders (i.e., both are increasing the size of
W), the magnitude of mass movements can be lessened with greater legitimacy of the regime
(Gurr, 1968a).13 Hence, if there is no such party to engage with nonvoters or potential
dissidents, the increased size of nonvoters should not be an ignorable thing to leaders.
resource of the advantaged, not alienated, citizens in more frequent involvement in the unconventional political participation (see also Inglehart, 1977, 1997; Dalton, 2000).
13 Gurr (1968) suggests three intervening societal variables of the loyalty of coercive forces,
institutionalization, and facilitation, and legitimacy of the political regime (1104–1106). Facilitation is the only variable that strengthens the magnitude of civil strife while all the other
lessen it.

137

In the context of restricted political rights, an election itself can be viewed as an institutionalized mechanism to provide a forum for political participation, thereby channeling
political dissent and discontent away from protest events (Powell, 1986a). For example, Hipsher (1996) suggests that mass de-mobilization can be ironically obtained by the patience of
the marginal population (the poor) so that democratic transition can be completed without
substantial provocation of the Right into reversing the transition process. Restoration of
democracy was well characterized by mass de-mobilization and institutionalization of the
protesters’ agenda in the Chilean shantytown movement in 1990 and the Spanish neighborhood movement in 1977. The close link between neighborhood organizations and the
Communist party in Spain in mid-1970s provide a favorable political opportunity structure
that allowed both to use contentious collective strategies by the movement.14
In the case of Chile, the shantytown dwellers had the political space to involve themselves
in mass protest when the 1982 economic crisis brought on elite discord in the authoritarian
regime and when the government was indecisive about how to deal with the mass movement.
Besides, the opposition party supported the shantytown dwellers’ protest either to restore
democracy or to at least bring the situation to the bargaining table with the authoritarian
incumbent. This eventually emboldened the movement and facilitated more mobilization.
From both cases, we can expect increased possibilities of a decline in mass movement
if nonvoters could be beneficiaries of institutional change such as expanded political and
civil rights and economic redistribution. Hence, if there is an increased size of the winning
coalition and thus more risk of deposition of leader exists, political rights expansion through
democratic elections works to the advantage of the dissenters or the marginal population
and leads to the mass de-mobilization of nonvoters.15
14 These

favorable opportunity structure included the death of Franco in November 1975, a
expanded militant labor movement, and the increase of elite opposition within both society
(mainly the church) and the state (Hipsher, 1996).
15 Bermeo (1990) highlights the marginal changes of electoral participation as a meaningful
and substantial changes for the marginal population. She argues that “even marginal changes
can have great meaning if you are living at the margin, and if you are given to dissent, the
difference between democracy and dictatorship can literally be the difference between life
138

So, the second hypothesis is:
H2: The increased winning coalition size will decrease the effect of nonvoter’s size change
on mass political movements.

5.3.3

Free Press as a ‘Mobilizer’

In this sub-section, I will describe the role of free press as a facilitator of nonvoters’ mass
mobilization in movements. The Frustration-Aggression hypothesis based on psychological
motivations of political alienation appears to explain whether and how nonvoters’ frustration can be expressed behaviorally in terms of aggression such as mass political movements,
especially anti-system movements. Along with other contextual variables of the winning
coalition size, the diverse resource structure for mobilization should be considered. Norris (2011) argues how excessively negative news, and extensive sexual scandals and financial
corruption, can suppress citizens’ satisfaction with the system which leads to a longer ‘democratic deficit’. In this context, a free press is regarded as one of public goods that facilitates
mass movements.
In pluralist democratic systems, a free press is mainly supported by political elites as a
main component of democratic principles. It is in nondemocratic systems that the role of
the free press has substantial outcomes in mass mobilization. In the Tarrow’s condition (3)
of influential allies, mass media is regarded as a key ally for the dissidents. As a well-known
history of the Reformation that was rapidly facilitated by Johan Gutenberg’s invention of
the printing press aptly describes, a free press helps to trigger contention by multiplying “the
impact of any one event throughout the society” (Huntington, 1974: 165). Rustow (1970)
has contended that the political transition is catalyzed as a result of the appearance of a
new elite that brings political arousal of the alienated and frustrated social groups into being
mobilized for concerted action. Then, free press takes a role of a ‘mobilizer’ for the new elites
to arouse these leaderless group in the transitional periods.
and death” (p. 374).

139

As for nonvoters, they represent the apolitical segment of population. Unlike election
losers, they are not largely represented by any political party, at least based on their electoral behavior. Until their agendas can be expressed through a new party or even a new
regime, electoral non-participation cannot easily result in protest and demonstration. Their
frustration can be translated into aggression and thus contention through a channel of free
mass media.
The aforementioned Leipzig case illustrates how mass protests facilitate the leader’s risk
of deposition. Why did the protest occur in Leipzig and lead to such substantial outcomes?
Lohmann (1994) listed four reasons; first, the citizens in Leipzig and its environs experienced
a more moribund industrial structure that was exacerbated by more severe environmental
problems. Second, there was no secret police to monitor, thus lack of state repression. Third,
thanks to the Leipzig trade fair, more access to international media coverage was available
for dramatic protest events in Leipzig. And, finally, Leipzig’s geographical location could
provide a focal point where frustrated and dissenting masses were willing to participate in
mass protest without being able to identify other participants (see also Opp and Gern, 1993).
As Lohmann shows in the third condition, news communications have had a major influence
on collective action such as the Monday demonstrations.
Regarding elections’ effects on mass political movements, Davenport (1997; 1998) argues
that national elections have significant association with less restriction on mass media as
well as those placed on citizens, especially in non-democratic systems. The role of relatively
free mass media seems to be critical in reducing the state’s repression, which can lead to
more protests by the dissidents in the society. This is also related to Tarrow’s fifth condition,
and it can be restated as follows: a decline in the state’s capacity or will to repress mass
media opens more chances of the dissidents’ rising. More recently, Bueno de Mesquita and
Smith (2010) consider free press as a special kind of public good, i.e., “coordination goods”.
When a leader has enough free resources such as oil and international aid and non-taxable
income, and at the same time when a leader faces mass threats, she is expected to contract

140

coordination goods (i.e., repress the free press) in a small size of W or to provide coordination
goods to buy off revolutionaries.
As shown in both works above, some limitations can be found in terms of the relations
between electoral politics and leader survival. For Davenport’s work, highlighted is the role
of elections in explaining the leader’s or the state’s repression of mass media for leadership or
regime stability. However, specific components of electoral politics are missing. Especially,
the role of potential dissidents composed of electoral losers and nonvoters was not dealt
with. For the Bueno de Mesquita and Smith’s work, while they extended the former work’s
argument by specifying the causal logic of endogenous institutional changes, i.e., leaders’
change of the winning coalition size depending on the levels of mass threats and the existing
winning coalition size, they did not give any attention to the role of electoral politics, which
critically and frequently determine the leader’s longevity.
Therefore, the third hypothesis follows:
H3a: The more freedom of press, the stronger the positive effect of nonvoter’s size change
on the level of mass political movements.
All the hypotheses regarding the determinants of mass political movements are summarized in Table 5.2.
[Table 5.2]

5.3.4

Leader Survival, Mass Threats and Nonvoters

Regarding the leaders’ responses to mass threats and elevated alienation and frustration,
Tarrow’s fourth condition of the elite’s disunity implies that breaches of the elite’s unity
may provide the mass movements’ legitimacy to protest against the party in power and
to counter-protest. For example, many Latin American countries’ successful democratic
transitions were characterized by pacted transitions in which the elite’s unity was maintained
between “seceding Rights” and “accommodating Left” (O’Donnell, Schmitter and Whitehead,
1986). This unity eventually increases the size of winning coalitions through the founding
141

elections, which includes the alienated citizens in the winning coalition, thereby leading to
less risk of mass threats.
In contrast, a leader under unpacted democracy may want to secure her tenure until
at least the second election, rather than instigating another wave of mass mobilization and
thus authoritarian reversals. As democratic breakdown studies showed, the greatest threat
to democratic transition derives from “a backlash by elements of a hard-line faction, most
commonly when the military executes a reactionary coup” (Bratton and Van de Walle, 1994;
460).16 Therefore, the disunity of elites is more likely to facilitate mass movements and to
risk leader deposition.
Related to the leader’s capacity to control the elite’s disunity, Tarrow’s last condition of
a decline of a state’s will or capacity to repress mass mobilization also represents a leader’s
responses to demands of mass movements such as expanding the winning coalitions size.
In this vein, Bueno de Mesquita and Smith (2010) highlight the role of free resources to
prevent breaches of unity among elites. By taking advantage of free resources such as oil,
international agencies’ aid, and nontax revenues, a leader is expected to either buy off revolutionaries’ demands when the leader cannot control the elite’s disunity (i.e., the previous
winning coalition is large) or to reduce coordination goods such as free press when the leader
is able to control the discord (i.e., the previous winning coalition is small).
From the previous analyses, nonvoters’ non-participation appears to be a significant key
predictor of mass mobilization for political movements. We now turn our focus to the
consequences of anti-system movements mobilized by alienated and frustrated citizens. Many
have tried to know the determinants of voter turnout and revealed that some institutional
mechanisms can increase turnout, such as less restrictive registration rules, proportional
representation systems, relatively infrequent elections, and compulsory voting (Gosnell, 1930;
Tingsten, 1975; Franklin, Eijk and Oppenhuis, 1996; Jackman, 1987; Jackman and Miller,
1995; Powell Jr, 1980; Powell, 1986a).
16 Many

unpacted democracies in Latin America have been reversed to authoritarian
regime, with the exception of Costa Rica (O’Donnell, Schmitter and Whitehead, 1986: 45).
142

Among them, Powell (1986a) notes that electoral participation provides “some channeling
of citizen discontent out of the street” and “legitimate political channels without loss of
executive stability” (37). As his dependent variable is voter turnout, he contends that less
frequent citizens’ riots, protests, and various kinds of turmoil are believed to increase voter
turnout. This association only focused on unidirectional causality without considering a
possibility of reverse causality. As discussed above, mass political movements can be seen
as citizens’ behavioral outcome originated from their deep-seated feelings of alienation and
dissent, and political alienation can be uncovered by looking at the variation of nonvoting
levels throughout elections. Therefore, voter turnout can tell leaders how stable the leaders’
tenure is. If we assume the interaction relations of mass movements and non-voting behavior
in terms of the chief executive’s stability from both the previous section’s analysis and
Powell’s implication, leader survival can be better understood.
Bueno de Mesquita et al. (2003) examine the causal relationship from revolutionary mass
movements to leaders’ institutional changes. Reducing the size of the selectorate can be well
found in the course of revolutionary regime changes. They show that each stage of revolution
(i.e., at the outset, during, and at the end) discourages an increase in selectorate size when
a leader has a larger coalition (378). The larger the winning coalition the leader has who is
facing (or heading) revolutionary stages, the less incentives she has to expand the selectorate
for her political survival.
While implicitly ignoring the nature of electoral participation by mass public or limited
number of the franchised citizen, however, Bueno de Mesquita et al. (2003) focus more on a
post-hoc variable of mass political movements, rather than paying attention to an (in)direct
but more fundamental effect than mass threats on leader survival by way of mass involvement
in the revolutionary movements (see also, Bueno de Mesquita and Smith, 2010).
So, the final hypotheses regarding the determinants of the leader’s deposition are described below and summarized in Table 5.3:
H4a: The more mass threats from mass political movements, the greater the leaders’ risk

143

of deposition.
H4b: The more nonvoters’ size change increases positively, the greater the leaders’ risk
of deposition.
H4c: The effect of mass threats on the risk of deposition will be strengthened by an
increased change of nonvoters’ size. (i.e., increased alienation and frustration)
[Table 5.3]

5.4

Methodology and Data

In this section, I will test the hypothesized effects on both the levels of mass political movements and leader survival. Thus, the dependent variables are separately composed of mass
political movements and leader’s risk of deposition. I will first describe each dependent
variable’s measure and how to construct the variables. Then, key independent variables of
nonvoter size, mass threats, and free press will be discussed.

5.4.1
5.4.1.1

Dependent Variables: Mass Political Movements and Leader Survival
Mass Political Movements

To test the first stage of relations of nonvoters and mass political movements, a measure of
the revolutionary threat should be first developed. I construct a measure of the occurrence
of antigovernment demonstrations, riots, strikes, and revolutions by using data drawn from
Banks (2000). Unlike the previous study of revolutionary threat found in selectorate theory
(Bueno de Mesquita and Smith, 2010; Bueno de Mesquita et al., 2003), I give different weights
on the four different components of a mass political movement measure in order to make
this variable compatible with the concept of anti-system revolutionary threat. When giving
weights, I follow Banks’s weights found in his combined variable of the domestic conflict
event data (the variable domestic9 ).

144

After giving weights, I create an Index of Mass Political Movements by borrowing Bueno de
Mesquita and Smith’s standardized version of the variable. That is, z = (ln(1 + x) −
mean(ln(1 + x)))/(std.dev.(ln1 + x))), where x = (demonstrations, riots, strikes, revolutions). By summing the four standardized variables and dividing by four, the weighted
political mass movement variable is created.

5.4.1.2

Leader Survival

I use Goemans, Gleditsch and Chiozza’s Archigos data set for a measure of leader survival
(2009). The existing leader longevity data suffer from some measurement problems such as
overlapping ruling time periods by two or more leaders, and significant gaps in the order of
leaders in power. The main variables from Archigos for the analysis are entry and exit dates
of each leader across countries.

5.4.2

Main Independent Variables

Main independent variables include nonvoter’s size, mass threat derived from mass political
movements, and free press as ‘mobilizer’ or ‘facilitator’. First, I include the variable of the
election loser’s size, conceptualizing electoral competition that inherently affects the risk
of deposition through electoral politics. Then, the operationalization of key independent
variables is described.

5.4.2.1

Election Loser

I use Vanhanen’s (2000) measure of democracy data. It consists of electoral participation
and electoral competition, based on Robert Dahl’s polyarchy concept (Dahl, 1971). I use
Vanhanen’s variable of electoral competition to measure the size of election losers. He conceptualizes electoral competition by using “the percentage share of the votes for the smaller
parties and independents in parliamentary elections, or of the seats in parliaments” (2000:
253).
145

The election loser’s size variable provides a measure of electoral competition in each
nation in each year.17 However, each nation has different electoral and party systems, and for
example, the proportional representation system with multiparty and coalition government
may be more associated with smaller size of election losers and nonvoters. In order to
ameliorate the country-specific differences, I construct an index on the changed size of election
losers (and also nonvoters below), rather than using the level of election loser’s size.
The changed size of election losers is largely incorporated into the electoral volatility
literature. The standard measure of electoral volatility is the index of deinstitutionalization
conceptualized by Przeworski (1975). This accounts for changes in votes for each party across
elections. The Pedersen index described below represents well the variation of electoral
volatility.
Vi = (1/2) ∗

|Vp(t+1) − Vpt |, where p represents party and t represents year.

And, this party-specific electoral volatility can be rewritten in a country-specific expression as follows:
2 ∗ Vi =

|LVi(t+1) − LVit | +

|W Vi(t+1) − W Vit |, where i represents country, t

represents year, LV represent loser’s vote, and WV represents winner’s vote.
From a revised expression of Przeworski’s deinstitutionalization (or volatility) index, if
the absolute value of the election loser’s component contributing to deinstitutionalization
increases, then the winner’s component will decrease, given the fact that deinstitutionalization of voting patterns (expressed as in 2 ∗ Vi ) is only affected by the both sizes of partisans.
By using an aggregate voter concept with the absolute terms, the concept of election volatility explicitly hides the election loser’s voting patterns as well as election winner’s patterns.
Specifically, when we are required to consider some interaction effects of voter’s behavior
17 If

there is no election in a nation in a certain year, the previous size of election losers is
used in the subsequent years. This also applies to the size of nonvoters below.

146

with institutional variables, the hidden components should be uncovered to separate out
different behaviors of voters. That is, ignoring the election loser’s (i.e., potential system
dissenter’s) voting patterns would result in confounding the effects of election volatility on
leadership stability as well as regime stability. Therefore, the changed size of election loser
can be written as follows:
LVit = LVit − LVi(t−3) 18
Moreover, we can see that election volatility is not the only cause of political decay and
deinstitutionalization. If election losers contributing to political decay with their volatility
may ‘exit’ the domain of electoral politics due to the low chance of being winners, they may
turn to being nonvoters in the next election cycle. Then, it seems that nonvoters’ role as
well as election losers’ role should affect significantly the levels of political decay in terms of
voting patterns.

5.4.2.2

Nonvoters

As we discussed in chapter 2, nonvoters are ones who cannot, are not willing to, or are not
asked to participate in the electoral process (Verba, Schlozman and Brady, 1995). Thus, it is
not easy to measure nonvoting electoral behavior. Considering the expected association with
our dependent variable of mass political movements and even the risk of leader’s deposition,
the nonvoting measure should reflect changed electoral behavior between elections.
Electoral participation can be revised from the earlier application of Przeworski (1975)
as follows:
Mi =

(V oteri /P opulationi ),i = 1, 2, ..., n, where i represents country.

18 Other

than three-year differences, I consider one- and five-year difference and obtain
similar substantive results. This applies to following variables of the changed size of nonvoter
and the changed level of mass political movement.

147

As the nonvoter’s size can be defined as 1 − Mi and the size varies in each election year,
the non-participation can be written as follows:
N Vit = 1 − Mit =

[1 − (V oterit /P opulationit )],i = 1, 2, ..., n; t = 1, 2, ..., m, where i
represents country and t represents year.

The ratio of voters to the total population, however, can be decomposed as follows:
V ote/P opulation = Eligible/P opulation × V ote/Eligible.19 As discussed above in the
election loser measure, the changed size of nonvoter is used to measure electoral nonparticipation:
N Vit = N Vit − N Vi(t−3)
The use of this changed size also ameliorates the aforementioned confounding effects of the
electoral system difference across countries. However, the electoral changes within a country
may cause another measurement problem. Thus, I use a big change of nonvoters as an
alternative measure to test the hypothesis (Appendix H). Not only for pluralist democracies,
but also in a state-controlled election systems, a newly introduced semi-competitive election
can change the incentive structure of electoral officials who used to coerce voters to vote.
19 Przeworski

(2009)’s recent revisit of earlier discussion on political mobilization deals
with potential biases, owing either to the variations derived from the registration issue in
electoral process or to the aging of the population (Przeworski, 1975). He admits that the
latter issue is related with the scarcity of the aging composition data. For the former, in the
context of a newly-introduced registration process in a country, the withdrawal of the extant
voters or the influx of the new voters to the elections may introduce a bias to my electoral
mobilization measure. This is because the additional registration process will make the
potential nonvoters easily withdraw from the election and prevent them from participating
in the election.
Yet, as we concern only the differences of nonvoter’s size in terms of their effects on mass
political movements in a country, the bias problem will only arise when we see a changed
electoral rules regarding the registration process. Therefore, the electoral mobilization measure that merges the registered and the eligible is not too vulnerable to this bias. Moreover,
if a country has high registration rate as in most countries (mostly above 90%) except for
the U.S. (68% in 2006), Bahamas (75% in 2007), and South Africa (77% in 2009), the bias
should not be serious (Rosenberg, Chen and for Justice, 2009).
148

Consequently, some drop of vote turnout can be expected with a little addition of democracy
(Shi, 1999). These problems can be ameliorated with the use of the big change of electoral
‘departicipation’, measured as a 5% increase.
When the voting-eligible population data is not available across countries and times,
voting-age population (VAP) can be considered (?). However, the historical records of VAP
are not reliable in any available data for a comprehensive set of country cases and time
periods, which makes the total adult population as an alternative measure for the purpose of
analysis.20 Further, Abramson, Aldrich and Rohde contend that correcting the denominator
only results in “relatively small differences in the overall estimate of turnout” (Abramson,
Aldrich and Rohde, 1998: 68). Therefore, I also use Vanhanen’s electoral participation
measure, which is “the percentage of the adult population that voted in elections” (2000:
253).
Specifically, a lowering of the voting age in most democracies occurred in the 1970s.
Powell (1986a) observed that ‘the age level increased turnout in the 1960s by a small amount,
but decreased it about 2% in the 1970s, with the American lowering of voting age and the
age bulge among the young” who are usually considered as being less interested in politics
(Niemi, Stanley and Evans, 1984). Other scholars suggests that the enfranchisement of 18,
19, and 20 year olds decreased overall turnout by only one percentage point (Wolfinger and
Rosenstone, 1980). Taking into consideration these concerns, I set a 5% of nonvoter increase
as a critical and substantial change of nonvoters in the later analysis.

5.4.2.3

Winning Coalition (W)

As discussed in the previous chapters, the winning coalition size can be derived from four
variables (Bueno de Mesquita et al., 2003). These are the regime type, the executive recruitment’s openness, the executive recruitment’s competitiveness, and a competitive party
20 Even the IDEA

(Institute for Democracy and Election Assistance) data seems to bear errors. For example, some country-eleciton-year cases show that voting-age population (VAP)
is smaller than registered voting population (Reynolds, Reilly and Ellis, 2005).
149

system. The regime type is from Arthur Banks’s Cross-National Time-Series Data Archive
and is measured on a 3-point scale (Civilian, Military-Civilian, Military regime). The other
three variables from Polity IV data (Marshall, Jaggers and Gurr, 2009).
Bueno de Mesquita and colleagues’ (2003) construction of the winning coalition size and
selectorate size is described in chapter 3. I use Bueno de Mesquita et al.’s original measure
for this chapter, rather than using the revised measure of W which is adopted in chapter 3.
The reason is twofold. First, with the more rough original measure of W, the empirical test
can be more conservative in terms of the effect of the winning coalition size. Second, as this
chapter relies only on the country-level variables, unlike chapter 3’s multilevel analysis, all
the country-level variables may not require more refined measure of W.21

5.4.2.4

Mass Threats

As Banks’s measure of domestic conflict events data rely on media coverage, the variable of
mass political movement suffers from reporting biases, societal norms, and regime stability
(cf. Bueno de Mesquita and Smith, 2010). Reporting biases mainly result from the variations
of media coverage across countries. Advanced democracies usually have better reporting
media to fully cover the occurrence of protest than new democracies where media coverage
favors the government or the party in power. Societal norms of more contentious behaviors in
citizens demands, such as French farmers, should be considered, too. Lastly, regime stability
involves the fact that some countries such as democratizing countries ‘from below’ are more
likely to see citizens protest in the streets more frequently than other advanced democracies.
Similar to the indexes of the changed sizes of election loser and nonvoter, I construct
a variable of the changed level in mass political movements. The changed level of mass
political movements, that is mass threat, can be formulated as follows:
21 In

contrast, chapter 3’s individual level dependent variable of political trust may need
more refined measure of W as described in chapter 3. This is so because each individual
respondent can be better understood in a more refined contextual measure of the winning
coalition than in a rough original measure of W.

150

mass = masst − masst−3
This variable of mass threat can tell us whether leaders face increasing or decreasing
mass threats from the changed level of mass political events.

5.4.2.5

Free Press

The free press measure is constructed from Freedom House’s (2007) measure of press freedom.
Like Bueno de Mesquita and Smith’s construct of coordination good22 , two separate time
periods of press freedom are combined into a single variable. From 1980 through 1988,
Freedom House has separate scores for print media and broadcast media. For the later years
1989 through 2000, more direct measure of press freedom can be found on a 3-point scale (0
= “not free,” 1 = “partially free,” 2 = “free”). For the purpose of analysis with an extended
time frame, I create averaged values of two separate scores of broadcast and print media
from the earlier time period and combine this average with the score from the later time
period of post-1988.

5.4.3

Covariates

Other than main independent variables, I need to control for some confounding effects on
both dependent variables of mass political movement and leader survival.

5.4.3.1

Mass Political Movements

When examining the causes of mass political events, I create a lagged variable of mass political movements since previous mass movements are expected to have a strong influence
on subsequent events. I control for unexpected natural disasters that are beyond a leader’s
control. Using data on earthquakes, I employ Brancati’s (2007) data complied from the
22 Coordination

goods is referred to as public goods that help people coordinate and organize for mass political movements (Bueno de Mesquita and Downs, 2006). Other coordination
goods include transparency and easy communication.
151

Centennial Earthquake Catalog (Engdahl and Villaseñor, 2002). From 0 for no major earthquake to 3 for the most disastrous quakes, the earthquakes beyond 5.5 on the Richter scale
are recorded since 1975. The direct measure of the quake’s intensity is independent of the
government’s preparedness for disaster and its efficiency in dealing with disasters. With the
independence of the effect of quakes, I control for population size, economic growth, and
income (per capita GDP). All these control variables come from the World Bank’s (2005)
World Development Indicators.

5.4.3.2

Leader Survival

When assessing the consequences of mass threat and electoral politics on leader survival, I
control for selectorate size since selectorate size is the denominator of the leader’s institution
of the loyalty norm (W/S). As the leader can pick her supporters from a larger pool, the
loyalty norm can be secured by her manipulation of the selectorate, thereby improving her
survival.
Other control variables include leader’s age, economic growth, and income (per capita
GDP). When autocrats age and become ill, their institution of small winning coalition cannot expect further private goods provision and thus the risk of deposition increases. The
Archigos data provides the information on leader’s age. Economic growth and income may
compensate for the age effect. Natural resources of oil exports and international agency’s aid
are also considered as control variables. These free resources are available from the World
Bank’s (2005) World Development Indicators. The variable Oil measures net fuel exports
and imports that is calculated as a percentage of merchandise exports and imports. The
variable Aid is measured by the Official Development Assistance from the World Bank’s
indicators.

152

5.5

Empirical Results

As this chapter consists of two stages of analysis to test the aforementioned hypotheses,
two sub-sections for each stage are discussed below. The first sub-section deals with the
causes of mass political movements, and the second sub-section covers the consequences of
mass threats and electoral politics on leader survival. To control for panel data problems, I
estimate fixed-effects models that include country-specific dummy variables.

5.5.1

Mass Political Movements, Nonvoters and Free Press

From Table 5.4, the hypotheses can be empirically confirmed. Compared to election winner’s
size, nonvoter’s increased size positively affects the level of mass political movements from
Model 1 through Model 3, which supports for the hypothesis H1.23
As Hypothesis H2 theorized regarding de-mobilized nonvoters, increased nonvoters do
not increase the level of mass political movement when faced by a large-coalition leadership.
The statistically significant and negative coefficients of the interaction term of the W and
23 I

also include a variable of election suspension as a dummy variable for the robust
tests. All the models including the dummy variable of election suspension show positive
effects of ∆Nonvoter on mass movement while some models turn out to have non-significant
∆Nonvoter and W. All the interaction terms appear to have the same directions and statistical significance as the results found in Table 5.4. The inclusion of the variable election
suspension, however, may raise some doubts because the variable Winning Coalition (W)
directly captures the definition of election suspension to a great extent (cf. Kennedy, 2009).
The correlation of the W and the variable Election Suspension is -0.630.
Theoretically speaking, including this variable in the models does not seem to be necessary. After military coups’ extended rule stops scheduled elections, for example, the increased
size of nonvoter reflects the previous winner’s frustration, and this alienation may become
non-voting behavior under the new elections by the military. Therefore, we can also expect
counter-revolutionary movements in the military’s extended rule. Another example can include small-coalition leader’s election suspension. For the loyal supporters for the leader in
the small W, the increased size of nonvoters up to 100% resulted from election suspension
may not mean that winner’s political alienation come into play regarding mass movement
immediately. Rather, election suspension reflects aggravated alienation and frustration of
existing nonvoters and losers who even now lose their future expectation of electoral participation as their civic expression. In both cases, the ∆Nonvoter measure does not cause any
biased measurement on political alienation due to the excluded variable of election suspension.

153

nonvoters’ size change imply that nonvoters’ dissent expressed by increased nonvoters turns
out to be dampened by the increased winning coalition size (H2 ). This is illustrated in Figure 5.1 below. For example, elections in China were based on a plebiscitary state-controlled
system in which the only serious dissidents or challengers expressed their discontent by abstaining from voting. This was a common phenomenon in other former communist countries
(Jacobs, 1970; Linz, 1978b; Hermet, Rose and Rouquie, 1978). After being transformed to a
limited-choice election system, the Chinese Communist Party election allows for a turnout
boost, departing from the dissidents’ vote avoidance (Shi, 1999). In doing so, these “departicipated” voters became ‘one-party’ voters, similar to the framework in Table 5.1 for
non-partisan voters (Roeder, 1989). Moreover, they could punish corrupt local officials,
thereby elevating their internal efficacy. And, the level of revolutionary movement could be
naturally suppressed in the electoral politics at least to a certain extent.
[Figure 5.1]
These empirically supported hypotheses imply that advanced democracies where voter
turnout drops often reflect political apathy and satisfaction that could secure leader’s tenure
(see Figure 5.1). However, small-coalition leaders would not ignore the increased size of
nonvoters and take these vanishing voters as a result of citizen’s frustration and relative
depravation (H1 ). This is because the frustrated citizens feel strong doubt about the effects of elections as a legitimizing tool. To many autocrats, (even nominal) elections can
‘buffer’ societal and political conflicts derived from frustration of the alienated either from
economic or from political issues (Blaydes, 2011; Brownlee, 2007). For example, the plebiscitary one-party system in sub-Saharan Africa encourages a high degree of electoral participation through which “the regime employs a party machine to distribute patronage to a fairly
wide array of economic and regional interests” (Bratton and Van de Walle, 1997: 78).
Regarding election loser’s increased size, which represents a more competitive election,
the increased size of election losers does not impact the level of mass political movements.

154

This suggests that losers may postpone their unachieved demands in either policies or in
patronage until the next round of elections. In doing so, election losers may prefer to stay
in the polls, rather than to protest in the street. This scenario is plausible when electoral
losers’ demands are considered as achievable either through the incumbent’s policy changes
or through substantial possibilities of the opposition’s winning in the next election. The
competitive elections faced by large-coalition leaders would make the incumbent leaders
and parties more accountable to the constituencies including election losers. Therefore,
mass political events are not much expected in this situation, as revealed by the negative
significant sign of the interaction term of the W and loser’s change in Model 2 in Table 5.4.
Another possible scenario regarding the effect of electoral competition on mass political
movements can be found in inter-ethnic conflict situations around electoral politics. In an
interesting work on votes and violence in India, Wilkinson (2004) argues that high electoral
competition and party fractionalization (at least 3 and more parties in the Indian electoral
context) results in less ethnic violence because the state governments and the majority party
politicians will try hard to appeal to the minorities’ votes to win upcoming elections by
providing public security. That is, “[i]f minorities are pivotal to electoral outcomes, politicians
will increase the supply of security and prevent riots in order to attract their votes.” The
importance of minority’s votes to the party in power indicates that a leader or the party
in power needs to consider increasing the size of the winning coalition (W). Increased size
of W with minorities’ inclusion in India, therefore, results from the leader’s responses in
policy changes in terms of protecting the minority’s security. Through the increased election
loser’s size, India’s Muslim minority is protected, and frustrated Hindu majority’s riots and
protests against state governments can be prevented.
Unexpected natural disasters such as earthquakes appear to positively influence antigovernmental activities. With increased chances of coordination through gathering in shelters and refugee camps, protests and anti-government demonstrations facilitate the downfall
of authoritarianism and incumbent presidents (Preston and Dillon, 2005; Bommer, 1985).

155

Achen and Bartels (2004) similarly contend that “voters regularly punish governments for
acts of God, including droughts, floods, and shark attacks . . . The electorate will take
out its frustrations on the incumbents and vote for out-parties” (1). Yet, the earthquake’s
coefficient does not appear to be significantly different from 0 at the 0.05 level in both Model
1.B and Model 3, holding three demeaned variables of Free Press, GDP per capita, and Population constant at their means while holding the other control variables constant at 0.24 In
the interaction terms of earthquakes and the W, large winning coalition countries such as
advanced democracies do not have any effects, and it suggests that they are not vulnerable
to these unexpected disasters that facilitate coordination of protests in the small winning
coalitions such as autocratic countries (Quiroz and Smith, 2010).
Population size appears to be positively associated with mass political movements in
Model 1. The more population a country has, the more likely a critical number of dissidents
can get a demonstration off the ground (cf. Marwell, 1970). This appears to contradict what
many collective action theorists argued for the positive relation of small group and collective
action, if one consider a country’s population as a whole (Olson, 1974); but empirically, more
population can provide more resources in terms of creating a critical mass. For example,
24 The

demeaning process is adopted here to avoid too much multicollinearity in the interaction models (Model 2 and Model 3) that seriously inflates the standard errors, and makes
the estimated regression coefficient unreliable. By using the command -collin- in Stata
and looking at reported Variance Inflation Factor (VIF), I detected these three variables that
cause multicollinearity. With many interaction terms, the models are inherently vulnerable
to the multicollinearity issue. After demeaning these three variables, most variables stay
under the 10 VIF, which indicates no serious collinearity. The VIFs are reported in the
Appendix I.
Unlike Bueno de Mesquita and Smith’s (2010) finding, the effect of Earthquake does not
appear to be significantly different from 0. It suggests that although the variable Earthquake
may become to be significant if all the control variables are hold constant at 0, the corrected
standard error by demeaning does not still make the coefficient of Earthquake be statistically
significant.
The standard error of Earthquake before demeaning is 0.082, which is greater than the
reported S.E. in Model 2.B. It suggests that Bueno de Mesquita and Smith’s (2010) models
for mass political movements may experience high multicollinearity issue. The potentially
inflated standard errors of the variable Earthquake in their models, however, still make the
variable significant.

156

more populous Afro-Americans have become involved in more social movements than other
minority groups in the United States (Oliver and Marwell, 1988). However, after controlling
for Free Press and Earthquake, Population becomes insignificant in Model 2.
[Table 5.4]
The key independent variable of Free Press appears to increase the level of mass political movements, as previous studies show (Bueno de Mesquita and Smith, 2010).25 The
interaction term of the W and Free Press suggests that increased freedom of press will not
lead to more mass political events to large-coalition leaders. As large-coalition leaders would
prevent mass threats by increasing the size of W that could include the alienated nonvoters
into the winning coalition, any, if not all, potential dissidents may have less incentive to get
coordinated by free press for mass mobilization.
In Model 3, the interaction term of Nonvoter and Free Press supports Hypothesis H3a.
While election losers as ‘allegiant voters’ are participating in electoral politics with the
expectation of being winners in the next round of election, the insignificant coefficient of the
interaction term of the changed size of loser and Free Press suggests that freedom of press
would help only to update loser’s information regarding the parties and candidates, rather
than facilitating political movements. Nonetheless, nonvoters’ size mainly consisting of the
alienated non-partisans is sensitive to the level of freedom of press in their country. With
the coordination good of a free press, the frustrated nonvoters can be politically aroused and
coordinated for mobilization of mass political movements.26
25 Free

press is regarded as an important coordination good for the citizens to coordinate
and organize. As the data covers from 1962 to 2000 (for Model 1) and from 1980 to 2000
(for Model 2 & Model 3), the role of social network coordination in revolutionary movements
that are currently observed in the middle east is not a concern for this chapter. Although
it seems important now and this study also has some implication for this, electoral politics
derived from nonvoters and potential dissidents from election losers is the focus of this study.
26 In an extra model, I could test how these three variables of freedom of press, alienated
nonvoters, and the winning coalition institution. If a free press is provided by small-coalition
leaders, the type of mass political movement is also likely to result in a revolutionary movement when the alienated nonvoters increase. This is so because the politically alienated mass

157

[Figure 5.2]
Finally, one possible concern is an endogeneity of electoral politics and the level of mass
political movements. As mass political events may affect the level of voter turnout, and then
the levels of election competition and participation may vary, electoral politics is influenced
by the level of mass political movements. To lessen this concern, I use lagged measures rather
than the contemporaneous measures for both the concepts of competition and participation.

5.5.2

Leader Survival, Revolutionary Threats and Nonvoters

As leader survival can be analyzed as the hazard rate over time in office and the decline
in the risk of deposition can also be expected in small-coalition institutions, a revised Cox
proportional hazard model is adopted in the second stage of the analysis. The Cox proportional hazard model is based on a subtle assumption of proportional hazard over time
that should be tested to reflect the effects of election loser’s and nonvoter’s changed sizes on
leader survival. The assumption of proportional hazard is that the effect of the independent
variables is constant over time. That is, over the time of a leader’s tenure, independent
variables have standard effects. However, we see the effects of both changed sizes should be
stronger over time, as the outsiders to her winning coalitions may be mobilized for protest
after elections at which moment the leader does not realize the changes seriously or even
in the small-coalition institution could prefer to have new regimes as well as new leaders if
they can threaten the leader’s deposition through the coordination good of free press. In
the case of a large-coalition institution, in contrast, the elevated level of the revolutionary
movement derived from the combination of increased nonvoters and free press is expected to
be weakened.
The three way interaction term for nonvoters is illustrated in Appendix J.1. As illustrated,
more substantial changes of the revolutionary movements can be found in a small-coalition
institution. This suggests that a small-coalition autocrat would tend to suppress any potential coordination goods that can facilitate revolutionary movements easily. The risk of
deposition, therefore, is expected to be significantly associated with the increased size of
the alienated nonvoters and its effect of revolutionary movement. This is the topic that the
second stage analysis in the next sub-section will cover.

158

think of the increased nonvoters as a favorable sign to her tenure (Powell, 1986a).27
I ran the global test of proportionality, based on Schoenfeld residuals (Box-Steffensmeier,
Reiter and Zorn, 2003; Box-Steffensmeier and Zorn, 2001). The model specifications in the
tables reflect a significant level of disproportionality. Harrell’s rho test on each covariate
suggests that the changed sizes of election losers and nonvoters have a changing impact over
time (p < 0.05). In order to control for the effect of disproportionality, time-by-covariate
interaction terms are included in the proportionality models in Table 5.5. The offending
covariates found in the Harrell’s rho test are interacted with the natural log of leader’s time
in office and both variables are specified to make additional disproportional models.
Three empirical questions are discussed concerning the relationships between mass threats
and leader survival. First, in order to examine leader survival, I show how the institution
of winning coalition and the nature of electoral competition and participation can be key
factors in determining whether leaders survive. As discussed earlier, the nature of electoral
competition and participation can be considered key origins for a country’s mass politics
that should be based upon its institutional contexts. The following tests show how electoral
politics affect a leader’s longevity and the extent to which mass threats moderates the effects.
I show that the effect of this mass threat measure on leader survival is conditioned by the mass
public’s electoral participation and competition as well as by “a leader’s winning coalition
institution and her access to free resources.” (Bueno de Mesquita and Smith, 2010: p.939)
First, an increase in the size of nonvoters increases the risk of deposition in small-coalition
institutions, but not in large-coalition ones. A one percentage point increase in the size of
nonvoters over the previous three years increases the leader’s risk of deposition by about 3%
in Model 4 and Model 5 (exp(2.703×.01) = 1.027 & exp(2.667×.01) = 1.027, respectively).28
Yet, an increasing size of nonvoters has decreasing effects on the risk of the leader’s survival
27 Among

the more industrialized countries in his data, Powell contends that high level of
voter turnout has association with less durable executive’s tenure (37).
28 As appeared in Appendix H, The big increase of 5% of nonvoter’s size as a dummy
variable results in 40% increase in the risk of leader’s deposition, on average. (exp(.337) =
1.401).

159

in large-coalition institutions. For example, when the winning coalition is at maximum as
1, Model 7 and Model 8 estimate that there is about 3.5% and 3.1% decrease in the risk
of deposition, respectively (exp(−8.206 × .01 + 4.643 × .01) = .965 & exp(−7.271 × .01 +
4.111 × .01) = .969). As Figure 5.2 illustrates, the winning coalition’s decreasing effect of
the hazard of deposition seems to be clear.
[Figure 5.2]
Regarding the effect of election losers’ changed size on leader survival, I expect an opposite
effect to the nonvoters’ effect. That is, increased size of election losers are expected to dampen
the effect of mass political movements on leader survival by broadening the political arena
in which the dissents from election losers make their “voice” at the voting booth, not in the
street. However, Model 4 in Table 5.5 indicates that a one percentage point increase in the
size of election losers (or electoral competition) over the previous three years increases the
risk of deposition by about 7% (exp(6.845 × .01) = 1.071). In Model 5-8, the effect is not
statistically significant at the 0.05 level.
Regarding free resources of Oil and Aid, the variable Oil only appears to be significant
across Model 5 through Model 8. With more free resources, a leader is expected to extend
her tenure more securely than one who lacks these resources. Income growth decreases the
risk of deposition to small-coalition leaders while the effect of income growth is not significant
to large-coalition leaders.
As Bueno de Mesquita and Smith (2010) contend from their analysis of mass threat to the
risk of deposition, my data also shows that a one standard deviation increase of revolutionary
political movements increases the risk of deposition by about 25-35%. Unlike the minimal
effects of W in reducing mass threats to leaders shown in their work, across Model 6 through
Model 8, the interaction term appears to be consistently significantly different from 0 and
shows a decreasing effect of the increased size of the W. Moreover, the decreasing effect
size is also distinguishable from 0. This is so because I used the revised variable of mass

160

political movements to have a revolution-weighted movement variable for my analysis, and
hence the variable

Mass Threat (now represented as revolutionary threat) seems to be

more sensitive to the size of W. The increased size of W, for example up to W=1, decreases
the risk of a leader’s deposition who faces revolutionary movement by 25%, according to
Model 8 (exp(.859 − 1.147 × 1) = .750). However, increasing the W’s size up to 0.75 does
not provide any benefits to the leader (exp(.859 − 1.147 ∗ .75) = .999).
[Table 5.5]

5.6

Conclusion

A leader faces many threats such as revolutionaries, protesters, and anti-government demonstrators. While considering these concerns, small-coalition leaders are more attentive to
legitimizing the existing regime by being re-elected in corrupt elections or even by providing
limited-choice electoral systems, rather than focusing resources on suppressing mass political events. This may be true especially if some election outcomes may indicate the mass
public’s frustration and even anger, thereby resulting in vanishing voters. Even in pluralist
democracies, large-coalition leaders pay heed to the effect of institutional reform when they
consider the significant interaction effect of the winning coalition in mass movements and
electoral politics.
From the two stages of leader survival analysis, mass threats to leader surviva, more
exactly revolutionary threats, are largely derived from the electoral (de)participation, rather
than electoral competition. While both dimensions of democracy defined by Dahl matter in
explaining democratic transition in the waves of democracy, revolutionary threats associated
with leader survival appear to have strong association with the number of nonvoters.
The empirical data shows that all of the listed hypotheses are supported. Electoral
(de)participation and revolutionary threats are key predictors to explain the leader’s survival
while free press affects indirectly via mass threats. And, these determinants have different

161

effects on leader’s survival according to the winning coalition institution. Increased size of
winning coalition dampens political alienation and mass threats in the short run. 29
29 Regarding

the long-term effects of increased size of winning coalition, an aggregate level
data is not appropriate to test how individuals perceive political institutions, which result
from a leader’s change of winning coalition change over time.

162

Tables & Figures for Chapter 5
Table 5.1: Party Identification and Voting Behavior in Elections
Partisan

Non-partisan

Winners or Losers

Non-partisan Voters

(Allegiant/Loyal Voters)

(Pragmatic/Policy-oriented Voters)

Partisan Nonvoters

Non-partisan Nonvoters

(Inefficacious/Boycotting Nonvoters)

(Alienated Nonvoters)

Voter

Nonvoter

163

Table 5.2: Hypothetical Effect of Voters/Nonvoters on Mass Political Movements with the
Low or High Press Freedom Interaction
Main
∆Winner

–

∆Loser

+

∆Nonvoter

+

Winning Coalition Size (W)

–

Free Press

2-way Interaction

+

W × ∆Loser

–

W × ∆Nonvoter

–

∆Loser × Free Press

+

∆Nonvoter × Free Press

+

164

Table 5.3: Hypothetical Relationship of Leader Survival: The Effect of Nonvoters and Mass
Threat on Leader Deposition
Effect on Leader Deposition
Mass Threats

+

∆ Nonvoter

+

Winning Coalition Size (W)

+

Mass Threats × Nonvoter’s Size

+

W × Mass Threats

–

W × ∆Nonvoter

–

165

Table 5.4: The Effect of Nonvoter Size on Mass Political Movements
Model 1

0.271∗∗

(0.020)

(0.020)

-0.265∗∗

-0.180†

-0.178†

(0.098)

(0.105)

0.353∗∗

0.301†

0.897∗∗

(0.156)

(0.316)

0.078

-0.103

(0.310)

(0.910)

0.026

0.277∗∗

(0.035)

(0.086)

0.066†

0.106

(0.034)

(0.079)

-0.062

-0.100

0.138

(0.038)

(0.211)

(0.092)

-0.009∗∗

-0.010∗∗

-0.010∗

(0.002)

(0.002)

(0.005)

0.184∗∗

0.134

-0.484∗

(0.045)

∆Loser

0.273∗∗

(0.112)

∆Nonvoter

0.282∗∗

(0.058)

Winning Coalition Size (W)

Model 3

(0.015)

Lagged Mass

Model 2

(0.101)

(0.214)

0.259
(0.245)

Free Press
Earthquake
Ln(GDPpc)
Growth
Ln(population)
2-Way Interaction
W * Ln(GDPpc)

-0.503
(0.330)

W * Growth

<0.000
(0.008)
1.443∗∗

W * Ln(population)

(0.528)
W * Earthquake

-0.059
(0.103)
Continued. . .

166

Table 5.4 (cont’d)

Model 1

Model 2

Model 3
-1.433∗∗

W * Nonvoter

(0.544)
W * Loser

-0.048
(1.247)
-0.440∗∗

W * Free Press

(0.138)
0.806∗

Nonvoter * Free Press

(0.342)
Loser * Free Press

0.134
(0.651)
0.133

0.133∗

(0.037)

Intercept

0.185∗∗

(0.062)

(0.066)

N

4185

2439

2439

Adj.-R2

0.280a

0.350a

0.361a

Fixed effects

149-country 144-country 144-country

a

Adjusted-R-squared was obtained by -areg- command in STATA (version
11.2).
†p<0.10, * p<0.05, ** p<0.01

167

Table 5.5: The Effect of Nonvoter Size and Mass Threat on the Risk of Leader’s Deposition
Model 4

-0.795

(0.363)

(2.339)

(2.313)

(2.358)

-1.442∗∗

-1.710∗∗

-1.364∗∗

-1.385∗∗

-1.346∗∗

(0.263)

(0.296)

(0.307)

(0.313)

0.024∗∗

0.026∗∗

0.038∗

0.033†

0.042∗

(0.007)

(0.018)

(0.018)

(0.018)

2.703∗∗

2.667∗∗

-0.078

4.643∗∗

4.111†

(0.912)

(0.885)

(1.360)

(2.164)

6.845∗∗

0.585

-7.355†

4.967

7.729†

(1.180)

∆Loser

-1.823

(0.666)

∆Nonvoter

-1.254

(0.005)

Age

2.022∗∗

(0.202)

Selectorate

1.751∗∗
(0.272)

Winning Coalition (W)

Model 5

(4.699)

(4.074)

(3.250)

(4.233)

4.031

7.041∗∗

∆Loser_ln(t)

Model 6 Model 7 Model 8

(2.554)
0.239∗∗

0.911∗∗

0.880∗∗

0.859∗∗

(0.068)

(0.082)

(0.181)

(0.181)

(0.190)

-0.065∗

-0.135∗

-0.119∗ -0.193∗∗

(0.028)

(0.056)

(0.054)

(0.058)

0.026∗

0.059∗

0.051∗

0.093∗∗

(0.013)

∆Mass Threat

(2.152)

0.311∗∗

(0.024)

(0.023)

(0.026)

-0.021

-0.093†

-0.102†

-0.069

Oil
Oil_ln(t)
Aid as %GDP

(0.016)

(0.055)

-0.066∗

-0.061∗

-0.068∗

(0.014)

(0.031)

(0.031)

(0.031)

0.003

0.037

-0.278

-0.299†

-0.251

(0.039)

Ln(GDPpc)

(0.054)

-0.037∗∗

(0.009)

Growth

(0.053)

-0.021∗

(0.071)

(0.171)

(0.171)

(0.177)

2-Way Interaction
W * Age

-0.020
(0.027)

-0.011 -0.104∗∗
(0.027)

(0.035)

Continued. . .

168

Table 5.5 (cont’d)

Model 4

Model 5

Model 6 Model 7 Model 8
0.043∗∗

W * Age_ln(t)

(0.010)
0.195∗

0.287∗∗

(0.098)

W * Oil

0.166†
(0.096)

(0.102)

-0.088†

W * Oil_ln(t)

-0.072 -0.141∗∗

(0.046)

0.082

(0.074)

(0.077)

0.064

0.053

0.070

(0.050)

(0.049)

(0.050)

0.546∗

0.555∗

0.563∗

(0.250)

W * Ln(GDPpc)

0.130†

(0.074)
W * Growth

(0.049)

0.114

W * Aid

(0.045)

(0.250)

(0.256)

-1.276∗∗ -1.288∗∗ -1.147∗∗

W * ∆Mass Threat

(0.280)

(6.073)

-8.206∗∗

-7.271∗

(2.338)
∆Loser * ∆Mass Threat

-6.118

(4.570)
W * ∆Nonvoter

(0.300)

-1.928

W * ∆Loser

(0.276)

(3.365)

-0.732
(1.331)

Log-likelihood
χ2

2894

1515

0.942

(0.646)

N

(1.854)

2.051∗∗

∆Nonvoter * ∆Mass Threat

-1.673

(0.979)

1515

1515

1515

-2021.998 -1043.064 -1004.333 -1007.46 -996.257
183.873

†p<0.10, * p<0.05, ** p<0.01

169

118.555

196.017

189.764

212.171

0.2
0.0
-0.2
-0.4

W=0
W=.25
W=.5
W=.75
W=1

-0.6

Level of Revolutionary Movements

0.4

0.6

Figure 5.1: Predicted Effects of Electoral Mobilization on Revolutionary Movements (2-way
interaction)

-0.4

-0.2

0.0

0.2

Changes in Nonvoter's Size (Changes in Electoral Mobilization)

170

0.4

0

.5

1

1.5

2

Figure 5.2: Estimated Baseline Cumulative Hazard (Stratified by W in Model 7)

0

10

20

30

40

_t
W=0
W=.50
W=1

171

W=.25
W=.75

50

Chapter 6
CONCLUSION

In the nineteenth century, Karl Marx forcefully contended that the proletariat, the modern
working class, were not only victims but also could be actors against the development of
the bourgeoisie, i.e., capital (Marx, Engels and Jones, 2002). Marx focused on the proletariat’s role as a historical force facilitating revolutionary movements against expanding
capitalist production and control of the state. As is widely known, Marx’s prediction was
not fulfilled, but his idea of the basic mode of dialectical analysis seems to be right. For
him, the struggles of ordinary people are constrained by institutional arrangements, but the
institutional logic can be challenged by mass threats. Historically, facilitating institutional
and societal arrangements favorable to revolutionary movements were lacking due to conditions that had minimized economic alienation and thus relative deprivation, at least within
industrial societies: “the spread of imperialism helped to produce the surpluses that would
raise working-class material standards in the mother countries; the balkanization of modern
industry helped to fractionalize the working class; new institutions such as public education
helped to ensure capitalistic ideological hegemony” (Piven and Cloward, 1977: x).
The dialectical process between people’s struggles and societal and economic institutions
during the course of capitalist development sheds light on our understanding of political
leaders’ stability. In this dissertation, I argue that mass political movements, and more
importantly, mass political attitudes and voting behavior, are fundamental causes to explain
leader survival. But, mass attitudes and behavior interact with institutional forms in complex
ways to affect leader survival.

172

6.1

Concluding Remarks

Interestingly, leader survival has been addressed mainly by students of international relations, while regime survival and government survival have been dealt with by many comparativists. In order to complement what is lost by ignoring domestic politics, political trust
and electoral behavior have been used here as intervening variables in determining leader
survival. In Chapter 3, mass political attitudes toward political institutions are affected by
electoral politics characterized by the election winners’, losers’, and non-partisans’ perceptions of the system. In hierarchical (or multi-level) models using multiple public opinion data
sets, political trust measured by confidence in the army, police, parliament, and president is
a function of a voter’s partisan status.
More specifically, I have theorized an interactive relationship of partisan status with winning coalition and selectorate size. For non-winning voters in elections, a small winning coalition size does not provide a satisfactory institutional arrangement, because a non-winning
voter’s consumption of state resources is limited by the leader’s support for her winning
group with private goods. Furthermore, the limited resources provided to non-winning voters in a small winning coalition divides the alienated groups into two types – election losers
and non-partisans – according to their different levels of partisan status and their different
perceptions of the winning coalition institution. I will discuss how election winners, losers,
and non-partisans perceive their political system in the context of winning coalition and
selectorate size.
• Election Winners: In a small winning coalition, winners get most benefits from a leader
who can provide substantial and stable private goods. Bueno de Mesquita et al. (2003)
contend that if with limited resources a leader can satisfy her coalition, she is able to
meet winners’ demands and thus secure the loyalty norms from her supporting group
in a small size of winning coalition given a fixed size of the selectorate. By providing
sufficient private goods to her coalition, the leader is able to prevent winners from

173

defecting to her challengers. Moreover, the leader who faces revolutionary threats can
utilize free resources such as foreign aid and natural resources in order to reduce the
extant small size of the coalition (Bueno de Mesquita and Smith, 2010). Hence, in
such a small coalition, winners’ support for the leader is directly associated with policy
benefits and the provision of private goods, rather than with diffuse consent to political
institutions.
In responding to demands for liberalization and democratization, a leader’s effort to
increase the size of winning coalition makes the leader’s tenure unstable. As the leader
has limited resources, election winners cannot obtain the same level of private goods
(per individual) in a large winning coalition institution. Decreased benefits raise the
chances that winners will defect to her challengers who promise to provide more welfare.
This leads winners to have lower specific loyalty to leaders, which is not necessarily
extended to diffuse political dissent from the system. In Hypothesis H1a, I hypothesized
that election winners have decreasing levels of political trust in the system as the shortterm loyalty norms (W/S) increase. This hypothesis is disconfirmed.
• Election Losers: Other than election winners in the winning coalition, non-winning
groups such as election losers and non-partisans are temporarily or persistently excluded or disadvantaged in accessing the benefits of private goods provided by a leader’s
public policy making. Election losers affiliate with opposition parties. With their affiliation with existing parties and thereby their tendency to accept the political institutions such as the electoral system, losers are well aware of which institutions work
for their benefits in terms of political representation and policy implementation (cf.
Anderson et al., 2005). That is, losers are more sensitive to the imminent and current
size of the winning coalition and the loyalty norms (i.e., the ratio of winning coalition
to the selectorate).
Consequently, a small size of short-term winning coalition arrangements aggravates

174

the extant losers’ dissent from the political system. This is so because losers’ strong
affiliation with opposition parties reduces their prospect of winning an election in a
small size of coalition and the losers’ perception of political institutions is mainly
dominated by the short-term institutional context of the winning coalition.
Hence, a more expanded size of a winning coalition allows short-sighted losers to have
more chances of becoming winners, which can be translated into diffuse trust in political
institutions at least in the short run. In other words, election losers are vulnerable to
a leader’s co-optation strategies such as increasing the winning coalition size. Losers’
short-sighted perception of the political system can be compatible with a short-term
winning coalition institution. Hypothesis H1b is that election losers/non-partisans have
increasing levels of political trust as the short-term loyalty norms increase. Empirically,
this hypothesis is confirmed, not for non-partisans, but for election losers.
• Non-partisans: The other non-winning group of non-partisans is prompted to politicize
when partisan independence is constrained by historical and long-term institutional
arrangements of the winning coalition, rather than by a short-term winning coalition
institution. Without having affiliated with a specific opposition party, non-partisans
tend to ‘exit the scene’ of electoral politics, and there exists no motivation for nonpartisans to be sensitive to the chances of becoming or joining the winning coalition
unless the perceived alienation is great enough for non-partisans to take remedial action
such as revolutionary movements (see Chapter 5). As non-partisans tend to be apathetic and indifferent to their leader and the political system from which they get few
of the spoils provided by the government, it is hard for non-partisans to be politicized
in the absence of persistent perception of political alienation.
In this context, non-partisans’ politicization depends on their sustained or persistent
political alienation resulting from a continuously increasing winning coalition and their
accessible memory of alienating coalition institutions, i.e., there being a large historical

175

winning coalition from which they are excluded. The cumulative size of the winning
coalition can activate non-partisans’ sense of exclusion and alienation, and the level of
non-partisans’ political alienation is worsened when the cumulative size of the winning
coalition expands: With a historically increased size of the winning coalition, nonpartisans feel more alienated from the political system that could have provided a
more inclusive system to losers with increased coalition size, and yet have excluded
non-partisans from it. Hypothesis H2b in chapter 3 was confirmed empirically, which
states that non-partisans have decreasing levels of political trust as historical loyalty
norms among electoral winners and losers increase. Non-partisans’ distrust of political
institutions is a direct effect of the interaction between partisan independence and
institutional memory, measured by accumulated size of the winning coalition.
In Chapter 4, South Korea’s impeachment process illustrates how selectorate politics
and partisan status affect differently the citizens’ perception of leader’s impeachment. With
individual-level survey data, this case study allows me to examine a more direct causal relationship between leader survival and the interaction effect of partisan status and individual
level loyalty norms. That is, as we may admit a potential limitation of the validity of the
loyalty norms measured only at the aggregate and institutional level, the individual level
loyalty norms of each partisan status address directly how winners, losers, and non-partisans
perceive the leader. South Korea’s unique experience of the impeachment process could
benefit from the experiment-like survey in 2004 South Korea that was done between the
17th general election and the final decision of the Constitutional Court on the impeached
President Roh. By employing a direct measure of the individual’s loyalty norms and the
perception of the impeachment from a timely-executed survey, I found that non-partisans’
politicization is not apparent in a sense that non-partisans’ impeachment support appears to
be moderately changing according to individual level’s loyalty norms toward the incumbent
leader and party.
I argue that as the conflict between the opposition parties of a conservative ‘old pol176

itics’ and the incumbent party of a progressive ‘young politics’ led to an unprecedented
impeachment process in South Korea’s constitutional history, this abnormal situation of an
impeachment process and the timing of the 2004 general election provided citizens with an
opportunity for voters to perceive increased internal efficacy. With increased internal efficacy in general, the impeachment process reduces non-partisans’ political alienation and
politicization, rather than aggravating them. Specifically, to citizens, the loyalty to the Uri
party allows for only a short-term loyalty norms since the party was founded in 2003. Consistent with the findings in disconfirming non-partisans’ politicization in Hypothesis H1b,
non-partisans who are constrained by individual-level loyalty norms in a short-term institutional context are not prompted to politicize. Thus, they are expected to appear less
responsive to the loyalty norms to the incumbent leader.
These findings are based on a specific understanding of the leader-party nexus in South
Korea. As South Korea’s case has revealed a strong nexus between a party and its leader,
partisans’ support for a party can translate well into the loyalty norms to their leader. Both
respondents who supported and did not support President Roh appear to support Roh’s
reinstatement when their loyalty toward Roh’s Uri party is at a higher level. However,
non-partisans’ de-politicization entails that their support for the reinstatement decreases
even when the non-partisans’ loyalty for the liberal and progressive Uri party increases.
Consequently, partisans, including both election winners and losers, show more extreme
changes of impeachment support while non-partisans appear to have a moderate change in
the perception of impeachment support.
Non-partisans’ sensitivity to the historical and long-term winning coalition institution
cannot be tested in Chapter 4 since an individual level survey data on the loyalty norms
does not provide an opportunity to measure the cumulative effect of the institutional arrangements on the loyalty norms. Without a measure of the historical arrangements, both
measures of the loyalty norm (one compared to all participating opposition parties) and
the comparative loyalty norm (one compared to two major opposition parties) show non-

177

partisans’ de-politicization in moderate changes of the impeachment support. Like the finding in Chapter 3’s multi-level cross-country analysis that non-partisans’ political trust does
not respond to the short-term institutional context, non-partisans’ impeachment support
in South Korea in 2004 appears to result in a de-politicization in the interaction of nonpartisans’ support for the short-lived incumbent party.
Chapter 5 concludes by examining the final causal relationship between political alienation and leader stability. This chapter tests whether and how nonvoting resulting from
non-partisan’s political alienation can affect the two key variables of mass political movements and leader survival. In this analysis, increased size of nonvoters induces more frequent
mass mobilization such as protests, anti-government demonstrations, riots, and revolutions.
More importantly, even controlling for mass threats to a leader resulting from mass mobilization, non-voter’s increase, as a proxy for increased political alienation, is associated with
greater risk of a leader’s deposition.
Consistent with the findings of non-partisans’ de-politicization in Chapter 3 and Chapter
4 in the short-term loyalty norms’ context, nonvoters appear to be less alienated in the presence of a larger winning coalition and small loyalty norms. Increased size of the short-term
winning coalition appears to dampen the positive effect of increasing the size of aggregated
nonvoters on mass political movements (Figure 5.1). In a more inclusive institutional arrangement of a large winning coalition, the alienated appear to suppress their persistent
perception of relative deprivation. In contrast, a free press can facilitate the effect of political alienation on mass threats to the leader, thereby leading to more politicization of
nonvoters. And its politicizing effect can be strengthened in the context of a small winning
coalition (Figure 5.2). That is, in a limited and constrained situation of information in a
small winning coalition, the alienated can mobilize with a small addition of freedom of the
press, thereby threatening the leader’s deposition.

178

6.2

Limitations and Extensions

Given that the relationship between political distrust/alienation and leader survival is hypothesized and tested here, I point out some related limitations below. Research on leader
survival mainly focuses on a cross-time analysis. In this context, individual-level data on
political trust is not easily available across the range of historical time in which leader survival can also be examined. Hence, my interest in the effect of citizens’ attitudes toward the
political system on leader survival required me to come up with an alternative measure of
political trust, which is non-voting as a proxy for political alienation. If we had more appropriate measures of political alienation in the light of voting behavior, they could provide
us with interesting new findings. Related to that, the measure of nonvoters should also be
revised with richer and more accurate information on election results. Especially, data on
the number of registered or eligible voters is not reliable across time and regions.
Another limitation involves data coverage on mass political movements. Since the years
covered range from 1980 through 2000 in the specific causal relation between political alienation and mass political movements in Chapter 5, the turmoils in the 1960s and the 1970s
could not be included. If free press data could be measured by appropriate proxies during this time period, more interesting results may be expected. Furthermore, since 2000,
the ‘internet revolution’ and ‘mobile revolution’ would result in different patterns of mass
mobilization. How to examine these effects in authoritarian regimes could be an appealing
subject in the future.
In this dissertation, I include only national elections where the “politically alienated"
would be more likely to be nonvoters. But in local-level elections it is more likely for the
alienated as well as voters on the whole to perceive the effect of their voting as relatively
substantial and thus to stimulate their participation in the decision-making process. In such
a situation, the alienated nonvoters are provided an opportunity to make their voice heard,
thereby being translated into voters. By focusing on local-level elections, I could examine
how nonvoters can turn into voters and compare both the changes of national elections and
179

local elections. Moreover, appropriate measures of nonvoters at both levels would provide a
chance to test a potential hypothesis that increased size of nonvoters at local level elections
is more associated with alienated citizens than one at national level elections.
Finally, the effect of leaders’ strategies on political alienation and mass threats could be
another extension of this dissertation. A leader can either reduce potential mass threats
by suppressing her constituents or buy off revolutionary threats by expanding public goods,
depending on the size of available free resources (Bueno de Mesquita and Smith, 2010: 946).
The effects of each strategy to manage mass threats end up either reducing or increasing the
size of winning coalition.
As was confirmed in Table 5.4 and Figure 5.1, this dissertation deals with variation of
nonvoters’ size, rather than the absolute size of nonvoters, considering the changes of the
nonvoter’s size as a variable to control for the contextual difference of the nonvoter’s size
in each country and each year. Yet, a closer examination of a direct effect of the size of
winning coalition on the absolute size of nonvoters could be useful in knowing how leaders’
strategies on political alienation and nonvoting work when facing mass threats. Possible
hypotheses include that as a leader increases the size of winning coalition, more electoral
participation is expected, and a decreasing number of nonvoters could lessen the intensity of
political alienation among the whole population.

180

APPENDICES

181

Appendix A
CHAPTER 3: CODING

• Political Trust. In the WVS and the East Asian Barometer, “(F)or each one, could you
tell me how much confidence you have in them: is it a great deal of confidence, quite
a lot of confidence, not very much confidence or none at all?”; In the AFB, “(H)ow
much do you trust each of the following, or haven’t you heard enough about them to
say?”; The EAB uses trust word instead of confidence. Respondents were coded on an
interval scale ranging 0 to 1.0.
• Partisan Status. In the WVS, “If there were a national/general election tomorrow,
for which party on this list would you vote?” For non-partisans, two answers of the
question, “I would not vote” and “I would cast a blank ballot”, were used; In the AFB,
“Do you feel close to any particular political party?” and “which party is that?” were
used, so non-partisans are unaffiliated citizens with any party; In the EAB, “Which
parties (or candidates for president if it was presidential race) did you vote for?” was
used. Then, among the missing observations, I used the questions of “Among the
political parties listed here (SHOWCARD), which party if any do you feel closest to?”
to impute the missing values.
• Perceived Personal Economy. In the WVS, “How satisfied are you with the financial
situation of your household?”; In the AFB, “Your own present living conditions?”; In
the EAB, “As for your own family, how do you rate your economic situation today?”
• Interest in Politics. In the WVS and the EAB, “How interested would you say you are
in politics?”; In the AFB, “How interested would you say you are in public affairs?”
Very (4); quite (3); hardly (2); not at all (1)
• Female. Male (0), female (1)

182

• Age. Actual age of respondent.
• Education. “What is the highest level of education you have achieved?” Respondents
were coded on a 0 to 8 scale, where 9 denotes the highest level of education.
• Historical W/S. See Appendix B
• Short-term W/S. 5-year mean of the recent adjusted W/S. The adjusted W was
calculated by refining the categories of all four indicators of the Polity data and
Arthur Banks’s data. Giving transitional period more weights, the W was re-coded:
Dual/Transition in XRCOMP was recoded as 0.5, rather than 1.0; Dual ExecutiveDesignation in XROPEN was recoded as 0.5, rather than 0; Factional and Transitional
in PARCOMP were recoded as 0.5, rather than 0; Military-Civilian regime type in
Banks’s polity02 variable was recoded as 0.5, rather than 0 (see Bueno de Mesquita
et al., 2003, p.134-5 for the original coding of W/S).
• ENPP. Effective Number of Parliamentary Parties. Source: Michael Gallagher’s Election Indices (http://www.tcd.ie/Political_Science/staff/michael_gallagher/ElSystems/)
and Matt Golder’s data (http://homepages.nyu.edu/~mrg217/elections.html).
• Disproportionality. The disparity between the distribution of votes at the election and
the allocation of seats. Source: Gallagher’s least square index
(http://www.tcd.ie/Political_Science/staff/michael_gallagher/ElSystems/Docts/lsq.php).
• Parliamentarism. Governance type. Direct presidential (0); Strong president elected
by assembly (1); Parliamentary (2). Source: The World Bank’s Database
(http://go.worldbank.org/2EAGGLRZ40).
• Unitarism. Average of Nonfederalism and Nonbicameralism. Federal (0); Semifederal
(1); Nonfederal (2). Strong bicameral (0); Weak bicameral (1); Unicameral (2). Source:
Gerring, Thacker and Moreno (2005) (http://www.bu.edu/sthacker/data.htm).

183

• Democracy Level. Average of Freedom House (0-10 scale) and revised Polity (0-10
scale). Source: Hadenius and Teorell (2005)
• Real GDP per capita (purchasing power parity US$), 2000. Source: Gleditsch (2002)
(http://privatewww.essex.ac.uk/~ksg/exptradegdp.html).

184

Appendix B
CHAPTER 3: HISTORICAL W/S FORMULA

The following formula is for the historical W/S measure;
t

Wt =
s=(t−99)

s − (t − 100)
100

For example, in order to calculate historical stock of W up to 1999 and up to 2003, I
used following equations;

1999

W1999 =
s=1900

2003

W2003 =
s=1904

s − 1899
100

∗ Rs =

1
2
100
R1900 +
R1900 + · · · +
R
100
100
100 1999

s − 1903
100

∗ Rs =

1
2
100
R1904 +
R1904 + · · · +
R
.
100
100
100 2003

185

Appendix C
CHAPTER 3: ROBUST CHECK FOR INFLUENCE CASE

The following models are estimated by using R package Influence.ME. This is originally a
MlwiN macro translated into R by R. Nieuwenhuis (Van der Meer, Te Grotenhuis and Pelzer,
2010).

Table C.1: Chapter 3: Multilevel Linear Model of Partisanship and W/S Controlling for
Influence Cases (Nigeria and Mali), Dependent Variable: Political Trust
Model C.1

Model C.2

-0.211***

-0.219***

(0.057)

(0.076)

-0.054

-0.040

(0.044)

(0.056)

-0.008***

-0.011***

(0.000)

(0.000)

0.001***

0.001***

(0.000)

(0.000)

0.001

-0.004†

(0.002)

(0.002)

0.018***

0.019***

(0.001)

(0.001)

-0.036

-0.135

(0.067)

(0.094)

-0.167

-0.005

(0.108)

(0.149)

0.003

0.022†

Individual-level Variable
Election Loser
Non-partisan
Education
Age
Female
Interest in Politics
Country-level Variable
Historical W/S
Short-term W/S
ENPP

Continued. . .
186

Table C.1 (cont’d)

Model C.1
(0.014)

(0.000)

0.022

-0.019

(0.047)

(0.093)

0.021

-0.023

(0.045)

(0.084)

-0.163***

-0.187**

(0.046)

(0.079)

-0.093*

-0.154

(0.041)

(0.097)

-0.062

-0.041

(0.050)

(0.064)

-0.084*

-0.107*

(0.039)

(0.047)

0.218***

0.202*

(0.070)

(0.089)

0.036

0.041

(0.053)

(0.065)

0.026***

0.027**

(0.008)

(0.011)

0.015**

0.019*

(0.006)

Post-Communist

-0.000

(0.000)

Latin America

(0.008)

0.000

Asia

0.000

(0.006)

Africa & Middle East

(0.003)

-0.005

Real GDP/cap

0.005†

(0.002)
Democracy level

(0.017)

0.002

Disproportionality

Model C.2

(0.008)

Cross-level Interaction Term
Loser*Historical W/S
Non-partisan*Historical W/S
Loser*Short-term W/S
Non-partisan*Short-term W/S
Loser*ENPP
Non-partisan*ENPP

Continued. . .

187

Table C.1 (cont’d)

Model C.1
0.003

(0.002)

0.731***

0.695***

(0.104)
Mali Influence

0.003**

(0.001)
Constant

(0.002)

0.003**

Non-partisan*Disproportionality

0.002

(0.002)

Loser*Disproportionality

Model C.2

(0.151)

0.929***
(0.128)

Nigeria Infuence

0.465***
(0.131)

USA Influence

0.941***
(0.177)

Variance of Random Effect
Country-level

0.004

0.003

Election Loser

0.002

0.002

Non-partisan

0.001

0.001

Country-level (alternative)

0.001

0.006

Election Loser

0.002

0.003

Non-partisan

0.001

0.002

Individual-level

0.040

0.043

AIC

-23,179

-13,608

BIC

-22,817

-22,000

Log-Likelihood

11,630

6,844

N. of Respondent

63,423

46,330

59

41

N. of Country
†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

188

Figure C.1: Influence Country Robust Check for Model 4

Two outliers, indicated as red triangle, are Nigeria (2003) and Mali (2002).

189

Country

Figure C.2: Influence Country Robust Check for Model 8

USA
NIG*
MLI
PHI
ZIM
ALB
JPN*
MAC
CHL
PAK
VEN
BOT
BNG
ROK
ARG
TAZ
SAF
TAW
JPN
SAF*
NAM
INS
SEN
MEX
ZAM
PER
IND
GHA
THI
MLD
TUR
ALG
CAN
ZIM*
SPN
TAN*
NIG
KEN
MZM
LES
MAW
0.00

0.02

0.04

0.06

0.08

0.10

Cook's Distance

* indicates additional country-year either from Afrobarometer or East Asian Barometer.

190

Appendix D
CHAPTER 3: COUNTRY CASES

Table D.1: Chapter 3: Countries (59 country-year samples)
Africa & Middle East

Asia

Latin America

Post-Communist

Industrial Society

Algeria

Bangladesh

Argentina

Albania

Austria†

Botswana

India

Chile

Bulgaria†

Canada

Ghana

Indonesia

Mexico

Czech Republic†

Denmark†

Kenya

Pakistan

Peru

Lithuania†

Finland†

Lesotho

Philippines

Venezuela

Moldova

France†

Malawi

South Korea

Poland†

Germany†

Mali

Taiwan

Slovenia†

Greece†

Mozambique

Thailand

Macedonia

Iceland†

Namibia

Ireland†

Nigeria*

Italy†

Senegal

Japan*

South Africa*

Luxembourg†

Tanzania*

Netherlands†

Turkey

Spain

Zambia

United Kingdom†

Zimbabwe*

United States
Portugal†

* includes two country-year samples;
† indicates missing countries due to the inclusion of perceived economy variable.

191

Appendix E
CHAPTER 4: BRANT TEST

Brant Test for the generalized ordered logit is shown below.
Table E.1: Chapter 4: Brant Test for Loyalty & Comparative Loyalty
Variable
chi2 p>chi2 df
All
22.06
0.055 13
Honam region
0.31
0.577 1
Youngnam region
9.31
0.002 1
Choongchung region
0.99
0.319 1
Kangwon region
0.92
0.337 1
Education
0.05
0.815 1
Female
0.45
0.501 1
Age
2.76
0.097 1
Trust in President
1.56
0.212 1
Trust in the Nat’l Assembly 0.70
0.404 1
Trust in Party
0.05
0.826 1
Loser
0.32
0.573 1
Non-partisan
1.43
0.231 1
Loyalty (Uri Party)
4.52
0.034 1

Variable
chi2 p>chi2 df
All
28.33
0.008 13
Honam region
0.05
0.826 1
Youngnam region
12.54
0.000 1
Choongchung region
0.28
0.595 1
kangwon region
1.12
0.289 1
Education
0.05
0.818 1
Fmale
0.76
0.385 1
Age
1.12
0.291 1
Trust in President
2.94
0.086 1
Trust in the Nat’l Assembly 1.34
0.247 1
Trust in Party
0.02
0.899 1
Loser
0.23
0.628 1
Non-partisan
1.85
0.174 1
Comparative Loyalty
8.21
0.004 1

192

Appendix F
CHAPTER 4: COMPREHENSIVE COMPARATIVE LOYALTY MODEL

All the parties are considered in the Comparative Loyalty variable by including all other
opposition parties, i.e., Democratic Labor Party, United Liberal Democrats, and People
Unity 21. The estimated models considering all 5 opposition parties are shown below.

Table F.1: Chapter 4: Partial Proportional Odds Models of Impeachment Support (Comparative Loyalty)
Model F.1

Model F.2

Model F.3

-0.624*

-0.613*

-0.622*

(0.293)

(0.293)

(0.293)

0.586*

0.589*

0.597*

(0.248)

(0.249)

(0.250)

0.243

0.272

0.283

(0.298)

(0.296)

(0.297)

0.429

0.416

0.412

(0.388)

(0.387)

(0.386)

0.004

0.001

-0.000

(0.053)

(0.054)

(0.055)

-0.111

-0.142

-0.129

(0.171)

(0.170)

(0.171)

-1.068***

-1.077***

-1.070***

(0.162)

(0.161)

(0.162)

0.455**

0.440*

0.428*

(0.173)

(0.173)

(0.174)

Opposed/(Strongly) Supportive
Honam regiona
Youngnam regiona
Choongchung

regiona

Kangwon regiona
Education
Female
Age

0.026**
(0.008)

Trust in President
Trust in Nat’l Assembly

Continued. . .
193

Table F.1 (cont’d)

Model F.1

(0.174)

(0.175)

0.533*

0.522*

(0.218)

(0.218)

(0.260)

0.299

0.222

-0.049

(0.236)

(0.231)

(0.289)

-0.307***

-0.315***

-0.342***

(0.041)

(0.042)

(0.054)

0.773***

0.782***

(0.195)

(0.196)

0.858**

0.839**

(0.280)

(0.283)

1.235

1.193

(0.755)

Comparative Loyalty (Uri Party)

0.267

0.515*

Non-partisan

0.256

(0.173)
Election Loser

Model F.3

0.233

Trust in Party

Model F.2

(0.762)

1987 Demo. Cohort (Born in 1953-68)b
1997 Fin. Crisis Cohort (Born in

1938-52)b

1950 War Cohort (Born in & before 1937)b
Loser*Comp. Loyalty

-0.036
(0.085)

Non-partisan*Comp. Loyalty

0.142†
(0.085)

Constant

0.503

1.143*

1.206*

(0.570)

(0.462)

(0.474)

-0.624*

-0.613*

-0.622*

(0.293)

(0.293)

(0.293)

-0.540†

-0.544†

-0.584†

(0.301)

(0.299)

(0.304)

0.243

0.272

0.283

Strongly Opposed/Opposed vs.
(Strongly) Supportive
Honam regiona
Youngnam

regiona

Choongchung regiona

Continued. . .

194

Table F.1 (cont’d)

Model F.1

0.412

(0.387)

(0.386)

0.001

-0.000

(0.053)

(0.054)

(0.055)

-0.111

-0.142

-0.129

(0.171)

(0.170)

(0.171)

-1.068***

-1.077***

-1.070***

(0.162)

(0.161)

(0.162)

0.455**

0.440*

0.428*

(0.173)

(0.173)

(0.174)

0.233

0.256

0.267

(0.173)

(0.174)

(0.175)

0.515*

0.533*

0.522*

(0.218)

(0.218)

(0.260)

0.299

0.222

-0.049

(0.236)

(0.231)

(0.289)

-0.307***

-0.315***

-0.342***

(0.041)

(0.042)

(0.054)

0.773***

0.782***

(0.195)

(0.196)

0.858**

0.839**

(0.280)

(0.283)

1.235

1.193

(0.755)

Age

0.416

0.004

Female

(0.297)

0.429

Education

(0.296)

(0.388)

Kangwon

Model F.3

(0.298)
regiona

Model F.2

(0.762)

0.026**
(0.008)

Trust in President
Trust in Nat’l Assembly
Trust in Party
Election Loser
Non-partisan
Comparative Loyalty (Uri Party)
1987 Demo. Cohort (Born in 1953-68)b
1997 Fin. Crisis Cohort (Born in 1938-52)b
1950 War Cohort (Born in & before

1937)b

Loser*Comp. Loyalty

-0.036
Continued. . .

195

Table F.1 (cont’d)

Model F.1

Model F.2

Model F.3
(0.085)

Non-partisan*Comp. Loyalty

0.142†
(0.085)

Constant

0.503

1.143*

1.206*

(0.570)

(0.462)

(0.474)

-0.624*

-0.613*

-0.622*

(0.293)

(0.293)

(0.293)

-0.540†

-0.544†

-0.584†

(0.301)

(0.299)

(0.304)

0.243

0.272

0.283

(0.298)

(0.296)

(0.297)

0.429

0.416

0.412

(0.388)

(0.387)

(0.386)

0.004

0.001

-0.000

(0.053)

(0.054)

(0.055)

-0.111

-0.142

-0.129

(0.171)

(0.170)

(0.171)

-1.068***

-1.077***

-1.070***

(0.162)

(0.161)

(0.162)

0.455**

0.440*

0.428*

(0.173)

(0.173)

(0.174)

0.233

0.256

0.267

(0.173)

(0.174)

(0.175)

Strongly Opposed/Opposed vs.
(Strongly) Supportive
Honam regiona
Youngnam regiona
Choongchung regiona
Kangwon regiona
Education
Female
Age

0.045***
(0.010)

Trust in President
Trust in Nat’l Assembly
Trust in Party

Continued. . .

196

Table F.1 (cont’d)

Model F.1

0.522*

(0.218)

(0.260)

0.299

0.222

-0.049

(0.236)

(0.231)

(0.289)

-0.452***

-0.464***

-0.483***

(0.058)

(0.059)

(0.076)

0.773***

0.782***

(0.195)

(0.196)

0.858**

0.839**

(0.280)

(0.283)

1.235

1.193

(0.755)

Comparative Loyalty (Uri Party)

0.533*

(0.218)
Non-partisan

Model F.3

0.515*

Election Loser

Model F.2

(0.762)

1987 Democracy Cohort (Born in 1953-68)
1997 Financial Crisis Cohort (Born in 1938-52)
1950 War Cohort (Born in & before 1937)
Loser*Comp. Loyalty

-0.036
(0.085)

Non-partisan*Comp. Loyalty

0.142†
(0.085)

Constant

-2.103***

-0.722

-0.664

(0.626)

(0.465)

(0.479)

Pseudo R-Square

0.297

0.294

0.297

Number of Cases

692

692

692

a Omitted
b Omitted

category is Seoul and Kyunggi Region
category is the 2002 Presidential Election Cohort (Born in & after 1969)
†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

197

Appendix G
CHAPTER 4: MULTINOMIAL LOGIT MODEL WITH IDEOLOGY
VARIABLE

Table G.1: Chapter 4: Multinomial Logit Models of Impeachment Support (Loyalty)
Strongly Opposed : Base Outcome
Strongly Opposed vs. Opposed
Model G.1
-0.635∗

Model G.2
(0.271)

-0.712∗∗

(0.276)

0.493∗

(0.223)

0.482∗

(0.224)

Choongchung

0.537

(0.341)

0.518

(0.346)

Kangwon

0.737

(0.471)

0.637

(0.478)

Education

-0.066

(0.059)

-0.076

(0.059)

Female

0.196

(0.181)

0.211

(0.183)

1987 Democracy Cohort

0.255

(0.205)

0.241

(0.208)

1997 Financial Crisis Cohort

0.180

(0.326)

0.158

(0.329)

1950 War Cohort

0.231

(0.866)

0.198

(0.854)

-0.823∗∗

(0.165)

-0.820∗∗

(0.166)

0.440∗

(0.187)

0.425∗

(0.189)

0.274

(0.183)

0.288

(0.185)

Loser

0.598∗

(0.282)

1.425

(0.935)

Non-partisan

0.351†

(0.205)

-0.642

(0.766)

(0.042)

-0.264∗∗

(0.076)

Loser*Loyalty(Uri)

-0.029

(0.128)

Non-partisan*Loyalty(Uri)

0.159†

(0.095)

0.162∗

(0.067)

-0.163

(0.121)

0.008

(0.095)

1.187

(0.737)

Honam
Youngnam

Trust in President
Trust in Nat’l Assembly
Trust in Party

-0.191∗∗

Loyalty to the Uri Party

0.116∗∗

Ideology(Conserv.)

(0.043)

Loser*Ideology
Non-partisan*Ideology
Intercept

0.842

(0.570)

Continued. . .
198

Table G.1 (cont’d)

Model G.1

Model G.2

Strongly Opposed vs. Supportive
-1.565∗∗

(0.551)

-1.571∗∗

(0.553)

-0.207

(0.334)

-0.212

(0.337)

Choongchung

0.230

(0.470)

0.266

(0.477)

Kangwon

0.861

(0.587)

0.814

(0.588)

Education

-0.094

(0.085)

-0.096

(0.086)

0.145

Honam
Youngnam

Female

(0.267)

0.142

(0.269)

1987 Democracy Cohort

0.927∗∗

(0.300)

0.908∗∗

(0.303)

1997 Financial Crisis Cohort

1.166∗∗

(0.425)

1.152∗∗

(0.428)

0.764

(1.283)

0.788

(1.238)

-1.845∗∗

(0.257)

-1.868∗∗

(0.261)

1.058∗∗

(0.264)

1.047∗∗

(0.267)

0.625∗

(0.277)

0.631∗

(0.278)

0.284

(0.373)

0.581

(1.144)

-0.581†

(0.320)

-1.681

(1.078)

(0.062)

-0.492∗∗

(0.109)

Loser*Loyalty(Uri)

-0.136

(0.167)

Non-partisan*Loyalty(Uri)

-0.016

(0.144)

0.198†

(0.102)

Loser*Ideology

0.076

(0.157)

Non-partisan*Ideology

0.214

(0.158)

1.989∗

(0.981)

1950 War Cohort
Trust in President
Trust in Nat’l Assembly
Trust in Party
Loser
Non-partisan

-0.525∗∗

Loyalty to the Uri Party

0.287∗∗

Ideology(Conserv.)

1.637∗

Intercept

(0.062)

(0.797)

Strongly Opposed vs. Strongly Supportive
Honam

-0.372

(0.738)

-0.237

(0.748)

0.380

(0.474)

0.373

(0.491)

1.254∗

(0.628)

1.498∗

(0.639)

Kangwon

0.003

(1.166)

0.064

(1.171)

Education

-0.052

(0.124)

-0.037

(0.127)

Youngnam
Choongchung

Continued. . .

199

Table G.1 (cont’d)

Model G.1
Female

-0.489

1987 Democracy Cohort

0.925†

Model G.2
(0.409)

-0.584

(0.417)
(0.500)

(0.488)

0.997∗

1.699∗∗

(0.582)

1.709∗∗

(0.604)

1.801

(1.384)

1.666

(1.387)

-2.220∗∗

(0.389)

-2.263∗∗

(0.390)

0.794∗

(0.395)

0.923∗

(0.407)

Trust in Party

0.440

(0.392)

0.331

(0.396)

Loser

0.845

(0.588)

-0.106

(1.727)

-0.064

(0.556)

-3.009†

(1.720)

-0.466∗∗

(0.090)

-0.312†

(0.172)

-0.459†

(0.247)

0.041

(0.215)

-0.242

(0.203)

Loser*Ideology

0.594∗

(0.255)

Non-partisan*Ideology

0.657∗

(0.260)

1.201

(1.523)

1997 Financial Crisis Cohort
1950 War Cohort
Trust in President
Trust in Nat’l Assembly

Non-partisan
Loyalty to the Uri Party
Loser*Loyalty(Uri)
Non-partisan*Loyalty(Uri)

0.272∗∗

Ideology(Conserv.)

Intercept

-0.103

N

(0.090)

(1.203)
835
-715.395

χ2
(48/60)
a Omitted
b Omitted

-703.575

525.222

Log-likelihood

835
548.864

category is Seoul and Kyunggi Region
category is the 2002 Presidential Election Cohort (Born in & after 1969)
†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

200

Appendix H
CHAPTER 5: MODELS FOR NONVOTER’S SUBSTANTIAL CHANGES
(OVER 5% INCREASE)

Table H.1: Chapter 5: Models for Nonvoter’s Substantial Changes
Model H.1 Model H.2 Model H.3 Model H.4 Model H.5
Winning Coalition (W)

2.465

(0.992)

(2.194)

(2.209)

(3.270)

0.848∗

1.468∗∗

-0.978

(0.533)

(1.562)

-1.405∗∗

-1.679∗∗

-1.383∗∗

-1.449∗∗

-1.411∗∗

(0.241)

(0.279)

(0.281)

(0.282)

-0.005

0.019∗∗

0.051∗∗

0.047∗∗

0.045∗∗

(0.016)

(0.006)

(0.016)

(0.016)

(0.017)

0.337∗∗

0.347∗

0.225

1.043∗∗

0.793∗

(0.115)

(0.142)

(0.152)

(0.344)

(0.391)

-0.265

0.223∗∗

0.954∗∗

1.027∗∗

0.898∗∗

(0.212)

(0.080)

(0.181)

(0.170)

(0.189)

-0.087∗∗

-0.072∗

-0.066∗

-0.070∗

(0.029)

(0.031)

(0.032)

(0.032)

0.036∗∗

0.034∗∗

0.032∗

0.033∗

(0.014)

(0.013)

(0.013)

(0.013)

-0.019

-0.019

-0.022

-0.021

(0.015)

Age_x

0.776

(0.189)
Age

1.062

(0.353)
Selectorate

-0.493

(0.720)
W_x

0.227

(0.046)

(0.047)

(0.047)

-0.035∗∗

-0.051†

-0.047

-0.044

0.014†
(0.008)

5% Nonvoter Dummy
∆Mass Threat
∆Mass Threat_x

0.301∗∗
(0.110)

Oil
Oil_x
Aid as %GDP
Growth

-0.032∗∗

Continued. . .
201

Table H.1 (cont’d)

Model H.1 Model H.2 Model H.3 Model H.4 Model H.5
(0.009)

(0.013)

(0.030)

(0.030)

(0.031)
-0.252

0.035

-0.259

(0.182)

(0.169)

(0.170)

(0.172)

-0.016

-0.113

(0.057)

(0.093)

-0.109∗∗

-0.099∗∗

-0.120∗∗

(0.029)

(0.029)

(0.041)

0.035∗∗

0.033∗∗

0.047∗

(0.008)

(0.008)

(0.022)

-0.007

-0.008

-0.007

(0.027)

(0.027)

(0.028)

0.012

0.019

0.018

(0.065)

(0.065)

(0.066)

0.034

0.027

0.023

(0.046)

(0.047)

(0.047)

0.524∗

0.552∗

0.516∗

(0.243)

(0.244)

(0.247)

-1.520∗∗

-1.481∗∗

-1.400∗∗

(0.254)

(0.261)

(0.270)

-1.162∗

-0.847

(0.497)

Ln(GDPpc)_x

0.281

(0.118)

Ln(GDPpc)

-0.281†

(0.547)

2-Way Interaction
W * Age
W * Age_x
W * Oil
W * Aid
W * Growth
W * Ln(GDPpc)
W * ∆Mass Threat
W * 5% Nonvoter
0.393∗

N
Log-likelihood
χ2

3198

1713

0.288†

(0.158)

5% Nonvoter * ∆Mass Threat

(0.173)

1713

1713

1713

-2286.499 -1220.863 -1185.853 -1186.237 -1184.535
199.314

128.479

198.501

Continued. . .

202

197.732

201.135

Table H.1 (cont’d)

Model H.1 Model H.2 Model H.3 Model H.4 Model H.5
†p<0.10, * p<0.05, ** p<0.01, *** p<0.001

203

Appendix I
CHAPTER 5: VARIANCE INFLATION FACTOR

Table I.1: Chapter 5: VIF from Model 2
(b)
After
(z_Free Press,
z_Population )

(a) Before Demeaning (Free
Press, GDP pc., Population)

Demeaning
z_GDP pc.,

Variable

VIF

Variable

VIF

Lag Mass

1.12

Lag Mass

1.06

W

183.87

W

1.51

∆Nonvoter

6.01

∆Nonvoter

5.96

∆Loser

13.42

∆Loser

12.54

Free Press(F.P.)

15.43

z_Free Press(F.P.)

5.58

Earthquake

7.63

Earthquake

6.83

GDP pc.

8.68

z_GDP pc.

5.66

Growth

4.35

Growth

4.33

Population

7.28

z_Population

4.41

W*GDP pc.

66.29

W*z_GDP pc.

5.97

W*Growth

4.70

W*Growth

4.70

W*Population

164.59

W*z_Population

4.74

W*Free Press

31.96

W*z_Free Press

5.67

W*Earthquake

7.56

W*Earthquake

7.03

W*∆Nonvoter

8.96

W*∆Nonvoter

5.77

W*∆Loser

19.28

W*∆Loser

12.80

∆Nonvoter*F.P.

4.97

∆Nonvoter*z_F.P.

1.78

∆Loser*F.P.

7.64

∆Loser*z_F.P.

2.02

Mean VIF

31.32

Mean VIF

5.46

204

Appendix J
CHAPTER 5: GOODNESS OF FIT

0.6

Figure J.1: Predicted Effects of Electoral Mobilization on Revolutionary Movements with
the Low or High Press Freedom Interaction (3-way interaction)

0.2
0.0
-0.2
-0.4
-0.6

Level of Revolutionary Movements

0.4

W=.25, Low Press Freedom
W=.25, High Press Freedom
W=.75, Low Press Freedom
W=.75, High Press Freedom

-0.4

-0.2

0.0

0.2

Changes in Nonvoter's Size (Changes in Electoral Mobilization)

205

0.4

0

1

2

3

4

Figure J.2: Goodness of Fit (Model 4)

0

1

2

3

4
Cox-Snell residual

Nelson-Aalen cumulative hazard

206

Cox-Snell residual

0

1

2

3

4

Figure J.3: Goodness of Fit (Model 5)

0

1

2

3
Cox-Snell residual

Nelson-Aalen cumulative hazard

207

4
Cox-Snell residual

0

1

2

3

4

Figure J.4: Goodness of Fit (Model 6)

0

1

2
Cox-Snell residual

Nelson-Aalen cumulative hazard

208

3

4
Cox-Snell residual

0

1

2

3

4

Figure J.5: Goodness of Fit (Model 7)

0

1

2
Cox-Snell residual

Nelson-Aalen cumulative hazard

209

3

4
Cox-Snell residual

0

1

2

3

4

Figure J.6: Goodness of Fit (Model 8)

0

1

2
3
Cox-Snell residual

Nelson-Aalen cumulative hazard

210

4
Cox-Snell residual

BIBLIOGRAPHY

211

BIBLIOGRAPHY

Aberbach, J.D. 1969. “Alienation and political behavior.” The American Political Science
Review 63(1):86–99.
Abramson, Paul R. 1983. Political attitudes in America: Formation and change. WH
Freeman, San Francisco.
Abramson, P.R. and A.W. Finifter. 1981. “On the meaning of political trust: new evidence
from items introduced in 1978.” American Journal of Political Science 25(2):297–307.
Abramson, P.R. and J.H. Aldrich. 1982. “The decline of electoral participation in America.”
The American Political Science Review 76(3):502–521.
Abramson, P.R., J.H. Aldrich and D.W. Rohde. 1998. Change and Continuity in the 1996
Elections. CQ Press.
Abramson, P.R. and W. Claggett. 1991. “Racial differences in self-reported and validated
turnout in the 1988 presidential election.” The Journal of Politics 53(01):186–197.
Achen, C.H. and L.M. Bartels. 2004. “Blind retrospection: Electoral Responses to Drought,
Flu, and Shark Attacks.”.
Adams, J. 2001a. “A theory of spatial competition with biased voters: Party policies viewed
temporally and comparatively.” British Journal of Political Science 31(01):121–158.
Adams, James. 2001b. Party competition and responsible party government: A theory of spatial competition based upon insights from behavioral voting research. University of Michigan
Press.
Akerlof, George A. and Rachel E. Kranton. 2000. “Economics and Identity.” Quarterly
Journal of Economics 115:715–753.
Aldrich, J.H. 1993. “Rational choice and turnout.” American Journal of Political Science
37(1):246–278.
Aldrich, J.H., J.L. Sullivan and E. Borgida. 1989. “Foreign affairs and issue voting: Do
presidential candidates" waltz before a blind audience?".” The American political science
review pp. 123–141.
Almond, G. and S. Verba. 1963. The Civic Culture: Political Attitudes and Democracy in
Five Nations. Princeton University Press.
Anderson, C., A. Blais, S. Bowler, T. Donovan and O. Listhaug. 2005. Losers’ consent:
elections and democratic legitimacy. Oxford University Press, USA.
Anderson, C.J. and A. Paskeviciute. 2009. Partisan Legitimacy: Parties, Partisanship, and
Support for the Political System. Vol. "Citizen Preferences, Political Institutions, and
Democratic Performance" Center for the Study of Democratic Performance, FSU.
212

Anderson, C.J. and A.J. LoTempio. 2002. “Winning, losing and political trust in America.”
British Journal of Political Science 32(02):335–351.
Anderson, C.J. and C.A. Guillory. 1997. “Political institutions and satisfaction with democracy: A cross-national analysis of consensus and majoritarian systems.” The American
Political Science Review 91(1):66–81.
Anderson, C.J. and M.M. Singer. 2008. “The sensitive left and the impervious right: multilevel models and the politics of inequality, ideology, and legitimacy in Europe.” Comparative Political Studies 41(4-5):564.
Anderson, C.J. and S.M. Mendes. 2006. “Learning to lose: Election outcomes, democratic
experience and political protest potential.” British Journal of Political Science 36(01):91–
111.
Atkeson, L.R., J.A. McCann, R.B. Rapoport and W.J. Stone. 1996. Citizens for Perot:
Assessing patterns of alienation and activism. In Broken Contract: Changing Relationships
Between Americans and Their Government, ed. S.C. Craig. Boulder, CO: Westview.
Bahry, D. and B.D. Silver. 1987. “Intimidation and the Symbolic Uses of Terror in the
USSR.” The American Political Science Review 81(4):1066–1098.
Bahry and Silver. 1990. “Soviet citizen participation on the eve of democratization.” American Political Science Review 84:821–847.
Banks, A.S. 2000. Cross-national time-series data archive: User’s manual.
Barnes, S.H., M. Kaase, K.R. Allerbeck et al. 1979. Political action: Mass participation in
five western democracies. Sage Publications.
Barry, B. 1970. Sociologists, economists, and democracy. University of Chicago Press.
Beaulieu, E.A. 2006. Protesting the contest: Election boycotts around the world, 1990-2002
PhD thesis University of California, San Diego.
Beetham, D. 1991. The legitimation of power. Macmillan.
Bendix, R. 1952. “Social stratification and political power.” The American Political Science
Review 46(2):357–375.
Berelson, B.R., P.F. Lazarsfeld and W.N. McPhee. 1954. Voting: A study of opinion formation in a presidential election. Chicago: Univ. of Chicago Pr.
Bermeo, N. 1990. “Rethinking regime change.” Comparative Politics 22(3):359–377.
Bernhard, M., C. Reenock and T. Nordstrom. 2003. “Economic Performance And Survival
In New Democracies.” Comparative Political Studies 36(4):404.
Bernhard, M., T. Nordstrom and C. Reenock. 2001. “Economic performance, institutional
intermediation, and democratic survival.” Journal of Politics 63(3):775–803.

213

Bianco, William T. 1994. Trust, Representatives and Constituents. Ann Arbor, MI: University of Michigan Press.
Black, Duncan. 1948. “On the rationale of group decision-making.” The Journal of Political
Economy 56(1):23–34.
Blaydes, L. 2011. Elections and Distributive Politics in Mubarak’s Egypt. Cambridge University Press.
Bommer, J. 1985. “The politics of disaster–Nicaragua.” Disasters 9(4):270–278.
Box-Steffensmeier, J.M. and C.J.W. Zorn. 2001. “Duration models and proportional hazards
in political science.” American Journal of Political Science pp. 972–988.
Box-Steffensmeier, J.M., D. Reiter and C. Zorn. 2003. “Nonproportional hazards and event
history analysis in international relations.” Journal of Conflict Resolution 47(1):33.
Brancati, D. 2007. “Political aftershocks: The impact of earthquakes on intrastate conflict.”
Journal of Conflict Resolution 51(5):715.
Bratton, M. 1997. “Deciphering Africa’s divergent transitions.” Political Science Quarterly
112(1):67–93.
Bratton, M. 1998. “Second elections in Africa.” Journal of Democracy 9(3):51–66.
Bratton, M. 2004. “The ’Alternation Effect’ in Africa.” Journal of Democracy 15:147–158.
Bratton, M. and G. Lambright. 2001. “Uganda’s Referendum 2000: The silent boycott.”
African Affairs 100(400):429.
Bratton, M. and N. Van de Walle. 1994. “Neopatrimonial regimes and political transitions
in Africa.” World Politics 46(4):453–489.
Bratton, M. and N. Van de Walle. 1997. Democratic experiments in Africa: Regime transitions in comparative perspective. Cambridge Univ Pr.
Brehm, J. and W. Rahn. 1997. “Individual-level evidence for the causes and consequences
of social capital.” American journal of political science pp. 999–1023.
Brownlee, J. 2007. Authoritarianism in an Age of Democratization. Cambridge Univ Pr.
Budge, I., Ivor Crewe and Dennis Farlie. 1976. Party identification and beyond. New York:
Wiley.
Bueno de Mesquita, B. and A. Smith. 2010. “Leader Survival, Revolutions, and the Nature
of Government Finance.” American Journal of Political Science 54(4):936–50.
Bueno de Mesquita, B., A. Smith, R.M. Siverson and J.D. Morrow. 2010. “Pilot Study on
Selectorate Institutions.”.
URL: http://alexanderhamilton.as.nyu.edu/page/pilotstudy

214

Bueno de Mesquita, B. and George Downs. 2006. “Strategic Coordination, Public Goods,
and Political Survival.”.
Bueno de Mesquita, B. and R.M. Siverson. 1995. “War and the survival of political leaders:
A comparative study of regime types and political accountability.” American Political
Science Review 89(4):841–855.
Bueno de Mesquita, B.B. and A. Smith. 2009. “Political survival and endogenous institutional
change.” Comparative Political Studies 42(2):167.
Bueno de Mesquita, Bruce, Alastair Smith, Randolph Siverson and James Morrow. 2003.
The Logic of Political Survival. The MIT Press.
Campbell, A., P.E. Converse, W.E. Miller and E. Donald. 1960. The American Voter. New
York: Wiley.
Caporaso, J. 2009. Is There a Quantitative-Qualitative Divide in Comparative Politics? The
Case of Process Tracing. Thousand Oaks, CA: Sage pp. 67–83.
Carlin, R.E. 2006. “The decline of citizen participation in electoral politics in postauthoritarian Chile.” Democratization 13(4):632–651.
Chacon, M., JA Robinson and R. Torvik. 2011. “When is Democracy an Equilibrium?: Theory and Evidence from Colombia’s La Violencia.” Journal of Conflict Resolution 55(3):366–
396.
Chang, E.C.C. 2008. “Electoral incentives and budgetary spending: Rethinking the role of
political institutions.” The Journal of Politics 70(04):1086–1097.
Chanley, V.A. 2002. “Trust in Government in the Aftermath of 9/11: Determinants and
Consequences.” Political Psychology 23(3):469–483.
Chanley, V.A., T.J. Rudolph and W.M. Rahn. 2000. “The origins and consequences of public
trust in government: A time series analysis.” Public Opinion Quarterly 64(3):239.
Chanley, V.A., T.J. Rudolph and W.M. Rahn. 2001. Public trust in government in the
Reagan years and beyond. In What is it about government that Americans dislike, ed.
J. R. Hibbing and E. Theiss-Morse. Cambridge: Cambridge University Press pp. 59–78.
Cheles, L., R. Ferguson and M. Vaughan. 1995. The far right in western and eastern Europe.
Longman Pub Group.
Cho, W. 2010. “Ethnic Fractionalization, Electoral Institutions, and Africans’ Political Attitudes.” Comparative Political Studies forthcoming.
Cho, W. and M. Bratton. 2006. “Electoral institutions, partisan status, and political support
in Lesotho.” Electoral Studies 25(4):731–750.
Citrin, J. 1974. “The political relevance of trust in government.” The American Political
Science Review 68(3):973–988.
215

Citrin, J. 1977. “Political alienation as a social indicator: Attitudes and action.” Social
Indicators Research 4(1):381–419.
Citrin, J. and D.P. Green. 1986. “Presidential leadership and the resurgence of trust in
government.” British Journal of Political Science 16(04):431–453.
Citrin, J., H. McClosky, J.M. Shanks and P.M. Sniderman. 1975. “Personal and political
sources of political alienation.” British Journal of Political Science 5(01):1–31.
Clarke, K.A. and R.W. Stone. 2008. “Democracy and the Logic of Political Survival.” American Political Science Review 102(03):387–392.
Cohen, Y. 1994. Radicals, reformers, and reactionaries: the prisoner’s dilemma and the
collapse of democracy in Latin America. University of Chicago Press.
Colaresi, M. 2004a. “When doves cry: International rivalry, unreciprocated cooperation, and
leadership turnover.” American Journal of Political Science 48(3):555–570.
Colaresi, M. 2004b. “Aftershocks: Postwar leadership survival, rivalry, and regime dynamics.”
International Studies Quarterly 48(4):713–728.
Colomer, J.M. 2000. Strategic Transitions: Game Theory and Democratization. The Johns
Hopkins University Press.
Colomer, J.M. 2003. Political institutions: democracy and social choice. Oxford University
Press, USA.
Cox, G.W. and M.C. Munger. 1989. “Closeness, expenditures, and turnout in the 1982 US
House elections.” The American Political Science Review pp. 217–231.
Craig, S.C. and M.A. Maggiotto. 1981. “Political discontent and political action.” The Journal of Politics 43(02):514–522.
Crozier, M., S.P. Huntington and J. Watanuki. 1975. The crisis of democracy: report on the
governability of democracies to the Trilateral Commission. New York University Press.
Dahl, R.A. 1954. A preface to democratic theory. University of Chicago Press.
Dahl, R.A. 1971. Polyarchy: participation and opposition. Vol. 54 Yale Univ Pr.
Dahl, R.A. 1991. Democracy and its Critics. Yale Univ Pr.
Dahl, R.A. 2006. On political equality. Yale Univ Pr.
Dalton, R.J. 2000. “Citizen attitudes and political behavior.” Comparative political studies
33(6-7):912.
Dalton, R.J. 2004. Democratic challenges, democratic choices: The erosion of political support in advanced industrial democracies. Oxford University Press, USA.

216

Dalton, R.J. 2006. Citizen politics: Public opinion and political parties in advanced industrial
democracies. CQ Press Washington, DC.
Dalton, Russell J. 1999. "Political Support in Advanced Industrial Democracies". In Critical
citizens: Global support for democratic government, ed. Pippa Norris. New York: Oxford
University Press.
Davenport, C. 1997. “From ballots to bullets: an empirical assessment of how national
elections influence state uses of political repression.” Electoral Studies 16(4):517–540.
Davenport, C. 1998. “Liberalking Event or Lethal Episode?: An Empirical Assessment of
How National Elections Affect the Suppression of Political and Civil Liberties.” Social
Science Quarterly 79(2):321–340.
Davenport, C. 2007. “State repression and political order.” Annual Review of Political Science
10:1–23.
Davies, J.C. 1962. “Toward a theory of revolution.” American sociological review 27(1):5–19.
Diamond, L.J. 2002. “Thinking about hybrid regimes.” Journal of democracy 13(2):21–35.
Downs, Anthony. 1957. An economic theory of democracy. Number Harper.
Dunning, T. 2011. “Fighting and Voting: Violent Conflict and Electoral Politics.” Journal
of Conflict Resolution .
Easton, D. 1965. A systems analysis of political life. New York: Wiley.
Easton, D. 1975. “A re-assessment of the concept of popular support.” British Journal of
Political Science 5:435–57.
Edelman, M.J. 1971. Politics as symbolic action: Mass arousal and quiescence. Markham
Pub. Co.
Edelman, Murray J. 1964. The symbolic uses of politics. Univ of Illinois Press.
Enders, C.K. and D. Tofighi. 2007. “Centering predictor variables in cross-sectional multilevel
models: A new look at an old issue.” Psychological Methods 12(2):121.
Engdahl, E.R. and A. Villaseñor. 2002. “Centennial earthquake catalog.” Washington, DC:
US Geological Survey .
Erikson, R.S. and D.W. Romero. 1990. “Candidate equilibrium and the behavioral model of
the vote.” The American Political Science Review 84(4):1103–1126.
Esaiasson, P. and S. Göteborg. 2009. "Why Politicians Don’t Protest Elections–Losers’
Consent in Established Democracies". Annual Meeting of the American Political Science
Association.
Fazio, R.H. 1986. “How do attitudes guide behavior.” Handbook of motivation and cognition:
Foundations of social behavior 1:204–243.
217

Feierabend, I.K. and R.L. Feierabend. 1966. “Aggressive behaviors within polities, 1948-1962:
a cross-national study.” Journal of Conflict Resolution pp. 249–271.
Ferejohn, J.A. and M.P. Fiorina. 1974. “The paradox of not voting: A decision theoretic
analysis.” The American political science review 68(2):525–536.
Ferson, W.E. and C.R. Harvey. 1991. “The variation of economic risk premiums.” Journal
of Political Economy 99(2):385–415.
Finifter, A.W. 1970. “Dimensions of political alienation.” The American Political Science
Review 64(2):389–410.
Finkel, S.E. 1985. “Reciprocal effects of participation and political efficacy: A panel analysis.”
American Journal of Political Science 29(4):891–913.
Franklin, M.N., C. Eijk and E. Oppenhuis. 1996. “The institutional context: Turnout.”
Choosing Europe? The European Electorate and National Politics in the Face of Union
pp. 306–331.
Freitag, M. and M. Bühlmann. 2009.
42(12):1537.

“Crafting Trust.” Comparative Political Studies

Fu, Vincent. 1998. “Estimating generalized ordered logit models.” Stata Technical Bulletin
8(44):160–64.
Fuchs, D., G. Guidorossi and P. Svensson. 1995. "Support for the democratic system".
In Citizens and the State, ed. H-D. Klingemann and D. Fuchs. Oxford University Press
pp. 323–53.
Gamson, W.A. 1968. Power and discontent. Dorsey Press.
Gamson, W.A. 1990. The strategy of social protest. Belmont, Calif.: Wadsworth.
Gandhi, J. and E. Lust-Okar. 2009. “Elections under authoritarianism.” Annual review of
political science 12:403–422.
Gandhi, Jennifer and Adam Przeworski. 2007. “Authoritarian institutions and the survival
of autocrats.” Comparative Political Studies 40(11):1279.
Geddes, B. 2005. Why parties and elections in authoritarian regimes? In Paper presented in
the Annual Meeting of the American Political Science Association.
Gerber, A.S. and G.A. Huber. 2009. “Partisanship and Economic Behavior: Do Partisan
Differences in Economic Forecasts Predict Real Economic Behavior?” American Political
Science Review 103(03):407–426.
Gerber, A.S. and G.A. Huber. 2010. “Partisanship, Political Control, and Economic Assessments.” American Journal of Political Science 54(1):153–173.
Gerring, J., P. Bond, W.T. Barndt and C. Moreno. 2005. “Democracy and economic growth.”
World Politics 57(3):323–364.
218

Gerring, J., S.C. Thacker and C. Moreno. 2005. “Centripetal democratic governance: A
theory and global inquiry.” American Political Science Review 99(04):567–581.
Gerring, J., S.C. Thacker and C. Moreno. 2008. A centripetal theory of democratic governance. Cambridge University Press.
Gibson, J.L., G.A. Caldeira and L.K. Spence. 2003. “Measuring Attitudes toward the United
States Supreme Court.” American Journal of Political Science 47(2):354–367.
Gleditsch, K.S. 2002. “Expanded trade and GDP data.” Journal of Conflict Resolution
46(5):712.
Goemans, H.E. 2000a. “Fighting for Survival.” Journal of Conflict Resolution 44(5):555.
Goemans, H.E., K.S. Gleditsch and G. Chiozza. 2009. “Introducing Archigos: A dataset of
political leaders.” Journal of Peace research 46(2):269.
Gosnell, H.F. 1930. Why Europe Votes. Number 19 The University of Chicago press.
Green, J.D. 1984. “Countermobilization as a revolutionary form.” Comparative Politics
16(2):153–169.
Group, World Bank. International Economics Dept. Development Data. 2005. World development indicators. World Bank.
Gurr, T. 1968a. “A causal model of civil strife: A comparative analysis using new indices.”
The American Political Science Review 62(4):1104–1124.
Gurr, T. 1968b. “Psychological factors in civil violence.” World Politics: A Quarterly Journal
of International Relations 20(2):245–278.
Gurr, T.R. 1970. Why men rebel. Princeton University Press.
Gurr, T.R. and W.H. Moore. 1997. “Ethnopolitical rebellion: A cross-sectional analysis
of the 1980s with risk assessments for the 1990s.” American Journal of Political Science
41(4):1079–1103.
Hadenius, A. and J. Teorell. 2005. "Assessing alternative indices of democracy". In Committee on Concepts and Methods Working Paper Series (August).
Hahm, Chaibong. 2008. “South Korea’s Miraculous Democracy.” Journal of Democracy
19(3):128–142.
Hahm, Chaihak and Sung Ho Kim. 2005. “Constitutionalism on Trial in South Korea.”
Journal of Democracy 16(2):28–42.
Hall, P.A. 1999. “Social capital in Britain.” British Journal of Political Science 29(03):417–
461.
Helmke, G. 2010. “The Origins of Institutional Crises in Latin America.” American Journal
of Political Science 54(3):737–750.
219

Heo, Uk and Hans Stockton. 2005. “The Impact of Democratic Transition on Elections and
Parties in South Korea.” Party Politics 11(6):674.
Hermet, G., R. Rose and A. Rouquie. 1978. Elections without choice. Halsted Press.
Hetherington, M.J. 1998. “The political relevance of political trust.” American Political
Science Review 92(4):791–808.
Hetherington, M.J. 1999. “The effect of political trust on the presidential vote, 1968-96.”
The American political science review 93(2):311–326.
Hetherington, M.J. 2005. Why trust matters. Princeton University Press.
Hetherington, M.J. and J.D. Nugent. 2001. Explaining public support for devolution: The
role of political trust. In What is it about government that Americans dislike?, ed. John R.
Hibbing and Elizabeth Theiss-Morse. New York: Cambridge University Press pp. 134–151.
Hetherington, M.J. and S. Globetti. 2002. “Political trust and racial policy preferences.”
American Journal of Political Science pp. 253–275.
Hibbs, D.A. 1973. Mass political violence: A cross-national causal analysis. Vol. 253 Wiley
New York.
Hicks, A. and D. Swank. 1992. “Politics, institutions, and welfare spending in industrialized
democracies, 1960-1982.” American Political Science Review 86(3):658–674.
Higgins, E.T. and G. King. 1981. Accessibility of social constructs: Information processing
consequences of individual and contextual variability. In Personality, cognition, and social
interaction, ed. N. Cantor and J. Kihlstrom. Hillsdale, NJ: Erlbaum pp. 69–122.
Hipsher, P.L. 1996. “Democratization and the decline of urban social movements in Chile
and Spain.” Comparative Politics 28(3):273–297.
Hirschman, A.O. 1970. Exit, voice, and loyalty: Responses to decline in firms, organizations,
and states. Harvard Univ Pr.
Hochstetler, K. 2006. “Rethinking presidentialism: challenges and presidential falls in South
America.” Comparative Politics pp. 401–418.
Holmberg, Soren. 2003. “Are Political parties necessary?” Electoral studies 22(2):287–299.
Hooghe, M. and D. Stolle. 2003. Generating social capital: civil society and institutions in
comparative perspective. Palgrave Macmillan.
Horiuchi, Y. and S. Lee. 2008. “The presidency, regionalism, and distributive politics in
South Korea.” Comparative Political Studies 41(6):861.
Huntington, S.P. 1968. Political order in changing societies. Yale Univ Pr.
Huntington, S.P. 1974. “Postindustrial politics: how benign will it be?” Comparative Politics
6(2):163–191.
220

Huntington, S.P. 1993. The third wave: Democratization in the late twentieth century. Univ
of Oklahoma Pr.
Hyman, Herbet H. 1959. Political socialization: A study in the psychology of political behavior. New York: Free Press.
Im, Hyug Baeg. 2004. “Faltering democratic consolidation in South Korea: democracy at
the end of the ’three Kims’ era.” Democratization 11(5):179–198.
Inglehart, R. 1977. The silent revolution. Princeton Univ. Press.
Inglehart, R. 1997. Modernization and postmodernization: Cultural, economic, and political
change in 43 societies. Princeton Univ Pr.
Jackman, R.W. 1987. “Political institutions and voter turnout in the industrial democracies.”
The American Political Science Review 81(2):405–423.
Jackman, R.W. and R.A. Miller. 1995. “Voter turnout in the industrial democracies during
the 1980s.” Comparative Political Studies 27(4):467.
Jackson, J.S. 1973. “Alienation and black political participation.” The Journal of Politics
35(04):849–885.
Jacobs, E.M. 1970. “Soviet local elections: what they are, and what they are not.” Soviet
Studies 22(1):61–76.
Jaggers, K. and T.R. Gurr. 1995. “Tracking democracy’s third wave with the Polity III data.”
Journal of Peace Research 32(4):469–482.
Javeline, Debra. 1999. “Response effects in polite cultures: A test of acquiescence in Kazakhstan.” Public Opinion Quarterly 63(1):1.
Kaase, M. 1988. Political alienation and protest. In Comparing Pluralist Democracies., ed.
M. Dogan. Westview Press, Boulder.
Kaase, M. 1990. Mass participation. In Continuities in Political Action: A Longitudinal Study
of Political Orientations in Three Western Democracies, ed. J. W. Van Deth M. K. Jennings. de Gruyter pp. 23–67.
Kaase, M. and S.H. Barnes. 1979. Conclusion: The future of political protest in Western
democracies. In Political Action: Mass participation in five Western democracies, ed. S.H.
Barnes and M. Kaase. Beverly Hills: Sage pp. 523–536.
Kang, Wontaek. 2005. Hanguk ui Jeongchi Gaehyeok gwa Minjujuui [Political Reform and
Democracy in Korea]. Seoul: Ingansarang.
Karl, T.L. 1990. “Dilemmas of democratization in Latin America.” Comparative Politics
23(1):1–21.
Karp, J.A. and S.A. Banducci. 2007. “Party mobilization and political participation in new
and old democracies.” Party Politics 13(2):217.
221

Kedar, O. and P.W. Shively. 2005. “Introduction: Special Issue on multilevel modeling for
large clusters.” Political Analysis 13(4):297–300.
Keele, L. 2005. “The authorities really do matter: Party control and trust in government.”
The Journal of Politics 67(03):873–886.
Kennedy, R. 2009. “Survival and Accountability: An Analysis of the Empirical Support for
“Selectorate Theory”.” International Studies Quarterly 53:695–714.
Kim, Byung-Kook. 1998. Korea’s Crisis of Success. In Democracy in East Asia, ed. Larry
Diamond and Marc Plattner. Johns Hopkins University Press pp. 113–32.
Kim, HeeMin. 1997. “Rational Choice Theory and Third World Politics: The 1990 Party
Merger in Korea.” Comparative Politics 30(1):83–100.
Kim, HeeMin, Jun Young. Choi and Jinman Cho. 2008. “Changing cleavage structure in
new democracies: An empirical analysis of political cleavages in Korea.” Electoral Studies
27(1):136–150.
Kim, Y.H. and D. Bahry. 2008. “Interrupted Presidencies in Third Wave Democracies.” The
Journal of Politics 70(3):807–822.
Koch, J.W. 2003. “Political cynicism and third party support in American presidential
elections.” American Politics Research 31(1):48.
Kravis, I.B., A.W. Heston and R. Summers. 1978. “Real GDP per capita for more than one
hundred countries.” The Economic Journal 88(350):215–242.
Kwon, H.Y. 2005. “Targeting public spending in a new democracy: Evidence from South
Korea.” British Journal of Political Science 35(02):321–341.
Lagos, M. 2001. “Between stability and crisis in Latin America.” Journal of Democracy
12(1):137–145.
Lazarsfeld, P., B. Berelson and H. and Gaudet. 1944. The People’s Choice. How the voter
makes up his mind in a presidential campaign. By Paul F. Lazarsfeld, Bernard Berelson,
Hazel Gaudet. Duell, Sloan & Pearce.
Lee, A.R. 1999. “What We Know about Party Nonidentifiers in South Korea.” Asian Affairs:
An American Review 25(4):209–219.
Lee, Gap-yoon. 1998. Hanguk ui Seongeo wa Jiyeokjuui [Elections and Regionalism in Korea].
Seoul: Oreum.
Lee, Yoonkyung. 2009. “Democracy without Parties? Political Parties and Social Movements
for Democratic Representation in Korea.” Korea Observer 40(1):27–52.
Levi, M., A. Sacks and T. Tyler. 2009. “Conceptualizing Legitimacy, Measuring Legitimating
Beliefs.” American Behavioral Scientist 53(3):354.

222

Levi, M. and L. Stoker. 2000. “Political Trust and Trustworthiness.” Annual Review of
Political Science 3:475–507.
Levitsky, S. and L.A. Way. 2002. “The rise of competitive authoritarianism.” Journal of
Democracy 13(2):51–65.
Lichbach, M.I. 1998. The rebel’s dilemma. Univ of Michigan Pr.
Lichbach, M.I. and T.R. Gurr. 1981. “The conflict process.” Journal of Conflict Resolution
25(1):3.
Linz, J. 1978a. The Breakdown of Democratic Regimes: Crisis, Breakdown, and ReEquilibration. Baltimore: Johns Hopkins University Press.
Linz, J. 1978b. “Non-competitive elections in Europe.” Elections without choice pp. 36–65.
Linz, J.J. 1976. “Some notes toward a comparative study of Fascism in sociological historical
perspective.” Fascism: A reader’s guide 3.
Lipset, S.M. 1959. “Some social requisites of democracy: Economic development and political
legitimacy.” The American Political Science Review 53(1):69–105.
Lipset, S.M. 1960. Political man: The social bases of politics. Doubleday Garden City, NY.
Lohmann, S. 1994. “The dynamics of informational cascades: The Monday demonstrations
in Leipzig, East Germany, 1989-91.” World Politics 47(1):42–101.
Long, J. Scott. 1997. Regression models for categorical and limited dependent variables. Sage
Publications, Inc.
Long, J. Scott. and Jeremy Freese. 2006. Regression models for categorical dependent variables using Stata. 2nd ed. Stata press College Station, TX.
Luke, D.A. 2005. Multilevel modeling. Sage.
Machado, F., C. Scartascini and M. Tommasi. 2011. “Political Institutions and Street
Protests in Latin America.” Journal of Conflict Resolution 55(3):340–365.
Mackenzie, G.C. and J.M. Labiner. 2002. “Opportunity lost: The rise and fall of trust
and confidence in government after September 11.” American Journal of Political Science
48(1):93–112.
Maeda, K. and M. Nishikawa. 2006. “Duration of Party Control in Parliamentary and Presidential Governments: A Study of 65 Democracies, 1950 to 1998.” Comparative Political
Studies 39(3):352.
Manin, B. 1997. The principles of representative government. Cambridge Univ Pr.
Marsh, A. and M. Kaase. 1979. Background of political action. In Political action: Mass
participation in five Western democracies, ed. S.H. Barnes and M. Kaase. Sage Beverly
Hills, CA pp. 97–136.
223

Marshall, M.G., K. Jaggers and T.R. Gurr. 2009. “Polity IV Project: Political Regime
Characteristics and Transitions, 1800-2009.” Polity IV .
Marwell, G. 1970. “Comment on Scott and El-Assal.” American Sociological Review 35(5).
Marx, K., F. Engels and G.S. Jones. 2002. The communist manifesto. Penguin Classics.
Mason, K. O., Mason W.M. Winsborough, H.H. and W. Kenneth Poole. 1973. “Some
methodological issues in cohort analysis of archival data.” American Sociological Review
38(2):242–258.
McDonald Michael, P. and L. Popkin Samuel. 2001. “The Myth of the Vanishing Voter.”
American Political Science Review 95(4):963–74.
McFaul, M. 2002. “The fourth wave of democracy and dictatorship: Noncooperative transitions in the postcommunist world.” World Politics 54(2):212–244.
Midlarsky, M.I. 1988. “Rulers and the ruled: patterned inequality and the onset of mass
political violence.” The American Political Science Review 82(2):491–509.
Migdal, J.S. 1988. Strong societies and weak states: state-society relations and state capabilities in the Third World. Princeton Univ Pr.
Milbrath, L.W. and M. L. Goel. 1977. Political participation: How and why do people get
involved in politics? 2nd ed. Chicago: Randy McNally.
Miller, A.H. 1974a. “Political issues and trust in government: 1964-1970.” The American
Political Science Review 68(3):951–972.
Miller, A.H. 1974b. “Rejoinder to" Comment" by Jack Citrin: Political Discontent or Ritualism?” The American Political Science Review 68(3):989–1001.
Miller, A.H., E.N. Goldenberg and L. Erbring. 1979. “Type-set politics: Impact of newspapers
on public confidence.” The American Political Science Review 73(1):67–84.
Miller, A.H. and O. Listhaug. 1990. “Political parties and confidence in government: A
comparison of Norway, Sweden and the United States.” British Journal of Political Science
20(03):357–386.
Miller, Warren E. and J.M. Shanks. 1996. The new American voter. Cambridge, MA:
Harvard University Press.
Miller, W.E. 1980. “Disinterest, disaffection, and participation in presidential politics.” Political Behavior 2(1):7–32.
Mishler, W. and R. Rose. 2005. “What are the political consequences of trust?” Comparative
Political Studies 38(9):1050.
Mo, Jongryn. 2001. “Political Culture and Legislative Gridlock.” Comparative Political Studies 34(5):467.

224

Modigliani, F. and R. Brumberg. 1954. Utility analysis and the consumption function: An
interpretation of cross-section data. In Post Keynesian Economics, ed. K. Kurihara. New
Brunswick, N.J.: Rutgers University Press.
Moehler, D.C. and S.I. Lindberg. 2009. “Narrowing the Legitimacy Gap: Turnovers as a
Cause of Democratic Consolidation.” The Journal of Politics 71(04):1448–1466.
Moore, B. 1966. Social origins of dictatorship and democracy: Lord and peasant in the
making of the modern world. Beacon Pr.
Muller, E.N. 1977. “Behavioral correlates of political support.” The American Political Science Review pp. 454–467.
Muller, E.N. 1979. Aggressive political participation. Princeton University Press.
Muller, E.N. 1985. “Income inequality, regime repressiveness, and political violence.” American Sociological Review 50(1):47–61.
Muller, E.N. and E. Weede. 1990. “Cross-national variation in political violence.” Journal of
Conflict Resolution 34(4):624.
Muller, E.N. and M.A. Seligson. 1987. “Inequality and insurgency.” The American Political
Science Review 81(2):425–451.
Muller, E.N., M.A. Seligson, H. Fu and M.I. Midlarsky. 1989. “Land inequality and political
violence.” The American Political Science Review 83(2):577–596.
Muller, E.N. and T.O. Jukam. 1983. “Discontent and aggressive political participation.”
British Journal of Political Science 13(02):159–179.
Muller, E.N., T.O. Jukam and M.A. Seligson. 1982. “Diffuse political support and antisystem political behavior: A comparative analysis.” American Journal of Political Science
26(2):240–264.
Murad, Havi, Anat Fleischman, Siegal Sadetzki, Orna Geyer and Laurence S. Freedman.
2003. “Small Samples and Ordered Logistic Regression: Does it help to collapse categories
of outcome?” The American Statistician 57(3):155–160.
Newton, K. and P. Norris. 2000. "Confidence in Public Institutions: Faith, Culture, or
Performance?". In Disaffected Democracies. What’s Troubling the Trilateral Countries,
ed. S. Pharr and R. Putnam. Princeton University Press.
Niemi, R.G., H.W. Stanley and C.L. Evans. 1984. “Age and Turnout Among the Newly Enfranchised: Life Cycle versus Experience Effects*.” European Journal of Political Research
12(4):371–386.
Norpoth, H. 1987. “Guns and butter and government popularity in Britain.” The American
political science review 81(3):949–959.
Norris, P. 2011. Democratic Deficit: Critical Citizens Revisited. Cambridge Univ Pr.
225

Norris, P., S. Walgrave and P. Van Aelst. 2005. “Who Demonstrates? Antistate rebels,
conventional Participants, or everyone?” Comparative politics 37(2):189–205.
North, D.C. 1990. Institutions, institutional change, and economic performance. Cambridge
Univ Pr.
Nye, J.S., P. Zelikow and D.C. King. 1997. Why people don’t trust government. Harvard
Univ Pr.
O’Donnell, G. 2007. “The perpetual crises of democracy.” Journal of Democracy 18(1):5.
O’Donnell, G., P. Schmitter and L. Whitehead. 1986. Transitions from Authoritarian Rule:
Tentative Conclusions about Uncertain Democracies. Vol. 4 Baltimore: Johns Hopkins
University Press: The Johns Hopkins University Press.
O’Lessker, K. 1968. “Who voted for Hitler? a new look at the class basis of Naziism.” The
American Journal of Sociology 74(1):63–69.
Oliver, P.E. and G. Marwell. 1988. “The Paradox of Group Size in Collective Action: A
Theory of the Critical Mass. II.” American Sociological Review 53(1):1–8.
Olson, M. 1974. The logic of collective action: Public goods and the theory of groups. Harvard
University Press.
Opp, K.D. and C. Gern. 1993. “Dissident groups, personal networks, and spontaneous cooperation: the East German revolution of 1989.” American Sociological Review 58(5):659–680.
Oppenhuis, E. 1995. Voting behavior in Europe: A comparative analysis of electoral participation and party choice. Het Spinhuis.
Orren, G. 1997. "Fall from grace: The publics loss of faith in government". In Why People
Don’t Trust Government, ed. P.D. Zelikow Nye, J.S. and D.C. King. Cambridge, Mass.:
Harvard University Press pp. 77–108.
Park, Chan Wook. 2002. Elections in Democratizing Korea. Seven Bridges Press pp. 118–46.
Peterson, B. and F.E. Harrell. 1990. “Partial proportional odds models for ordinal response
variables.” Applied Statistics 39(2):205–217.
Pharr, Susan J. and Robert D. Putnam. 2000. Disaffected democracies: what’s troubling the
trilateral countries? Princeton University Press.
Pierce, R. and E. Converse Philip. 1989. “Attitudinal roots of popular protest: The French
upheaval of May 1968.” International Journal of Public Opinion Research 1(3):221–241.
Piven, F.F. and R.A. Cloward. 1977. Poor people’s movements: Why they succeed, how they
fail. Pantheon Books.
Powell, G.B. 1981. “Party systems and political system performance: Voting participation,
government stability and mass violence in contemporary democracies.” The American
Political Science Review 75(4):861–879.
226

Powell, G.B. 1986a. “American voter turnout in comparative perspective.” The American
Political Science Review 80(1):17–43.
Powell, G.B. 1986b. “Extremist parties and political turmoil: Two puzzles.” American Journal of Political Science 30(2):357–378.
Powell Jr, G.B. 1980. Voting turnout in thirty democracies: Partisan, legal, and socioeconomic influences. In Electoral participation: A comparative analysis. Sage.
Preston, J. and S. Dillon. 2005. Opening Mexico: The making of a democracy. Farrar Straus
& Giroux.
Przeworski, A. 1975. “Institutionalization of voting patterns, or is mobilization the source of
decay?” The American Political Science Review 69(1):49–67.
Przeworski, A. 1991. Democracy and the market: Political and economic reforms in Eastern
Europe and Latin America. Cambridge Univ Pr.
Przeworski, A. 1992. “The games of transition.” Issues in democratic consolidation: The new
South American democracies in comparative perspective pp. 105–52.
Przeworski, A. 2005. “Democracy as an Equilibrium.” Public Choice 123(3):253–273.
Przeworski, A. 2009. “Constraints and Choices.” Comparative Political Studies 42(1):4.
Putnam, R.D. 1995. “Tuning in, tuning out: The strange disappearance of social capital in
America.” PS: Political Science and Politics 28(4):664–683.
Putnam, R.D., R. Leonardi and R.Y. Nanetti. 1994. Making democracy work. Princeton
Univ. Press.
Quiroz, A. F. and A. Smith. 2010. “Surviving Disasters.” Working Paper .
Ragsdale, L. and J.G. Rusk. 1993. “Who are nonvoters? Profiles from the 1990 Senate
elections.” American Journal of Political Science 37(3):721–746.
Raudenbush, S.W. and A.S. Bryk. 2002. Hierarchical linear models: Applications and data
analysis methods. Sage Publications, Inc.
Reynolds, A., B. Reilly and A. Ellis. 2005. Electoral System Design: The New International
IDEA Handbook. International IDEA.
Riker, W.H. and P.C. Ordeshook. 1968. “A Theory of the Calculus of Voting.” The American
Political Science Review 62(1):25–42.
Roeder, P.G. 1989. “Electoral avoidance in the Soviet Union.” Soviet Studies 41(3):462–483.
Rosenberg, J.S., M. Chen and Brennan Center for Justice. 2009. Expanding Democracy:
Voter Registration Around the World. Brennan Center for Justice.

227

Rosenstone, S. and J.M. Hansen. 1993. Mobilization, participation and democracy in America. MacMillan Publishing.
Rosenstone, S.J. and R.E. Wolfinger. 1978. “The effect of registration laws on voter turnout.”
The American Political Science Review 72(1):22–45.
Rothstein, B. 2009. “Creating Political Legitimacy: Electoral Democracy Versus Quality of
Government.” American Behavioral Scientist 53(3):311.
Rothstein, B. and D. Stolle. 2008. “The state and social capital: An institutional theory of
generalized trust.” Comparative Politics 40(4):441–60.
Rothstein, Bo. 2000. “Trust, Social Dilemmas and Collective Memories.” Journal of Theoretical Politics 12(4):477–501.
Rothstein, Bo. 2002. "Sweden: Social Capital in the Social Democratic State". In Democracies in Flux, ed. R.D. Putnam. Oxford Scholarship Online Monographs pp. 289–333.
Rudolph, T.J. and J. Evans. 2005. “Political trust, ideology, and public support for government spending.” American Journal of Political Science pp. 660–671.
Rustow, D.A. 1970. “Transitions to democracy: Toward a dynamic model.” Comparative
Politics 2(3):337–363.
Schattschneider, E.E. 1960. The Semi-Sovereign People: A Realist’s View of Democracy in
America. New York: Holt, Rinehart and Winston Inc.
Schumpeter, J.A. 1952. Capitalism, socialism and democracy. 4th ed. London: Goerge Allen
& Unwin.
Schwartz, D.C. 1973. Political alienation and political behavior. Chicago: Aldine Publishing
Co.
Schwartz, M.A. 2000. “Continuity strategies among political challengers: the case of social
credit.” American Review of Canadian Studies 30.
Scott, J.C. 1985. Weapons of the weak: Everyday forms of peasant resistance. Yale Univ Pr.
Scott, J.C. 1990. Domination and the arts of resistance: Hidden transcripts. Yale Univ Pr.
Seeman, M. 1959. “On the meaning of alienation.” American sociological review 24(6):783–
791.
Seligson, M.A. 1983. “On the measurement of diffuse support: some evidence from Mexico.”
Social Indicators Research 12(1):1–24.
Seligson, M.A. and J.F. Carrión. 2002. “Political support, political skepticism, and political
stability in new democracies.” Comparative Political Studies 35(1):58.
Shepsle, K.A. 1979. “Institutional arrangements and equilibrium in multidimensional voting
models.” American Journal of Political Science 23(1):27–59.
228

Shi, T. 1999. “Voting and nonvoting in China: voting behavior in plebiscitary and limitedchoice elections.” The Journal of Politics 61(04):1115–1139.
Silver, B.D., B.A. Anderson and P.R. Abramson. 1986. “Who overreports voting?” The
American Political Science Review 80(2):613–624.
Skocpol, T. 1979. States and social revolutions: a comparative analysis of France, Russia,
and China. Cambridge University Press.
Snijders, T.A.B. and R.J. Bosker. 1999. Multilevel analysis: An introduction to basic and
advanced multilevel modeling. SAGE publications Ltd.
Snyder, J. 2000. “From voting to violence: democratization and nationalist violence.” New
York and London: WW Norton and Company .
Steenbergen, M.R. and B.S. Jones. 2002. “Modeling multilevel data structures.” American
Journal of Political Science 46(1):218–237.
Stokes, D.E. 1962. Popular Evaluations of Government: An Empirical Assessment. In Ethics
and Bigness: Scientific, Academic, Religious, Political and Military, ed. Harlan Cleveland
and Harold D. Lasswell (New York: Harper & Brothers, 1962). New York: Harper and
Brothers pp. 61–72.
Stokes, S.C. 2005. “Perverse accountability: A formal model of machine politics with evidence
from Argentina.” American Political Science Review 99(03):315–325.
Strom, K. 1990. “A behavioral theory of competitive political parties.” American Journal of
Political Science 34(2):565–598.
Svolik, M. 2008. “Authoritarian reversals and democratic consolidation.” American Political
Science Review 102(02):153–168.
Tarrow, S. 1998. Power in movement. Cambridge University Press Cambridge.
Tarrow, S. 2000. Mad Cows and Social Activists. In Disaffected democracies: What’s troubling the trilateral countries, ed. Susan J. Pharr and Robert D. Putnam. Princeton University Press pp. 270–90.
Tavits, Margit. 2007. “American Journal of Political Science; Clarity of Responsibility and
Corruption.” 51(1):218–229.
Tilly, C. 1978. From mobilization to revolution. McGraw-Hill.
Tilly, C. 1996. European revolutions, 1492-1992. Wiley-Blackwell.
Tingsten, H. 1975. Political behavior: Studies in election statistics. Arno Press.
Torcal, M., Richard Gunther and Jose R. Montero. 2002. Anti-party sentiments in Southern
Europe. In Political Parties: Old Concepts and New Challenges. New York: Oxford
University Press.

229

Uslaner, E.M. 2002. The moral foundations of trust. Cambridge Univ Pr.
Uslaner, E.M. 2003. “Trust, Democracy and Governance.” Generating Social Capital: Civil
society and institutions in comparative perspective p. 171.
Valenzuela, A. 2004. “Latin American presidencies interrupted.” Journal of Democracy 15:5–
19.
Van der Meer, T., M. Te Grotenhuis and B. Pelzer. 2010. “Influential Cases in Multilevel
Modeling: A Methodological Comment.” American Sociological Review 75(1):173.
Vanhanen, T. 2000. “A new dataset for measuring democracy, 1810-1998.” Journal of Peace
Research 37(2):251.
Ventura, Raphael. 2001. “Family political socialization in multiparty systems.” Comparative
Political Studies 34(6):666.
Verba, S., K.L. Schlozman and H.E. Brady. 1995. Voice and equality: Civic voluntarism in
American politics. Harvard Univ Pr.
Verba, S. and N.H. Nie. 1972. “Participation in America: Social equality and political
democracy.” New York: Harper& Row .
Verba, S., N.H. Nie and J. Kim. 1987. Participation and political equality: A seven-nation
comparison. University of Chicago Press.
Vreeland, J.R. 2008. “The effect of political regime on civil war.” Journal of Conflict Resolution 52(3):401.
Walker, J.L. 1966. “A critique of the elitist theory of democracy.” The American Political
Science Review 60(2):285–295.
Walt, S.M. 1996. Revolution and war. Cornell Univ Pr.
Warren, M.E. 1999. Democracy and trust. Cambridge Univ Pr.
Weingast, B.R. 1997. “The political foundations of democracy and the rule of law.” The
American Political Science Review 91(2):245–263.
Wilkinson, S. 2004. Votes and violence. Cambridge University Press.
Williams, Richard. 2006. “Generalized ordered logit/partial proportional odds models for
ordinal dependent variables.” The Stata Journal 6(1):58–82.
Winship, Christopher and David J. Harding. 2008. “A General Strategy for the Identification
of Age, Period, Cohort Models: A Mechanism Based Approach.” Sociological Methods and
Research 36(3):362–401.
Wolfinger, R.E. and S.J. Rosenstone. 1980. Who votes? Yale Univ Pr.

230

Wright, J.D. 1976. The dissent of the governed: Alienation and democracy in America.
Academic Press, New York.
Wrong, D.H. 1979. Power: Its forms, bases, and uses. Transaction Pub.
Zmerli, S. and K. Newton. 2008. “Social trust and attitudes toward democracy.” Public
Opinion Quarterly .

231