\begin{titlepage}
\begin{center}
\vspace*{2cm}
\Huge{\textbf{Leadership Transitions and Survival: Coups, Autocoups, and Power Dynamics}}
\vspace{1.5cm}
\Large{Zhu Qi}
\vspace{5cm}
\large{A thesis submitted for the degree of \\ Doctor of Philosophy in Political Science}
\vspace{0.8cm}
\large{Department of Government}
\vspace{0.5cm}
\large{University of Essex}
\vspace{1.5cm}
\large{September 2024}
\vspace{2cm}
\end{center}
\end{titlepage}
Leadership Transitions and Survival: Coups, Autocoups, and Power Dynamics
The completion of this thesis marks the end of a significant journey, one filled with hard work, continuous learning, and moments of profound joy. This dissertation would not have been possible without the support and encouragement of numerous individuals to whom I am deeply indebted.
First and foremost, I extend my heartfelt gratitude to my esteemed supervisor, Professor Kristian Skrede Gleditsch. His invaluable guidance, unwavering support, and insightful feedback have been instrumental in shaping this research. His expertise and encouragement have truly been the cornerstone of this academic endeavor.
I am equally grateful to Professor Han Dorussen, the chair of my board panel, for his continuous support and constructive criticism. His thoughtful input has significantly enhanced the quality and depth of my research.
I would like to acknowledge the vital contributions of my initial co-supervisors, Dr. Saurabh Pant and Professor David Siroky. Their instruction and guidance during the early stages of my research, prior to their departure from the University of Essex, laid a strong foundation for this work.
This dissertation has benefited immensely from the expertise of several esteemed scholars in the field. I am particularly indebted to Dr. Brian J Phillips, Dr. Prabin Khadka, and Dr. Winnie Xia for their insightful comments, advice, and suggestions. Their contributions have undoubtedly enriched this research.
On a personal note, I owe a debt of gratitude to my family for their unwavering support and love. To my beloved wife, Ji Zhi, who has been my steadfast pillar of strength throughout this journey - your patience and encouragement have been immeasurable. To my dear children, Siyan and Sisheng, your joy and curiosity have been a constant source of motivation and inspiration.
I am profoundly thankful to my father for his enduring support and belief in my abilities. To the cherished memory of my late mother - your love, guidance, and the values you instilled in me continue to shape my path and inspire my endeavors. And to my three brothers, whose support enabled me to pursue my PhD without worries - your contribution to this achievement cannot be overstated.
While many have contributed to the success of this work, any errors or shortcomings remain entirely my own.
Abstract
This dissertation investigates the dynamics of irregular power transitions, focusing on coups and autocoups, and their impact on leader survival. Through a comprehensive analysis of historical data, case studies, and advanced quantitative modelling, this research enhances our understanding of political power dynamics and the complex factors influencing leadership longevity following irregular transitions.
The study begins by examining the crucial role of power dynamics, shaped by regime type, in determining coup success rates and attempt frequency. Employing a double probit model with sample selection, the research demonstrates that the anticipated likelihood of coup success significantly affects coup attempts, with military regimes exhibiting particular vulnerability due to their inherent power structure.
While often overlooked in existing literature, autocoups are revealed to have a profound impact on democratic backsliding. This study introduces a refined definition of autocoups and presents a novel dataset covering events from 1945 to 2024, facilitating more robust quantitative analysis of this phenomenon.
Utilizing survival analysis techniques, the research compares the tenure of leaders who ascend to power through coups versus autocoups. The findings indicate that coup-installed leaders experience significantly shorter tenures and face higher risks of removal. In contrast, autocoup leaders, who manipulate existing systems to extend their rule, tend to remain in power longer. This disparity suggests that autocoups may incentivize power grabs and contribute to democratic erosion.
This dissertation makes several significant contributions to political science literature:
Conceptual clarity: Establishes a precise definition of autocoups, addressing a gap in the current understanding of this understudied phenomenon.
Data innovation: Introduces a comprehensive dataset on autocoups, enabling more extensive quantitative analyses and fostering future research in this area.
Analytical framework: Develops a unified approach for studying both coups and autocoups, allowing for comparative analysis of these critical political events and their potential effects on democratic backsliding.
Keywords: Coups, Autocoups, Power transitions, Leadership survival, Democratic backsliding, Regime type, Political instability
1 Introduction
1.1 Research Question and Context
Irregular power transitions, characterised by a disregard for constitutional procedures, represent a critical area of study in political science. These transitions not only disrupt established rules but also often employ unconstitutional tactics to consolidate power post-transition. Moreover, such events can inspire emulation among other ambitious leaders, potentially triggering a cascade of similar actions across different political contexts.
Despite extensive research on irregular power transitions, a fundamental question continues to challenge political scientists:
Why do some leaders face premature ousting while others extend their tenure beyond constitutionally mandated limits?
This inquiry extends to understanding the stark disparities in leadership longevity, where some rulers maintain power for decades while others’ tenures are measured in years, months, or even days.
This dissertation focuses on this central question, aiming to provide a comprehensive analysis dedicated to understanding:
How leaders ascend to or overstay in power through unconstitutional means.
What factors determine the duration of a leader’s rule following an irregular ascent to power.
1.2 Coups and Autocoups: A Comparative Unified Framework for Irregular Power Transitions
In the study of irregular power transitions, coups have traditionally dominated academic discourse due to their frequency and visibility. According to the Archigos dataset (Goemans, Gleditsch, and Chiozza 2009), coups1 accounted for more than half of the approximately 145 irregular leader exits from 1945 to 2015. The Global Instances of Coups (GIC) dataset (Powell and Thyne 2011) records an even higher number, with 245 coup-related removals from 1950 to 2024.
The prevalence of coups has led to a well-established definition and extensive research, particularly since the turn of the millennium (Thyne and Powell 2019). This scholarly focus has resulted in a general consensus on the definition of coups, as well as the development of comprehensive datasets for quantitative analyses. While debates on the precise definition persist, most scholars, including this study, adhere to the definition proposed by Powell and Thyne (2011). This widely accepted definition encompasses two fundamental elements:
Perpetrators and Victims: Perpetrators of coups are elites within the existing power structure and the victims are incumbent executive leaders.
Strategy and Aim: The primary objective is the complete removal of incumbents from power.
It is important to note that this definition excludes actions that only partially challenge the incumbent’s authority or seek policy changes without aiming for a complete transfer of power.
The scholarly consensus on coup definition has facilitated the development of several comprehensive datasets, enabling rigorous quantitative analyses in political science research. Notable among these are: GIC Dataset, Cline Centre Coup d’État Project Dataset (Peyton et al. 2024), and Colpus Dataset (Chin, Carter, and Wright 2021). These datasets have become invaluable resources in political science, enabling researchers to conduct large-scale, comparative studies on the causes, dynamics, and consequences of coups across different political contexts and historical periods.
However, another form of irregular power transition has been largely overlooked in power transition studies: leaders who refuse to relinquish power and extend their mandated terms. At least three challenges still exist in the analyses of this type of irregular power transition:
Terminology: Unlike the more widely accepted term “coup,” various terms (e.g., self-coup, autogolpe, executive coup, incumbent takeover) are used to describe this phenomenon.
Definition: Existing definitions often conflate power expansion and power extension2. For instance, Maxwell A. Cameron (1998a) defines an autogolpe primarily in terms of power expansion, which does not align with the definition of coup.
Data: A consensus dataset for autocoups is lacking, with existing datasets varying in terminology, definitions, and coverage years Baturo and Tolstrup (2022).
To address this gap in the literature and provide a framework for analysis, we propose the term “autocoup” to describe this type of transition. This concept will be extensively discussed and defined in Chapter 3, providing a basis for integrating these events into the broader study of irregular power transitions and their impact on political systems and leader longevity.
Analysing coups and autocoups within a unified framework is crucial for a comprehensive understanding of irregular power transitions and leader survival for three reasons:
Both phenomena significantly influence democratic backsliding and represent the most frequent means of irregular power transition.
Their similar nomenclature reflects a fundamental similarity: while a coup aims to replace the current leader, an autocoup seeks to prevent the succession of a future leader.
This comparative approach will enable a more nuanced analysis of irregular power transitions, contributing to our understanding of political stability, democratic erosion, and leadership longevity in various regime types.
1.3 Academic Contributions
This study addresses a critical gap in the literature by offering a unified framework for analysing both coups and autocoups. My contributions are threefold:
Emphasis on Power Dynamics and Regime Types: This study highlights the significant role of power dynamics, particularly the influence of regime types, in determining the success and frequency of coup attempts, underscores how the expected chances of coup success motivate such attempts, with military regimes being notably susceptible.
Refined Definition and Novel Dataset for Autocoups: I introduce a refined definition of autocoups, develop a novel dataset covering events from 1945 to 2024, and fill a significant gap in the existing literature and enable a comparative analysis with classic coups.
Survival Analysis of Leaders from Different Entry Modes: This research applys survival analysis to existing coup data and the new autocoup dataset, demonstrates how different modes of entry into power significantly affect leader survival, and reveals that leaders who come to power through coups typically have shorter tenures and face higher removal risks compared to those who extend their rule through autocoups.
These contributions collectively advance the field by providing a more holistic understanding of irregular power transitions, offering new tools and data for quantitative analysis of autocoups, and demonstrating the interconnectedness of power acquisition methods and leadership survival.
This work lays the foundation for future research on the dynamics of political power, regime stability, and democratic backsliding, offering both theoretical insights and practical implications for policy-makers and scholars alike.
1.4 Implications
The examination of irregular power transitions provides a crucial perspective on the interrelated phenomena of democratic backsliding, breakdown, and autocratic intensification. The findings of this study provide logical explanations for several political phenomena:
Regression of Global Democracy Levels: This study can explain why global democracy levels have regressed to pre-2000 levels. For example, Freedom House reports an 18th consecutive year of global freedom decline in 2023 (Freedom House 2024). Irregular power transitions inevitably violate democratic or constitutional norms and disrupt the trajectory towards stable democracies or democratization from autocracies. Leaders who ascend through irregular means often undermine constitutional norms to seize or overstay in power, creating a vicious cycle of eroding democratic institutions.
Within-Regime Democratic Erosion: This study explains why democratic backsliding often occurs within regimes, with democracies becoming less liberal and autocracies becoming less competitive (Mechkova, Lührmann, and Lindberg 2017). This is particularly because autocoups have become more prevalent than traditional coups since 2000 (Bermeo 2016). Importantly, autocoups typically do not result in immediate regime change, contributing to the observed pattern of within-regime democratic erosion.
Prevalence of Autocoups Since 2000: This research explains why autocoups are more prevalent since 2000. Firstly, autocoups have a significantly higher probability of success compared to traditional coups. Additionally, even when unsuccessful, the repercussions of autocoups are generally less severe than those of failed traditional coups. Moreover, successful autocoup leaders tend to maintain power for substantially longer periods than leaders who enter through traditional coups.
1.5 Overview of the Thesis
This dissertation investigates the complex dynamics of irregular power transitions and their implications for leadership survival and democratic processes. The study examines three key aspects: the determinants of classic coup attempts, the conceptualization and analysis of autocoups, and the impact of power acquisition methods on leadership longevity.
Chapter 2: Determinants of Classic Coup Attempts
This chapter delves into the factors influencing classic coup attempts, with a novel focus on the less observable but crucial factor of expected coup success rates. Key features include the utilization of a double probit model with sample selection and the analysis of how expected success rates significantly influence coup attempts. It also examines how regime types shape the balance of power between incumbents and challengers, determining the chances of success of coups. Findings indicate a substantially higher coup risk in military regimes compared to dominant-party regimes.
Chapter 3: Conceptualising and Analysing Autocoups
This chapter refines and analyses the concept of autocoups, with a specific focus on power extensions by incumbent leaders. Notable elements include the redefinition of autocoups as instances where incumbent leaders refuse mandated power transitions. Introduction of a novel dataset covering autocoup events from 1945 to 2024, encompassing 110 attempts and 87 successes. Presentation of case studies and empirical analyses demonstrating the utility of this dataset for more quantitative research in political science.
Chapter 4: Impact of Power Acquisition Methods on Leadership Longevity
This chapter investigates how the method of power acquisition affects the tenure of leaders who ascend through coups versus those who extend their tenure via autocoups. Key aspects include hypothesis testing on the significant impact of the accession method on leader tenure. Evidence of differing survival times between coup-entry and autocoup leaders, employing the Cox proportional hazards model and time-dependent Cox model. Analysis of how the risk-reward profile of autocoups might motivate democratic backsliding and power personalization.
Chapter 5: Conclusion and Future Directions
The final chapter synthesizes the summary of key findings from each substantive chapter. Acknowledges study limitations and their potential impact on findings. Outlines future research directions, emphasizing the need for further exploration of irregular power transitions, particularly coups and autocoups.
2 Power Dynamics and Coup Attempts: A Selection Mechanism Analysis
Abstract
Despite extensive research identifying around one hundred potential determinants of coup attempts, no consensus has been reached. This study introduces a novel approach that prioritizes determinants based on their impact on coup success. By analysing coup success rates, the study hypothesizes that the expected outcomes of coups are critical determinants of their occurrence. Utilizing a double probit model with sample selection, the research investigates the relationship between regime types and coup attempts. The findings confirm that regime type, by shaping internal power dynamics, is a crucial determinant of coup likelihood. Military and personalist regimes, characterized by weaker institutional frameworks and higher vulnerability, demonstrate significantly higher susceptibility to coups compared to dominant-party regimes.
Keywords: Coups, Power transitions, Sample selection
Coups d’état, defined by Powell and Thyne (2011) as “illegal and overt attempts by the military or other elites within the state apparatus to unseat the sitting executive” (p. 252), represent a critical challenge to political stability and democratic governance worldwide. This chapter examines the complex dynamics of coup attempts, their success rates, and the factors that influence both their occurrence and outcomes.
Coups occur with varying frequency across countries and regions. For instance: In Latin America, Bolivia experienced 23 coups between 1950 and 1984, while Argentina saw 20 during a similar period. In Africa, Sudan endured 17 coups between 1955 and 2023. Contrastingly, countries like Mexico (1917-2000) and South Africa (since 1950) have not experienced any coups.
This variability raises a fundamental question: Why are coups more frequent in some countries than others?
Despite decades of research, political scientists have yet to reach a consensus on the key determinants of coups. Gassebner, Gutmann, and Voigt (2016) highlight this challenge, noting that approximately 100 potential factors have been proposed, with their study testing 66 of these across three million model permutations. This proliferation of variables presents a critical issue: How can we establish a framework that allows scholars to focus on the most relevant factors, rather than navigating an ever-expanding list of potential determinants?
My analysis reveals a significant oversight in previous research: the exclusive focus on pre-coup conditions without adequate consideration of post-coup factors. This approach has neglected a critical element in coup dynamics – the expected probability of success. Key observations supporting this perspective, as shown in Table 2, include:
High-stakes nature of coups: Failed coups often result in severe consequences for perpetrators, including imprisonment, exile, or death.
Selectivity in coup attempts: Despite 491 coup attempts since 1950, they represent only about 4% of over 12,000 country-years in the same period (GIC dataset).
Satisfactory success rate: Approximately half of all coup attempts succeed, suggesting careful selection of opportunities by plotters.
Table 2: Top 10 countries with the most coup attempts
Country
Coup Attempted
Coup Succeeded
Success Rate
Bolivia
23
11
47.8%
Argentina
20
7
35.0%
Sudan
17
6
35.3%
Haiti
13
9
69.2%
Venezuela
13
0
0.0%
Iraq
12
4
33.3%
Syria
12
8
66.7%
Thailand
12
8
66.7%
Ecuador
11
5
45.5%
Burundi
11
5
45.5%
Guatemala
10
5
50.0%
Total
491
245
49.9%
Source: GIC dataset
The low occurrence rate and high success rate clearly indicate that the initiation of coups is highly selective. In other words, the likelihood of a coup occurring depends greatly on its probability of success. However, this probability is not directly observable to outsiders, including researchers, prior to a coup attempt.
Given the limitations in directly observing coup success probabilities, I propose focusing on regime type as a crucial proxy for coup outcomes. This approach is based on the following reasoning: Coup outcomes are ultimately determined by the balance of power within a regime and regime types are classified based on control over the military, policy-making authority, and appointment of officials. By analysing power dynamics across different regime types, we can gain valuable insights into the structural factors shaping coup attempts and their outcomes.
To address the selection bias inherent in studying coup attempts, I employ a double probit model with sample selection. This approach allows for simultaneous analysis of factors influencing both coup initiation and success.
Through this comprehensive analysis, I aim to provide a more nuanced understanding of the factors driving coup attempts and shaping their outcomes, contributing to both scholarly discourse and practical efforts in two key ways:
Emphasis on expected chances of success: By focusing on the probability of success as a driver of coup attempts, I offer a more targeted approach to understanding coup dynamics.
Highlighting the significance of regime type: I demonstrate how regime type influences coup likelihood, even when researchers lack perfect knowledge of internal balance of power.
The remainder of this chapter is organized as follows: Section 2 explores the dynamics of coup attempts and their outcomes. Section 3 outlines the research design, methodology, and variables. Section 4 presents and discusses the empirical findings. Section 5 concludes with key insights and their implications for understanding and potentially mitigating coup risks.
2.2 Dynamics of coup attempts and outcomes
Coup attempts are driven by a complex interplay of factors, encompassing both the motivations of potential challengers (disposition) and the resources and opportunities available to them (capability). This section delves into the dynamics of coup attempts, exploring the motivations behind them, the factors influencing their success, and the role of regime types in shaping coup susceptibility.
2.2.1 Motivations for coups
This section focuses on the motivations that compel actors to undertake coups, categorizing them into three main types:
Personal Ambition: The allure of absolute power, prestige, and wealth serves as a significant motivator for many coup plotters. Seizing control promises the ability to shape national policies, control resources, and make impactful decisions without constraints. The pursuit of prestige, recognition, potential economic gain, and the desire to leave a lasting legacy can further incentivize individuals to undertake this risky endeavour.
Purported National Interest: Coups are sometimes justified as necessary interventions to address national crises, uphold the constitution, or facilitate a transition to democracy. While such claims require scrutiny, genuine examples exist. For instance, the 2010 coup in Niger ousted President Tandja, who attempted an unconstitutional third term by dissolving the opposing court and calling a self-serving referendum (Ginsburg and Elkins 2019).
Self-Preservation: In certain cases, coups act as pre-emptive strikes against perceived threats. Coup leaders might not necessarily seek power for themselves but rather fear elimination or political persecution by the incumbent leaders. An example is Idi Amin’s 1971 coup against Ugandan President Obote, who was attempting to remove Amin from his military command position (Sudduth 2017).
These motivations are often most prevalent in autocratic regimes, where justifications under the guise of national interest or self-preservation can mask personal agendas. Stable democracies rarely face the same level of constitutional crises or political persecution that might necessitate a coup. However, newly established democracies can be vulnerable to instability, economic downturns, and democratic backsliding, creating opportunities for coup plotters to exploit these weaknesses and justify their actions.
Despite these potential motivations, coups remain relatively uncommon, occurring in only about 4% of country-years since 1950. This rarity highlights the importance of capability – even the most motivated actors need the resources and opportunities to succeed.
2.2.2 Capability for coups
The decision to attempt a coup hinges not only on motivation but also on a calculated assessment of the chances of success. Several factors influence this assessment:
Military Strength: A clear advantage in military capabilities compared to the incumbent regime significantly increases the odds of a successful coup (Powell et al. 2018; Choulis et al. 2022).
Internal Divisions within the Regime: Existing fractures within the government’s power structure can be exploited by coup plotters to gain support from disgruntled factions.
Public Support: Widespread discontent with the incumbent regime, especially within the military or key sectors of society, can create a ripe environment for a successful coup.
Foreign Backing: External support from powerful nations can provide resources, legitimacy, and even direct military intervention to tip the scales in favor of the coup plotters.
While historical data might suggest a high success rate for coups, it is crucial to consider selection bias. We only observe attempted coups, not the numerous plots that never materialize. Analysing launched coup data alone can be misleading. To understand coup attempts and their likelihood comprehensively, we need a theoretical framework that accounts for this selection bias.
2.2.3Effect of Expected Coup Outcomes on Coup Attempts
A frequently cited framework (Gassebner, Gutmann, and Voigt 2016; Aidt and Leon 2019) offers a structured approach to assess the disposition and capability of coup attempts by evaluating the anticipated benefits for coup plotters. The expected pay-off of a coup can be represented by the equation:
\[
\begin{aligned}
E(U) = p \times B + (1 - p) \times (-C)
\end{aligned}
\tag{1}\]
Where:
\(E(U)\): Expected utility or pay-off of the coup attempt;
\(B\) represents the return of a successful coup;
\(C\) signifies the cost of a failed coup;
\(p\) represents the probability of coup success.
The condition for staging a coup is when the expected benefit is positive (\(E(U) > 0\)). Rearranging the equation, we get:
\[
\begin{aligned}
p \times B > (1 - p) \times C
\end{aligned}
\tag{2}\]
This implies that for a coup to be attempted, the expected benefits of success must outweigh the expected costs of failure.
Quantifying \(B\) and \(C\) is inherently difficult. The loss of life, freedom, or loved ones after a failed coup, as well as the value of assuming leadership after a successful coup, are challenging to measure precisely.
However, the framework’s core logic remains valuable. Given the difficulty in precisely quantifying \(B\) and \(C\), we can treat them as roughly equal. This allows us to shift our focus to the probability of success (\(p\)). The simplified Equation 2 becomes:
\[
\begin{aligned}
p > (1-p)
\end{aligned}
\tag{3}\]
This suggests that a success probability greater than 50% is necessary for a coup to be attempted. While empirical data shows a slightly lower overall success rate for coups since 1950 (49.9%, as shown in Table 2), it is crucial to remember that this is an average and does not reflect the specific probabilities assessed by coup plotters beforehand.
Therefore, in this study, unlike the majority of previous research that focuses on the prerequisites for triggering a coup, I consider the expected coup outcome as the fundamental determinant of a coup. Based on this discussion, I propose the first hypothesis:
H1: The fundamental determinant of a coup attempt is the perceived chance of success. Coup plotters likely require a success threshold of at least 50%.
This leads to the next question: What factors determine coup success and consequently influence the decision to attempt one? The answer lies in understanding the balance of power inherent in the different power dynamics across various regime types.
2.2.4 Regime types and power dynamics
Historical coup success rates do not dictate individual coup attempts as each one faces different situation and conditions. Coup plotters assess their chances based on their unique context. Although many factors determine the success of a coup, ultimately, the decisive factor remains the comparison of strength between incumbents and coup perpetrators. More precisely, it is the comparison of their military power.
While military strength is often seen as the primary factor in coup attempts, leading to an oversimplification of coups as solely military events, the reality is far more complex (Singh 2016). The clandestine nature of coups necessitates small, secretive groups, making it difficult to gauge the stance of other factions within the military. As Geddes (1999) points out, the success of a coup often hinges on the reactions of these other factions. This uncertainty creates a complex decision-making environment for potential coup plotters. Furthermore, incumbents often employ strategic division within the military as a coup-proofing measure. Böhmelt and Pilster (2014) describe how leaders intentionally create rival groups within the armed forces, establishing an artificial balance and structural obstacles. In addition, factors beyond military force, such as internal divisions within the ruling elites, public support, and foreign backing, significantly shape the balance of power.
Therefore, a more holistic approach involves analyzing the overall balance of power within a political system. However, as noted in the introduction of this chapter, directly observing this balance of power presents significant challenges for outside observers. In stead, an alternative approach is to analyze the factors that are decisive in shaping power dynamics. This method aligns more closely with academic research goals, which are less concerned with identifying specific power challengers capable of launching successful coups and more focused on understanding the patterns and underlying factors of coup attempts.
Among the various factors that influence power dynamics, regime type emerges as one of the most crucial. First of all, regime types are fundamentally classifications based on power structures within a political system. Secondly, different regime types allocate authority differently for critical decisions such as: deployment of military forces, appointment of key officials and military officers, and formulation and implementation of main policies.
provide a classification of autocratic regime types based on their power structures. This classification offers valuable insights into the power dynamics within different autocracies and their relative susceptibility to coups. The three main types of autocratic regimes, their characteristics, and their vulnerability to coup attempts are explored as following (Table 3).
Geddes, Wright, and Frantz (2014) provide a classification of autocratic regime types based on their power structures. The three main types in autocracies and their power dynamics are:
Military Regimes: Characterized by a junta – a group of military officers controlling leadership selection and policy formulation. Examples include regimes in Brazil (1964-1985), Argentina (1976-1983), and El Salvador (1948-1984) (Geddes 1999).
Coup Susceptibility: Despite their military nature, these regimes are surprisingly unstable due to internal power struggles within the junta. The absence of a clear final authority and the presence of multiple military factions increase the likelihood of resorting to force to resolve disputes, making these regimes the most vulnerable to coups.
Personalist Regimes: Power is concentrated in a single, charismatic leader who controls the military, policy, and succession. Examples include Rafael Trujillo’s regime in the Dominican Republic (1930-1961), Idi Amin’s regime in Uganda (1971-1979), and Jean-Bédel Bokassa’s regime in the Central African Republic (1966-1979) (Geddes 1999).
Coup Susceptibility: Personalist regimes are relatively stable during the leader’s tenure. However, they face a higher risk of coups due to unclear succession plans and vulnerabilities associated with the leader’s personal weaknesses, health, and mortality.
Dominant-Party Regimes: Power resides within a well-organized ruling party, with leaders acting as its representatives. The party structure and ideology foster internal cohesion and a long-term vision. Examples include the Partido Revolucionario Institutional (PRI) in Mexico, the Revolutionary Party of Tanzania (CCM), and Leninist parties in various Eastern European countries (Geddes 1999).
Coup Susceptibility: Dominant-party regimes exhibit the greatest resilience against coups due to their institutionalized structures, unified leadership, clear ideology, and internal discipline.
Empirical data also supports this theoretical framework. While military regimes represent only 5.6% of country-years since 1950, they experience a disproportionate share of coups (over 22%). Personalist regimes, constituting 13% of country-years, account for 23% of coups. Conversely, dominant-party regimes, representing 22.6% of country-years, account for only 16.7% of coups (Table 4). These statistics demonstrate the disproportionate vulnerability of military and personalist regimes to coups, while highlighting the relative stability of dominant-party regimes.
The distinct power dynamics exhibited by these regime types significantly influence their susceptibility to coups. This analysis leads to our second hypothesis:
The susceptibility to coups varies significantly among different types of autocratic regimes, with military regimes being the most vulnerable, followed by personalist regimes, and dominant-party regimes being the least susceptible.
H2: The susceptibility to coups varies significantly among different types of autocratic regimes, with military regimes being the most vulnerable, followed by personalist regimes, and dominant-party regimes being the least susceptible.
This study employs a sophisticated statistical approach to account for the selective nature of coup attempts. While coup attempt rates vary across regimes, success rates tend to be surprisingly consistent, hovering around 50% (as shown in Table 4). This suggests that coup attempts are not random acts, but rather strategically planned and undertaken only when the odds of success appear favorable. A standard statistical model would not account for this selectivity, potentially leading to biased results.
To address this issue, we utilize a double probit with sample selection model, similar to the approach used by J. Powell (2012) and Böhmelt and Pilster (2014). This model, known as a Heckman probit model or bivariate probit model with sample selection, consists of two parts:
Selection Equation (Stage 1): This stage analyses the factors influencing whether a coup attempt occurs in a particular country-year.
Outcome Equation (Stage 2): This stage focuses on the probability of success for those coup attempts that actually take place.
The selection equation (first stage) models the probability that a coup attempt occurs:
Where \(\Phi(\cdot)\) is the cumulative distribution function of the standard normal distribution.
2.3.2 Variables
2.3.2.1 Dependent variable
The analysis utilizes data on coup attempts and outcomes from Powell and Thyne (2011). A successful coup is defined as one where the incumbent leader is removed from power for more than seven days. The dataset covers the period from 1950 to 2023 and includes information on 491 coup attempts, with roughly half (245) being successful. Descriptive statistics for these coup attempts and regime types can be found in Table 2 and Table 4.
Coup Attempt: Binary variable indicating whether a coup attempt occurred (1) or not (0) in a given country-year.
Coup Success: Binary variable indicating whether a coup attempt was successful (1) or failed (0), conditional on a coup attempt occurring.
2.3.2.2 Key Independent Variable: Regime Type
I categorize regime types following Geddes, Wright, and Frantz (2014) (GWF), focusing on military, personalist, and dominant-party regimes, with democracies and monarchies included for comparison. Descriptive statistics for regime types are presented in Table 4.
Control variables
The control variables are chosen based on the research of Gassebner, Gutmann, and Voigt (2016). They analysed 66 factors potentially influencing coups and found that slow economic growth, prior coup attempts, and other forms of political violence are particularly significant factors. Therefore, we include economic performance, political violence, and the number of previous coups as our main control variables.
Economic Level: Represented by GDP per capita. This measure provides an indication of the overall economic health and standard of living in a country. We use GDP per capita data (in constant 2017 international 1000 dollars, PPP) from the V-Dem dataset by Fariss et al. (2022).
Economic Performance: Measured using the current-trend (\(CT\)) ratio developed by Krishnarajan (2019). This ratio compares a country’s current GDP per capita to the average GDP per capita over the previous five years. A higher \(CT\) ratio indicates stronger economic performance. For a country \(i\) at year \(t\), the \(CT\) ratio is calculated as follows:
Political Stability: This variable captures overall regime stability by including a violence index that encompasses all types of internal and interstate wars and violence. The data for this index is sourced from the variable “actotal” in the Major Episodes of Political Violence dataset (Marshall 2005), with 0 representing the most stable conditions (no violence at all) and 18 representing the most unstable.
Previous coups: Included in the selection equation as either: a) The number of previous coups in a country (Model 1), or b) The time since the last coup attempt (Model 2 for robustness check).
Table 5: Sample Selection Model of Regime Type and Coup Success, 1950-2019
The double probit model with sample selection, estimated using the sampleSelection package (Toomet and Henningsen 2008) in R, provides valuable insights into the factors influencing coup attempts and their outcomes across different regime types from 1950 to 2019 (Table 5). I present two models that differ slightly in their treatment of previous coups: Model 1 incorporates the number of previous coups, while Model 2 utilizes the time elapsed since the last coup.
2.4.1 Selection Model: Coup Attempts
In the selection model (Column 1), military and personalist regimes exhibit significant positive coefficients at the 1% level, indicating a higher likelihood of experiencing coup attempts compared to dominant-party regimes. This aligns with our theoretical expectations regarding internal power struggles within military juntas and succession vulnerabilities in personalist regimes.
Table 6: Average marginal effects of coup attempts (Selection of Model 1)
Term
Contrast
AME1
Ratio Percent
Regime: Democracy
mean(democracy - dominant-party)
0.003
13.040
Military
mean(military - dominant-party)
0.070
277.730
Monarchy
mean(monarchy - dominant-party)
0.020
80.280
Personal
mean(personal - dominant-party)
0.024
93.980
Economic trend
mean(+1)
−0.001
−2.850
GDP per capita
mean(+1)
−0.002
−5.400
Political violence
mean(+1)
0.003
6.550
Previous coups
mean(+1)
0.002
5.930
1 AME: Average Marginal Effect
Table 6 clarifies the regime effects using Average Marginal Effects (AME) and ratio percentages. The military regime’s marginal effect of 0.07 indicates that the probability of coup attempts in military regimes is 7 percentage points (pp) higher than in dominant-party regimes, ceteris paribus. This translates to military regimes being about 277.7% more likely to encounter coups than dominant-party regimes. Similarly, personalist regimes show a 2.4 pp higher probability, about 94% more likely compared to dominant-party regimes.
Control variables show effects in expected directions but with weaker magnitudes. Stronger economic performance, indicated by higher economic growth trends and GDP per capita levels, correlates with a lower risk of coup attempts. Political violence shows a positive effect, indicating that higher levels of instability increase the likelihood of coups. The positive coefficient for the number of previous coups suggests a “copycat” effect from earlier incidents.
2.4.2 Outcome Model: Coup Success
The outcome model (Columns 2 and 4 in Table 5) reveals determinants of coup success. Military regimes demonstrate a higher probability of coup success compared to dominant-party regimes, aligning with expectations that military regimes face higher coup risks due to their increased chances of success. Personalist and monarchical regimes show slight positive effects on coup success, but these effects are not statistically significant.
Control variables exhibit different patterns in the outcome model compared to the selection model. Both GDP per capita and political violence maintain a weak influence, similar to their effects in the selection model. However, the economic trend shows a less significant negative effect on coup success.
2.4.3 Model Comparison (Model 1 vs Model 2)
Model 2 employs years since the last coup instead of the number of previous coups, with an interaction term between previous coups (as a binary variable) and years since the last coup. Generally, Model 2 shows results in the same direction as Model 1, albeit with relatively lower coefficients (Table 7).
The differences between Model 1 and Model 2 suggest that while the recency of coups matters, the overall history of coups in a country may have a stronger influence on future coup attempts.
Table 7: Average marginal effects of coup attempts (Selection of Model 2)
Term
Contrast
AME1
Ratio Percent
Regime: Democracy
mean(democracy - dominant-party)
0.003
8.920
Military
mean(military - dominant-party)
0.028
91.630
Monarchy
mean(monarchy - dominant-party)
0.018
56.730
Personal
mean(personal - dominant-party)
0.009
30.080
Economic trend
mean(+1)
−0.001
−2.530
GDP per capita
mean(+1)
−0.001
−2.890
Political violence
mean(+1)
0.003
7.330
Previous coups (P)
mean(1 - 0)
0.023
92.090
NA
mean(+1)
−0.002
−5.050
1 AME: Average Marginal Effect
2.4.4 Discussion of key findings
The \(\rho\) values of 0.898 in Model 1 and 0.386 in Model 2, significant at 1% and 10% levels respectively, indicate strong correlation between unobserved factors influencing coup attempts and coup success. This supports the appropriateness of the sample selection model and underscores the importance of considering both stages in the analysis.
The significant coefficients with theoretically consistent directions suggest the model effectively captures key aspects of coup dynamics. The observed disparity between coup attempt rates and success rates across regimes points towards selection bias, further validating the use of the sample selection model.
2.4.5 Theoretical Implications
These results strongly support my theoretical framework, highlighting the crucial role of regime structure in determining coup vulnerability. The findings underscore that coups are strategic actions undertaken when odds appear favourable, rather than random events.
2.5 Conclusion
This study addresses the lack of consensus in empirical research on coup predictors by introducing a novel approach that prioritizes determinants based on their impact on coup success. By analysing coup success rates, I posit that the expected outcomes of coups are critical determinants of their occurrence. Employing a double probit model with sample selection, the research investigates and confirms a strong and robust relationship between regime types and coup attempts.
The main findings reveal that regime type plays a pivotal role in the likelihood of coup attempts. Military and personalist regimes, characterized by weaker institutional frameworks and higher vulnerability, demonstrate significantly higher susceptibility to coups compared to dominant-party regimes. This underscores the importance of supporting initiatives that establish regimes with constitutional institutions rather than those dependent on military power or personal authority, as the latter prove more volatile and coup-prone.
The research also indicates that stronger economic performance, while not as influential as regime type, is associated with a lower risk of coups. This suggests that policies promoting economic development can be effective in reducing coup risk. Given that regime type is often determined during regime formation and is difficult to alter, promoting economic growth might be the most viable coup-proofing strategy available to incumbents.
However, these findings present a paradox for autocratic leaders, particularly dictators and military juntas. While institutionalization of regimes could enhance stability and reduce coup risk, few such leaders are willing to implement these reforms. This reluctance stems from the potential constraints on their power or shortened terms that such changes might entail. Thus, while institutions may benefit the regime’s longevity, they do not necessarily serve the immediate interests of individual leaders.
This study opens avenues for further research, including:
Investigating the long-term effects of regime types on political stability, including democratization or personalization.
Examining how international factors might influence regime choices and coup dynamics when internal factors are too weak to prevent coups.
In conclusion, this chapter provides robust evidence for the critical role of regime type in determining coup risk, while also acknowledging the complex realities of political power. By illuminating the strategic calculus behind coup attempts, it offers valuable insights for scholars seeking to understand political instability and policymakers working to promote democratic transitions and political stability in diverse global contexts.
3 Autocoups: Conceptual Clarification and Analysis of Power Extensions by Incumbent Leaders
Abstract
This study aims to clarify the concept of autocoups, specifically focusing on power extensions by incumbent leaders. By distinguishing autocoups from the broader and more ambiguous concepts of self-coups or executive takeovers, which encompass both executive power aggrandizement and power extension, this research redefines the concept of autocoups. Based on this refined definition, I introduce a novel dataset of autocoup events from 1945 to 2024. Using this newly compiled dataset, the research includes three types of case studies that provide qualitative insights into the dynamics of autocoups. Additionally, an empirical analysis of the determinants of autocoup attempts and success is offered to demonstrate how the autocoup dataset can be employed for more quantitative research. This study contributes to the existing literature by providing a clearer conceptual framework and a novel dataset of autocoups. It enhances our understanding of the mechanisms and motivations behind power extensions by incumbent leaders and examines the implications for democratic backsliding, democratic breakdown, and autocratic deterioration. The insights gained from this study could draw more attention to the effects of autocoups on power transitions, political stability, and democratic resilience.
keywords: Autocoups, Coups, Power transitions
3.1 Introduction
The study of irregular power transitions has predominantly focused on coups due to their frequency and significant impact. This emphasis has led to substantial scholarly attention and the creation of comprehensive datasets. However, as noted by Gassebner, Gutmann, and Voigt (2016), the multitude of potential factors proposed to explain coups has often resulted in increased confusion rather than a clearer understanding of coup dynamics.
In contrast, another form of irregular power transition—the incumbent leader’s refusal to relinquish power—has received relatively less attention despite its significance. Recent decades have witnessed a decline in traditional coups and a rise in this “incumbent retention or overstay” type of irregular power transition, particularly since the end of the Cold War (Ginsburg, Melton, and Elkins (2010); Baturo (2014); Versteeg et al. (2020)).
This chapter will redefine and clarify this type of irregular power transition, where leaders overstay their mandated term limits, as “autocoup.” Although analyses related to autocoups are not rare, the existing literature exhibits several notable shortcomings:
Terminological Ambiguity: Terms such as self-coups, autocoups, autogolpes, incumbent takeovers, executive aggrandizement, overstay, and continuismo are used without a clear, universally accepted definition. This lack of standardization leads to confusion and inconsistent application (Marsteintredet and Malamud 2019; Baturo and Tolstrup 2022).
Limited Data: Due to the conceptual ambiguity surrounding autocoups, data collection remains in its early stages compared to the rich datasets available for traditional coups.
Moreover, analyses of autocoups are often not integrated with those of traditional coups. As a distinct category of coup, autocoups lack a corresponding definition in relation to classic coups. Consequently, traditional coups and autocoups are frequently analysed separately, despite their related nature, and there is a paucity of comparative analyses.
Studying autocoups is important for several reasons. Firstly, autocoups can undermine the rule of law, weaken institutions, and contribute to democratic backsliding or authoritarian personalization. Secondly, like traditional coups, successful autocoups increase the likelihood of future irregular power transitions. For example, since 1945, approximately 62% of leaders who extended their terms through autocoups in non-democratic countries were either ousted or assassinated while in office (Baturo 2019). Thirdly, failed autocoups often lead to instability, inciting protests, violence, and even civil wars.
This chapter aims to address these gaps by focusing on autocoups, aiming to clarify terminology, refine concepts and definitions, enhance data collection, and explore determinants through empirical analysis, contributing in three key areas:
Conceptual Clarification: The term “autocoup” will be redefined and clarified, with a focus on power extension.
Data Collection: A new dataset of autocoups since 1945 will be introduced based on this refined definition.
Empirical Analysis: Utilizing this dataset, a quantitative analysis of the factors influencing leaders’ decisions to attempt autocoups will be conducted.
The structure of this chapter is as follows: Section 2 will review definitions related to power expansions and extensions, leading to a precise definition of autocoups. Section 3 will present the new autocoup dataset. Sections 4 and 5 will explore the determinants of autocoup attempts through case studies and demonstrate the application of the dataset in empirical analysis. The conclusion will summarize key findings and suggest directions for future research.
Definitions of power expansion and power extension in political science can often be ambiguous or overlapping. To ensure clarity in this study, we propose distinct definitions for these concepts:
Power Expansion: This refers to the process by which an incumbent leader acquires additional authority or control over state apparatuses beyond their original mandate. This may involve centralizing power, reducing checks and balances, or assuming roles typically held by other branches of government.
Power Extension: This concept specifically relates to the temporal aspect of an incumbent’s rule. It describes situations where a leader prolongs their tenure beyond the originally mandated term in office, often through constitutional amendments, cancellation of elections, or other means of circumventing term limits.
By distinguishing between these two concepts, we aim to provide a more nuanced analysis of how leaders consolidate and maintain their positions. While power expansion and extension can often occur simultaneously, differentiating between them allows for a more precise examination of the strategies employed by leaders to entrench their rule.
This distinction is crucial for our study as it enables us to: 1. Analyze the different mechanisms leaders use to solidify their power 2. Examine the potential consequences of each type of power consolidation 3. Explore the relationship between power expansion and extension in various political contexts
Understanding these nuances is essential for comprehending the complex dynamics of political power and leadership longevity in both democratic and authoritarian systems.”
3.2 Autocoups: A literature review and clarification of definitions
The concept of autocoups, or the illegitimate seizure of power by incumbent leaders, remains a complex and contested area of study. Unlike coups, which are generally understood as illegal attempts by elites to overthrow the government, autocoups lack a clear and consistent definition. This ambiguity hinders our ability to accurately identify, analyze, and compare instances of this phenomenon.
3.2.1 Terminology
The most common term in autocoup literature is “self-coup,” or “autogolpe” in Spanish (Przeworski et al. 2000; Maxwell A. Cameron 1998a; Bermeo 2016; Helmke 2017; Marsteintredet and Malamud 2019). This term gained academic prominence after Peruvian President Alberto Fujimori dissolved Congress, temporarily suspended the constitution, and ruled by decree in 1992 (Mauceri 1995; Maxwell A. Cameron 1998b). However, as Marsteintredet and Malamud (2019) point out, the term “self-coup” can be misleading, as it implies a coup against oneself, which is inaccurate since it targets other state institutions or apparatus.
Another approach to describe coups staged by incumbents is to use terms with adjectives or modifiers, such as “presidential coup,” “executive coup,” “constitutional coup,” “electoral coup,” “judicial coup,” “slow-motion coup,” “soft coup,” and “parliamentary coup” (Marsteintredet and Malamud 2019). While these terms can be helpful in specific contexts, their proliferation often adds to the overall confusion rather than providing clarification. Most of these terms focus on the specific methods used by coup perpetrators but fail to clearly identify the perpetrator, necessitating further explanation. In fact, many of these methods could be employed either by or against executive leaders.
A third alternative involves terms like “incumbent takeover,” “executive takeover,” or “overstay.” Incumbent takeover refers to “an event perpetuated by a ruling executive that significantly reduces the formal and/or informal constraints on his/her power” (Baturo and Tolstrup 2022, 374), based on earlier research (Svolik 2014). Meanwhile, overstay is defined as “staying longer than the maximum term as it stood when the candidate originally came into office” (Ginsburg, Melton, and Elkins 2011, 1844). These terms identify the perpetrator (the incumbent) and/or the nature of the event (overstaying/extending power). However, they do not highlight the illegality or illegitimacy of these actions. Therefore, they cannot serve as a direct counterpart to “coup,” which clearly denotes the illegality of leadership ousters, while “takeover” or “overstay” diminish the severity.
As these terms often lack precision, focusing on specific methods rather than the core act of power usurpation, this study proposes “autocoup” as the most suitable term for this phenomenon. Unlike other terms, ‘autocoup’ clearly identifies the perpetrator and the illegitimate nature of the power grab, distinguishing it from traditional coups.
3.2.2 Definition
While terminology is important, another issue arises with the previous definition of autocoups: What is the emphasis of an autocoup—power expansion, power extension, or both?
Definitions of power expansion and power extension in political science can often be ambiguous or overlapping. To ensure clarity in the study of autocoups, I propose distinct definitions for these concepts:
Power Expansion: This refers to the process by which an incumbent leader acquires additional authority or control over state apparatuses beyond their original mandate. This may involve centralizing power, reducing checks and balances, or encroaching on the authority of other branches like the legislature or judiciary.
Power Extension: This describes situations where a leader prolongs their tenure beyond the originally mandated term in office, often through constitutional amendments, cancellation of elections, or other means of circumventing term limits.
Existing definitions of autocoups or related concepts often either are ambiguous between power expansion and extension, or focus more on power expansion, which has several drawbacks.
Firstly, defining autocoups primarily in terms of power expansion does not align well with the definition of a coup. The focus of a classical coup is clearly on the ouster of the current leader, not merely a limitation or restriction on their power. Using the same logic, a more appropriate definition of an autocoup should prioritize the tenure extension of executive leadership. Power restriction on incumbents would not be coded as a coup as long as they remain in office. Similarly, an executive leader acquiring more power from other branches could be coded as power aggrandizement, but not an autocoup, as long as they step down when their term expires.
Secondly, emphasizing power expansion in autocoups often neglects the ultimate purpose of incumbents. It is irrational for an incumbent to expand executive power and then pass the powerful role to future leaders. Although the term “self-coup” gained prominence from the 1992 Fujimori case in Peru, which initially involved seizing power from other institutions, it is important to note that Fujimori ultimately extended his term limits through constitutional amendments. The 1993 Constitution allowed Fujimori to run for a second term, which he won with popularity in April 1995. Shortly after Fujimori began his second term, his supporters in Congress passed a law of “authentic interpretation” which effectively allowed him to run for another term in 2000, which he won amid suspicions and rumors. However, he did not survive the third term. In 2000, facing charges of corruption and human rights abuses, Fujimori fled Peru and took refuge in Japan (Ezrow 2019).
Thirdly, measuring the extent of power expansion to qualify as an autocoup can be challenging. As Maxwell A. Cameron (1998a) defined, a self-coup is “a temporary suspension of the constitution and dissolution of congress by the executive, who rules by decree until new legislative elections and a referendum can be held to ratify a political system with broader executive power” (p. 220). “Broader executive power” is difficult to define, and it would be problematic and disputable no matter how it is defined.
Therefore, this study argues that a more accurate definition of autocoups should prioritize power extension as the core characteristic. It is easy to identify the event and the outcome in the first place. In most cases, autocoups in terms of power extension involve power expansions as their prerequisite and foreshadowing.
Based on these criteria, I define an autocoup as the illegitimate extension of an incumbent leader’s term in office beyond the originally mandated limits through unconstitutional means. This definition emphasizes the core characteristic of power extension while acknowledging the potential for power expansion as a related phenomenon. Three key points need to be highlighted for this definition.
Firstly, this definition refers to the actual leaders of the country, regardless of their official titles. Typically, this would be the president; however, in some cases, such as in Germany, the primary leader is the premier, as the president is a nominal head of state.
Secondly, while the primary characteristic of an autocoup is extending the term in office, this definition does not exclude instances of power expansion. Both aspects can coexist, but the extension of the term is the central element.
Thirdly, autocoups, by their nature, subvert legal norms and established power transfer mechanisms. No matter how legitimate they claim to be, their illegitimacy is not beyond a reasonable doubt as long as the incumbents are the direct beneficiaries. This critical aspect will be explored further in Section 3.
3.3 Introduction to the Autocoup Dataset
3.3.1 Defining the scope
Categorizing political events as autocoups inevitably involves challenging borderline cases. To maintain consistency and avoid ambiguity, this study adopts a broad coding approach: All instances of incumbents extending their original mandated term in office are coded as autocoups, regardless of the apparent legality of the extension.
This approach is justified because truly legitimate amendments to power transition institutions should apply only to subsequent leaders, not the incumbent. Even when extension procedures appear legal, the legitimacy is questionable when the incumbent is the direct beneficiary.
3.3.2 Classifying autocoups
Autocoups manifest in various forms. I categorize them based on several key factors:
Methods Employed: Specific strategies used by incumbents (e.g., constitutional amendments, election cancellation).
Degree of Legality: Extent of deviation from established legal norms.
Duration of Extension: Length of time the incumbent remains in office beyond designated term limits.
Outcomes: Whether the autocoup attempt succeeds or fails.
This study primarily focuses on the methods employed, while coding for other aspects when information is available.
Evasion of term limits
Evasion of term limits is a common tactic employed in autocoups. Incumbents often resort to seemingly legal manoeuvres to extend their hold on power. These manoeuvres primarily involve manipulating constitutional provisions through various means. The incumbents may pressure legislative bodies (congress) or judicial institutions (Supreme Court) to reinterpret existing term limits, amend the constitution to extend terms, or even replace the constitution altogether. This might also involve popular vote through referendums, or a combination of these approaches. The extension can range from a single term to indefinite rule.
These manoeuvres primarily involve manipulating constitutional provisions through various means.
Changing Term Length: Incumbents might lengthen the official term duration (e.g., from 4 to 6 years) to stay in office longer, even if the number of allowed terms remains unchanged. Examples, in the dataset, include Presidents Dacko (CAR, 1962), Kayibanda (Rwanda, 1973), and Pinochet (Chile, 1988).
Enabling re-election: This approach involves incumbents modifying legal or constitutional frameworks to permit themselves to run for leadership again, despite initial restrictions. These restrictions might include prohibitions on re-election, bans on immediate re-election, or term limits that the incumbents have already reached. An illustrative example is President Menem of Argentina in 1993, who leveraged this tactic to extend his tenure.
Removing Term Limits Altogether: This approach, as seen with President Xi Jinping of China in 2018, technically allows the leader to rule for life, although they may still need to participate in elections (a formality in such cases).
DeclaringLeader for Life: This differs from removing term limits as the leader still faces elections (although potentially rigged or uncontested). An example is Indonesia’s President Sukarno, who attempted to declare himself president for life in 1963 (ultimately unsuccessful).
These methods are often used in combination. Initially, the duration of a term is extended, followed by amendments to allow re-election, then the removal of term limits, and finally, the declaration of the leader for life. For example, Haitian President François Duvalier amended the constitution in 1961 to permit immediate re-election and then declared himself president for life in 1964.
Election Manipulation or Rigging
Election manipulation or rigging is the second most commonly used tactic to extend an incumbent’s tenure.
Delaying or Removing Elections: Delaying or removing scheduled elections without legitimate justification is a frequent method used by incumbents to maintain power. For instance, Chadian President François Tombalbaye delayed general elections until 1969 after assuming power in 1960. Similarly, Angolan President José Eduardo dos Santos suspended elections throughout his rule from 1979 to 2017.
Refusing Unfavourable Election Results: Incumbents may refuse to accept unfavourable election results and attempt to overturn them through illegitimate means. For example, President Donald Trump of the United States refused to accept the results of the 2020 election and tried to overturn them.
Rigging Elections: Winning elections with an extraordinarily high percentage of votes is highly questionable. This study will code elections where the incumbent wins more than 90% of the vote as autocoups. For instance, President Teodoro Obiang of Equatorial Guinea has consistently won elections with over 95% of the vote in multi-party elections since 1996, indicating election rigging.
Excluding Opposition in Elections: Manipulating the electoral process by excluding opposition parties or candidates from participation, effectively creating a one-candidate race, clearly signifies an autocoup.
Use of Figurehead
To circumvent term limits, some incumbents might choose a close associate to act as a figurehead, taking the office publicly while the incumbent retains real power behind the scenes. This can be achieved through seemingly subordinate positions.
One example is Russia in 2008. Facing term limits, President Putin selected Dmitry Medvedev to run for president. After the election, Medvedev appointed Putin as Prime Minister. However, most analysts believe Putin wielded the true power throughout this period.
Reassigning supreme authority to a new role
This tactic involves an incumbent leader manipulating the constitution or legal framework to create a new position of power, or elevate an existing one, before stepping down from their current role. They then strategically take on this new position, effectively retaining significant control despite appearing to relinquish power. For example, in 2017, Recep Tayyip Erdoğan, the Prime Minister of Turkey, spearheaded a constitutional referendum that transitioned the country from a parliamentary system to a presidential one. This new system concentrated significant executive power in the presidency. Following the referendum’s approval, Erdoğan successfully ran for the newly established presidency, effectively retaining control under a different title.
One-Time Arrangement for Current Leaders
This strategy involves special arrangements that extend the term or tenure of current leaders without altering the underlying institutions. For example, Lebanon extended President Émile Lahoud’s term by three years in 2004 through a one-time arrangement.
3.3.3 Data Coding
The autocoup dataset is built upon existing studies and datasets, ensuring a comprehensive and reliable foundation. Table 8 outlines the main sources used for coding the autocoup dataset.
The Archigos dataset (Goemans, Gleditsch, and Chiozza 2009) and the Political Leaders’ Affiliation Database (PLAD) (Bomprezzi et al. 2024) provide comprehensive data on all leaders from 1875 to 2023, although our coding only includes autocoups since 1945. These datasets are invaluable for identifying actual rulers, distinguishing them from nominal heads of state.
The Incumbent Takeover dataset (Baturo and Tolstrup 2022) integrates data from 11 related datasets, offering a broad spectrum of cases where leaders significantly reduced constraints on their power. This dataset includes both power expansions and extensions, necessitating cross-referencing with Archigos to verify qualifications for autocoups.
Table 8: Main Data Sources for Coding the Autocoup Dataset
Dataset
Authors
Coverage
Obervations
Archigos
Goemans et al (2009)
1875-2015
3409
PLAD
Bomprezzi et al. (2024)
1989-2023
1334
Incumbent Takeover
Baturo and Tolstrup (2022)
1913-2019
279
In total, 110 observations were coded, with 95 overlapping with the candidate data from Incumbent Takeover. The remaining 15 events were newly coded by the author through verification with other sources such as Archigos, PLAD and news reports.
The main deviation from the Incumbent Takeover dataset arises from excluding power expansions that do not involve attempts to extend tenure.
The dataset encompasses a total of 14 variables along with the notes field.
Country Identification: Country code (ccode) and country name (country) from Correlates of War project (Stinnett et al. 2002).
Leader Information: Name of the de facto leader (leader_name,coded following Archigos and PLAD datasets).
Timeline Variables: Date the leader assumed power (entry_date), date the leader left office(exit_date), date of the significant event marking the autocoup (autocoup_date), and Start date of the leader’s additional term acquired through the autocoup (extending_date).
Power Transition Methods: Categorical variable for how the leader entered power (entry_method), categorical variable for how the leader exited power (exit_method), dummy variable indicating regular (1) or irregular (0) entry (entry_regular), and dummy variable indicating regular (1) or irregular (0) exit (exit_regular).
Autocoup Details: Key variable capturing methods used to extend power (autocoup_method) and outcome of the autocoup attempt (autocoup_outcome, “fail and lose power”, “fail but complete original tenure”, or “successful”). For successful coups, the additional term length can be calculated from the difference between exit_date and extending_date.
Data Source: Identifies the dataset source used for coding (source).
Additional Notes: Provides context for exceptional cases (notes).
There are a few coding challenges and decisions worth mention. For cases where extensions happen incrementally, the autocoup_date reflects a significant event marking the extension, such as a legislative vote or successful referendum. In cases where a leader undertook multiple autocoup attempts, details are recorded in the notes field. Care was taken to differentiate between cases of power expansion and actual attempts to extend tenure, which required cross-referencing multiple sources. Determining the success or failure of an autocoup attempt often required in-depth research, especially for less documented cases.
The primary coding has identified 110 autocoup cases from 1945 to 2024, involving 73 countries. This comprehensive dataset provides a rich source of information for analysing trends and patterns in autocoup attempts across different political contexts.
Table 9: Autocoup methods and success rates (1945-2021)
Autocoup Method
Autocoup Attempted
Autocoup Succeeded
Success Rate
Enabling re-election
46
33
71.74%
Removing term limits
14
14
100.00%
Delaying elections
9
9
100.00%
Leader for life
9
9
100.00%
Changing term length
7
5
71.43%
Figurehead
6
5
83.33%
One-time arrangement
5
4
80.00%
Refusing election results
4
1
25.00%
Reassigning power role
4
2
50.00%
Rigging elections
3
2
66.67%
Cancelling elections
3
3
100.00%
Total
110
87
79.09%
Table 9 presents a breakdown of the autocoup methods employed by leaders:
The most common autocoup method is “enabling re-election”, accounting for 46 events. This is followed by “removing term limits altogether” (14 cases), and then “delaying elections” and “declaring the leader for life” (each with 9 cases).
The overall success rate of autocoups is 79%, which is significantly higher than the approximately 50% success rate of classical coups. This high success rate can be attributed to several factors:
Incumbent Advantage: Leaders already in power have access to resources and institutional mechanisms that can be leveraged to their advantage.
Gradual Implementation: Unlike sudden coups, autocoups can be implemented gradually, allowing leaders to build support and legitimacy over time.
Legal Facade: Many autocoup methods operate within a veneer of legality, making them harder to oppose openly.
Control of State Apparatus: Incumbents often have significant control over state institutions, which can be used to facilitate their autocoup attempts.
However, success rates vary significantly across different methods.
100% Success Rate: Removing term limits, delaying elections, declaring the leader for life, and cancelling elections all have perfect success rates. This suggests that once these processes are set in motion, they are difficult to reverse.
Lower Success Rates: Refusing to accept election results has the lowest success rate, with only 1 out of 4 attempts succeeding. Although the sample size is limited (only 4 cases in total), this trend might suggest several factors at play. These include greater democratic resilience in systems where general elections are regularly held, heightened international scrutiny and pressure in response to blatant manipulation of election results, and stronger domestic opposition to such overt power grabs.
3.4 Determinants of Autocoup Attempts: Case Studies
3.4.1 High Frequency and Success Rate of Autocoups in Post-Communist Countries
Analysis of our dataset reveals a notably high frequency and success rate of autocoups in post-communist countries. These nations, formerly communist regimes prior to the collapse of the Soviet Union, have largely evolved into ‘hybrid regimes’ (Nurumov and Vashchanka 2019), with only a few retaining their communist status. The data documents 12 cases of autocoups aimed at prolonging incumbency in these countries, with only two attempts failing. Examination of these cases highlights several distinctive characteristics:
Inherited Authoritarian Systems: Despite most of these 12 countries transitioning from communist to non-communist governments (with the exception of China), they retained many authoritarian systems from their communist past.
Continuity of Former Elites: The transitions did not result in the removal or overthrow of previous ruling groups. Instead, former communist elites often maintained their positions of power.
Subverted Democratic Processes: While general elections and term limits were introduced in most of these countries, the legacy of former communist regimes frequently led to the circumvention of term limits and manipulation of elections (Nurumov and Vashchanka 2019).
Case 1: Lifelong Ruler–Alexander Lukashenko in Belarus
Alexander Lukashenko, a former member of the Supreme Soviet of the Byelorussian Soviet Socialist Republic, became the head of the interim anti-corruption committee of the Supreme Council of Belarus following the dissolution of the Soviet Union. Elected as Belarus’s first president in 1994, he has maintained this position ever since. Initially, the 1994 constitution limited presidents to two successive terms. However, Lukashenko removed this restriction in 2004. International monitors have not regarded Belarusian elections as free and fair since his initial victory. Despite significant protests, Lukashenko has consistently claimed to win with a high vote share, often exceeding 80% in each election. This pattern is evident across all five Central Asian countries of the former Soviet Union, where post-dissolution leaders were typically high officials or heads of the former Soviet republics who continued their leadership in the presidency.
Case 2: Transferring Power to a Handpicked Successor–Nursultan Nazarbayev in Kazakhstan
Nursultan Nazarbayev served as the first president of Kazakhstan from 1991 until 2019. Prior to the dissolution of the Soviet Union, he held de facto leadership as the First Secretary of the Communist Party of Kazakhstan. Following independence, he was elected as the first president and retained office until 2019 through various means, including resetting term limits due to the implementation of new constitutions. Notably, Nazarbayev did not officially eliminate term limits but instead created an exemption for the “First President” (Nurumov and Vashchanka 2019). Unlike Lukashenko, who remains the incumbent of Belarus, Nazarbayev transferred the presidency to a designated successor, Kassym-Jomart Tokayev, in 2019. However, he retained significant influence as the Chairman of the Security Council of Kazakhstan until 2022.
3.4.2 Autocoups for immediate re-election: Cases of Latin America
Latin America has a long-standing tradition of maintaining term limit conventions. Simón Bolívar, the founding father of Bolivia, was initially a strong advocate for term limits, stating in 1819, “Nothing is as dangerous as allowing the same citizen to remain in power for a long time… That’s the origin of usurpation and tyranny” (Ginsburg and Elkins 2019, 38). Although Bolívar eventually modified his stance, arguing in his 1826 Constitution Assembly speech that “a president for life with the right to choose the successor is the most sublime inspiration for the republican order,” term limits became a convention in Latin America. Approximately 81% of Latin American constitutions between independence and 1985 imposed some form of term limits on the presidency (Marsteintredet 2019).
An analysis of cases in Latin American countries reveals two notable patterns.
Often Successful at Breaking Non-re-election or Non-immediate Re-election Restrictions
Unlike other presidential systems where two terms are more common, non-re-election or non-immediate re-election used to be prevalent in Latin America. According to Marsteintredet (2019), non-consecutive re-election was mandated in about 64.9% of all constitutions between independence and 1985, while 5.9% banned re-election entirely.
However, adherence to these conventions has varied across the region. Since Mexico introduced non-re-election institutions in 1911 at the start of the Mexican Revolution, they have remained inviolate (Klesner 2019). Similarly, Panama and Uruguay have never altered their re-election rules, and Costa Rica has only experienced a brief period (1897-1913) permitting immediate presidential re-election since prohibiting it in 1859 (Marsteintredet 2019). In many other countries, however, constitutions have been frequently amended or violated.
The pursuit of re-election or consecutive re-election, therefore, has been a significant trigger for autocoups aimed at power extension in this region. Our research documents 32 autocoup cases, with over 50% (17 cases) attempting to enable re-election or immediate re-election, and about 59% (10 cases out of 17) being successful.
Unlike those who attempt to overstay in office indefinitely, many Latin American leaders exit after their second term expires. Examples include President Fernando Henrique Cardoso of Brazil (1995-2003), President Danilo Medina of the Dominican Republic (2012-2020), and President Juan Orlando Hernández of Honduras (2014-2022) (Ginsburg and Elkins 2019; Marsteintredet 2019; Landau, Roznai, and Dixon 2019; Baturo 2019; Neto and Acácio 2019).
Failing to Further Extend Tenure
This trend does not imply that none of these leaders attempted further extensions, but rather that most accepted their unsuccessful outcomes without abusing their power to manipulate the process. While autocoups aimed at securing one additional term are often successful, attempts to overstay beyond this are frequently unsuccessful.
Two contrasting cases illustrate the varied outcomes of term limit challenges:
Unsuccessful Extension – Carlos Menem (Argentina): President Menem successfully extended his tenure by one term through a 1994 constitutional amendment allowing one executive re-election. He was subsequently re-elected in 1995. However, his attempt to reset his term count, arguing that his first term (1988-1995) should not count as it was under previous constitutions, was unanimously rejected by the Supreme Court in March 1999 (Llanos 2019). A similar scenario unfolded with President Álvaro Uribe of Colombia (2002-2010) (Baturo 2019).
Successful Extension – Daniel Ortega (Nicaragua): In contrast, Daniel Ortega, the incumbent president of Nicaragua, successfully extended his presidency. In 2009, the Supreme Court of Justice of Nicaragua permitted his re-election in 2011. Subsequently, in 2014, the National Assembly of Nicaragua approved constitutional amendments abolishing presidential term limits, allowing Ortega to run for an unlimited number of five-year terms. As a result, he has held the presidency since 2007 (Close 2019).
3.4.3 As common as classical coups: Cases of African countries
Classical coups have been prevalent in Africa, accounting for approximately 45% of all global coups (219 out of 491 cases) since 1950, involving 45 out of 54 African countries (GIC dataset). While autocoups are less frequent compared to traditional coups, they maintain a significant presence in Africa. Among 110 documented autocoup cases globally, 46% (51 cases) occurred in Africa, involving 36 countries. Notably, the success rate of autocoups in Africa is over 84% (43 out of 51 attempts), which surpasses both the success rate of classical coups in the region (roughly 50%) and the global average success rate of autocoups (79%).
Identifying a clear pattern of autocoups in Africa is challenging, mirroring the complexity observed with classical coups. Various factors have been proposed to explain this phenomenon:
Quality of Democracy: The quality of democracy is a critical factor influencing respect for term limits (Reyntjens 2016).
International Influence: International aid or donor influence can play a significant role in discouraging attempts at power extension (Brown 2001; Tangri and Mwenda 2010).
Organized Opposition and Party Unity: The extent of organized opposition and the president’s ability to enforce unity within the ruling party are crucial factors (Cheeseman 2019).
Utilizing the Africa Executive Term Limits (AETL) dataset, Cassani (2020) highlights human rights abuses and the desire for impunity as main drivers for incumbents to cling to power. The more authoritarian a leader, the more likely they are to attempt to break term limits and overstay in office. A leader’s ability to secure the loyalty of the armed forces through public investment increases the chances of success in overstaying.
Despite both coups and autocoups being prevalent, there has been a noticeable shift since the end of the Cold War in 1991: Traditional coups have decreased in frequency while autocoups have become more prevalent.
This trend can be partially attributed to the introduction of multi-party elections in Africa in the 1990s, which also brought in term limits for executives (Cassani 2020; Cheeseman 2019). Before 1991, personal or military rule was more common, and term limits were less frequent. Post-1991, with more term limits introduced, challenges to these limits have increased. However, it is crucial to note that this increase in challenges does not necessarily imply that violations are more common than adherence to term limits, because total power transitions have increased compared to the past.
3.5 Empirical Analysis: An Example of Utilizing the Autocoup Dataset
The availability of the autocoup dataset has made it feasible to conduct quantitative analyses that extend beyond traditional case studies. This section provides a straightforward example of how to utilize this dataset effectively. To analyze the determinants of autocoup attempts, I employ a probit regression model. This approach differs from the double probit model with sample selection used in Chapter 2 for coup attempts and success analyses. Instead, I use two separate probit models. Due to the high probability of success in autocoups, they do not exhibit the typical sample selection characteristics that necessitated the use of a sample selection model in our earlier analysis of traditional coups.
3.5.1 Dependent Variables
Autocoup Attempt: Binary variable indicating whether an autocoup attempt occurred (1) or not (0) during the tenure of an incumbent leader.
Autocoup Success: Binary variable indicating whether an autocoup attempt was successful (1) or failed (0), conditional on an autocoup attempt occurring.
3.5.2 Independent Variables
The selection of independent variables are consistent with the coup analysis in Chapter 2, plus the population size and the leader’s age.
Population Size: To account for its potential impact on leaders’ tenures, we consider the log of the population size. This transformation helps in managing the wide range of population sizes across different countries. The data is sourced from the V-Dem dataset and is evaluated to understand its influence on power transitions. Larger populations may present more governance challenges and potential sources of opposition, thereby affecting the stability and longevity of a leader’s tenure.
Leader’s Age: The age of the leader is included as an additional variable in the analysis, offering insights into potential correlations with leadership strength. Older leaders may have different experiences, networks, and health considerations that could influence their ability to maintain power. This data is sourced from Archigos and PLAD datasets.
Unlike the analysis of coup determinants, which could theoretically occur in any given year, I assume that an autocoup happens only once during an incumbent leader’s tenure, as a successful autocoup negates the need for another attempt. However, this assumption does not always reflect reality, as leaders might attempt further extensions or try again after a failed attempt. For simplicity, I overlook these possibilities in our analysis.
Therefore, in our probit model, the unit of analysis for autocoups is the entire tenure of a leader, rather than a country-year. I establish a base year for the variables: for leaders who staged an autocoup, we use the year of their first attempt as the base year; for leaders who did not attempt to overstay, I use the middle year of their tenure as the base year.
3.5.3 Results and discussions
Table 10: Determinants of autocoup attempts and success (1945-2018)
Table 10 summarizes the findings from the probit regression models based on our analysis of the determinants of autocoup attempts and their success.
Model 1, which examines autocoup attempts, reveals only one significant predictor besides the constant term. Among the regime types, personalist regimes significantly increase the likelihood of autocoup attempts, all else being equal. This suggests that leaders in personalist regimes are more prone to attempt to extend their power through autocoups compared to leaders in democratic regimes (reference regime). Leaders in dominant-party and military regimes, however, show no significant difference in the likelihood of attempting an autocoup compared to democratic leaders.
The model for autocoup success (Model 2) shows similar dynamics. Personalist regimes again have a strong positive and significant effect on the success of autocoups compared to democratic leaders. Dominant-party regimes also show a positive and marginally significant effect. However, a detailed examination reveals that about half of the successful autocoups in dominant-party regimes (9 out of 20) exhibit a personalist style, such as “party-personal-military” regimes.
This outcome is logical since personalist leaders are typically much more powerful than other types of leaders, making them more inclined and capable of overstaying in power.
Other factors play an insignificant role in determining the attempts and outcomes of autocoups. This aligns with our conclusions on the determinants of classic coups. Both coups and autocoups are significantly affected by power dynamics. As power transitions involve the struggle between seizing and maintaining power, the balance of power status quo inevitably matters in both coups and autocoups. This also explains the high success rate of autocoups. Compared to power challengers, incumbents are in an obviously advantageous position. Incumbent leaders can use state power to their benefit, which is difficult to counteract. Even the abuse of power is often unchecked under a powerful leader’s rule.
The empirical analysis of autocoups yields significant implications for real-world politics. In particular, the high overall success rate of autocoups highlights the vulnerability of democratic institutions to gradual erosion by incumbent leaders. The threshold for ousting or impeaching an incumbent leader through constitutional means is exceptionally high, with success often requiring more than a simple majority and substantial support across various sectors. Resorting to illegal means, such as a coup, presents even greater challenges due to high costs, severe consequences, and a low likelihood of success.
Conversely, political dynamics, whether in democracies or autocracies, tend to favour incumbents even when they act unconstitutionally. Incumbents can leverage state resources to achieve their political ambitions, benefiting from a high probability of success and minimal consequences in case of failure. This asymmetry in power and risk creates a concerning scenario: for incumbents who do not respect constitutional institutions, the opportunity to launch an autocoup appears sufficiently low-risk to warrant an attempt.
3.6 Conclusion
This chapter presents a comprehensive analysis of autocoups, focusing specifically on political events where incumbent leaders illegitimately extend their tenure in power. By refining the concept and distinguishing it from broader definitions such as ‘self-coups’, ‘autogolpes’, and ‘executive takeovers’, I introduce a novel dataset of autocoups spanning from 1945 to 2024. Through this more precise definition and dataset, this research expands the analysis of irregular power dynamics from irregular transition to irregular retention, offering a broader and more nuanced review of the phenomenon.
The findings reveal that personalist regimes are significantly more likely to experience autocoup attempts and succeed in these attempts compared to democracies. Dominant-party systems, often exhibiting personalist characteristics, also show an association with successful autocoups. While regime type significantly influences autocoups, other factors appear less impactful, mirroring classic coups where the balance of power is a more essential determinant. The high success rate of autocoups can be attributed to the inherent advantages incumbents possess, such as control over or abuse of state power and the difficulty of removing or impeaching them through legal or illegal means.
However, several limitations warrant consideration for future research. Firstly, the definition of an autocoup requires further commentary and discussion to gain wider acceptance in the academic community. Despite efforts to maintain objectivity, some coding decisions may involve subjective judgements, particularly in borderline cases. Secondly, due to the nature of autocoups, which are less frequent than classic coups (491 coups versus 110 autocoups during the same period), the quantitative analysis cannot be conducted as a country-year variable as in coup studies. This raises the issue of choosing an appropriate base year. For instance, when analysing how GDP level or growth rate, political stability, or the age of the leader affects autocoup attempts, we must decide which year’s value should be used. In this study, I chose the middle year of a leader’s tenure or the year they staged the autocoup, but determining the most appropriate year requires further discussion and potentially sensitivity analyses.
Despite these limitations, this research significantly enhances our understanding of the mechanisms and motivations behind autocoups, contributing to the literature on political stability and democratic resilience. The findings highlight the vulnerability of political systems, particularly democracies, to erosion from within by incumbent leaders.
Future studies could build on this work by employing the dataset to explore more nuanced power dynamics or examine the long-term impacts of these events on political systems. Particularly fruitful areas for investigation include the relationship between autocoups and democratic backsliding, democratic breakdown, and the personalization of power. Additionally, comparative analyses between autocoups and traditional coups could yield insights into the evolving nature of power consolidation strategies in different political contexts.
In conclusion, this study not only provides a valuable resource for future research but also contributes to our understanding of the complex interplay between leadership, institutional structures, and political stability. As autocratic tendencies continue to challenge democratic norms globally, the insights gained from this analysis of autocoups become increasingly relevant for both scholars and policy-makers concerned with preserving and strengthening democratic institutions.
4 Power Acquisition and Leadership Survival: A Comparative Analysis of Coup-Entry and Autocoup Leaders
Abstract
This chapter examines the relationship between methods of power acquisition and the longevity of leaders who assume power through irregular means, specifically coup-entry and autocoup leaders. The central hypothesis posits that the mode of accession significantly influences leader tenure. Employing Cox proportional hazards and time-dependent Cox models, this study provides robust evidence of divergent survival times between these two leader types. Findings reveal that coup-entry leaders face a substantially higher risk of removal from office compared to their autocoup counterparts. These results have significant implications for political stability and democratic processes, suggesting that the perceived low costs and high rewards associated with autocoups may incentivize incumbents to extend their tenure through this mechanism, potentially contributing to democratic backsliding. This research makes a notable contribution to the academic literature by leveraging a newly developed dataset on autocoups. By doing so, it offers valuable insights into the dynamics of irregular leadership transitions and enhances our understanding of the complex interplay between power acquisition methods and political longevity.
keywords: Coups, Autocoups, Leadership Survival, Cox Model
4.1 Introduction
The longevity of political leaders has long captivated scholars and observers. Why do some rulers endure for decades while others are ousted within months or even days? This question has spurred extensive research in political science, primarily focusing on the survival of political leaders. However, within this broader framework, a specific subset of leaders—those who ascend to power through coups or extend their tenure through autocoups—has received comparatively less attention. Examining the tenures of these leaders is crucial, as it illuminates the dynamics of irregular leadership transitions and their implications for political stability and democratic processes.
Leaders who attain power through conventional means often follow predictable patterns of ascension, tenure, and departure. In contrast, those who rise through irregular channels, such as coups or autocoups, present more complex and intriguing cases for study. Archigos dataset underscores the prevalence of such irregular transitions: between 1945 and 2015, over half of leaders who assumed power irregularly (158 out of 308) also exited irregularly. This rate significantly surpasses that of leaders who accessed office through regular channels, of whom only 14.5% (213 out of 1,472) experienced irregular departures.
Coup-entry and autocoup leaders constitute a substantial portion of these irregular cases. Archigos dataset notes that of 374 leaders who exited irregularly, 246 (65.8%) were ousted through coups. Frantz and Stein (2016) further demonstrate that coup-related exits account for approximately one-third of all exits in autocracies, surpassing any other transition type. Additionally, the autocoup dataset, introduced in Chapter 3, documents 110 autocoup attempts between 1945 and 2024, of which 87 were successful.
Measuring the survival tenure of coup-entry and autocoup leaders presents challenges due to the inherent irregularity and uncertainty of their positions. Nevertheless, a comparative analysis reveals that leaders who extend power through autocoups tend to have longer average post-autocoup tenures (approximately 11 years) compared to coup-entry leaders (approximately 5.7 years), suggesting a potential tenure gap of over five years.
Code
# Fit survival curvesfit <-survfit(Surv(survival_year, status) ~ group, data = survival_cox_ph_model)# Create log-rank plot (Kaplan-Meier plot)ggsurvplot(fit, data = survival_cox_ph_model, risk.table =TRUE, risk.table.height =0.3, pval =TRUE,surv.median.line ="hv",break.x.by =5, legend.title ="Leader Type",legend.labs =c("Autocoup leaders", "Coup-entry leaders"),xlab ="Survival Years",#title = "Survival Curves of overstay and coup-entry leaders",font.title =12)
Figure 1: Survival curves of overstaying and coup-entry leaders
A preliminary log-rank test in survival analysis, as illustrated in Figure 1, demonstrates a statistically significant difference between the tenures of autocoup and coup-entry leaders. The survival curve for autocoup leaders consistently exceeds that of coup-entry leaders, indicating longer survival times and a reduced risk of ouster for autocoup leaders.
This study posits that the method of accession significantly influences leadership longevity. Coup-entry leaders likely confront greater challenges to their rule, resulting in shorter average tenures compared to autocoup leaders. The analysis, employing Cox proportional hazards and time-dependent Cox models, supports this hypothesis, demonstrating that autocoup leaders generally experience longer tenures than coup-entry leaders.
This research offers two primary contributions to the field. First, it highlights an understudied factor in leadership survival analysis: the impact of the method of accession to power. My findings suggest that leader survival is influenced not only by ruling strategies but also by the initial method of acquiring power. Second, by employing survival models, this study provides empirical evidence of the significant difference in tenure duration between autocoup and coup-entry leaders. This insight may explain the increasing prevalence of tenure extensions through autocoups since 2000, as more incumbents observe and potentially emulate successful precedents.
The remainder of this chapter is structured as follows: Section 2 provides a comprehensive literature review on political survival, establishing the context for this research. Section 3 explores the factors influencing the survival of coup and autocoup leaders. Section 4 outlines the methodology and data used, including the application of survival models to analyse the determinants of leadership longevity. Section 5 presents the analysis findings and a detailed discussion of the results. Finally, Section 6 concludes by synthesizing key takeaways and exploring their broader implications for political stability and democratic processes.
4.2 Literature review
The longevity of political leaders has been a central focus in political science research for decades, driven by the wide-ranging variations observed across regimes, countries, and historical periods. This field encompasses two interconnected aspects: regime survival and individual leader survival. Regime survival focuses on the endurance of political systems, such as monarchies, political parties, or specific ideological structures, while leader survival concerns the duration of individual leaders’ time in office.
These concepts often exhibit contrasting patterns. For instance, parliamentary democracies like Japan or the United Kingdom may experience prolonged periods of party dominance coupled with frequent leadership changes, whereas communist regimes typically demonstrate enduring party rule with more frequent leadership transitions. Presidential systems, such as the United States or many military regimes, tend to exhibit more frequent changes in both ruling party or junta and leader. This study specifically investigates the dynamics of individual leader survival, focusing on factors influencing the duration of leaders’ time in office.
The existing literature on leader survival is extensive and multifaceted. Some studies explore specific mechanisms influencing leadership longevity within particular regimes, such as democracies (Svolik 2014) or autocracies (Davenport, RezaeeDaryakenari, and Wood 2021), while others aim to develop more generalizable theoretical frameworks explaining leader survival across different political systems (Bueno de Mesquita et al. 2003). Although a universal theory remains an aspirational goal, the complexities of leadership survival across diverse regime types present significant challenges.
Power transition mechanisms vary substantially across different regimes, particularly between democracies and autocracies. Autocratic systems often feature closed leadership selection processes, restricted to a narrow pool of individuals such as royal families, military elites, or ruling party members. While some autocracies may hold elections, significant barriers to entry for legitimate challengers typically persist. Potential rivals may face threats like assassination, imprisonment, or exile. Moreover, the opacity of selection processes makes it difficult to assess genuine levels of public support compared to democracies. Consequently, conceptualizing selectorates or winning coalitions, as proposed by Bueno de Mesquita et al. (2003), becomes problematic in many autocratic contexts.
Given these complexities, focusing research on specific regimes or leader types may be more fruitful. While regular leadership changes offer valuable insights, they provide limited opportunities to explore the dynamics of leader survival. In contrast, the study of irregular leaders, such as those who ascend to power through coups or overstay in office through autocoups, offers a more compelling avenue for research due to the inherent complexities and uncertainties surrounding their leadership trajectories.
Coups have received considerable scholarly attention, with research examining coup prevention strategies, the impact of coups on leadership, and the subsequent actions of coup leaders. Existing research delves into strategies for thwarting coups (J. Powell 2017; Sudduth 2017; De Bruin 2020) and how leaders extend their tenures after surviving coup attempts (Easton and Siverson 2018). Sudduth (2017) examines post-coup actions of dictators, focusing on purge strategies, while Sudduth and Bell (2018) investigates how leaders’ entry methods affect their removal in dictatorships.
However, a significant gap exists in the literature regarding the comparison of leadership survival between coup-entry and autocoup leaders. This study aims to address this gap by investigating and comparing the duration of leadership survival for these two leader types, contributing to a more nuanced understanding of political survival in irregular leadership transitions.
4.3 Survival dynamics of autocoup and coup-entry leaders
The study of leadership survival in political systems presents inherent challenges due to the opacity and diverse mechanisms of power transitions. However, these challenges underscore the significance of this research, as it illuminates understudied dynamics in political leadership. While the survival of political leaders exhibits complexity and variation, it is not entirely devoid of patterns. Leaders of similar types often display significant comparability.
4.3.1 Key Definitions and Scope
Before delving into the comparison, it is essential to clarify several key terminologies:
Coup and Autocoup: These terms adhere to the definitions established in previous chapters.
Tenure Length Threshold: To ensure meaningful analysis, this study focuses on leaders with substantial periods in power, applying a six-month threshold to both autocoup and coup-entry leaders.
Autocoup Leader: An incumbent leader who successfully uses illegitimate or unconstitutional means to extend their tenure in power.
Coup-Entry Leader: The individual who assumes power after a successful coup, regardless of their role in the coup itself.
This study focuses on comparing the post-autocoup tenure of autocoup leaders with the post-coup tenure of coup-entry leaders, motivated by the relevance and similarity of these leader types in terms of illegitimacy, uncertainty, and instability.
4.3.2 Challenges in Power Consolidation
Both autocoup and coup-entry leaders face distinct challenges in consolidating their power, primarily due to differences in the intensity of issues related to illegitimacy, uncertainty, and instability. This disparity creates an uneven playing field in terms of power dynamics, with coup-entry leaders at a significant disadvantage. Table 11 compares the main features between autocoup and coup-entry leaders.
Table 11: Main features of autocoup and coup-entry leaders
Feature
Autocoup Leader
Coup Entry Leader
Illegitimacy
Normally attained through lawful procedures, but lacking consensus legitimacy
Blatantly illegal
Uncertainty
Initially with some certainty, but decreases as the leader's age grows or health worsens
Significant uncertainty initially
Instability
Relatively stable
Unstable except when a strongman emerges or constitutional institutions are established
Balance of Power
Generally in a better position of power
Initially unclear and challenging to establish a balance
Illegitimacy
While both types of leaders suffer from a legitimacy deficit, the nature of this deficit differs:
Coup-Entry Leaders: Their illegitimacy is explicit due to the open seizure of power.
Autocoup Leaders: They employ a deceptive strategy, manipulating legal processes to create a façade of democratic legitimacy.
Uncertainty
The irregular paths to power create uncertainty regarding their reigns and eventual departures. However, the levels of uncertainty differ.
Coup-entry leaders face three major uncertainties: Unclear who will assume leadership after the coup, uncertain tenure length, and ambiguity regarding future successors.
Autocoup leaders present a clearer picture: No ambiguity about who will rule after an autocoup, many seek to extend their rule indefinitely or incrementally.
Instability
The awareness of shaky legitimacy and persistent uncertainty breeds insecurity and a sense of crisis.
Coup-entry leaders must reshape power dynamics and often purge potential adversaries. They need to create a new equilibrium, often disrupting established structures. They also face potential backlash even from close allies and hence must compromise with internal or external power challengers.
Autocoup leaders encounter fewer abrupt changes in their regimes. They face less pressure to dismantle existing ruling paradigms and have more time to implement changes gradually.
4.3.3 Empirical Evidence and Hypothesis
Empirical evidence supports the disadvantage faced by coup-entry leaders. Data reveals a correlation between the frequency of coup attempts in a country and the likelihood of future coups. Over a third of coups occur in the top ten countries with the most attempts since 1950 (Table 2). The average survival period following an autocoup is approximately five years longer than that of coup-entry leaders (Figure 1).
The distinct challenges faced by autocoup and coup-entry leaders in consolidating power create a self-perpetuating cycle that influences their tenure length. Coup-entry leaders, facing greater challenges, struggle to attract and retain strong support, making them more vulnerable to internal and external challenges. Conversely, autocoup leaders, often benefiting from a veneer of legitimacy and a stronger initial position, are better able to consolidate power and attract supporters, leading to potentially longer tenures.
Based on these observations, I propose the following hypothesis:
H1: Political leaders who successfully extend their tenure through autocoups are more likely to survive longer compared to coup-entry leaders.
4.4 Research Design
This study employs survival analysis to test the hypothesis that autocoup leaders have longer survival times in office compared to coup-entry leaders. The research design utilizes Cox models to analyze the survival tenures of these two types of leaders, accounting for multiple factors that may influence their time in power.
4.4.1 Methodology: Survival analysis
Two Cox models will be employed to analyze the survival tenures of coup-entry and autocoup leaders:
Cox Proportional Hazards (PH) Model: This model uses only the variables present at the entry year, without considering changes over time.
Time-Dependent Cox Model: This model accounts for variations in time-dependent control variables such as economic performance and political stability.
The Cox model is preferred over the Kaplan-Meier model as it allows for the estimation of multiple factors’ impacts. While it does not directly estimate the duration of tenure in office, it evaluates the hazard rate associated with being ousted from power. This approach captures different facets of the same phenomenon: as a leader’s cumulative hazard of being ousted increases, their probability of survival in office decreases.
4.4.2 Data and Variables
The dependent variables include survival time and end point status:
Survival Time: The duration of a leader’s tenure, measured in days. For coup-entry leaders, the survival time begins on the day they assume power through a coup. For autocoup leaders, the survival time starts on the expiration date of their original legitimate term. For example, Xi Jinping assumed power in 2013 and removed term limits in 2018. His original legitimate tenure was set to end in 2023, so his survival time begins in 2023, marking the start of his post-autocoup tenure. The survival time concludes on the day the leader exits office, applicable to both coup-entry and autocoup leaders.
End point status: This variable indicates the manner in which the leader’s tenure concluded, categorized as follows:
0 = Censored: This status is assigned to leaders who leave office through regular means other than being ousted. This includes leaders transferring power to their designated successors, leaving office as their terms expire, losing in general elections, voluntarily leaving office due to health issues, or dying of natural causes.
1 = Ousted: This status is assigned to leaders who are forced to leave office. This includes leaders resigning under pressure, being ousted by coups or other forces, or being assassinated.
The key independent variable is the leader type, which categorizes leaders into two distinct groups:
Group A = Autocoup Leader: Leaders who extend their tenure through autocoups.
Group B = Coup-Entry Leader: Leaders who assume power through coups.
This variable is the primary independent variable of interest, serving as the basis for comparing the survival time between these two types of leaders.
The data for both dependent and independent variables are sourced from the autocoup dataset introduced in this study, Archigos, and PLAD.
Control variables include economic performance, political stability, population size, and the leader’s age, which are consistent with the autocoup analysis in Chapter 3.
4.5 Results and discussion
4.5.1 Model results
Using the survival package in R (Therneau 2024), I present the regression results for both the Cox Proportional Hazards (Cox PH) model and the time-dependent Cox model in Table 12.
Table 12: Cox models for survival time of different types of leaders
Characteristic
Cox PH Model
Time-dependent Cox Model
N
Event N
HR
1,2
SE
2
N
Event N
HR
1,2
SE
2
Leader Type
Autocoup leaders
76
31
1.00
—
737
29
1.00
—
Coup-entry leaders
148
73
2.71***
0.252
853
73
2.23***
0.246
GDP Growth Trend
224
104
1.94
1.08
1,590
102
0.20*
0.981
GDP per capita
224
104
0.97*
0.020
1,590
102
0.95**
0.023
Population: log
224
104
0.98
0.083
1,590
102
0.90
0.079
Polity 5
224
104
0.99
0.025
1,590
102
1.01
0.023
Political stability
224
104
1.00
0.053
1,590
102
1.11*
0.049
Age
224
104
1.01
0.010
1,590
102
1.00
0.011
1
p<0.1; p<0.05; p<0.01
2
HR = Hazard Ratio, SE = Standard Error
Both the Cox PH model and the time-dependent Cox model analyses revealed a statistically significant association between leadership type and the hazard of removal from power. Since time-dependent Cox model use the control variables which change over time, I interpret the main findings based on time-dependent model.
Coup-entry leaders were found to have a hazard ratio of 2.23 in the time-dependent model compared to autocoup leaders (reference group), assuming all other variables in the model are held constant. This suggests that coup-entry leaders face a significantly greater risk of removal from power compared to autocoup leaders. At any given time during their tenure, coup-entry leaders are 2.23 times more likely to be ousted from power compared to autocoup leaders, all else being equal in the model.
The control variables perform differently in the two models. Economic level (GDP per capita) exhibits statistically significant effects in both models. In the time-dependent model, the hazard ratio of 0.95 indicates that for each unit increase in GDP per capita (measured in units of $10,000), the hazard (or risk) of being ousted at any given time is reduced by 5%, assuming all other variables in the model are held constant.
GDP growth trend demonstrates a more substantial effect in reducing the risk of coups. Specifically, a 1 percentage point higher economic growth trend is associated with an 80% reduction in the risk of being ousted, although this effect is only statistically significant at the 10% level. This suggests a possible trend where positive economic performance might mitigate the risk of removal from power, but the evidence is not robust enough to confirm this conclusively.
Political stability, as measured by the violence index, shows that a 1-point increase in the index correlates with an 11% higher risk of being ousted. However, this effect is also only statistically significant at the 10% level, indicating a weaker but potentially important relationship between increased violence and the risk of removal from office.
4.5.2 Discussion
(a) Cox PH Model
(b) Time-dependent Cox Model
Figure 2: Survival curves for Cox Model
The survival curves depicted in Figure 2 illustrate the survival rates for leaders of both types. Both the Cox PH model and the time-dependent Cox model produce similar plots. Notably, the survival curve for coup-entry leaders exhibits a significantly lower trajectory compared to that of autocoup leaders. The steeper drop at the early stage for coup-entry leaders indicates they are more likely to be ousted shortly after assuming power. Additionally, the survival curve for coup-entry leaders crosses the median survival line much earlier (about 3,000 days) than that of autocoup leaders (about 8,500 days). This disparity suggests that autocoup leaders tend to remain in power for longer durations than their coup-entry counterparts.
(a) Cox PH Model
(b) Time-dependent Cox Model
Figure 3: Hazard ratios and 95% CIs for Leader Ousting
Figure 3 displays the hazard ratios and corresponding 95% confidence intervals for the variables incorporated in the Cox model. Both the Cox Proportional Hazards (PH) model and the time-dependent model produce similar plots, reinforcing the robustness of the findings. Key points to note include:
The closer the hazard ratio (represented by the dots) is to 1, the less impact the variable has on the risk of being ousted. A hazard ratio of 1 indicates no effect.
The whiskers extending from the dots represent the 95% confidence intervals. If these whiskers cross the vertical blue line at 1, it indicates that the variable is not statistically significant at the 5% level.
The hazard ratio for coup-entry leaders is significantly greater than 1 and statistically significant at the 5% level. This indicates that coup-entry leaders face a substantially higher risk of being ousted compared to autocoup leaders.
Most other variables have hazard ratios close to 1, suggesting that a one-unit increase in these variables does not significantly affect the risk of being ousted.
Although the hazard ratio for GDP growth trend is considerably less than 1 in the time-dependent model, indicating a potential protective effect, it is not statistically significant at the 5% level. However, it is statistically significant at the 10% level, suggesting that better economic performance may help to consolidate the rule of the incumbents to some extent, albeit the evidence is not as strong.
4.5.3 Assessing the Proportional Hazards Assumption
Assessing the proportional hazards assumption is crucial for the validity of the Cox model results. To evaluate this, we used the chi-square test based on Schoenfeld residuals to determine whether the covariate effects remain constant (proportional) over time. Although the Cox PH model violates the proportional hazards assumption, our primary analysis relies on the time-dependent Cox model, which does not show strong evidence of violating the proportional hazards assumption for any covariate. The global p-value of 0.416 is much greater than the 5% significance level, indicating that the proportional hazards assumption is reasonably met for the time-dependent Cox model.
4.6 Conclusion
This chapter examined the survival durations of political leaders who come to power through irregular means, specifically coups and autocoups. I hypothesized that the mode of accession significantly influences leader tenure. Employing survival analysis techniques, including the Cox proportional hazards model and a time-dependent Cox model, I found strong evidence that autocoup leaders generally enjoy longer tenures than coup-entry leaders.
The findings revealed a significant difference in average tenure, with post-autocoup leaders averaging approximately 11 years in power compared to 5.6 years for coup-entry leaders. The time-dependent Cox model further indicated that coup-entry leaders are 2.23 times more likely to be ousted from power at any given time compared to autocoup leaders, all else being equal.
These results highlight the importance of understanding the phenomenon of autocoups, where leaders extend their rule by manipulating legal frameworks. Due to the relative ease and potential benefits of autocoups, this method of power retention might incentivize more leaders to employ it. Consequently, democratic backsliding could become more prevalent as autocoups weaken democratic institutions and constitutional norms, particularly in nascent democracies or those transitioning from autocracy.
This study contributes to the field of leadership survival by demonstrating that the mode of accession significantly impacts leader tenure, a factor previously under-explored in the literature. By utilizing both Cox models, the research offers robust analytical techniques for studying political leadership survival and provides strong evidence of divergent tenure lengths between these two types of irregular-entry leaders.
However, limitations exist. The study relies heavily on the autocoup dataset collected and coded by the author. The concept and data itself are relatively novel within academia. Future research should refine and establish wider recognition for the term “autocoup,” leading to more accurate and comprehensive data collection efforts. Expanding the dataset to include more cases and integrating it with data on other irregular leadership transitions could yield a more holistic understanding of political survival in such contexts.
Overall, this chapter underscores the need for more nuanced approaches to studying political tenure and the mechanisms of irregular power retention, contributing valuable insights into the dynamics of political stability and the risks associated with different forms of non-democratic leadership succession.
5 Conclusion
5.1 Main Findings
This study delves into the dynamics and implications of irregular power transitions, focusing on coups and autocoups. The findings illuminate the complex interplay between incumbents and challengers fighting for power.
Firstly, our analysis reveals that the expected success rate of a coup attempt significantly influences its likelihood. This success rate is heavily influenced by the balance of power between the incumbent regime and challengers, which is largely determined by regime type. We find that military regimes, although with more control over their own military forces, face a higher risk of coups compared to dominant-party regimes.
Secondly, the study introduces a redefined concept: the autocoup. Defined as an incumbent leader’s refusal to relinquish power as mandated, this research distinguishes autocoups from broader terms like self-coups. Based on this definition, we present the first publicly available dataset of autocoup events from 1945 to 2022, encompassing 110 attempts and 87 successful autocoups. Case studies and empirical analyses demonstrate the dataset’s utility for quantitative research, providing a robust foundation for further analysis on autocoups.
Thirdly, employing survival analysis techniques, the study finds clear differences in leader longevity between those who come to power through coups and those who extend their rule through autocoups. The results indicate that coup-installed leaders face a significantly higher risk of removal compared to autocoup leaders who manipulate the system to extend their rule.
5.2 Limitations and directions for future research
This study offers a novel framework for analysing irregular power transitions, but some limitations require further exploration:
Data refinement: Defining and classifying autocoups is a new approach. Future research should validate this classification system through additional studies and expert evaluations.
Data harmonization: The current analysis faces challenges due to mismatched units (country-year vs. leader) between coup and autocoup datasets. Future efforts should explore data harmonization techniques for more robust comparisons.
Democratic backsliding: While this study establishes a connection between irregular power transitions and democratic backsliding, further empirical evidence is needed to solidify this link.
Several avenues exist for future research:
Terminology and data collection: Refining the “autocoup” concept and achieving wider recognition will facilitate more accurate and comprehensive data collection.
Dataset expansion: Expanding the autocoup dataset with more cases and integrating it with data on other irregular leadership transitions can provide a more holistic view of political survival after these events.
Power dynamics and long-term impacts: Utilizing this dataset, future studies can delve deeper into power dynamics at play and explore the long-term consequences of irregular transitions on political systems, particularly regarding democratic backsliding, breakdown, and personalization of power.
In conclusion, this study sheds light on the dynamics of irregular power transitions, specifically focusing on coups and autocoups. By redefining autocoups, classifying the dataset, analysing determinants, and comparing leader longevity, we establish a framework for understanding irregular transitions and leader survival. This work contributes to a deeper understanding of democratic resilience and political stability. Future research can build upon this foundation by conducting further empirical analyses based on the novel autocoup dataset and continuing to refine the framework.
Antonio, Robert J. 2021. “Democracy and Capitalism in the Interregnum: Trump’s Failed Self-Coup and After.”Critical Sociology 48 (6): 937–65. https://doi.org/10.1177/08969205211049499.
Arriola, Leonardo R. 2009. “Patronage and Political Stability in Africa.”Comparative Political Studies 42 (10): 1339–62. https://doi.org/10.1177/0010414009332126.
Böhmelt, Tobias, and Ulrich Pilster. 2014. “The Impact of Institutional Coup-Proofing on Coup Attempts and Coup Outcomes.”International Interactions 41 (1): 158–82. https://doi.org/10.1080/03050629.2014.906411.
Bomprezzi, Pietro, Axel Dreher, Andreas Fuchs, Teresa Hailer, Andreas Kammerlander, Lennart Kaplan, Silvia Marchesi, Tania Masi, Charlotte Robert, and Kerstin Unfried. 2024. “Wedded to Prosperity? Informal Influence and Regional Favoritism.” Discussion Paper. CEPR.
Brown, Stephen. 2001. “Authoritarian Leaders and Multiparty Elections in Africa: How Foreign Donors Help to Keep Kenya’s Daniel Arap Moi in Power.”Third World Quarterly 22 (5): 725–39. https://doi.org/10.1080/01436590120084575.
Bueno de Mesquita, Bruce, Alastair Smith, Randolph M. Siverson, and James D. Morrow. 2003. The Logic of Political Survival. The MIT Press. https://doi.org/10.7551/mitpress/4292.001.0001.
Cameron, Maxwell A. 1998a. “Latin American Autogolpes : Dangerous Undertows in the Third Wave of Democratisation.”Third World Quarterly 19 (2): 219–39. https://doi.org/10.1080/01436599814433.
———. 2019. “Should I Stay or Should I Go? Term Limits, Elections, and Political Change in Kenya, Uganda, and Zambia.” In, 311–38. Oxford University PressOxford. https://doi.org/10.1093/oso/9780198837404.003.0016.
Chin, John J, David B Carter, and Joseph G Wright. 2021. “The Varieties of Coups D’état: Introducing the Colpus Dataset.”International Studies Quarterly 65 (4): 1040–51. https://doi.org/10.1093/isq/sqab058.
Choulis, Ioannis, Marius Mehrl, Abel Escribà-Folch, and Tobias Böhmelt. 2022. “How Mechanization Shapes Coups.”Comparative Political Studies 56 (2): 267–96. https://doi.org/10.1177/00104140221100194.
Davenport, Christian, Babak RezaeeDaryakenari, and Reed M Wood. 2021. “Tenure Through Tyranny? Repression, Dissent, and Leader Removal in Africa and Latin America, 19902006.”Journal of Global Security Studies 7 (1). https://doi.org/10.1093/jogss/ogab023.
Easton, Malcolm R, and Randolph M Siverson. 2018. “Leader Survival and Purges After a Failed Coup d’état.”Journal of Peace Research 55 (5): 596–608. https://doi.org/10.1177/0022343318763713.
Escribà-Folch, Abel. 2013. “Repression, Political Threats, and Survival Under Autocracy.”International Political Science Review 34 (5): 543–60. https://doi.org/10.1177/0192512113488259.
Fariss, Christopher J., Therese Anders, Jonathan N. Markowitz, and Miriam Barnum. 2022. “New Estimates of Over 500 Years of Historic GDP and Population Data.”Journal of Conflict Resolution 66 (3): 553–91. https://doi.org/10.1177/00220027211054432.
Frantz, Erica, and Elizabeth A. Stein. 2016. “Countering Coups: Leadership Succession Rules in Dictatorships.”Comparative Political Studies 50 (7): 935–62. https://doi.org/10.1177/0010414016655538.
Gandhi, Jennifer, and Adam Przeworski. 2007. “Authoritarian Institutions and the Survival of Autocrats.”Comparative Political Studies 40 (11): 1279–1301. https://doi.org/10.1177/0010414007305817.
Gassebner, Martin, Jerg Gutmann, and Stefan Voigt. 2016. “When to Expect a Coup d’état? An Extreme Bounds Analysis of Coup Determinants.”Public Choice 169 (3-4): 293–313. https://doi.org/10.1007/s11127-016-0365-0.
Geddes, Barbara, Joseph Wright, and Erica Frantz. 2014. “Autocratic Breakdown and Regime Transitions: A New Data Set.”Perspectives on Politics 12 (2): 313–31. https://doi.org/10.1017/s1537592714000851.
Ginsburg, Tom, James Melton, and Zachary Elkins. 2010. “On the Evasion of Executive Term Limits.”Wm. & Mary L. Rev. 52: 1807.
———. 2011. “On the Evasion of Executive Term Limits.”William and Mary Law Review 52: 1807.
Goemans, Henk E., Kristian Skrede Gleditsch, and Giacomo Chiozza. 2009. “Introducing Archigos: A Dataset of Political Leaders.”Journal of Peace Research 46 (2): 269–83. https://doi.org/10.1177/0022343308100719.
Krishnarajan, Suthan. 2019. “Economic Crisis, Natural Resources, and Irregular Leader Removal in Autocracies.”International Studies Quarterly 63 (3): 726–41. https://doi.org/10.1093/isq/sqz006.
Landau, David, Yaniv Roznai, and Rosalind Dixon. 2019. “Term Limits and the Unconstitutional Constitutional Amendment Doctrine.” In, 53–74. Oxford University PressOxford. https://doi.org/10.1093/oso/9780198837404.003.0004.
Licht, Amanda A. 2009. “Coming into Money: The Impact of Foreign Aid on Leader Survival.”Journal of Conflict Resolution 54 (1): 58–87. https://doi.org/10.1177/0022002709351104.
Marsteintredet, Leiv, and Andrés Malamud. 2019. “Coup with Adjectives: Conceptual Stretching or Innovation in Comparative Research?”Political Studies 68 (4): 1014–35. https://doi.org/10.1177/0032321719888857.
Mauceri, Philip. 1995. “State Reform, Coalitions, and The Neoliberal Autogolpe in Peru.”Latin American Research Review 30 (1): 7–37. https://doi.org/10.1017/s0023879100017155.
Mechkova, Valeriya, Anna Lührmann, and Staffan I. Lindberg. 2017. “How Much Democratic Backsliding?”Journal of Democracy 28 (4): 162–69. https://doi.org/10.1353/jod.2017.0075.
Morrison, Kevin M. 2009. “Oil, Nontax Revenue, and the Redistributional Foundations of Regime Stability.”International Organization 63 (1): 107–38. https://doi.org/10.1017/s0020818309090043.
Neto, Octavio Amorim, and Igor P. Acácio. 2019. “Presidential Term Limits as a Credible-Commitment Mechanism.” In, 123–40. Oxford University PressOxford. https://doi.org/10.1093/oso/9780198837404.003.0007.
Palmer, Harvey D., and Guy D. Whitten. 1999. “The Electoral Impact of Unexpected Inflation and Economic Growth.”British Journal of Political Science 29 (4): 623–39. https://doi.org/10.1017/s0007123499000307.
Peyton, Buddy, Joseph Bajjalieh, Dan Shalmon, Michael Martin, and Emilio Soto. 2024. “Cline Center Coup d’état Project Dataset.” University of Illinois at Urbana-Champaign. https://doi.org/10.13012/B2IDB-9651987_V7.
Pion-Berlin, David, Thomas Bruneau, and Richard B. Goetze. 2022. “The Trump Self-Coup Attempt: Comparisons and CivilMilitary Relations.”Government and Opposition 58 (4): 789–806. https://doi.org/10.1017/gov.2022.13.
Posner, Daniel N., and Daniel J. Young. n.d. “Term Limits: Leadership, Political Competition and the Transfer of Power.” In, 260–78. Cambridge University Press. https://doi.org/10.1017/9781316562888.011.
Powell, Jonathan. 2012. “Determinants of the Attempting and Outcome of Coups d’état.”Journal of Conflict Resolution 56 (6): 1017–40. https://doi.org/10.1177/0022002712445732.
———. 2017. “Leader Survival Strategies and the Onset of Civil Conflict: A Coup-Proofing Paradox.”Armed Forces & Society 45 (1): 27–44. https://doi.org/10.1177/0095327x17728493.
Powell, Christopher Faulkner, William Dean, and Kyle Romano. 2018. “Give Them Toys? Military Allocations and Regime Stability in Transitional Democracies.”Democratization 25 (7): 1153–72. https://doi.org/10.1080/13510347.2018.1450389.
Powell, and Thyne. 2011. “Global Instances of Coups from 1950 to 2010: A New Dataset.”Journal of Peace Research 48 (2): 249–59. https://doi.org/10.1177/0022343310397436.
Przeworski, Adam, Michael E. Alvarez, Jose Antonio Cheibub, and Fernando Limongi. 2000. “Democracy and Development,” August. https://doi.org/10.1017/cbo9780511804946.
Quiroz Flores, Alejandro, and Alastair Smith. 2012. “Leader Survival and Natural Disasters.”British Journal of Political Science 43 (4): 821–43. https://doi.org/10.1017/s0007123412000609.
Smith, Benjamin. 2004. “Oil Wealth and Regime Survival in the Developing World, 19601999.”American Journal of Political Science 48 (2): 232–46. https://doi.org/10.1111/j.0092-5853.2004.00067.x.
Stinnett, Douglas M., Jaroslav Tir, Paul F. Diehl, Philip Schafer, and Charles Gochman. 2002. “The Correlates of War (Cow) Project Direct Contiguity Data, Version 3.0.”Conflict Management and Peace Science 19 (2): 59–67. https://doi.org/10.1177/073889420201900203.
Sudduth, Jun Koga. 2017. “Strategic Logic of Elite Purges in Dictatorships.”Comparative Political Studies 50 (13): 1768–1801. https://doi.org/10.1177/0010414016688004.
Sudduth, Jun Koga, and Curtis Bell. 2018. “The Rise Predicts the Fall: How the Method of Leader Entry Affects the Method of Leader Removal in Dictatorships.”International Studies Quarterly 62 (1): 145–59. https://doi.org/10.1093/isq/sqx075.
Svolik, Milan W. 2014. “Which Democracies Will Last? Coups, Incumbent Takeovers, and the Dynamic of Democratic Consolidation.”British Journal of Political Science 45 (4): 715–38. https://doi.org/10.1017/s0007123413000550.
Tangri, Roger, and Andrew M. Mwenda. 2010. “President Museveni and the Politics of Presidential Tenure in Uganda.”Journal of Contemporary African Studies 28 (1): 31–49. https://doi.org/10.1080/02589000903542574.
Thyne, Clayton, Powell, Sarah Parrott, and Emily VanMeter. 2017. “Even Generals Need Friends.”Journal of Conflict Resolution 62 (7): 1406–32. https://doi.org/10.1177/0022002716685611.
Versteeg, Mila, Timothy Horley, Anne Meng, Mauricio Guim, and Marilyn Guirguis. 2020. “The Law and Politics of Presidential Term Limit Evasion.”Colum. L. Rev. 120: 173.
Williams, Laron K. 2011. “Pick Your Poison: Economic Crises, International Monetary Fund Loans and Leader Survival.”International Political Science Review 33 (2): 131–49. https://doi.org/10.1177/0192512111399006.
Wright, Joseph, Erica Frantz, and Barbara Geddes. 2013. “Oil and Autocratic Regime Survival.”British Journal of Political Science 45 (2): 287–306. https://doi.org/10.1017/s0007123413000252.
Yu, Shu, and Richard Jong-A-Pin. 2016. “Political Leader Survival: Does Competence Matter?”Public Choice 166 (1-2): 113–42. https://doi.org/10.1007/s11127-016-0317-8.
Footnotes
According to the Archigos dataset, “Removed by Military, without Foreign Support” and “Removed by Other Government Actors, without Foreign Support” in the variable exitcode are classified as coups.↩︎
The definitions and concepts of power expansion and power extension can often be ambiguous. In this study, we define power expansion as an incumbent acquiring additional authority from other branches or apparatuses of the state. Conversely, power extension refers to an incumbent prolonging their tenure beyond the originally mandated term in office.↩︎
Source Code
---title: "Leadership Transitions and Survival:\n Coups, Autocoups, and Power Dynamics"author: - name: Zhu Qi affiliations: University of Essex email: qz21485@essex.ac.ukformat: aps-html: default aps-pdf: documentclass: report keep-tex: true colorlinks: true urlcolor: blue link-citations: true linkcolor: blue fig-asp: 0.618 number-sections: true shift-heading-level-by: -1 toc: true toc-title: Contents lof: true lot: true crossref: lof-title: Figures lot-title: Tables linestretch: 1.618 indent: true fontsize: 12pt mainfont: "Times New Roman" sansfont: "Arial" monofont: "Courier New" #latex-engine: pdflatex pdf-engine: xelatex geometry: # - showframe - top = 3cm - bottom = 3cm - left = 3cm - right = 2.7cm header-includes: - \usepackage{sectsty} - \chapterfont{\centering} - \usepackage{lscape} - \newcommand{\blandscape}{\begin{landscape}} - \newcommand{\elandscape}{\end{landscape}} docx: defaultcategories: [Article, Coups, Autocoups, Leadership Survival]bibliography: references.bibinclude-before-body: title.texexecute: warning: false message: false---```{r}#| label: setup#| include: falseformat <- knitr::opts_knit$get("rmarkdown.pandoc.to")pacman::p_load(tidyverse, data.table, gt, gtsummary, gtExtras, janitor, scales, zoo, democracyData, vdemdata, readxl, here, stargazer, survival, survminer, ggsurvfit, sampleSelection, marginaleffects, glue)```## Acknowledgements {.unnumbered}The completion of this thesis marks the end of a significant journey, one filledwith hard work, continuous learning, and moments of profound joy. Thisdissertation would not have been possible without the support and encouragementof numerous individuals to whom I am deeply indebted.First and foremost, I extend my heartfelt gratitude to my esteemed supervisor,Professor Kristian Skrede Gleditsch. His invaluable guidance, unwavering support,and insightful feedback have been instrumental in shaping this research. Hisexpertise and encouragement have truly been the cornerstone of this academicendeavor.I am equally grateful to Professor Han Dorussen, the chair of my board panel, forhis continuous support and constructive criticism. His thoughtful input hassignificantly enhanced the quality and depth of my research.I would like to acknowledge the vital contributions of my initial co-supervisors,Dr. Saurabh Pant and Professor David Siroky. Their instruction and guidanceduring the early stages of my research, prior to their departure from theUniversity of Essex, laid a strong foundation for this work.This dissertation has benefited immensely from the expertise of several esteemedscholars in the field. I am particularly indebted to Dr. Brian J Phillips, Dr.Prabin Khadka, and Dr. Winnie Xia for their insightful comments, advice, andsuggestions. Their contributions have undoubtedly enriched this research.On a personal note, I owe a debt of gratitude to my family for their unwaveringsupport and love. To my beloved wife, Ji Zhi, who has been my steadfast pillar ofstrength throughout this journey - your patience and encouragement have beenimmeasurable. To my dear children, Siyan and Sisheng, your joy and curiosity havebeen a constant source of motivation and inspiration.I am profoundly thankful to my father for his enduring support and belief in myabilities. To the cherished memory of my late mother - your love, guidance, andthe values you instilled in me continue to shape my path and inspire myendeavors. And to my three brothers, whose support enabled me to pursue my PhDwithout worries - your contribution to this achievement cannot be overstated.While many have contributed to the success of this work, any errors orshortcomings remain entirely my own.## Abstract {.unnumbered}This dissertation investigates the dynamics of irregular power transitions,focusing on coups and autocoups, and their impact on leader survival. Through acomprehensive analysis of historical data, case studies, and advancedquantitative modelling, this research enhances our understanding of politicalpower dynamics and the complex factors influencing leadership longevity followingirregular transitions.The study begins by examining the crucial role of power dynamics, shaped byregime type, in determining coup success rates and attempt frequency. Employing adouble probit model with sample selection, the research demonstrates that theanticipated likelihood of coup success significantly affects coup attempts, withmilitary regimes exhibiting particular vulnerability due to their inherent powerstructure.While often overlooked in existing literature, autocoups are revealed to have aprofound impact on democratic backsliding. This study introduces a refineddefinition of autocoups and presents a novel dataset covering events from 1945 to2024, facilitating more robust quantitative analysis of this phenomenon.Utilizing survival analysis techniques, the research compares the tenure ofleaders who ascend to power through coups versus autocoups. The findings indicatethat coup-installed leaders experience significantly shorter tenures and facehigher risks of removal. In contrast, autocoup leaders, who manipulate existingsystems to extend their rule, tend to remain in power longer. This disparitysuggests that autocoups may incentivize power grabs and contribute to democraticerosion.This dissertation makes several significant contributions to political scienceliterature:- Conceptual clarity: Establishes a precise definition of autocoups, addressing a gap in the current understanding of this understudied phenomenon.- Data innovation: Introduces a comprehensive dataset on autocoups, enabling more extensive quantitative analyses and fostering future research in this area.- Analytical framework: Develops a unified approach for studying both coups and autocoups, allowing for comparative analysis of these critical political events and their potential effects on democratic backsliding.***Keywords:** Coups, Autocoups, Power transitions, Leadership survival,Democratic backsliding, Regime type, Political instability*## Introduction### Research Question and ContextIrregular power transitions, characterised by a disregard for constitutionalprocedures, represent a critical area of study in political science. Thesetransitions not only disrupt established rules but also often employunconstitutional tactics to consolidate power post-transition. Moreover, suchevents can inspire emulation among other ambitious leaders, potentiallytriggering a cascade of similar actions across different political contexts.Despite extensive research on irregular power transitions, a fundamental questioncontinues to challenge political scientists:**Why do some leaders face premature ousting while others extend their tenurebeyond constitutionally mandated limits?**This inquiry extends to understanding the stark disparities in leadershiplongevity, where some rulers maintain power for decades while others' tenures aremeasured in years, months, or even days.This dissertation focuses on this central question, aiming to provide acomprehensive analysis dedicated to understanding:- How leaders ascend to or overstay in power through unconstitutional means.- What factors determine the duration of a leader's rule following an irregular ascent to power.### Coups and Autocoups: A Comparative Unified Framework for Irregular Power TransitionsIn the study of irregular power transitions, coups have traditionally dominatedacademic discourse due to their frequency and visibility. According to theArchigos dataset [@goemans2009], coups[^1] accounted for more than half of theapproximately 145 irregular leader exits from 1945 to 2015. The Global Instancesof Coups (GIC) dataset [@powell2011] records an even higher number, with 245coup-related removals from 1950 to 2024.The prevalence of coups has led to a well-established definition and extensiveresearch, particularly since the turn of the millennium [@thyne2019]. Thisscholarly focus has resulted in a general consensus on the definition of coups,as well as the development of comprehensive datasets for quantitative analyses.While debates on the precise definition persist, most scholars, including thisstudy, adhere to the definition proposed by @powell2011. This widely accepteddefinition encompasses two fundamental elements:- **Perpetrators and Victims**: Perpetrators of coups are elites within the existing power structure and the victims are incumbent executive leaders.- **Strategy and Aim**: The primary objective is the complete removal of incumbents from power.It is important to note that this definition excludes actions that only partiallychallenge the incumbent's authority or seek policy changes without aiming for acomplete transfer of power.The scholarly consensus on coup definition has facilitated the development ofseveral comprehensive datasets, enabling rigorous quantitative analyses inpolitical science research. Notable among these are: GIC Dataset, Cline CentreCoup d'État Project Dataset [@peyton2024], and Colpus Dataset [@chin2021]. Thesedatasets have become invaluable resources in political science, enablingresearchers to conduct large-scale, comparative studies on the causes, dynamics,and consequences of coups across different political contexts and historicalperiods.However, another form of irregular power transition has been largely overlookedin power transition studies: leaders who refuse to relinquish power and extendtheir mandated terms. At least three challenges still exist in the analyses ofthis type of irregular power transition:- **Terminology**: Unlike the more widely accepted term "coup," various terms (e.g., self-coup, autogolpe, executive coup, incumbent takeover) are used to describe this phenomenon.- **Definition**: Existing definitions often conflate power expansion and power extension[^2]. For instance, @cameron1998a defines an autogolpe primarily in terms of power expansion, which does not align with the definition of coup.- **Data**: A consensus dataset for autocoups is lacking, with existing datasets varying in terminology, definitions, and coverage years @baturo2022.To address this gap in the literature and provide a framework for analysis, wepropose the term "autocoup" to describe this type of transition. This conceptwill be extensively discussed and defined in Chapter 3, providing a basis forintegrating these events into the broader study of irregular power transitionsand their impact on political systems and leader longevity.Analysing coups and autocoups within a unified framework is crucial for acomprehensive understanding of irregular power transitions and leader survivalfor three reasons:- Both phenomena significantly influence democratic backsliding and represent the most frequent means of irregular power transition.- Their similar nomenclature reflects a fundamental similarity: while a coup aims to replace the current leader, an autocoup seeks to prevent the succession of a future leader.- This comparative approach will enable a more nuanced analysis of irregular power transitions, contributing to our understanding of political stability, democratic erosion, and leadership longevity in various regime types.### Academic ContributionsThis study addresses a critical gap in the literature by offering a unifiedframework for analysing both coups and autocoups. My contributions are threefold:- **Emphasis on Power Dynamics and Regime Types**: This study highlights the significant role of power dynamics, particularly the influence of regime types, in determining the success and frequency of coup attempts, underscores how the expected chances of coup success motivate such attempts, with military regimes being notably susceptible.- **Refined Definition and Novel Dataset for Autocoups**: I introduce a refined definition of autocoups, develop a novel dataset covering events from 1945 to 2024, and fill a significant gap in the existing literature and enable a comparative analysis with classic coups.- **Survival Analysis of Leaders from Different Entry Modes**: This research applys survival analysis to existing coup data and the new autocoup dataset, demonstrates how different modes of entry into power significantly affect leader survival, and reveals that leaders who come to power through coups typically have shorter tenures and face higher removal risks compared to those who extend their rule through autocoups.These contributions collectively advance the field by providing a more holisticunderstanding of irregular power transitions, offering new tools and data forquantitative analysis of autocoups, and demonstrating the interconnectedness ofpower acquisition methods and leadership survival.This work lays the foundation for future research on the dynamics of politicalpower, regime stability, and democratic backsliding, offering both theoreticalinsights and practical implications for policy-makers and scholars alike.### ImplicationsThe examination of irregular power transitions provides a crucial perspective onthe interrelated phenomena of democratic backsliding, breakdown, and autocraticintensification. The findings of this study provide logical explanations forseveral political phenomena:- **Regression of Global Democracy Levels**: This study can explain why global democracy levels have regressed to pre-2000 levels. For example, Freedom House reports an 18th consecutive year of global freedom decline in 2023[@freedomhouse2024freedom]. Irregular power transitions inevitably violate democratic or constitutional norms and disrupt the trajectory towards stable democracies or democratization from autocracies. Leaders who ascend through irregular means often undermine constitutional norms to seize or overstay in power, creating a vicious cycle of eroding democratic institutions.- **Within-Regime Democratic Erosion**: This study explains why democratic backsliding often occurs within regimes, with democracies becoming less liberal and autocracies becoming less competitive [@mechkova2017]. This is particularly because autocoups have become more prevalent than traditional coups since 2000 [@bermeo2016]. Importantly, autocoups typically do not result in immediate regime change, contributing to the observed pattern of within-regime democratic erosion.- **Prevalence of Autocoups Since 2000**: This research explains why autocoups are more prevalent since 2000. Firstly, autocoups have a significantly higher probability of success compared to traditional coups. Additionally, even when unsuccessful, the repercussions of autocoups are generally less severe than those of failed traditional coups. Moreover, successful autocoup leaders tend to maintain power for substantially longer periods than leaders who enter through traditional coups.### Overview of the ThesisThis dissertation investigates the complex dynamics of irregular powertransitions and their implications for leadership survival and democraticprocesses. The study examines three key aspects: the determinants of classic coupattempts, the conceptualization and analysis of autocoups, and the impact ofpower acquisition methods on leadership longevity.#### Chapter 2: Determinants of Classic Coup Attempts {.unnumbered}This chapter delves into the factors influencing classic coup attempts, with anovel focus on the less observable but crucial factor of expected coup successrates. Key features include the utilization of a double probit model with sampleselection and the analysis of how expected success rates significantly influencecoup attempts. It also examines how regime types shape the balance of powerbetween incumbents and challengers, determining the chances of success of coups.Findings indicate a substantially higher coup risk in military regimes comparedto dominant-party regimes.#### Chapter 3: Conceptualising and Analysing Autocoups {.unnumbered}This chapter refines and analyses the concept of autocoups, with a specific focuson power extensions by incumbent leaders. Notable elements include theredefinition of autocoups as instances where incumbent leaders refuse mandatedpower transitions. Introduction of a novel dataset covering autocoup events from1945 to 2024, encompassing 110 attempts and 87 successes. Presentation of casestudies and empirical analyses demonstrating the utility of this dataset for morequantitative research in political science.#### Chapter 4: Impact of Power Acquisition Methods on Leadership Longevity {.unnumbered}This chapter investigates how the method of power acquisition affects the tenureof leaders who ascend through coups versus those who extend their tenure viaautocoups. Key aspects include hypothesis testing on the significant impact ofthe accession method on leader tenure. Evidence of differing survival timesbetween coup-entry and autocoup leaders, employing the Cox proportional hazardsmodel and time-dependent Cox model. Analysis of how the risk-reward profile ofautocoups might motivate democratic backsliding and power personalization.#### Chapter 5: Conclusion and Future Directions {.unnumbered}The final chapter synthesizes the summary of key findings from each substantivechapter. Acknowledges study limitations and their potential impact on findings.Outlines future research directions, emphasizing the need for further explorationof irregular power transitions, particularly coups and autocoups.## Power Dynamics and Coup Attempts: A Selection Mechanism Analysis {#sec-chapter2}### Abstract {.unnumbered}Despite extensive research identifying around one hundred potential determinantsof coup attempts, no consensus has been reached. This study introduces a novelapproach that prioritizes determinants based on their impact on coup success. Byanalysing coup success rates, the study hypothesizes that the expected outcomesof coups are critical determinants of their occurrence. Utilizing a double probitmodel with sample selection, the research investigates the relationship betweenregime types and coup attempts. The findings confirm that regime type, by shapinginternal power dynamics, is a crucial determinant of coup likelihood. Militaryand personalist regimes, characterized by weaker institutional frameworks andhigher vulnerability, demonstrate significantly higher susceptibility to coupscompared to dominant-party regimes.***Keywords:** Coups, Power transitions, Sample selection*\newpage### Introduction```{r}#| label: tbl-coup_year_counts#| tbl-cap: "Coups by year and by country"#| tbl-subcap:#| - "by year frequency"#| - "by country frequency"#| layout-ncol: 2#| echo: !expr format == "html"#| include: !expr format == "html"#load coup dataCoup <-fread("https://www.uky.edu/~clthyn2/coup_data/powell_thyne_coups_final.txt")attempts <-length(Coup$country)succeeded <-length(Coup$country[Coup$coup ==2])coups <- Coup |>summarise(`Coup attempted`=n(),`Coup succeeded`=sum(coup ==2),.by = country )Coup |>summarise(counts =n(),.by = year) |>arrange(desc(counts)) |>head(15) |>clean_names(case ="title") |>gt() |>tab_options(table.width =pct(95),column_labels.font.weight ="bold") |>tab_source_note(source_note =md("***Source: GIC dataset***"))Coup |>summarise(counts =n(),.by = country) |>arrange(desc(counts)) |>head(15) |>clean_names(case ="title") |>gt() |>cols_align(align ="right",columns =2 ) |>tab_options(table.width =pct(95),column_labels.font.weight ="bold") |>tab_source_note(source_note =md("***Source: GIC dataset***"))```Coups d'état, defined by @powell2011 as "illegal and overt attempts by themilitary or other elites within the state apparatus to unseat the sittingexecutive" (p. 252), represent a critical challenge to political stability anddemocratic governance worldwide. This chapter examines the complex dynamics ofcoup attempts, their success rates, and the factors that influence both theiroccurrence and outcomes.Coups occur with varying frequency across countries and regions. For instance: InLatin America, Bolivia experienced`r coups |> filter(country == "Bolivia") |> pull(2)` coups between 1950 and 1984,while Argentina saw `r coups |> filter(country == "Argentina") |> pull(2)` duringa similar period. In Africa, Sudan endured`r coups |> filter(country == "Sudan") |> pull(2)` coups between 1955 and 2023.Contrastingly, countries like Mexico (1917-2000) and South Africa (since 1950)have not experienced any coups.This variability raises a fundamental question: **Why are coups more frequent insome countries than others?**Despite decades of research, political scientists have yet to reach a consensuson the key determinants of coups. @gassebner2016 highlight this challenge, notingthat approximately 100 potential factors have been proposed, with their studytesting 66 of these across three million model permutations. This proliferationof variables presents a critical issue: How can we establish a framework thatallows scholars to focus on the most relevant factors, rather than navigating anever-expanding list of potential determinants?My analysis reveals a significant oversight in previous research: the exclusivefocus on pre-coup conditions without adequate consideration of post-coup factors.This approach has neglected a critical element in coup dynamics – the expectedprobability of success. Key observations supporting this perspective, as shown in@tbl-coups, include:- **High-stakes nature of coups**: Failed coups often result in severe consequences for perpetrators, including imprisonment, exile, or death.- **Selectivity in coup attempts**: Despite 491 coup attempts since 1950, they represent only about 4% of over 12,000 country-years in the same period (GIC dataset).- **Satisfactory success rate**: Approximately half of all coup attempts succeed, suggesting careful selection of opportunities by plotters.```{r}#| label: tbl-coups#| tbl-cap: Top 10 countries with the most coup attempts#| echo: !expr format == "html"coups |># add one row which adds up the totalarrange(desc(`Coup attempted`)) |>head(11) |>add_row(country ="Total",`Coup attempted`=sum(coups[, 2]),`Coup succeeded`=sum(coups[, 3]) ) |>mutate(Success_rate =percent(`Coup succeeded`/`Coup attempted`, accuracy =0.1 )) |>clean_names(case ="title") |>gt() |>tab_options(table.width =pct(90),column_labels.font.weight ="bold",table.layout ="fixed" ) |># cols_width(1 ~ px(120),# 2:4 ~ px(88)# ) |>cols_align(align ="center",columns =2:3 ) |>tab_source_note(source_note =md("*Source: GIC dataset*"))```The low occurrence rate and high success rate clearly indicate that theinitiation of coups is highly selective. In other words, the likelihood of a coupoccurring depends greatly on its probability of success. However, thisprobability is not directly observable to outsiders, including researchers, priorto a coup attempt.Given the limitations in directly observing coup success probabilities, I proposefocusing on regime type as a crucial proxy for coup outcomes. This approach isbased on the following reasoning: Coup outcomes are ultimately determined by thebalance of power within a regime and regime types are classified based on controlover the military, policy-making authority, and appointment of officials. Byanalysing power dynamics across different regime types, we can gain valuableinsights into the structural factors shaping coup attempts and their outcomes.To address the selection bias inherent in studying coup attempts, I employ adouble probit model with sample selection. This approach allows for simultaneousanalysis of factors influencing both coup initiation and success.Through this comprehensive analysis, I aim to provide a more nuancedunderstanding of the factors driving coup attempts and shaping their outcomes,contributing to both scholarly discourse and practical efforts in two key ways:- **Emphasis on expected chances of success**: By focusing on the probability of success as a driver of coup attempts, I offer a more targeted approach to understanding coup dynamics.- **Highlighting the significance of regime type**: I demonstrate how regime type influences coup likelihood, even when researchers lack perfect knowledge of internal balance of power.The remainder of this chapter is organized as follows: Section 2 explores thedynamics of coup attempts and their outcomes. Section 3 outlines the researchdesign, methodology, and variables. Section 4 presents and discusses theempirical findings. Section 5 concludes with key insights and their implicationsfor understanding and potentially mitigating coup risks.### Dynamics of coup attempts and outcomesCoup attempts are driven by a complex interplay of factors, encompassing both themotivations of potential challengers (**disposition**) and the resources andopportunities available to them (**capability**). This section delves into thedynamics of coup attempts, exploring the motivations behind them, the factorsinfluencing their success, and the role of regime types in shaping coupsusceptibility.#### Motivations for coupsThis section focuses on the motivations that compel actors to undertake coups,categorizing them into three main types:- **Personal Ambition:** The allure of absolute power, prestige, and wealth serves as a significant motivator for many coup plotters. Seizing control promises the ability to shape national policies, control resources, and make impactful decisions without constraints. The pursuit of prestige, recognition, potential economic gain, and the desire to leave a lasting legacy can further incentivize individuals to undertake this risky endeavour.- **Purported National Interest:** Coups are sometimes justified as necessary interventions to address national crises, uphold the constitution, or facilitate a transition to democracy. While such claims require scrutiny, genuine examples exist. For instance, the 2010 coup in Niger ousted President Tandja, who attempted an unconstitutional third term by dissolving the opposing court and calling a self-serving referendum [@ginsburg2019].- **Self-Preservation:** In certain cases, coups act as pre-emptive strikes against perceived threats. Coup leaders might not necessarily seek power for themselves but rather fear elimination or political persecution by the incumbent leaders. An example is Idi Amin's 1971 coup against Ugandan President Obote, who was attempting to remove Amin from his military command position [@sudduth2017].These motivations are often most prevalent in autocratic regimes, wherejustifications under the guise of national interest or self-preservation can maskpersonal agendas. Stable democracies rarely face the same level of constitutionalcrises or political persecution that might necessitate a coup. However, newlyestablished democracies can be vulnerable to instability, economic downturns, anddemocratic backsliding, creating opportunities for coup plotters to exploit theseweaknesses and justify their actions.Despite these potential motivations, coups remain relatively uncommon, occurringin only about 4% of country-years since 1950. This rarity highlights theimportance of capability – even the most motivated actors need the resources andopportunities to succeed.#### Capability for coupsThe decision to attempt a coup hinges not only on motivation but also on acalculated assessment of the chances of success. Several factors influence thisassessment:- **Military Strength**: A clear advantage in military capabilities compared to the incumbent regime significantly increases the odds of a successful coup[@powell2018; @choulis2022].- **Internal Divisions within the Regime**: Existing fractures within the government's power structure can be exploited by coup plotters to gain support from disgruntled factions.- **Public Support**: Widespread discontent with the incumbent regime, especially within the military or key sectors of society, can create a ripe environment for a successful coup.- **Foreign Backing**: External support from powerful nations can provide resources, legitimacy, and even direct military intervention to tip the scales in favor of the coup plotters.While historical data might suggest a high success rate for coups, it is crucialto consider selection bias. We only observe attempted coups, not the numerousplots that never materialize. Analysing launched coup data alone can bemisleading. To understand coup attempts and their likelihood comprehensively, weneed a theoretical framework that accounts for this selection bias.#### **Effect of Expected Coup Outcomes on Coup Attempts**A frequently cited framework [@gassebner2016; @aidt2019] offers a structuredapproach to assess the disposition and capability of coup attempts by evaluatingthe anticipated benefits for coup plotters. The expected pay-off of a coup can berepresented by the equation:$$\begin{aligned}E(U) = p \times B + (1 - p) \times (-C)\end{aligned}$$ {#eq-eq1}Where:- $E(U)$**:** Expected utility or pay-off of the coup attempt;- $B$ represents the return of a successful coup;- $C$ signifies the cost of a failed coup;- $p$ represents the probability of coup success.The condition for staging a coup is when the expected benefit is positive($E(U) > 0$). Rearranging the equation, we get:$$\begin{aligned}p \times B > (1 - p) \times C\end{aligned}$$ {#eq-eq2}This implies that for a coup to be attempted, the expected benefits of successmust outweigh the expected costs of failure.Quantifying $B$ and $C$ is inherently difficult. The loss of life, freedom, orloved ones after a failed coup, as well as the value of assuming leadership aftera successful coup, are challenging to measure precisely.However, the framework's core logic remains valuable. Given the difficulty inprecisely quantifying $B$ and $C$, we can treat them as roughly equal. Thisallows us to shift our focus to the probability of success ($p$). The simplified@eq-eq2 becomes:$$\begin{aligned}p > (1-p)\end{aligned}$$ {#eq-eq3}This suggests that a success probability greater than 50% is necessary for a coupto be attempted. While empirical data shows a slightly lower overall success ratefor coups since 1950 (49.9%, as shown in @tbl-coups), it is crucial to rememberthat this is an average and does not reflect the specific probabilities assessedby coup plotters beforehand.Therefore, in this study, unlike the majority of previous research that focuseson the prerequisites for triggering a coup, I consider the expected coup outcomeas the fundamental determinant of a coup. Based on this discussion, I propose thefirst hypothesis:> ***H1: The fundamental determinant of a coup attempt is the perceived chance of> success. Coup plotters likely require a success threshold of at least 50%.***This leads to the next question: What factors determine coup success andconsequently influence the decision to attempt one? The answer lies inunderstanding the balance of power inherent in the different power dynamicsacross various regime types.#### Regime types and power dynamicsHistorical coup success rates do not dictate individual coup attempts as each onefaces different situation and conditions. Coup plotters assess their chancesbased on their unique context. Although many factors determine the success of acoup, ultimately, the decisive factor remains the comparison of strength betweenincumbents and coup perpetrators. More precisely, it is the comparison of theirmilitary power.While military strength is often seen as the primary factor in coup attempts,leading to an oversimplification of coups as solely military events, the realityis far more complex [@singh2016]. The clandestine nature of coups necessitatessmall, secretive groups, making it difficult to gauge the stance of otherfactions within the military. As @geddes1999 points out, the success of a coupoften hinges on the reactions of these other factions. This uncertainty creates acomplex decision-making environment for potential coup plotters. Furthermore,incumbents often employ strategic division within the military as a coup-proofingmeasure. @böhmelt2014 describe how leaders intentionally create rival groupswithin the armed forces, establishing an artificial balance and structuralobstacles. In addition, factors beyond military force, such as internal divisionswithin the ruling elites, public support, and foreign backing, significantlyshape the balance of power.Therefore, a more holistic approach involves analyzing the overall balance ofpower within a political system. However, as noted in the introduction of thischapter, directly observing this balance of power presents significant challengesfor outside observers. In stead, an alternative approach is to analyze thefactors that are decisive in shaping power dynamics. This method aligns moreclosely with academic research goals, which are less concerned with identifyingspecific power challengers capable of launching successful coups and more focusedon understanding the patterns and underlying factors of coup attempts.Among the various factors that influence power dynamics, regime type emerges asone of the most crucial. First of all, regime types are fundamentallyclassifications based on power structures within a political system. Secondly,different regime types allocate authority differently for critical decisions suchas: deployment of military forces, appointment of key officials and militaryofficers, and formulation and implementation of main policies.provide a classification of autocratic regime types based on their powerstructures. This classification offers valuable insights into the power dynamicswithin different autocracies and their relative susceptibility to coups. Thethree main types of autocratic regimes, their characteristics, and theirvulnerability to coup attempts are explored as following (@tbl-regimes1).@geddes2014 provide a classification of autocratic regime types based on theirpower structures. The three main types in autocracies and their power dynamicsare:- **Military Regimes:** Characterized by a junta – a group of military officers controlling leadership selection and policy formulation. Examples include regimes in Brazil (1964-1985), Argentina (1976-1983), and El Salvador (1948-1984) [@geddes1999]. **Coup Susceptibility:** Despite their military nature, these regimes are surprisingly unstable due to internal power struggles within the junta. The absence of a clear final authority and the presence of multiple military factions increase the likelihood of resorting to force to resolve disputes, making these regimes the most vulnerable to coups.- **Personalist Regimes:** Power is concentrated in a single, charismatic leader who controls the military, policy, and succession. Examples include Rafael Trujillo's regime in the Dominican Republic (1930-1961), Idi Amin's regime in Uganda (1971-1979), and Jean-Bédel Bokassa's regime in the Central African Republic (1966-1979) [@geddes1999]. **Coup Susceptibility:** Personalist regimes are relatively stable during the leader's tenure. However, they face a higher risk of coups due to unclear succession plans and vulnerabilities associated with the leader's personal weaknesses, health, and mortality.- **Dominant-Party Regimes:** Power resides within a well-organized ruling party, with leaders acting as its representatives. The party structure and ideology foster internal cohesion and a long-term vision. Examples include the Partido Revolucionario Institutional (PRI) in Mexico, the Revolutionary Party of Tanzania (CCM), and Leninist parties in various Eastern European countries [@geddes1999]. **Coup Susceptibility:** Dominant-party regimes exhibit the greatest resilience against coups due to their institutionalized structures, unified leadership, clear ideology, and internal discipline.Empirical data also supports this theoretical framework. While military regimesrepresent only 5.6% of country-years since 1950, they experience adisproportionate share of coups (over 22%). Personalist regimes, constituting 13%of country-years, account for 23% of coups. Conversely, dominant-party regimes,representing 22.6% of country-years, account for only 16.7% of coups(@tbl-regimes). These statistics demonstrate the disproportionate vulnerabilityof military and personalist regimes to coups, while highlighting the relativestability of dominant-party regimes.The distinct power dynamics exhibited by these regime types significantlyinfluence their susceptibility to coups. This analysis leads to our secondhypothesis:The susceptibility to coups varies significantly among different types ofautocratic regimes, with military regimes being the most vulnerable, followed bypersonalist regimes, and dominant-party regimes being the least susceptible.> ***H2: The susceptibility to coups varies significantly among different types> of autocratic regimes, with military regimes being the most vulnerable,> followed by personalist regimes, and dominant-party regimes being the least> susceptible.***\newpage\blandscape```{r}#| label: tbl-regimes1#| tbl-cap: Main features of different types of regimes #| echo: !expr format == "html"read.csv(here("data/regimes_comparison.csv")) |>clean_names(case ="title") |>gt() |>tab_options(column_labels.font.weight ="bold",table.width =pct(100),table.align ="center",page.orientation ="landscape" ) |>cols_width(1:6~px(100) ) |>tab_source_note(source_note =md("*Source: GWF & Author*"))```\newpage```{r}#| label: tbl-regimes#| tbl-cap: Regime types and coups since 1950 #| echo: !expr format == "html"regimes <- REIGN |>distinct(reign_cowcode, year, .keep_all = T) |>mutate(regime_type =case_when( gwf_regimetype %in%c("party-based","party-personal","party-military","party-personal-military","oligarchy" ) ~"dominant-party", gwf_regimetype =="personal"~"personal", gwf_regimetype %in%c("military", "military personal", "indirect military") ~"military", gwf_regimetype =="monarchy"~"monarchy", gwf_regimetype %in%c("presidential", "parliamentary") ~"democracy",TRUE~"other" ) ) |>filter(year >1949)coup_model <-fread("https://raw.githubusercontent.com/reddylee/Datasets/main/coup_model.csv")regime_total <- coup_model |>select(regime_type = regime) |>tabyl(regime_type) |>arrange(-n) |>adorn_totals("row") |>mutate(country_year =n,share =percent(percent),.keep ="unused") coup1 <- Coup |>select(country, ccode, year, coup)coup_by_regime <- regimes |>select(ccode = reign_cowcode, extended_country_name, year, regime_type) |>full_join(coup1, by =join_by(ccode, year)) |>mutate(regime =na.locf(regime_type, .by = ccode)) |>filter(!is.na(coup)) |>tabyl(regime) |>adorn_totals("row") |>mutate(num_of_coups = n,percent_of_coups =percent(percent,accuracy =0.1),.keep ="unused") coup_by_succeed <- regimes |>select(ccode = reign_cowcode, extended_country_name, year, regime_type) |>full_join(coup1, by =join_by(ccode, year)) |>mutate(regime =na.locf(regime_type, .by = ccode)) |>filter(!is.na(coup), coup ==2) |>tabyl(regime) |>adorn_totals("row") |>select(regime, succeeded_coup = n) coup_regimes <- coup_by_regime |>left_join(coup_by_succeed) |>mutate(success_rate =percent(succeeded_coup / num_of_coups, accuracy =0.1))regime_total |>left_join(coup_regimes, join_by(regime_type == regime)) |>mutate(regime_type =str_to_title(regime_type),coup_likelihood =percent(num_of_coups / country_year, accuracy =0.1)) |>select(-succeeded_coup) |>clean_names(case ="title") |>gt() |>tab_options(column_labels.font.weight ="bold",table.width =pct(100),table.align ="center") |>tab_source_note(source_note =md("*Source: REIGN and GIC Datasets*"))```\elandscape### Research Design#### Double probit with sample selection model {#sec-coup231}This study employs a sophisticated statistical approach to account for theselective nature of coup attempts. While coup attempt rates vary across regimes,success rates tend to be surprisingly consistent, hovering around 50% (as shownin @tbl-regimes). This suggests that coup attempts are not random acts, butrather strategically planned and undertaken only when the odds of success appearfavorable. A standard statistical model would not account for this selectivity,potentially leading to biased results.To address this issue, we utilize a double probit with sample selection model,similar to the approach used by @powell2012 and @böhmelt2014. This model, knownas a Heckman probit model or bivariate probit model with sample selection,consists of two parts:- **Selection Equation (Stage 1)**: This stage analyses the factors influencing whether a coup attempt occurs in a particular country-year.- **Outcome Equation (Stage 2)**: This stage focuses on the probability of success for those coup attempts that actually take place.The selection equation (first stage) models the probability that a coup attemptoccurs:$$\begin{aligned}y_{1i}^*&=\alpha_0 + \alpha_1 Regime_i + \mathbf{X}_i \boldsymbol{A} + \mu_{1i}\\\\y_{1i} &= \begin{cases} 1 &\text{if $y_{1i}^*>0$ (coup attempt occurs)} \\\\0 &\text{if $y_{1i}^*\le0$ (no coup attempt)}\end{cases}\end{aligned}$$ {#eq-eq4}The outcome equation (second stage) models the probability that a coup attempt issuccessful, given that it occurs:$$\begin{aligned}y_{2i}^*&=\beta_0 + \beta_1 Regime_i + \mathbf{Z}_i \boldsymbol{B} + \mu_{2i}\\\\y_{2i} &= \begin{cases} 1 &\text{if $y_{2i}^*>0$ (coup succeeds)} \\\\0 &\text{if $y_{2i}^*\le0$ (coup failes)}\end{cases}\end{aligned}$$ {#eq-eq5}Where:- $y_{1i}^*$ *and* $y_{2i}^*$ are latent variables- $\text{Regime}_i$ is a categorical variable (military, personalist, or dominant-party)- $\mathbf{X}_i$ and $\mathbf{Z}_i$ are vectors of control variables- $\mu_{1i}$ and $\mu_{2i}$ are error terms, assumed to follow a bivariate normal distribution with correlation $\rho$The model assumes:$$\binom{\mu_{1 i}}{\mu_{2 i}} \sim N\left(\binom{0}{0},\left(\begin{array}{ll}1 & \rho \\\rho & 1\end{array}\right)\right)$$The probability equations are:$$\begin{aligned}P\left(y_{1 i}=1\right) & =\Phi\left(\alpha_0+\alpha_1 \operatorname{Regime}_i+\mathbf{X}_i \boldsymbol{A}\right)\end{aligned}$$ {#eq-eq7}$$\begin{aligned}P\left(y_{2 i}=1 \mid y_{1 i}=1\right) & =\Phi\left(\beta_0+\beta_1 \operatorname{Regime}_i+\mathbf{Z}_i \boldsymbol{B}\right)\end{aligned}$$ {#eq-eq8}Where $\Phi(\cdot)$ is the cumulative distribution function of the standardnormal distribution.#### Variables##### Dependent variableThe analysis utilizes data on coup attempts and outcomes from @powell2011. Asuccessful coup is defined as one where the incumbent leader is removed frompower for more than seven days. The dataset covers the period from 1950 to 2023and includes information on 491 coup attempts, with roughly half (245) beingsuccessful. Descriptive statistics for these coup attempts and regime types canbe found in @tbl-coups and @tbl-regimes.- **Coup Attempt**: Binary variable indicating whether a coup attempt occurred (1) or not (0) in a given country-year.- **Coup Success**: Binary variable indicating whether a coup attempt was successful (1) or failed (0), conditional on a coup attempt occurring.##### Key Independent Variable: Regime Type {#sec-chapter2322}I categorize regime types following @geddes2014 (GWF), focusing on military,personalist, and dominant-party regimes, with democracies and monarchies includedfor comparison. Descriptive statistics for regime types are presented in@tbl-regimes.##### Control variables {.unnumbered}The control variables are chosen based on the research of @gassebner2016. Theyanalysed 66 factors potentially influencing coups and found that slow economicgrowth, prior coup attempts, and other forms of political violence areparticularly significant factors. Therefore, we include economic performance,political violence, and the number of previous coups as our main controlvariables.```{r}#| label: variables#| results: asis#| echo: falsegdppc <-"- **Economic Level:** Represented by GDP per capita. This measure provides an indication of the overall economic health and standard of living in a country. We use GDP per capita data (in constant 2017 international 1000 dollars, PPP) from the V-Dem dataset by @fariss2022."ct <-"- **Economic Performance:** Measured using the current-trend ($CT$) ratio developed by @krishnarajan2019. This ratio compares a country's current GDP per capita to the average GDP per capita over the previous five years. A higher $CT$ ratio indicates stronger economic performance. For a country $i$ at year $t$, the $CT$ ratio is calculated as follows:"glue(gdppc)glue(ct)```$$ \begin{aligned} CT_{i,t} = {GDP/cap_{i,t} \over {1 \over 5} {\sum_{k=1}^5GDP/cap_{i,t-k}}} \end{aligned}$$ {#eq-eq6}- **Political Stability:** This variable captures overall regime stability by including a violence index that encompasses all types of internal and interstate wars and violence. The data for this index is sourced from the variable “actotal” in the Major Episodes of Political Violence dataset[@marshall2005current], with 0 representing the most stable conditions (no violence at all) and 18 representing the most unstable.- **Previous coups:** Included in the selection equation as either: a) The number of previous coups in a country (Model 1), or b) The time since the last coup attempt (Model 2 for robustness check).### Results and Discussion```{r}#| label: tbl-coupmodel#| tbl-cap: "Sample Selection Model of Regime Type and Coup Success, 1950-2019"#| results: asis#| echo: !expr format == "html"model_1 <- coup_model|>filter(regime !="other") |>select(coup, regime, GDP_trend, GDP_pc, violence, coup_dummy = pre_coups) |>mutate(regime =fct_relevel(regime, "dominant-party")) |>selection(coup >0~ regime + GDP_trend + GDP_pc + violence + coup_dummy, coup >1~ regime + GDP_trend + GDP_pc + violence,#method = "2step",data = _) model_1_selection <- model_1model_1_selection$param$index$betaO <- model_1$param$index$betaSmodel_1_selection$param$index$betaS <- model_1$param$index$betaOmodel_2 <- coup_model |>filter(regime !="other") |>select(Coup = coup, regime, GDP_trend, GDP_pc, violence, coup_dummy, time_since_last_coup) |>mutate(regime =fct_relevel(regime, "dominant-party")) |>selection(Coup >0~ regime + GDP_trend + GDP_pc + violence + coup_dummy*time_since_last_coup, Coup >1~ regime + GDP_trend + GDP_pc + violence,data = _) model_2_selection <- model_2model_2_selection$param$index$betaO <- model_2$param$index$betaSmodel_2_selection$param$index$betaS <- model_2$param$index$betaOstargazer(model_1_selection, model_1, model_2_selection, model_2,#type = "text",float = F,header = F,dep.var.caption ="",dep.var.labels =c("Model 1", "Model 2"),multicolumn = T,dep.var.labels.include = T, column.labels =c("Coup Attempts", "Coup Outcome","Coup Attempts", "Coup Outcome"),intercept.bottom = F,covariate.labels =c("Constant", "Regime: Democracy", "\\hspace{1.6cm}Military", "\\hspace{1.6cm}Monarchy", "\\hspace{1.6cm}Personalist", "Economic trend", "GDP per capita", "Political violence", "Previous coups (P)", "Yrs since coup (Y)", "Interaction term: P * Y"),summary.logical = F,column.sep.width ="7pt",font.size ="small" )```The double probit model with sample selection, estimated using the***sampleSelection*** package [@sampleSelection-2] in R, provides valuableinsights into the factors influencing coup attempts and their outcomes acrossdifferent regime types from 1950 to 2019 (@tbl-coupmodel). I present two modelsthat differ slightly in their treatment of previous coups: Model 1 incorporatesthe number of previous coups, while Model 2 utilizes the time elapsed since thelast coup.#### Selection Model: Coup AttemptsIn the selection model (Column 1), military and personalist regimes exhibitsignificant positive coefficients at the 1% level, indicating a higher likelihoodof experiencing coup attempts compared to dominant-party regimes. This alignswith our theoretical expectations regarding internal power struggles withinmilitary juntas and succession vulnerabilities in personalist regimes.```{r}#| label: tbl-mfx1#| tbl-cap: Average marginal effects of coup attempts (Selection of Model 1)#| echo: falseavg_comparisons(model_1) |>select(1:3) |>select(1:2, ame = estimate) |>cbind(avg_comparisons(model_1, comparison ="ratio") |>select(3)) |>mutate('ratio(%)'=round((estimate -1) *100, 2),ame =round(ame, 3)) |>select(-estimate) |>mutate(term =case_when(term =="coup_dummy"~"Previous coups", term =="GDP_trend"~"Economic trend", term =="GDP_pc"~"GDP per capita", term =="violence"~"Political violence",str_detect(contrast, "democracy") ~"Regime: Democracy",str_detect(contrast, "military") ~"Military",str_detect(contrast, "monarchy") ~"Monarchy",str_detect(contrast, "personal") ~"Personal",TRUE~ term )) |>mutate(term =factor(term, levels =c("Regime: Democracy","Military","Monarchy","Personal","Economic trend","GDP per capita","Political violence","Previous coups")),contrast =str_replace( contrast, "mean\\((.+?)\\) - mean\\((.+?)\\)", "mean(\\1 - \\2)" )) |>arrange(term) |>clean_names(case ="title") |>rename(AME = Ame) |>gt() |>cols_align(align ="left", columns = Term) |>tab_style(style =cell_text(indent =px(63)),locations =cells_body(columns = Term,rows =2:4 ) ) |>tab_options(table.width =pct(95),# table.font.size = px(14),column_labels.font.weight ="bold" ) |>fmt_number(decimals =3# specify the number of decimal places ) |>tab_footnote(footnote ="AME: Average Marginal Effect",locations =cells_column_labels(columns = AME) )```@tbl-mfx1 clarifies the regime effects using Average Marginal Effects (AME) andratio percentages. The military regime's marginal effect of 0.07 indicates thatthe probability of coup attempts in military regimes is 7 percentage points (pp)higher than in dominant-party regimes, ceteris paribus. This translates tomilitary regimes being about 277.7% more likely to encounter coups thandominant-party regimes. Similarly, personalist regimes show a 2.4 pp higherprobability, about 94% more likely compared to dominant-party regimes.Control variables show effects in expected directions but with weaker magnitudes.Stronger economic performance, indicated by higher economic growth trends and GDPper capita levels, correlates with a lower risk of coup attempts. Politicalviolence shows a positive effect, indicating that higher levels of instabilityincrease the likelihood of coups. The positive coefficient for the number ofprevious coups suggests a "copycat" effect from earlier incidents.#### Outcome Model: Coup SuccessThe outcome model (Columns 2 and 4 in @tbl-coupmodel) reveals determinants ofcoup success. Military regimes demonstrate a higher probability of coup successcompared to dominant-party regimes, aligning with expectations that militaryregimes face higher coup risks due to their increased chances of success.Personalist and monarchical regimes show slight positive effects on coup success,but these effects are not statistically significant.Control variables exhibit different patterns in the outcome model compared to theselection model. Both GDP per capita and political violence maintain a weakinfluence, similar to their effects in the selection model. However, the economictrend shows a less significant negative effect on coup success.#### Model Comparison (Model 1 vs Model 2)Model 2 employs years since the last coup instead of the number of previouscoups, with an interaction term between previous coups (as a binary variable) andyears since the last coup. Generally, Model 2 shows results in the same directionas Model 1, albeit with relatively lower coefficients (@tbl-mfx2).The differences between Model 1 and Model 2 suggest that while the recency ofcoups matters, the overall history of coups in a country may have a strongerinfluence on future coup attempts.\clearpage```{r}#| label: tbl-mfx2#| tbl-cap: Average marginal effects of coup attempts (Selection of Model 2)#| echo: falseavg_comparisons(model_2) |>select(1:3) |>select(1:2, ame = estimate) |>cbind(avg_comparisons(model_2, comparison ="ratio") |>select(3)) |>mutate('ratio(%)'=round((estimate -1) *100, 2),ame =round(ame, 3)) |>select(-estimate) |>mutate(term =case_when(term =="coup_dummy"~"Previous coups (P)", term =="time_since_last_coup"~"Yrs since coup (P)" , term =="coup_dummy*time_since_last_coup"~"Interaction term: P * Y", term =="GDP_trend"~"Economic trend", term =="GDP_pc"~"GDP per capita", term =="violence"~"Political violence",str_detect(contrast, "democracy") ~"Regime: Democracy",str_detect(contrast, "military") ~"Military",str_detect(contrast, "monarchy") ~"Monarchy",str_detect(contrast, "personal") ~"Personal",TRUE~ term )) |>mutate(term =factor(term, levels =c("Regime: Democracy","Military","Monarchy","Personal","Economic trend","GDP per capita","Political violence","Previous coups (P)","Yrs since coup (Y)","Interaction term: P * Y")),contrast =str_replace( contrast, "mean\\((.+?)\\) - mean\\((.+?)\\)", "mean(\\1 - \\2)" )) |>arrange(term) |>clean_names(case ="title") |>rename(AME = Ame) |>gt() |>cols_align(align ="left", columns = Term) |>tab_style(style =cell_text(indent =px(63)),locations =cells_body(columns = Term,rows =2:4 ) ) |>tab_options(table.width =pct(95),# table.font.size = px(14),column_labels.font.weight ="bold" ) |>fmt_number(decimals =3# specify the number of decimal places ) |>tab_footnote(footnote ="AME: Average Marginal Effect",locations =cells_column_labels(columns = AME) )```#### Discussion of key findingsThe $\rho$ values of 0.898 in Model 1 and 0.386 in Model 2, significant at 1% and10% levels respectively, indicate strong correlation between unobserved factorsinfluencing coup attempts and coup success. This supports the appropriateness ofthe sample selection model and underscores the importance of considering bothstages in the analysis.The significant coefficients with theoretically consistent directions suggest themodel effectively captures key aspects of coup dynamics. The observed disparitybetween coup attempt rates and success rates across regimes points towardsselection bias, further validating the use of the sample selection model.#### Theoretical ImplicationsThese results strongly support my theoretical framework, highlighting the crucialrole of regime structure in determining coup vulnerability. The findingsunderscore that coups are strategic actions undertaken when odds appearfavourable, rather than random events.### ConclusionThis study addresses the lack of consensus in empirical research on couppredictors by introducing a novel approach that prioritizes determinants based ontheir impact on coup success. By analysing coup success rates, I posit that theexpected outcomes of coups are critical determinants of their occurrence.Employing a double probit model with sample selection, the research investigatesand confirms a strong and robust relationship between regime types and coupattempts.The main findings reveal that regime type plays a pivotal role in the likelihoodof coup attempts. Military and personalist regimes, characterized by weakerinstitutional frameworks and higher vulnerability, demonstrate significantlyhigher susceptibility to coups compared to dominant-party regimes. Thisunderscores the importance of supporting initiatives that establish regimes withconstitutional institutions rather than those dependent on military power orpersonal authority, as the latter prove more volatile and coup-prone.The research also indicates that stronger economic performance, while not asinfluential as regime type, is associated with a lower risk of coups. Thissuggests that policies promoting economic development can be effective inreducing coup risk. Given that regime type is often determined during regimeformation and is difficult to alter, promoting economic growth might be the mostviable coup-proofing strategy available to incumbents.However, these findings present a paradox for autocratic leaders, particularlydictators and military juntas. While institutionalization of regimes couldenhance stability and reduce coup risk, few such leaders are willing to implementthese reforms. This reluctance stems from the potential constraints on theirpower or shortened terms that such changes might entail. Thus, while institutionsmay benefit the regime's longevity, they do not necessarily serve the immediateinterests of individual leaders.This study opens avenues for further research, including:- Investigating the long-term effects of regime types on political stability, including democratization or personalization.- Examining how international factors might influence regime choices and coup dynamics when internal factors are too weak to prevent coups.In conclusion, this chapter provides robust evidence for the critical role ofregime type in determining coup risk, while also acknowledging the complexrealities of political power. By illuminating the strategic calculus behind coupattempts, it offers valuable insights for scholars seeking to understandpolitical instability and policymakers working to promote democratic transitionsand political stability in diverse global contexts.## Autocoups: Conceptual Clarification and Analysis of Power Extensions by Incumbent Leaders {#sec-chapter3}### Abstract {.unnumbered}This study aims to clarify the concept of autocoups, specifically focusing onpower extensions by incumbent leaders. By distinguishing autocoups from thebroader and more ambiguous concepts of self-coups or executive takeovers, whichencompass both executive power aggrandizement and power extension, this researchredefines the concept of autocoups. Based on this refined definition, I introducea novel dataset of autocoup events from 1945 to 2024. Using this newly compileddataset, the research includes three types of case studies that providequalitative insights into the dynamics of autocoups. Additionally, an empiricalanalysis of the determinants of autocoup attempts and success is offered todemonstrate how the autocoup dataset can be employed for more quantitativeresearch. This study contributes to the existing literature by providing aclearer conceptual framework and a novel dataset of autocoups. It enhances ourunderstanding of the mechanisms and motivations behind power extensions byincumbent leaders and examines the implications for democratic backsliding,democratic breakdown, and autocratic deterioration. The insights gained from thisstudy could draw more attention to the effects of autocoups on power transitions,political stability, and democratic resilience.***keywords**: Autocoups, Coups, Power transitions*\newpage### IntroductionThe study of irregular power transitions has predominantly focused on coups dueto their frequency and significant impact. This emphasis has led to substantialscholarly attention and the creation of comprehensive datasets. However, as notedby @gassebner2016, the multitude of potential factors proposed to explain coupshas often resulted in increased confusion rather than a clearer understanding ofcoup dynamics.In contrast, another form of irregular power transition—the incumbent leader'srefusal to relinquish power—has received relatively less attention despite itssignificance. Recent decades have witnessed a decline in traditional coups and arise in this "incumbent retention or overstay" type of irregular powertransition, particularly since the end of the Cold War (@ginsburg2010evasion;@baturo2014; @versteeg2020law).This chapter will redefine and clarify this type of irregular power transition,where leaders overstay their mandated term limits, as "autocoup." Althoughanalyses related to autocoups are not rare, the existing literature exhibitsseveral notable shortcomings:- **Terminological Ambiguity:** Terms such as self-coups, autocoups, autogolpes, incumbent takeovers, executive aggrandizement, overstay, and continuismo are used without a clear, universally accepted definition. This lack of standardization leads to confusion and inconsistent application[@marsteintredet2019; @baturo2022].- **Limited Data:** Due to the conceptual ambiguity surrounding autocoups, data collection remains in its early stages compared to the rich datasets available for traditional coups.- **Methodological Gaps:** Research on autocoups has largely relied on case studies [@cameron1998; @antonio2021; @pion-berlin2022], with few studies employing quantitative analysis.Moreover, analyses of autocoups are often not integrated with those oftraditional coups. As a distinct category of coup, autocoups lack a correspondingdefinition in relation to classic coups. Consequently, traditional coups andautocoups are frequently analysed separately, despite their related nature, andthere is a paucity of comparative analyses.Studying autocoups is important for several reasons. Firstly, autocoups canundermine the rule of law, weaken institutions, and contribute to democraticbacksliding or authoritarian personalization. Secondly, like traditional coups,successful autocoups increase the likelihood of future irregular powertransitions. For example, since 1945, approximately 62% of leaders who extendedtheir terms through autocoups in non-democratic countries were either ousted orassassinated while in office [@baturo2019]. Thirdly, failed autocoups often leadto instability, inciting protests, violence, and even civil wars.This chapter aims to address these gaps by focusing on autocoups, aiming toclarify terminology, refine concepts and definitions, enhance data collection,and explore determinants through empirical analysis, contributing in three keyareas:- **Conceptual Clarification:** The term "autocoup" will be redefined and clarified, with a focus on power extension.- **Data Collection:** A new dataset of autocoups since 1945 will be introduced based on this refined definition.- **Empirical Analysis:** Utilizing this dataset, a quantitative analysis of the factors influencing leaders' decisions to attempt autocoups will be conducted.The structure of this chapter is as follows: Section 2 will review definitionsrelated to power expansions and extensions, leading to a precise definition ofautocoups. Section 3 will present the new autocoup dataset. Sections 4 and 5 willexplore the determinants of autocoup attempts through case studies anddemonstrate the application of the dataset in empirical analysis. The conclusionwill summarize key findings and suggest directions for future research.Definitions of power expansion and power extension in political science can oftenbe ambiguous or overlapping. To ensure clarity in this study, we propose distinctdefinitions for these concepts:- Power Expansion: This refers to the process by which an incumbent leader acquires additional authority or control over state apparatuses beyond their original mandate. This may involve centralizing power, reducing checks and balances, or assuming roles typically held by other branches of government.- Power Extension: This concept specifically relates to the temporal aspect of an incumbent's rule. It describes situations where a leader prolongs their tenure beyond the originally mandated term in office, often through constitutional amendments, cancellation of elections, or other means of circumventing term limits.By distinguishing between these two concepts, we aim to provide a more nuancedanalysis of how leaders consolidate and maintain their positions. While powerexpansion and extension can often occur simultaneously, differentiating betweenthem allows for a more precise examination of the strategies employed by leadersto entrench their rule.This distinction is crucial for our study as it enables us to: 1. Analyze thedifferent mechanisms leaders use to solidify their power 2. Examine the potentialconsequences of each type of power consolidation 3. Explore the relationshipbetween power expansion and extension in various political contextsUnderstanding these nuances is essential for comprehending the complex dynamicsof political power and leadership longevity in both democratic and authoritariansystems."### Autocoups: A literature review and clarification of definitionsThe concept of autocoups, or the illegitimate seizure of power by incumbentleaders, remains a complex and contested area of study. Unlike coups, which aregenerally understood as illegal attempts by elites to overthrow the government,autocoups lack a clear and consistent definition. This ambiguity hinders ourability to accurately identify, analyze, and compare instances of thisphenomenon.#### TerminologyThe most common term in autocoup literature is "self-coup," or "autogolpe" inSpanish [@przeworski2000; @cameron1998a; @bermeo2016; @helmke2017;@marsteintredet2019]. This term gained academic prominence after PeruvianPresident Alberto Fujimori dissolved Congress, temporarily suspended theconstitution, and ruled by decree in 1992 [@mauceri1995; @cameron1998]. However,as @marsteintredet2019 point out, the term "self-coup" can be misleading, as itimplies a coup against oneself, which is inaccurate since it targets other stateinstitutions or apparatus.Another approach to describe coups staged by incumbents is to use terms withadjectives or modifiers, such as "presidential coup," "executive coup,""constitutional coup," "electoral coup," "judicial coup," "slow-motion coup,""soft coup," and "parliamentary coup" [@marsteintredet2019]. While these termscan be helpful in specific contexts, their proliferation often adds to theoverall confusion rather than providing clarification. Most of these terms focuson the specific methods used by coup perpetrators but fail to clearly identifythe perpetrator, necessitating further explanation. In fact, many of thesemethods could be employed either by or against executive leaders.A third alternative involves terms like "incumbent takeover," "executivetakeover," or "overstay." Incumbent takeover refers to "an event perpetuated by aruling executive that significantly reduces the formal and/or informalconstraints on his/her power" [@baturo2022, p.374], based on earlier research[@svolik2014]. Meanwhile, overstay is defined as "staying longer than the maximumterm as it stood when the candidate originally came into office"[@ginsburg2011evasion, p.1844]. These terms identify the perpetrator (theincumbent) and/or the nature of the event (overstaying/extending power). However,they do not highlight the illegality or illegitimacy of these actions. Therefore,they cannot serve as a direct counterpart to "coup," which clearly denotes theillegality of leadership ousters, while "takeover" or "overstay" diminish theseverity.As these terms often lack precision, focusing on specific methods rather than thecore act of power usurpation, this study proposes "autocoup" as the most suitableterm for this phenomenon. Unlike other terms, 'autocoup' clearly identifies theperpetrator and the illegitimate nature of the power grab, distinguishing it fromtraditional coups.#### DefinitionWhile terminology is important, another issue arises with the previous definitionof autocoups: What is the emphasis of an autocoup—power expansion, powerextension, or both?Definitions of power expansion and power extension in political science can oftenbe ambiguous or overlapping. To ensure clarity in the study of autocoups, Ipropose distinct definitions for these concepts:- **Power Expansion:** This refers to the process by which an incumbent leader acquires additional authority or control over state apparatuses beyond their original mandate. This may involve centralizing power, reducing checks and balances, or encroaching on the authority of other branches like the legislature or judiciary.- **Power Extension:** This describes situations where a leader prolongs their tenure beyond the originally mandated term in office, often through constitutional amendments, cancellation of elections, or other means of circumventing term limits.Existing definitions of autocoups or related concepts often either are ambiguousbetween power expansion and extension, or focus more on power expansion, whichhas several drawbacks.Firstly, defining autocoups primarily in terms of power expansion does not alignwell with the definition of a coup. The focus of a classical coup is clearly onthe ouster of the current leader, not merely a limitation or restriction on theirpower. Using the same logic, a more appropriate definition of an autocoup shouldprioritize the tenure extension of executive leadership. Power restriction onincumbents would not be coded as a coup as long as they remain in office.Similarly, an executive leader acquiring more power from other branches could becoded as power aggrandizement, but not an autocoup, as long as they step downwhen their term expires.Secondly, emphasizing power expansion in autocoups often neglects the ultimatepurpose of incumbents. It is irrational for an incumbent to expand executivepower and then pass the powerful role to future leaders. Although the term"self-coup" gained prominence from the 1992 Fujimori case in Peru, whichinitially involved seizing power from other institutions, it is important to notethat Fujimori ultimately extended his term limits through constitutionalamendments. The 1993 Constitution allowed Fujimori to run for a second term,which he won with popularity in April 1995. Shortly after Fujimori began hissecond term, his supporters in Congress passed a law of "authenticinterpretation" which effectively allowed him to run for another term in 2000,which he won amid suspicions and rumors. However, he did not survive the thirdterm. In 2000, facing charges of corruption and human rights abuses, Fujimorifled Peru and took refuge in Japan [@ezrow2019].Thirdly, measuring the extent of power expansion to qualify as an autocoup can bechallenging. As @cameron1998a defined, a self-coup is "a temporary suspension ofthe constitution and dissolution of congress by the executive, who rules bydecree until new legislative elections and a referendum can be held to ratify apolitical system with broader executive power" (p. 220). "Broader executivepower" is difficult to define, and it would be problematic and disputable nomatter how it is defined.Therefore, this study argues that a more accurate definition of autocoups shouldprioritize power extension as the core characteristic. It is easy to identify theevent and the outcome in the first place. In most cases, autocoups in terms ofpower extension involve power expansions as their prerequisite and foreshadowing.Based on these criteria, I define **an autocoup as the illegitimate extension ofan incumbent leader's term in office beyond the originally mandated limitsthrough unconstitutional means**. This definition emphasizes the corecharacteristic of power extension while acknowledging the potential for powerexpansion as a related phenomenon. Three key points need to be highlighted forthis definition.Firstly, this definition refers to the actual leaders of the country, regardlessof their official titles. Typically, this would be the president; however, insome cases, such as in Germany, the primary leader is the premier, as thepresident is a nominal head of state.Secondly, while the primary characteristic of an autocoup is extending the termin office, this definition does not exclude instances of power expansion. Bothaspects can coexist, but the extension of the term is the central element.Thirdly, autocoups, by their nature, subvert legal norms and established powertransfer mechanisms. No matter how legitimate they claim to be, theirillegitimacy is not beyond a reasonable doubt as long as the incumbents are thedirect beneficiaries. This critical aspect will be explored further in Section 3.### Introduction to the Autocoup Dataset#### Defining the scopeCategorizing political events as autocoups inevitably involves challengingborderline cases. To maintain consistency and avoid ambiguity, this study adoptsa broad coding approach: All instances of incumbents extending their originalmandated term in office are coded as autocoups, regardless of the apparentlegality of the extension.This approach is justified because truly legitimate amendments to powertransition institutions should apply only to subsequent leaders, not theincumbent. Even when extension procedures appear legal, the legitimacy isquestionable when the incumbent is the direct beneficiary.#### Classifying autocoups {#sec-classify}Autocoups manifest in various forms. I categorize them based on several keyfactors:- **Methods Employed**: Specific strategies used by incumbents (e.g., constitutional amendments, election cancellation).- **Degree of Legality**: Extent of deviation from established legal norms.- **Duration of Extension**: Length of time the incumbent remains in office beyond designated term limits.- **Outcomes**: Whether the autocoup attempt succeeds or fails.This study primarily focuses on the methods employed, while coding for otheraspects when information is available.##### Evasion of term limits {.unnumbered}Evasion of term limits is a common tactic employed in autocoups. Incumbents oftenresort to seemingly legal manoeuvres to extend their hold on power. Thesemanoeuvres primarily involve manipulating constitutional provisions throughvarious means. The incumbents may pressure legislative bodies (congress) orjudicial institutions (Supreme Court) to reinterpret existing term limits, amendthe constitution to extend terms, or even replace the constitution altogether.This might also involve popular vote through referendums, or a combination ofthese approaches. The extension can range from a single term to indefinite rule.These manoeuvres primarily involve manipulating constitutional provisions throughvarious means.- **Changing Term Length:** Incumbents might lengthen the official term duration (e.g., from 4 to 6 years) to stay in office longer, even if the number of allowed terms remains unchanged. Examples, in the dataset, include Presidents Dacko (CAR, 1962), Kayibanda (Rwanda, 1973), and Pinochet (Chile, 1988).- **Enabling re-election:** This approach involves incumbents modifying legal or constitutional frameworks to permit themselves to run for leadership again, despite initial restrictions. These restrictions might include prohibitions on re-election, bans on immediate re-election, or term limits that the incumbents have already reached. An illustrative example is President Menem of Argentina in 1993, who leveraged this tactic to extend his tenure.- **Removing Term Limits Altogether:** This approach, as seen with President Xi Jinping of China in 2018, technically allows the leader to rule for life, although they may still need to participate in elections (a formality in such cases).- **Declaring** **Leader for Life:** This differs from removing term limits as the leader still faces elections (although potentially rigged or uncontested). An example is Indonesia's President Sukarno, who attempted to declare himself president for life in 1963 (ultimately unsuccessful).These methods are often used in combination. Initially, the duration of a term isextended, followed by amendments to allow re-election, then the removal of termlimits, and finally, the declaration of the leader for life. For example, HaitianPresident François Duvalier amended the constitution in 1961 to permit immediatere-election and then declared himself president for life in 1964.##### Election Manipulation or Rigging {.unnumbered}Election manipulation or rigging is the second most commonly used tactic toextend an incumbent's tenure.- **Delaying or Removing Elections:** Delaying or removing scheduled elections without legitimate justification is a frequent method used by incumbents to maintain power. For instance, Chadian President François Tombalbaye delayed general elections until 1969 after assuming power in 1960. Similarly, Angolan President José Eduardo dos Santos suspended elections throughout his rule from 1979 to 2017.- **Refusing Unfavourable Election Results:** Incumbents may refuse to accept unfavourable election results and attempt to overturn them through illegitimate means. For example, President Donald Trump of the United States refused to accept the results of the 2020 election and tried to overturn them.- **Rigging Elections:** Winning elections with an extraordinarily high percentage of votes is highly questionable. This study will code elections where the incumbent wins more than 90% of the vote as autocoups. For instance, President Teodoro Obiang of Equatorial Guinea has consistently won elections with over 95% of the vote in multi-party elections since 1996, indicating election rigging.- **Excluding Opposition in Elections:** Manipulating the electoral process by excluding opposition parties or candidates from participation, effectively creating a one-candidate race, clearly signifies an autocoup.##### **Use of Figurehead** {.unnumbered}To circumvent term limits, some incumbents might choose a close associate to actas a figurehead, taking the office publicly while the incumbent retains realpower behind the scenes. This can be achieved through seemingly subordinatepositions.One example is Russia in 2008. Facing term limits, President Putin selectedDmitry Medvedev to run for president. After the election, Medvedev appointedPutin as Prime Minister. However, most analysts believe Putin wielded the truepower throughout this period.##### Reassigning supreme authority to a new role {.unnumbered}This tactic involves an incumbent leader manipulating the constitution or legalframework to create a new position of power, or elevate an existing one, beforestepping down from their current role. They then strategically take on this newposition, effectively retaining significant control despite appearing torelinquish power. For example, in 2017, Recep Tayyip Erdoğan, the Prime Ministerof Turkey, spearheaded a constitutional referendum that transitioned the countryfrom a parliamentary system to a presidential one. This new system concentratedsignificant executive power in the presidency. Following the referendum'sapproval, Erdoğan successfully ran for the newly established presidency,effectively retaining control under a different title.##### One-Time Arrangement for Current Leaders {.unnumbered}This strategy involves special arrangements that extend the term or tenure ofcurrent leaders without altering the underlying institutions. For example,Lebanon extended President Émile Lahoud's term by three years in 2004 through aone-time arrangement.#### Data CodingThe autocoup dataset is built upon existing studies and datasets, ensuring acomprehensive and reliable foundation. @tbl-source outlines the main sources usedfor coding the autocoup dataset.The Archigos dataset [@goemans2009] and the Political Leaders' AffiliationDatabase (PLAD) [@bomprezzi2024wedded] provide comprehensive data on all leadersfrom 1875 to 2023, although our coding only includes autocoups since 1945. Thesedatasets are invaluable for identifying actual rulers, distinguishing them fromnominal heads of state.The Incumbent Takeover dataset [@baturo2022] integrates data from 11 relateddatasets, offering a broad spectrum of cases where leaders significantly reducedconstraints on their power. This dataset includes both power expansions andextensions, necessitating cross-referencing with Archigos to verifyqualifications for autocoups.```{r}#| label: tbl-source#| tbl-cap: Main Data Sources for Coding the Autocoup Dataset#| echo: false#| results: asisread.csv(here("data/autocoup_source.csv")) |>clean_names(case ="title") |>gt() |>tab_options(column_labels.font.weight ="bold",table.width =pct(100),table.align ="center") |>cols_width(2~px(200) )``````{r}#| label: data-autocoups#| echo: falseAutocoup <-fread("https://raw.githubusercontent.com/reddylee/Datasets/main/autocoup_clean.csv")# nrow(Autocoup)# nrow(Autocoup |> filter(!is.na(source)))```In total, `r nrow(Autocoup)` observations were coded, with`r nrow(Autocoup |> filter(!is.na(source)))` overlapping with the candidate datafrom Incumbent Takeover. The remaining`r nrow(Autocoup) - nrow(Autocoup |> filter(!is.na(source)))` events were newlycoded by the author through verification with other sources such as Archigos,PLAD and news reports.The main deviation from the Incumbent Takeover dataset arises from excludingpower expansions that do not involve attempts to extend tenure.The dataset encompasses a total of 14 variables along with the *notes* field.- **Country Identification:** Country code (*ccode*) and country name (*country*) from Correlates of War project [@stinnett2002].- **Leader Information:** Name of the de facto leader (*leader_name*,coded following Archigos and PLAD datasets).- **Timeline Variables:** Date the leader assumed power (*entry_date*), date the leader left office(*exit_date*), date of the significant event marking the autocoup (*autocoup_date*), and Start date of the leader's additional term acquired through the autocoup (*extending_date*).- **Power Transition Methods:** Categorical variable for how the leader entered power (*entry_method*), categorical variable for how the leader exited power (*exit_method*), dummy variable indicating regular (1) or irregular (0) entry (*entry_regular*), and dummy variable indicating regular (1) or irregular (0) exit (*exit_regular*).- **Autocoup Details:** Key variable capturing methods used to extend power (*autocoup_method*) and outcome of the autocoup attempt (*autocoup_outcome*, "fail and lose power", "fail but complete original tenure", or "successful"). For successful coups, the additional term length can be calculated from the difference between *exit_date* and *extending_date*.- **Data Source:** Identifies the dataset source used for coding (*source*).- **Additional Notes:** Provides context for exceptional cases (*notes*).There are a few coding challenges and decisions worth mention. For cases whereextensions happen incrementally, the *autocoup_date* reflects a significant eventmarking the extension, such as a legislative vote or successful referendum. Incases where a leader undertook multiple autocoup attempts, details are recordedin the notes field. Care was taken to differentiate between cases of powerexpansion and actual attempts to extend tenure, which required cross-referencingmultiple sources. Determining the success or failure of an autocoup attempt oftenrequired in-depth research, especially for less documented cases.#### Data descriptions```{r}#| label: data_descriptions#| echo: !expr format == "html"country <- Autocoup |>distinct(ccode) |>summarise(count =n())autocoup_desc <- Autocoup |>summarise(autocoup_attempted =n(),autocoup_succeeded =sum(autocoup_outcome =="Yes"),.by = autocoup_method) |>arrange(-autocoup_attempted)```The primary coding has identified `r nrow(Autocoup)` autocoup cases from 1945 to2024, involving `r country$count` countries. This comprehensive dataset providesa rich source of information for analysing trends and patterns in autocoupattempts across different political contexts.\clearpage```{r}#| label: tbl-autocoup_method#| tbl-cap: Autocoup methods and success rates (1945-2021)#| echo: !expr format == "html"autocoup_desc |>adorn_totals() |>mutate(Success_rate =percent(autocoup_succeeded/autocoup_attempted)) |>clean_names(case ="title") |>gt() |>tab_options(column_labels.font.weight ="bold",table.width =pct(100),table.align ="center") |>cols_width(1~px(180),2:3~px(140),4~px(100) )#autocoup_desc$autocoup_attempted[1]```@tbl-autocoup_method presents a breakdown of the autocoup methods employed byleaders:The most common autocoup method is "enabling re-election", accounting for`r autocoup_desc$autocoup_attempted[1]` events. This is followed by "removingterm limits altogether" (`r autocoup_desc$autocoup_attempted[2]` cases), and then"delaying elections" and "declaring the leader for life" (each with`r autocoup_desc$autocoup_attempted[3]` cases).The overall success rate of autocoups is 79%, which is significantly higher thanthe approximately 50% success rate of classical coups. This high success rate canbe attributed to several factors:- Incumbent Advantage: Leaders already in power have access to resources and institutional mechanisms that can be leveraged to their advantage.- Gradual Implementation: Unlike sudden coups, autocoups can be implemented gradually, allowing leaders to build support and legitimacy over time.- Legal Facade: Many autocoup methods operate within a veneer of legality, making them harder to oppose openly.- Control of State Apparatus: Incumbents often have significant control over state institutions, which can be used to facilitate their autocoup attempts.However, success rates vary significantly across different methods.- **100% Success Rate**: Removing term limits, delaying elections, declaring the leader for life, and cancelling elections all have perfect success rates. This suggests that once these processes are set in motion, they are difficult to reverse.- **Lower Success Rates:** Refusing to accept election results has the lowest success rate, with only 1 out of 4 attempts succeeding. Although the sample size is limited (only 4 cases in total), this trend might suggest several factors at play. These include greater democratic resilience in systems where general elections are regularly held, heightened international scrutiny and pressure in response to blatant manipulation of election results, and stronger domestic opposition to such overt power grabs.### Determinants of Autocoup Attempts: Case Studies#### High Frequency and Success Rate of Autocoups in Post-Communist CountriesAnalysis of our dataset reveals a notably high frequency and success rate ofautocoups in post-communist countries. These nations, formerly communist regimesprior to the collapse of the Soviet Union, have largely evolved into 'hybridregimes' [@nurumov2019], with only a few retaining their communist status. Thedata documents 12 cases of autocoups aimed at prolonging incumbency in thesecountries, with only two attempts failing. Examination of these cases highlightsseveral distinctive characteristics:- **Inherited Authoritarian Systems**: Despite most of these 12 countries transitioning from communist to non-communist governments (with the exception of China), they retained many authoritarian systems from their communist past.- **Continuity of Former Elites**: The transitions did not result in the removal or overthrow of previous ruling groups. Instead, former communist elites often maintained their positions of power.- **Subverted Democratic Processes**: While general elections and term limits were introduced in most of these countries, the legacy of former communist regimes frequently led to the circumvention of term limits and manipulation of elections [@nurumov2019].##### Case 1: Lifelong Ruler–Alexander Lukashenko in Belarus {.unnumbered}Alexander Lukashenko, a former member of the Supreme Soviet of the ByelorussianSoviet Socialist Republic, became the head of the interim anti-corruptioncommittee of the Supreme Council of Belarus following the dissolution of theSoviet Union. Elected as Belarus's first president in 1994, he has maintainedthis position ever since. Initially, the 1994 constitution limited presidents totwo successive terms. However, Lukashenko removed this restriction in 2004.International monitors have not regarded Belarusian elections as free and fairsince his initial victory. Despite significant protests, Lukashenko hasconsistently claimed to win with a high vote share, often exceeding 80% in eachelection. This pattern is evident across all five Central Asian countries of theformer Soviet Union, where post-dissolution leaders were typically high officialsor heads of the former Soviet republics who continued their leadership in thepresidency.##### Case 2: Transferring Power to a Handpicked Successor–Nursultan Nazarbayev in Kazakhstan {.unnumbered}Nursultan Nazarbayev served as the first president of Kazakhstan from 1991 until2019. Prior to the dissolution of the Soviet Union, he held de facto leadershipas the First Secretary of the Communist Party of Kazakhstan. Followingindependence, he was elected as the first president and retained office until2019 through various means, including resetting term limits due to theimplementation of new constitutions. Notably, Nazarbayev did not officiallyeliminate term limits but instead created an exemption for the "First President"[@nurumov2019]. Unlike Lukashenko, who remains the incumbent of Belarus,Nazarbayev transferred the presidency to a designated successor, Kassym-JomartTokayev, in 2019. However, he retained significant influence as the Chairman ofthe Security Council of Kazakhstan until 2022.```{r}#| label: latin#| include: falselibrary(countrycode)latin <- codelist |>filter(region =="Latin America & Caribbean") |>select(region, cown) |>drop_na() |>pull(cown)Autocoup |>filter(ccode %in% latin) |>summarise(count =n(),.by = country) |>adorn_totals()Autocoup |>filter(ccode %in% latin) |>summarise(count =n(),.by = autocoup_method) |>adorn_totals()Autocoup |>filter(ccode %in% latin) |>summarise(count =n(),.by =c(autocoup_method,autocoup_outcome)) |>adorn_totals()```#### Autocoups for immediate re-election: Cases of Latin AmericaLatin America has a long-standing tradition of maintaining term limitconventions. Simón Bolívar, the founding father of Bolivia, was initially astrong advocate for term limits, stating in 1819, "Nothing is as dangerous asallowing the same citizen to remain in power for a long time... That's the originof usurpation and tyranny" [@ginsburg2019, p.38]. Although Bolívar eventuallymodified his stance, arguing in his 1826 Constitution Assembly speech that "apresident for life with the right to choose the successor is the most sublimeinspiration for the republican order," term limits became a convention in LatinAmerica. Approximately 81% of Latin American constitutions between independenceand 1985 imposed some form of term limits on the presidency[@marsteintredet2019a].An analysis of cases in Latin American countries reveals two notable patterns.##### Often Successful at Breaking Non-re-election or Non-immediate Re-election Restrictions {.unnumbered}Unlike other presidential systems where two terms are more common,non-re-election or non-immediate re-election used to be prevalent in LatinAmerica. According to @marsteintredet2019a, non-consecutive re-election wasmandated in about 64.9% of all constitutions between independence and 1985, while5.9% banned re-election entirely.However, adherence to these conventions has varied across the region. SinceMexico introduced non-re-election institutions in 1911 at the start of theMexican Revolution, they have remained inviolate [@klesner2019]. Similarly,Panama and Uruguay have never altered their re-election rules, and Costa Rica hasonly experienced a brief period (1897-1913) permitting immediate presidentialre-election since prohibiting it in 1859 [@marsteintredet2019a]. In many othercountries, however, constitutions have been frequently amended or violated.The pursuit of re-election or consecutive re-election, therefore, has been asignificant trigger for autocoups aimed at power extension in this region. Ourresearch documents 32 autocoup cases, with over 50% (17 cases) attempting toenable re-election or immediate re-election, and about 59% (10 cases out of 17)being successful.Unlike those who attempt to overstay in office indefinitely, many Latin Americanleaders exit after their second term expires. Examples include President FernandoHenrique Cardoso of Brazil (1995-2003), President Danilo Medina of the DominicanRepublic (2012-2020), and President Juan Orlando Hernández of Honduras(2014-2022) [@ginsburg2019; @marsteintredet2019a; @landau2019; @baturo2019;@neto2019].##### Failing to Further Extend Tenure {.unnumbered}This trend does not imply that none of these leaders attempted furtherextensions, but rather that most accepted their unsuccessful outcomes withoutabusing their power to manipulate the process. While autocoups aimed at securingone additional term are often successful, attempts to overstay beyond this arefrequently unsuccessful.Two contrasting cases illustrate the varied outcomes of term limit challenges:- **Unsuccessful Extension – Carlos Menem (Argentina)**: President Menem successfully extended his tenure by one term through a 1994 constitutional amendment allowing one executive re-election. He was subsequently re-elected in 1995. However, his attempt to reset his term count, arguing that his first term (1988-1995) should not count as it was under previous constitutions, was unanimously rejected by the Supreme Court in March 1999 [@llanos2019]. A similar scenario unfolded with President Álvaro Uribe of Colombia (2002-2010)[@baturo2019].- **Successful Extension – Daniel Ortega (Nicaragua)**: In contrast, Daniel Ortega, the incumbent president of Nicaragua, successfully extended his presidency. In 2009, the Supreme Court of Justice of Nicaragua permitted his re-election in 2011. Subsequently, in 2014, the National Assembly of Nicaragua approved constitutional amendments abolishing presidential term limits, allowing Ortega to run for an unlimited number of five-year terms. As a result, he has held the presidency since 2007 [@close2019].#### As common as classical coups: Cases of African countries```{r}#| label: africa#| include: falseafrica <- codelist |>filter(continent =="Africa") |>select(continent, cown) |>drop_na() |>pull(cown)Autocoup |>filter(ccode %in% africa) |>summarise(count =n(),.by = country) |>adorn_totals()Autocoup |>filter(ccode %in% africa) |>distinct(country)Autocoup |>filter(ccode %in% africa) |>summarise(count =n(),.by = autocoup_method) |>adorn_totals()Autocoup |>#filter(ccode %in% africa) |>summarise(count =n(),.by = autocoup_outcome) |>adorn_totals()```Classical coups have been prevalent in Africa, accounting for approximately 45%of all global coups (219 out of 491 cases) since 1950, involving 45 out of 54African countries (GIC dataset). While autocoups are less frequent compared totraditional coups, they maintain a significant presence in Africa. Among 110documented autocoup cases globally, 46% (51 cases) occurred in Africa, involving36 countries. Notably, the success rate of autocoups in Africa is over 84% (43out of 51 attempts), which surpasses both the success rate of classical coups inthe region (roughly 50%) and the global average success rate of autocoups (79%).Identifying a clear pattern of autocoups in Africa is challenging, mirroring thecomplexity observed with classical coups. Various factors have been proposed toexplain this phenomenon:- **Natural Resources**: Countries rich in natural resources, particularly oil or diamonds, may see leaders more likely to attempt and succeed in extending their terms [@posner; @cheeseman2015; @cheeseman2019a].- **Quality of Democracy**: The quality of democracy is a critical factor influencing respect for term limits [@reyntjens2016].- **International Influence**: International aid or donor influence can play a significant role in discouraging attempts at power extension [@brown2001; @tangri2010].- **Organized Opposition and Party Unity**: The extent of organized opposition and the president's ability to enforce unity within the ruling party are crucial factors [@cheeseman2019].Utilizing the Africa Executive Term Limits (AETL) dataset, @cassani2020highlights human rights abuses and the desire for impunity as main drivers forincumbents to cling to power. The more authoritarian a leader, the more likelythey are to attempt to break term limits and overstay in office. A leader'sability to secure the loyalty of the armed forces through public investmentincreases the chances of success in overstaying.Despite both coups and autocoups being prevalent, there has been a noticeableshift since the end of the Cold War in 1991: Traditional coups have decreased infrequency while autocoups have become more prevalent.This trend can be partially attributed to the introduction of multi-partyelections in Africa in the 1990s, which also brought in term limits forexecutives [@cassani2020; @cheeseman2019]. Before 1991, personal or military rulewas more common, and term limits were less frequent. Post-1991, with more termlimits introduced, challenges to these limits have increased. However, it iscrucial to note that this increase in challenges does not necessarily imply thatviolations are more common than adherence to term limits, because total powertransitions have increased compared to the past.### Empirical Analysis: An Example of Utilizing the Autocoup DatasetThe availability of the autocoup dataset has made it feasible to conductquantitative analyses that extend beyond traditional case studies. This sectionprovides a straightforward example of how to utilize this dataset effectively. Toanalyze the determinants of autocoup attempts, I employ a probit regressionmodel. This approach differs from the double probit model with sample selectionused in [Chapter @sec-chapter2] for coup attempts and success analyses. Instead,I use two separate probit models. Due to the high probability of success inautocoups, they do not exhibit the typical sample selection characteristics thatnecessitated the use of a sample selection model in our earlier analysis oftraditional coups.#### Dependent Variables- **Autocoup Attempt**: Binary variable indicating whether an autocoup attempt occurred (1) or not (0) during the tenure of an incumbent leader.- **Autocoup Success**: Binary variable indicating whether an autocoup attempt was successful (1) or failed (0), conditional on an autocoup attempt occurring.#### Independent VariablesThe selection of independent variables are consistent with the coup analysis in[Chapter @sec-chapter2], plus the population size and the leader's age.- **Population Size:** To account for its potential impact on leaders' tenures, we consider the log of the population size. This transformation helps in managing the wide range of population sizes across different countries. The data is sourced from the V-Dem dataset and is evaluated to understand its influence on power transitions. Larger populations may present more governance challenges and potential sources of opposition, thereby affecting the stability and longevity of a leader's tenure.- **Leader's Age:** The age of the leader is included as an additional variable in the analysis, offering insights into potential correlations with leadership strength. Older leaders may have different experiences, networks, and health considerations that could influence their ability to maintain power. This data is sourced from Archigos and PLAD datasets.Unlike the analysis of coup determinants, which could theoretically occur in anygiven year, I assume that an autocoup happens only once during an incumbentleader's tenure, as a successful autocoup negates the need for another attempt.However, this assumption does not always reflect reality, as leaders mightattempt further extensions or try again after a failed attempt. For simplicity, Ioverlook these possibilities in our analysis.Therefore, in our probit model, the unit of analysis for autocoups is the entiretenure of a leader, rather than a country-year. I establish a base year for thevariables: for leaders who staged an autocoup, we use the year of their firstattempt as the base year; for leaders who did not attempt to overstay, I use themiddle year of their tenure as the base year.#### Results and discussions```{r}#| label: tbl-autocoupmodel#| tbl-cap: "Determinants of autocoup attempts and success (1945-2018)"#| results: asis#| echo: falseautocoup_model <-fread("https://raw.githubusercontent.com/reddylee/Datasets/main/autocoup_model.csv")model_attempt <- autocoup_model |>filter(!regime %in%c("monarchy", "other")) |>mutate(regime =fct_relevel(regime, "presidential")) |>glm(autocoup_attempt ~ regime + GDP_pc + GDP_trend + violence + age + pop_log, family =binomial(link ="probit"),data = _)model_success <- autocoup_model |>filter(!regime %in%c("monarchy", "other")) |>mutate(regime =fct_relevel(regime, "presidential"),autocoup_outcome =factor(autocoup_outcome)) |>filter(autocoup_attempt ==1) |>glm(autocoup_outcome ~ regime + GDP_pc + GDP_trend + violence + age + pop_log, family =binomial(link ="probit"),data = _)stargazer(model_attempt, model_success, header = F,float = F,column.labels =c("Autocoup Attempts", "Autocoup Outcome"),intercept.bottom = F,covariate.labels =c("Constant","Regime: Dominant-party","\\hspace{1.6cm}Military","\\hspace{1.6cm}Personalist","GDP per capita","Economic trend","Political stability","Age","Population(log)"),dep.var.caption ="",dep.var.labels.include = F,summary.logical = T,column.sep.width ="50pt" )#view(autocoup_model)```@tbl-autocoupmodel summarizes the findings from the probit regression modelsbased on our analysis of the determinants of autocoup attempts and their success.Model 1, which examines autocoup attempts, reveals only one significant predictorbesides the constant term. Among the regime types, personalist regimessignificantly increase the likelihood of autocoup attempts, all else being equal.This suggests that leaders in personalist regimes are more prone to attempt toextend their power through autocoups compared to leaders in democratic regimes(reference regime). Leaders in dominant-party and military regimes, however, showno significant difference in the likelihood of attempting an autocoup compared todemocratic leaders.The model for autocoup success (Model 2) shows similar dynamics. Personalistregimes again have a strong positive and significant effect on the success ofautocoups compared to democratic leaders. Dominant-party regimes also show apositive and marginally significant effect. However, a detailed examinationreveals that about half of the successful autocoups in dominant-party regimes (9out of 20) exhibit a personalist style, such as "party-personal-military"regimes.This outcome is logical since personalist leaders are typically much morepowerful than other types of leaders, making them more inclined and capable ofoverstaying in power.Other factors play an insignificant role in determining the attempts and outcomesof autocoups. This aligns with our conclusions on the determinants of classiccoups. Both coups and autocoups are significantly affected by power dynamics. Aspower transitions involve the struggle between seizing and maintaining power, thebalance of power status quo inevitably matters in both coups and autocoups. Thisalso explains the high success rate of autocoups. Compared to power challengers,incumbents are in an obviously advantageous position. Incumbent leaders can usestate power to their benefit, which is difficult to counteract. Even the abuse ofpower is often unchecked under a powerful leader's rule.The empirical analysis of autocoups yields significant implications forreal-world politics. In particular, the high overall success rate of autocoupshighlights the vulnerability of democratic institutions to gradual erosion byincumbent leaders. The threshold for ousting or impeaching an incumbent leaderthrough constitutional means is exceptionally high, with success often requiringmore than a simple majority and substantial support across various sectors.Resorting to illegal means, such as a coup, presents even greater challenges dueto high costs, severe consequences, and a low likelihood of success.Conversely, political dynamics, whether in democracies or autocracies, tend tofavour incumbents even when they act unconstitutionally. Incumbents can leveragestate resources to achieve their political ambitions, benefiting from a highprobability of success and minimal consequences in case of failure. Thisasymmetry in power and risk creates a concerning scenario: for incumbents who donot respect constitutional institutions, the opportunity to launch an autocoupappears sufficiently low-risk to warrant an attempt.### ConclusionThis chapter presents a comprehensive analysis of autocoups, focusingspecifically on political events where incumbent leaders illegitimately extendtheir tenure in power. By refining the concept and distinguishing it from broaderdefinitions such as 'self-coups', 'autogolpes', and 'executive takeovers', Iintroduce a novel dataset of autocoups spanning from 1945 to 2024. Through thismore precise definition and dataset, this research expands the analysis ofirregular power dynamics from irregular transition to irregular retention,offering a broader and more nuanced review of the phenomenon.The findings reveal that personalist regimes are significantly more likely toexperience autocoup attempts and succeed in these attempts compared todemocracies. Dominant-party systems, often exhibiting personalistcharacteristics, also show an association with successful autocoups. While regimetype significantly influences autocoups, other factors appear less impactful,mirroring classic coups where the balance of power is a more essentialdeterminant. The high success rate of autocoups can be attributed to the inherentadvantages incumbents possess, such as control over or abuse of state power andthe difficulty of removing or impeaching them through legal or illegal means.However, several limitations warrant consideration for future research. Firstly,the definition of an autocoup requires further commentary and discussion to gainwider acceptance in the academic community. Despite efforts to maintainobjectivity, some coding decisions may involve subjective judgements,particularly in borderline cases. Secondly, due to the nature of autocoups, whichare less frequent than classic coups (491 coups versus 110 autocoups during thesame period), the quantitative analysis cannot be conducted as a country-yearvariable as in coup studies. This raises the issue of choosing an appropriatebase year. For instance, when analysing how GDP level or growth rate, politicalstability, or the age of the leader affects autocoup attempts, we must decidewhich year's value should be used. In this study, I chose the middle year of aleader's tenure or the year they staged the autocoup, but determining the mostappropriate year requires further discussion and potentially sensitivityanalyses.Despite these limitations, this research significantly enhances our understandingof the mechanisms and motivations behind autocoups, contributing to theliterature on political stability and democratic resilience. The findingshighlight the vulnerability of political systems, particularly democracies, toerosion from within by incumbent leaders.Future studies could build on this work by employing the dataset to explore morenuanced power dynamics or examine the long-term impacts of these events onpolitical systems. Particularly fruitful areas for investigation include therelationship between autocoups and democratic backsliding, democratic breakdown,and the personalization of power. Additionally, comparative analyses betweenautocoups and traditional coups could yield insights into the evolving nature ofpower consolidation strategies in different political contexts.In conclusion, this study not only provides a valuable resource for futureresearch but also contributes to our understanding of the complex interplaybetween leadership, institutional structures, and political stability. Asautocratic tendencies continue to challenge democratic norms globally, theinsights gained from this analysis of autocoups become increasingly relevant forboth scholars and policy-makers concerned with preserving and strengtheningdemocratic institutions.## Power Acquisition and Leadership Survival: A Comparative Analysis of Coup-Entry and Autocoup Leaders### Abstract {.unnumbered}This chapter examines the relationship between methods of power acquisition andthe longevity of leaders who assume power through irregular means, specificallycoup-entry and autocoup leaders. The central hypothesis posits that the mode ofaccession significantly influences leader tenure. Employing Cox proportionalhazards and time-dependent Cox models, this study provides robust evidence ofdivergent survival times between these two leader types. Findings reveal thatcoup-entry leaders face a substantially higher risk of removal from officecompared to their autocoup counterparts. These results have significantimplications for political stability and democratic processes, suggesting thatthe perceived low costs and high rewards associated with autocoups mayincentivize incumbents to extend their tenure through this mechanism, potentiallycontributing to democratic backsliding. This research makes a notablecontribution to the academic literature by leveraging a newly developed dataseton autocoups. By doing so, it offers valuable insights into the dynamics ofirregular leadership transitions and enhances our understanding of the complexinterplay between power acquisition methods and political longevity.***keywords**: Coups, Autocoups, Leadership Survival, Cox Model*\newpage### IntroductionThe longevity of political leaders has long captivated scholars and observers.Why do some rulers endure for decades while others are ousted within months oreven days? This question has spurred extensive research in political science,primarily focusing on the survival of political leaders. However, within thisbroader framework, a specific subset of leaders—those who ascend to power throughcoups or extend their tenure through autocoups—has received comparatively lessattention. Examining the tenures of these leaders is crucial, as it illuminatesthe dynamics of irregular leadership transitions and their implications forpolitical stability and democratic processes.Leaders who attain power through conventional means often follow predictablepatterns of ascension, tenure, and departure. In contrast, those who rise throughirregular channels, such as coups or autocoups, present more complex andintriguing cases for study. Archigos dataset underscores the prevalence of suchirregular transitions: between 1945 and 2015, over half of leaders who assumedpower irregularly (158 out of 308) also exited irregularly. This ratesignificantly surpasses that of leaders who accessed office through regularchannels, of whom only 14.5% (213 out of 1,472) experienced irregular departures.Coup-entry and autocoup leaders constitute a substantial portion of theseirregular cases. Archigos dataset notes that of 374 leaders who exitedirregularly, 246 (65.8%) were ousted through coups. @frantz2016 furtherdemonstrate that coup-related exits account for approximately one-third of allexits in autocracies, surpassing any other transition type. Additionally, theautocoup dataset, introduced in [Chapter 3](#sec-chapter3), documents 110autocoup attempts between 1945 and 2024, of which 87 were successful.```{r}#| label: data#| include: falsesurvival_cox_ph_model <-fread("https://raw.githubusercontent.com/reddylee/Datasets/main/survival_cox_ph_model.csv") |>mutate(survival_year = survival_days /365)survival_cox_td_model <-fread("https://raw.githubusercontent.com/reddylee/Datasets/main/survival_cox_td_model.csv")avg <- survival_cox_ph_model |>summarise(avg=round(mean(survival_year),1),.by = group)```Measuring the survival tenure of coup-entry and autocoup leaders presentschallenges due to the inherent irregularity and uncertainty of their positions.Nevertheless, a comparative analysis reveals that leaders who extend powerthrough autocoups tend to have longer average post-autocoup tenures(approximately `r avg$avg[1]` years) compared to coup-entry leaders(approximately `r avg$avg[2]` years), suggesting a potential tenure gap of overfive years.```{r}#| label: fig-logrank#| fig-width: 7#| fig-height: 5#| fig-align: center#| fig-cap: Survival curves of overstaying and coup-entry leaders#| echo: !expr format == "html"# Fit survival curvesfit <-survfit(Surv(survival_year, status) ~ group, data = survival_cox_ph_model)# Create log-rank plot (Kaplan-Meier plot)ggsurvplot(fit, data = survival_cox_ph_model, risk.table =TRUE, risk.table.height =0.3, pval =TRUE,surv.median.line ="hv",break.x.by =5, legend.title ="Leader Type",legend.labs =c("Autocoup leaders", "Coup-entry leaders"),xlab ="Survival Years",#title = "Survival Curves of overstay and coup-entry leaders",font.title =12)```A preliminary log-rank test in survival analysis, as illustrated in @fig-logrank,demonstrates a statistically significant difference between the tenures ofautocoup and coup-entry leaders. The survival curve for autocoup leadersconsistently exceeds that of coup-entry leaders, indicating longer survival timesand a reduced risk of ouster for autocoup leaders.This study posits that the method of accession significantly influencesleadership longevity. Coup-entry leaders likely confront greater challenges totheir rule, resulting in shorter average tenures compared to autocoup leaders.The analysis, employing Cox proportional hazards and time-dependent Cox models,supports this hypothesis, demonstrating that autocoup leaders generallyexperience longer tenures than coup-entry leaders.This research offers two primary contributions to the field. First, it highlightsan understudied factor in leadership survival analysis: the impact of the methodof accession to power. My findings suggest that leader survival is influenced notonly by ruling strategies but also by the initial method of acquiring power.Second, by employing survival models, this study provides empirical evidence ofthe significant difference in tenure duration between autocoup and coup-entryleaders. This insight may explain the increasing prevalence of tenure extensionsthrough autocoups since 2000, as more incumbents observe and potentially emulatesuccessful precedents.The remainder of this chapter is structured as follows: Section 2 provides acomprehensive literature review on political survival, establishing the contextfor this research. Section 3 explores the factors influencing the survival ofcoup and autocoup leaders. Section 4 outlines the methodology and data used,including the application of survival models to analyse the determinants ofleadership longevity. Section 5 presents the analysis findings and a detaileddiscussion of the results. Finally, Section 6 concludes by synthesizing keytakeaways and exploring their broader implications for political stability anddemocratic processes.### Literature reviewThe longevity of political leaders has been a central focus in political scienceresearch for decades, driven by the wide-ranging variations observed acrossregimes, countries, and historical periods. This field encompasses twointerconnected aspects: regime survival and individual leader survival. Regimesurvival focuses on the endurance of political systems, such as monarchies,political parties, or specific ideological structures, while leader survivalconcerns the duration of individual leaders' time in office.These concepts often exhibit contrasting patterns. For instance, parliamentarydemocracies like Japan or the United Kingdom may experience prolonged periods ofparty dominance coupled with frequent leadership changes, whereas communistregimes typically demonstrate enduring party rule with more frequent leadershiptransitions. Presidential systems, such as the United States or many militaryregimes, tend to exhibit more frequent changes in both ruling party or junta andleader. This study specifically investigates the dynamics of individual leadersurvival, focusing on factors influencing the duration of leaders' time inoffice.The existing literature on leader survival is extensive and multifaceted. Somestudies explore specific mechanisms influencing leadership longevity withinparticular regimes, such as democracies [@svolik2014] or autocracies[@davenport2021], while others aim to develop more generalizable theoreticalframeworks explaining leader survival across different political systems[@buenodemesquita2003]. Although a universal theory remains an aspirational goal,the complexities of leadership survival across diverse regime types presentsignificant challenges.Power transition mechanisms vary substantially across different regimes,particularly between democracies and autocracies. Autocratic systems oftenfeature closed leadership selection processes, restricted to a narrow pool ofindividuals such as royal families, military elites, or ruling party members.While some autocracies may hold elections, significant barriers to entry forlegitimate challengers typically persist. Potential rivals may face threats likeassassination, imprisonment, or exile. Moreover, the opacity of selectionprocesses makes it difficult to assess genuine levels of public support comparedto democracies. Consequently, conceptualizing selectorates or winning coalitions,as proposed by @buenodemesquita2003, becomes problematic in many autocraticcontexts.Given these complexities, focusing research on specific regimes or leader typesmay be more fruitful. While regular leadership changes offer valuable insights,they provide limited opportunities to explore the dynamics of leader survival. Incontrast, the study of irregular leaders, such as those who ascend to powerthrough coups or overstay in office through autocoups, offers a more compellingavenue for research due to the inherent complexities and uncertaintiessurrounding their leadership trajectories.Two primary perspectives have emerged to explain the dynamics of leader survival.The first emphasizes objective factors and resources, such as personal competence[@yu2016], societal stability [@arriola2009], economic development [@palmer1999;@williams2011], natural resource endowments [@smith2004; @quirozflores2012;@wright2013], and external support [@licht2009; @wright2008; @thyne2017]. Thesecond focuses on subjective factors and strategies, including politicalpolicies, responses to opposition, and tactics for consolidating power[@gandhi2007; @morrison2009; @escribà-folch2013; @davenport2021].Coups have received considerable scholarly attention, with research examiningcoup prevention strategies, the impact of coups on leadership, and the subsequentactions of coup leaders. Existing research delves into strategies for thwartingcoups [@powell2017; @sudduth2017; @debruin2020] and how leaders extend theirtenures after surviving coup attempts [@easton2018]. @sudduth2017 examinespost-coup actions of dictators, focusing on purge strategies, while @sudduth2018investigates how leaders' entry methods affect their removal in dictatorships.However, a significant gap exists in the literature regarding the comparison ofleadership survival between coup-entry and autocoup leaders. This study aims toaddress this gap by investigating and comparing the duration of leadershipsurvival for these two leader types, contributing to a more nuanced understandingof political survival in irregular leadership transitions.### Survival dynamics of autocoup and coup-entry leadersThe study of leadership survival in political systems presents inherentchallenges due to the opacity and diverse mechanisms of power transitions.However, these challenges underscore the significance of this research, as itilluminates understudied dynamics in political leadership. While the survival ofpolitical leaders exhibits complexity and variation, it is not entirely devoid ofpatterns. Leaders of similar types often display significant comparability.#### Key Definitions and ScopeBefore delving into the comparison, it is essential to clarify several keyterminologies:- **Coup and Autocoup**: These terms adhere to the definitions established in previous chapters.- **Tenure Length Threshold**: To ensure meaningful analysis, this study focuses on leaders with substantial periods in power, applying a six-month threshold to both autocoup and coup-entry leaders.- **Autocoup Leader**: An incumbent leader who successfully uses illegitimate or unconstitutional means to extend their tenure in power.- **Coup-Entry Leader**: The individual who assumes power after a successful coup, regardless of their role in the coup itself.This study focuses on comparing the post-autocoup tenure of autocoup leaders withthe post-coup tenure of coup-entry leaders, motivated by the relevance andsimilarity of these leader types in terms of illegitimacy, uncertainty, andinstability.#### Challenges in Power ConsolidationBoth autocoup and coup-entry leaders face distinct challenges in consolidatingtheir power, primarily due to differences in the intensity of issues related toillegitimacy, uncertainty, and instability. This disparity creates an unevenplaying field in terms of power dynamics, with coup-entry leaders at asignificant disadvantage. @tbl-leaders compares the main features betweenautocoup and coup-entry leaders.\clearpage```{r}#| label: tbl-leaders#| tbl-cap: Main features of autocoup and coup-entry leaders #| echo: false#| results: asisread.csv(here("data/leader_comparison.csv")) |>clean_names(case ="title") |>gt() |>tab_options(column_labels.font.weight ="bold",table.width =pct(100),table.align ="center") |>cols_width(1~px(100),2:3~px(250) )# |> data_color(rows = c(2, 4), colors = "#edede9")```##### Illegitimacy {.unnumbered}While both types of leaders suffer from a legitimacy deficit, the nature of thisdeficit differs:- **Coup-Entry Leaders**: Their illegitimacy is explicit due to the open seizure of power.- **Autocoup Leaders**: They employ a deceptive strategy, manipulating legal processes to create a façade of democratic legitimacy.##### Uncertainty {.unnumbered}The irregular paths to power create uncertainty regarding their reigns andeventual departures. However, the levels of uncertainty differ.Coup-entry leaders face three major uncertainties: Unclear who will assumeleadership after the coup, uncertain tenure length, and ambiguity regardingfuture successors.Autocoup leaders present a clearer picture: No ambiguity about who will ruleafter an autocoup, many seek to extend their rule indefinitely or incrementally.##### Instability {.unnumbered}The awareness of shaky legitimacy and persistent uncertainty breeds insecurityand a sense of crisis.Coup-entry leaders must reshape power dynamics and often purge potentialadversaries. They need to create a new equilibrium, often disrupting establishedstructures. They also face potential backlash even from close allies and hencemust compromise with internal or external power challengers.Autocoup leaders encounter fewer abrupt changes in their regimes. They face lesspressure to dismantle existing ruling paradigms and have more time to implementchanges gradually.#### Empirical Evidence and HypothesisEmpirical evidence supports the disadvantage faced by coup-entry leaders. Datareveals a correlation between the frequency of coup attempts in a country and thelikelihood of future coups. Over a third of coups occur in the top ten countrieswith the most attempts since 1950 (@tbl-coups). The average survival periodfollowing an autocoup is approximately five years longer than that of coup-entryleaders (@fig-logrank).The distinct challenges faced by autocoup and coup-entry leaders in consolidatingpower create a self-perpetuating cycle that influences their tenure length.Coup-entry leaders, facing greater challenges, struggle to attract and retainstrong support, making them more vulnerable to internal and external challenges.Conversely, autocoup leaders, often benefiting from a veneer of legitimacy and astronger initial position, are better able to consolidate power and attractsupporters, leading to potentially longer tenures.Based on these observations, I propose the following hypothesis:***H1: Political leaders who successfully extend their tenure through autocoupsare more likely to survive longer compared to coup-entry leaders.***### Research DesignThis study employs survival analysis to test the hypothesis that autocoup leadershave longer survival times in office compared to coup-entry leaders. The researchdesign utilizes Cox models to analyze the survival tenures of these two types ofleaders, accounting for multiple factors that may influence their time in power.#### Methodology: Survival analysisTwo Cox models will be employed to analyze the survival tenures of coup-entry andautocoup leaders:- **Cox Proportional Hazards (PH) Model**: This model uses only the variables present at the entry year, without considering changes over time.- **Time-Dependent Cox Model**: This model accounts for variations in time-dependent control variables such as economic performance and political stability.The Cox model is preferred over the Kaplan-Meier model as it allows for theestimation of multiple factors' impacts. While it does not directly estimate theduration of tenure in office, it evaluates the hazard rate associated with beingousted from power. This approach captures different facets of the samephenomenon: as a leader's cumulative hazard of being ousted increases, theirprobability of survival in office decreases.#### Data and VariablesThe dependent variables include survival time and end point status:- **Survival Time:** The duration of a leader's tenure, measured in days. For coup-entry leaders, the survival time begins on the day they assume power through a coup. For autocoup leaders, the survival time starts on the expiration date of their original legitimate term. For example, Xi Jinping assumed power in 2013 and removed term limits in 2018. His original legitimate tenure was set to end in 2023, so his survival time begins in 2023, marking the start of his post-autocoup tenure. The survival time concludes on the day the leader exits office, applicable to both coup-entry and autocoup leaders.- **End point status:** This variable indicates the manner in which the leader's tenure concluded, categorized as follows: **0 = Censored:** This status is assigned to leaders who leave office through regular means other than being ousted. This includes leaders transferring power to their designated successors, leaving office as their terms expire, losing in general elections, voluntarily leaving office due to health issues, or dying of natural causes. **1 = Ousted:** This status is assigned to leaders who are forced to leave office. This includes leaders resigning under pressure, being ousted by coups or other forces, or being assassinated.The key independent variable is the leader type, which categorizes leaders intotwo distinct groups:- **Group A = Autocoup Leader**: Leaders who extend their tenure through autocoups.- **Group B = Coup-Entry Leader**: Leaders who assume power through coups.This variable is the primary independent variable of interest, serving as thebasis for comparing the survival time between these two types of leaders.The data for both dependent and independent variables are sourced from theautocoup dataset introduced in this study, Archigos, and PLAD.Control variables include economic performance, political stability, populationsize, and the leader's age, which are consistent with the autocoup analysis in[Chapter @sec-chapter3].### Results and discussion#### Model resultsUsing the `survival` package in R [@survival], I present the regression resultsfor both the Cox Proportional Hazards (Cox PH) model and the time-dependent Coxmodel in @tbl-cox.```{r}#| label: model#| echo: !expr format == "html"Mydata1 <- survival_cox_ph_model |>mutate(group =factor(group, labels =c("Autocoup leaders", "Coup-entry leaders")))model_cox1 <- Mydata1 |>coxph(Surv(survival_days, status) ~ group + GDP_trend + GDP_pc + pop_log + polity5 + violence + age,data = _ ) tbl1 <- model_cox1|>tbl_regression(exponentiate = T,conf.int = F,add_estimate_to_reference_rows = T,label =list(group ~"Leader Type", GDP_trend ~"GDP Growth Trend", GDP_pc ~"GDP per capita", pop_log ~"Population: log", polity5 ~"Polity 5", violence ~"Political stability", age ~"Age")) |>add_n(location ="level") |>add_nevent(location ="level") |>add_significance_stars(hide_p = T,hide_se = F, hide_ci = T,thresholds =c(0.01, 0.05, 0.1), ) |>bold_p(t =0.05) |>bold_labels() Mydata2 <- survival_cox_td_model |>mutate(group =factor(group, labels =c("Autocoup leaders", "Coup-entry leaders")),survt = (tstop - tstart) *365)model_cox2 <- Mydata2 |>coxph(Surv(T1, T2, status) ~ group + GDP_trend + GDP_pc + pop_log + polity5 + violence + age +cluster(id),data = _ ) tbl2 <- model_cox2|>tbl_regression(exponentiate = T,conf.int = F,add_estimate_to_reference_rows = T,label =list(group ~"Leader Type", GDP_trend ~"GDP Growth Trend", GDP_pc ~"GDP per capita", pop_log ~"Population: log", polity5 ~"Polity 5", violence ~"Political stability", age ~"Age")) |>add_n(location ="level") |>add_nevent(location ="level") |>add_significance_stars(hide_p = T,hide_se = F, hide_ci = T,thresholds =c(0.01, 0.05, 0.1), ) |>bold_p(t =0.05) |>bold_labels() ``````{r}#| label: tbl-cox#| tbl-cap: Cox models for survival time of different types of leaders#| echo: !expr format == "html"tbl_merge(list(tbl1, tbl2),tab_spanner =c("**Cox PH Model**", "**Time-dependent Cox Model**")) |>as_gt()```Both the Cox PH model and the time-dependent Cox model analyses revealed astatistically significant association between leadership type and the hazard ofremoval from power. Since time-dependent Cox model use the control variableswhich change over time, I interpret the main findings based on time-dependentmodel.Coup-entry leaders were found to have a hazard ratio of 2.23 in thetime-dependent model compared to autocoup leaders (reference group), assuming allother variables in the model are held constant. This suggests that coup-entryleaders face a significantly greater risk of removal from power compared toautocoup leaders. At any given time during their tenure, coup-entry leaders are2.23 times more likely to be ousted from power compared to autocoup leaders, allelse being equal in the model.The control variables perform differently in the two models. Economic level (GDPper capita) exhibits statistically significant effects in both models. In thetime-dependent model, the hazard ratio of 0.95 indicates that for each unitincrease in GDP per capita (measured in units of \$10,000), the hazard (or risk)of being ousted at any given time is reduced by 5%, assuming all other variablesin the model are held constant.GDP growth trend demonstrates a more substantial effect in reducing the risk ofcoups. Specifically, a 1 percentage point higher economic growth trend isassociated with an 80% reduction in the risk of being ousted, although thiseffect is only statistically significant at the 10% level. This suggests apossible trend where positive economic performance might mitigate the risk ofremoval from power, but the evidence is not robust enough to confirm thisconclusively.Political stability, as measured by the violence index, shows that a 1-pointincrease in the index correlates with an 11% higher risk of being ousted.However, this effect is also only statistically significant at the 10% level,indicating a weaker but potentially important relationship between increasedviolence and the risk of removal from office.#### Discussion```{r}#| label: fig-coxSurv#| fig-cap: Survival curves for Cox Model#| fig-subcap: #| - Cox PH Model#| - Time-dependent Cox Model#| layout-ncol: 2#| echo: falsetestdata1 <- Mydata1 |>summarise(.by = group,GDP_trend =mean(GDP_trend),GDP_pc =mean(GDP_pc),pop_log =mean(pop_log),polity5 =mean(polity5),violence =mean(violence),age =mean(age) )model_cox1 |>survfit(newdata = testdata1) |>ggsurvplot(data = testdata1,xlab ="Days",surv.median.line ="hv",legend.title ="Leader Type",legend.labs =c("Autocoup leaders", "Coup-entry leaders"))testdata2 <- Mydata2 |>summarise(.by = group,GDP_trend =mean(GDP_trend),GDP_pc =mean(GDP_pc),pop_log =mean(pop_log),polity5 =mean(polity5),violence =mean(violence),age =mean(age) )model_cox2 |>survfit(newdata = testdata2) |>ggsurvplot(data = testdata2,xlab ="Days",surv.median.line ="hv",legend.title ="Leader Type",legend.labs =c("Autocoup leaders", "Coup-entry leaders"))```The survival curves depicted in @fig-coxSurv illustrate the survival rates forleaders of both types. Both the Cox PH model and the time-dependent Cox modelproduce similar plots. Notably, the survival curve for coup-entry leadersexhibits a significantly lower trajectory compared to that of autocoup leaders.The steeper drop at the early stage for coup-entry leaders indicates they aremore likely to be ousted shortly after assuming power. Additionally, the survivalcurve for coup-entry leaders crosses the median survival line much earlier (about3,000 days) than that of autocoup leaders (about 8,500 days). This disparitysuggests that autocoup leaders tend to remain in power for longer durations thantheir coup-entry counterparts.```{r}#| label: fig-coxHR#| fig-cap: Hazard ratios and 95% CIs for Leader Ousting#| fig-subcap: #| - Cox PH Model#| - Time-dependent Cox Model#| layout-ncol: 2#| echo: falsecoxplottidy <-tidy(model_cox1, exponentiate = T, conf.int = T) |>mutate(term =c("Coup-entry leaders","GDP Growth Trend","GDP per capita","Population: log","Polity 5","Political stability","Age"))coxplottidy |>ggplot(aes(y = term,x = estimate,xmin = conf.low,xmax = conf.high )) +geom_pointrange() +geom_vline(xintercept =1,linetype ="dashed",color ="skyblue") +labs(x ="Hazard ratios", y =NULL, title ="") +theme_bw() coxplottidy <-tidy(model_cox2, exponentiate = T, conf.int = T) |>mutate(term =c("Coup-entry leaders","GDP Growth Trend","GDP per capita","Population: log","Polity 5","Political stability","Age"))coxplottidy |>ggplot(aes(y = term,x = estimate,xmin = conf.low,xmax = conf.high )) +geom_pointrange() +geom_vline(xintercept =1,linetype ="dashed",color ="skyblue") +labs(x ="Hazard ratios", y ="", title ="") +theme_bw() ```@fig-coxHR displays the hazard ratios and corresponding 95% confidence intervalsfor the variables incorporated in the Cox model. Both the Cox ProportionalHazards (PH) model and the time-dependent model produce similar plots,reinforcing the robustness of the findings. Key points to note include:- The closer the hazard ratio (represented by the dots) is to 1, the less impact the variable has on the risk of being ousted. A hazard ratio of 1 indicates no effect.- The whiskers extending from the dots represent the 95% confidence intervals. If these whiskers cross the vertical blue line at 1, it indicates that the variable is not statistically significant at the 5% level.- The hazard ratio for coup-entry leaders is significantly greater than 1 and statistically significant at the 5% level. This indicates that coup-entry leaders face a substantially higher risk of being ousted compared to autocoup leaders.- Most other variables have hazard ratios close to 1, suggesting that a one-unit increase in these variables does not significantly affect the risk of being ousted.- Although the hazard ratio for GDP growth trend is considerably less than 1 in the time-dependent model, indicating a potential protective effect, it is not statistically significant at the 5% level. However, it is statistically significant at the 10% level, suggesting that better economic performance may help to consolidate the rule of the incumbents to some extent, albeit the evidence is not as strong.#### Assessing the Proportional Hazards AssumptionAssessing the proportional hazards assumption is crucial for the validity of theCox model results. To evaluate this, we used the chi-square test based onSchoenfeld residuals to determine whether the covariate effects remain constant(proportional) over time. Although the Cox PH model violates the proportionalhazards assumption, our primary analysis relies on the time-dependent Cox model,which does not show strong evidence of violating the proportional hazardsassumption for any covariate. The global p-value of 0.416 is much greater thanthe 5% significance level, indicating that the proportional hazards assumption isreasonably met for the time-dependent Cox model.```{r}#| label: robust_test#| include: falsetest <-cox.zph(model_cox2)```### ConclusionThis chapter examined the survival durations of political leaders who come topower through irregular means, specifically coups and autocoups. I hypothesizedthat the mode of accession significantly influences leader tenure. Employingsurvival analysis techniques, including the Cox proportional hazards model and atime-dependent Cox model, I found strong evidence that autocoup leaders generallyenjoy longer tenures than coup-entry leaders.The findings revealed a significant difference in average tenure, withpost-autocoup leaders averaging approximately 11 years in power compared to 5.6years for coup-entry leaders. The time-dependent Cox model further indicated thatcoup-entry leaders are 2.23 times more likely to be ousted from power at anygiven time compared to autocoup leaders, all else being equal.These results highlight the importance of understanding the phenomenon ofautocoups, where leaders extend their rule by manipulating legal frameworks. Dueto the relative ease and potential benefits of autocoups, this method of powerretention might incentivize more leaders to employ it. Consequently, democraticbacksliding could become more prevalent as autocoups weaken democraticinstitutions and constitutional norms, particularly in nascent democracies orthose transitioning from autocracy.This study contributes to the field of leadership survival by demonstrating thatthe mode of accession significantly impacts leader tenure, a factor previouslyunder-explored in the literature. By utilizing both Cox models, the researchoffers robust analytical techniques for studying political leadership survivaland provides strong evidence of divergent tenure lengths between these two typesof irregular-entry leaders.However, limitations exist. The study relies heavily on the autocoup datasetcollected and coded by the author. The concept and data itself are relativelynovel within academia. Future research should refine and establish widerrecognition for the term "autocoup," leading to more accurate and comprehensivedata collection efforts. Expanding the dataset to include more cases andintegrating it with data on other irregular leadership transitions could yield amore holistic understanding of political survival in such contexts.Overall, this chapter underscores the need for more nuanced approaches tostudying political tenure and the mechanisms of irregular power retention,contributing valuable insights into the dynamics of political stability and therisks associated with different forms of non-democratic leadership succession.## Conclusion### Main FindingsThis study delves into the dynamics and implications of irregular powertransitions, focusing on coups and autocoups. The findings illuminate the complexinterplay between incumbents and challengers fighting for power.Firstly, our analysis reveals that the expected success rate of a coup attemptsignificantly influences its likelihood. This success rate is heavily influencedby the balance of power between the incumbent regime and challengers, which islargely determined by regime type. We find that military regimes, although withmore control over their own military forces, face a higher risk of coups comparedto dominant-party regimes.Secondly, the study introduces a redefined concept: the autocoup. Defined as anincumbent leader's refusal to relinquish power as mandated, this researchdistinguishes autocoups from broader terms like self-coups. Based on thisdefinition, we present the first publicly available dataset of autocoup eventsfrom 1945 to 2022, encompassing 110 attempts and 87 successful autocoups. Casestudies and empirical analyses demonstrate the dataset's utility for quantitativeresearch, providing a robust foundation for further analysis on autocoups.Thirdly, employing survival analysis techniques, the study finds cleardifferences in leader longevity between those who come to power through coups andthose who extend their rule through autocoups. The results indicate thatcoup-installed leaders face a significantly higher risk of removal compared toautocoup leaders who manipulate the system to extend their rule.### Limitations and directions for future researchThis study offers a novel framework for analysing irregular power transitions,but some limitations require further exploration:- **Data refinement:** Defining and classifying autocoups is a new approach. Future research should validate this classification system through additional studies and expert evaluations.- **Data harmonization:** The current analysis faces challenges due to mismatched units (country-year vs. leader) between coup and autocoup datasets. Future efforts should explore data harmonization techniques for more robust comparisons.- **Democratic backsliding:** While this study establishes a connection between irregular power transitions and democratic backsliding, further empirical evidence is needed to solidify this link.Several avenues exist for future research:- **Terminology and data collection:** Refining the "autocoup" concept and achieving wider recognition will facilitate more accurate and comprehensive data collection.- **Dataset expansion:** Expanding the autocoup dataset with more cases and integrating it with data on other irregular leadership transitions can provide a more holistic view of political survival after these events.- **Power dynamics and long-term impacts:** Utilizing this dataset, future studies can delve deeper into power dynamics at play and explore the long-term consequences of irregular transitions on political systems, particularly regarding democratic backsliding, breakdown, and personalization of power.In conclusion, this study sheds light on the dynamics of irregular powertransitions, specifically focusing on coups and autocoups. By redefiningautocoups, classifying the dataset, analysing determinants, and comparing leaderlongevity, we establish a framework for understanding irregular transitions andleader survival. This work contributes to a deeper understanding of democraticresilience and political stability. Future research can build upon thisfoundation by conducting further empirical analyses based on the novel autocoupdataset and continuing to refine the framework.```{r}#| label: references#| results: asis#| eval: !expr format == "latex"#| echo: falsecat("\\newpage## References")```[^1]: According to the Archigos dataset, "Removed by Military, without Foreign Support" and "Removed by Other Government Actors, without Foreign Support" in the variable exitcode are classified as coups.[^2]: The definitions and concepts of power expansion and power extension can often be ambiguous. In this study, we define power expansion as an incumbent acquiring additional authority from other branches or apparatuses of the state. Conversely, power extension refers to an incumbent prolonging their tenure beyond the originally mandated term in office.
