Income Inequality s Impact on the. Occurrence of Coup D états. Suheyla Cavdar

Transcription

1 Income Inequality s Impact on the Occurrence of Coup D états Suheyla Cavdar New York University International Relations Honors Thesis Professor Alastair Smith Spring 2017

2 Cavdar 2 Abstract Coups, or the illegal overthrow of governments by the states military or elites, have been a frequent occurrence throughout world history. Although the number of attempts has decreased since the 1960s, recent coups in the developing world prove that they are still possible. The recent literature on coups generally focuses on a specific time period or region of the world. This paper is on the global occurrences of successful and attempted coups since the 1960s and whether a country s income inequality impacts the likelihood of coup occurrences. The data and results empirically support my theory that inequality, during civil unrest and economic decline, creates incentives and opportunities for the state s elites to attempt a coup to protect their own interests and resources. Due to the systematic coup proofing and risks involved, democratic countries are less likely to experience both attempted and successful coups compared to non-democratic ones. Thus, non-democratic countries with high levels of income inequality, civil unrest, and economic stagnation are more vulnerable to coups than countries that are more democratic with lower levels of inequality, less unrest, and economic growth. Coups are not new phenomena. Yet, they still remain unpredictable and difficult to explain. Income inequality is one factor that clarifies the circumstances that can lead to a coup. Further equality might have the potential to decrease domestic instability in nations and global uncertainty in terms of foreign policies and international affairs.

3 Cavdar 3 I. Introduction The illegal overthrow of a government by its elite or military has been a common occurrence throughout world history. However, the number of coup d états around the world has gradually decreased since the 1960s (see Figure 1). The recent slight increase in the percentage of success could mean the environment and situations that generate coups are changing or elites, the small but dominant groups of people, especially in politics, the economy, and military, that control a disproportionate amount of wealth and power in a country, are better at predicting their probability of success, and only initiating coups when they are certain they will succeed. Either way, recent events in Turkey, Thailand, and Egypt prove that coups remain a current issue that is very difficult to predict and explain. Figure 1: Global Occurrences of Coups Coups could be caused by numerous factors and analyzed through many interpretations. Income inequality, or the gap between the rich elite and the poorer masses not part of the especially wealthy or influential, is one element that can shed light on the conditions and circumstances that could influence the chances of coups. This paper is on

4 Cavdar 4 whether a country s income inequality impacts the likelihood of a coup d état attempt and its success, and how the impact differs in a variety of environments and situations. Recent research on coups are usually on a specific time period, region of the world, or only on autocracies. My thesis contributes to the existing literature by focusing on the global occurrences of both successful and attempted coups since the 1960s. The results empirically demonstrate that income inequality, during times of civil unrest and economic decline increases the likelihood of coups. Moving from the 25th percentile to the 75th percentile of income inequality, the probability of a coup will increase by 6.4 percentage points. Inequality, unrest, and economic problems cause the leadership to lose legitimacy and create an incentive for the state s elites to attempt to overthrow the current government in order to protect their own interests and resources. When the poorer masses rise to demand more equality and resources, the executive, who wants to remain in power, might be willing to appease them through redistribution of wealth or more power. Fearing losing their wealth and influence, the elite might attempt a coup to prevent or react to a shift in status quo while the government is already weakened. 1 Moreover, the level of democracy decreases the likelihood of coups by 0.2%. Democratic countries have fewer coups most likely due to the risks and systematic coup proofing. Elites often topple the establishment through the country s armed forces and, in particular, senior officers who command them. Coup-proofing is structures put in place, usually in democratic nations, that prevent small groups from seizing power and decrease the likelihood of coups, such as splitting the military or security agencies into different 1 Gurr and Marshall s Polity IV data measure level of democracy based on their elections, competitiveness, openness, political participation, and checks on executive authority. The score ranges from -10 to 10, with -10 to -6 defined as autocracies, -5 to 5 anocracies, and 6 to 10 democracies.

5 Cavdar 5 divisions that check and balance each other's powers (Quinlivan 1999). My results support the theory that non-democratic countries with more income inequality are more vulnerable to the occurrence of coup d états than democratic countries. II. Background A coup d état is illegal and overt attempts by the military or other elites within the state apparatus to unseat the sitting executive, the chief leader that has the authority and responsibility to govern a state and execute the law, while a successful coup is when the elites seize and hold power for at least seven days (Powell & Thyne 2011). The overthrowing of a regime by the military has immense impacts that create instability both domestically and internationally. For one, they can hurt the economy of a country years after a coup, affecting its GDP, development, manufacturing, and investment. More importantly, they are a great setback for democracy, civil liberties, and human rights. The military often suspends existing constitutions and laws and declares martial law with curfews, bans, and censorship of the press. Coups often lead to mass imprisonment and violence, especially against the old regime and its supporters or the perpetrators of an attempt. In Egypt, Sisi targeted the Muslim Brotherhood, jailing its leaders, members, and supporters. It is reported that more than 3,000 people have been killed during the months of unrest following the coup, including the Rabaa Square massacre, where forces killed about a thousand pro-morsi protesters. The single night of the coup attempt in Turkey led to about 300 people s death. Fadel (2016) and others argued the Turks unpleasant history with coups that involved killings, executions, imprisonment, and torture, contributed greatly to the coup not

6 Cavdar 6 succeeding. Afterwards, President Erdogan began an aggressive purge, firing or arresting more than 100,000 officials, officers, judges, journalists, and teachers who he claimed had ties to the coup attempt. The aftermath of both successful and failed coups can be very ruthless and repressive. Internationally, coups can hurt foreign relations between allies. For instance, the U.S. is legally required to cut off aid to countries where a coup has happened. This is mostly the reason why the government insisted on not acknowledging the events in Egypt as a coup. On another point, the 2016 coup attempt in Turkey, a U.S. ally and NATO member, in 2016 and the uncertainty it caused threatened the fight against ISIS and the Syrian refugee agreement. Coups have an impact inside and outside the state s borders, and can negatively impact the lives of thousands of people. III. Literature Review The main question and theory are drawn and extended from existing literature on inequality and coups. They can be organized into certain common arguments and themes, such as the ideal environments for coups, role of poor masses, role of the sitting executive, role of elites, ability of elites to carry out coups through the military, and the impact of democracy on the likelihood of coups. One strand of literature contributes to the analysis that governments that are illegitimate due to inequality, instability, or economic problems are more convenient for elites to overthrow. Alesina and Perotti (1996) indicate inequality increases the likelihood of social unrest. Houle (2016) highlights that class inequality, by inciting instability and distributive conflicts, presents an incentive for different classes to seize power for their

7 Cavdar 7 own purposes. According to Acemoglu and Robinson (2001 & 2006), inequality encourages social unrest and political instability, which creates an opportunity for elites within the state to plot a coup against the government. Inequality can lead to instability and illegitimacy, while the regime s political and economic illegitimacy increases the chances of coups. Thyne and Powell (2014) state that coups are precipitated by mass discontent over the regime s legitimacy. Johnson and Thyne (2016) further argue instability increases the occurrences of coups. They explain how protests, a sign of illegitimacy, develop an environment and incentive for coups, while also making it easier for coup-plotters to coordinate. By analyzing a range of sources, Belkin and Schofer (2003) conclude that coup vulnerability is based on the strength of civil society, legitimacy, and past coups. Additionally, Alesina and Perotti s study suggests that income inequality increases the chances of instability and uncertainty, which decreases investment and reduces growth. Recessions and periods of economic difficulties are an opportunity for coups, as Acemoglu and Robinson note. According to Londregan and Poole (1990), a country s level of economic well being can determine the probability of coups. More specifically, less economic growth increases the probability of a government being overthrown by a coup. Overall, there seems to be consensus that inequality is a sign of illegitimacy, which creates an opportunity for elites to plot coups. The occurrences of coups depend on masses creating unrest and elites taking up the moment. Acemoglu and Robinson and Houle support the arguments regarding the roles of the masses, the government, and the elites in coups. Coups are a preventative or reactionary measure for elites to protect their wealth in countries with high inequality.

8 Cavdar 8 Acemoglu and Robinson infer that when there is inequality in non-democratic countries, the regime represents the elite s interests. Thus, elites prefer non-democracies so they can choose policies that are most beneficial to them, such as lower taxes and no redistribution of wealth. When high inequality exists, the poor masses favor high taxes for the rich and redistribution if they have political influence. In non-democracies, the masses pose a threat to the elites, especially if there is unrest and instability. The government in power can try to stabilize the situation through redistribution of assets or promises of reform. Therefore, elites have an incentive to prevent democratization or radical policies such as land reform due to the threat to their wealth and power. Houle presents that inequality creates a threat to the current regime. It generates grief, providing a coalition on which coup plotters can rely on to initiate and maintain their regime once the coup occurs. The elites can gain the support of the people through promises of new policies based on their preferences, while the masses are more prone to join insurgencies. The elites need not only the opportunity and incentive, but also the ability to overthrow the executive. A group of authors assert that states with inequality and less political participation tend to have stronger militaries, less coup proofing, and, consecutively, fewer coups. Conrad, Kim, and Souva (2013) demonstrate how institutions that restrict political participation, such as non-democracies, increase the influence and resources of narrow political, bureaucratic interests. Because bureaucracies are interested in maximizing the budget of the government and military, these militaries are strong enough to carry out coup attempts. Coups are not caused by the military seeking power, but a result of them already having influence in the government. As regimes age, their

9 Cavdar 9 policies become rooted in the system, so older regimes tend to spend more on their militaries. Powell (2012), who defines coup proofing as changes to the structure of the military which inhibit its capacity to stage coups, explains how coup proofing reduces the likelihood of coup occurrences by weakening the military s power. The second half of Houle s paper focuses on how higher inequality creates a greater threat to rulers. Hence, rulers have a motive to not establish coup proofing structures and, instead, allow militaries to increase their strength in order to quell any mass uprisings. Inequality decreases structural coup proofing, which can make the military s strength enough to coordinate and carry out coups. Although the threat from the masses has decreased, the ruler is now bound to the military that has the power to overthrow him if undermined. Houle draws greatly from Svolik (2012), who claims that inequality increases the threat of the masses for the dictators in autocracies, forcing rulers to give more power to the military. Moreover, Svolik s analysis reveals the effect of inequality on military intervention in politics within autocracies is an inverted U-shape. Meaning, countries that are neither too equal nor too unequal are most likely to experience coups. Svolik indicates that in equal dictatorships, the military is too weak to intervene because there is not a real threat to the establishment that requires a powerful military. However, in unequal dictatorships, a military intervention is not necessary for the military to get its way the military enjoys great autonomy and strength due to the threat the government faces from the masses. Yet, in the mid-ranges, the military is not powerful enough to always determine policy, but is powerful enough to initiate a coup for its interests.

10 Cavdar 10 Finally, an assortment of works makes conclusions on the different political situations that manifest in democracies and autocracies. It has been established that democracies have fewer coups than non-democracies. Acemoglu and Robinson note that coups are possible in democracies as long as an opportunity arises. Bell (2016) asserts that less coup attempts happen in democracies, but are more likely to succeed when they do happen. Established democracies and some autocracies have succession rules. When succession rules exist, Frantz and Stein (2016) allege, the elite plotters would be more likely to wait it out than risk failing with a coup attempt. Because elites would have less incentive to strike preemptively, succession rules make coordination among coup plotters more difficult. Hiroi and Omori (2013) state that hybrid intermediate regimes are the most vulnerable to coups because of the lack of legal means to gaining power, plotters ease of coordination, and shortage of a credible punishment threat against plotters. Furthermore, autocracies have a lower risk of coups during times of political and social instabilities than democracies. IV. Theoretical Intuitions Income inequality, especially during times of civil uprisings and economic decline, will lead to the breakdown of the existing system s legitimacy, which will create opportunities for elites to overthrow the incumbent government in order to protect their own interests and resources. The overall hypothesis is that income inequality will increase the likelihood of coup occurrences. The three central actors in this theory the poor masses, the current executive government, and the rich elites have their own set of interests and roles. The incentives will change according to the circumstances and

11 Cavdar 11 environment of the country. The interests of the actors differ in non-democracies and democracies, mostly due to differences in the military s structure and strength. My thesis question is based on the theory that income inequality, civil unrest, and economic stagnation have a negative impact on a leader s legitimacy, thereby increasing the likelihood of coups. A government has legitimacy, derived from the consent of the governed, when it is the accepted authority to execute power and has influence over the people (Locke 1689). Without legitimacy, the government is at risk of collapsing, especially if the influential elite group does not support it. Civil unrest a disturbed state or range of events in the form of public expression or motivated attacks by the masses for a political reason or against the status quo is an indicator of illegitimacy. Illegitimacy can be caused by inequality, turmoil due to changes or lack of changes in policies, or economic stagnation and decline. When the economy is not growing, the masses are less likely to be satisfied and the government is more likely to be perceived as illegitimate. Londregan and Poole argue that slow levels of economic growth increase the probability that a government is overthrown by a coup. Likewise, Alesina and Perotti depict how often income inequality leads to social discontent, which increases socio-political instability and uncertainty in the environment. Political uncertainty reduces investments, the primary engine of growth, and thereby reducing growth. Even if inequality does not necessarily impede economic growth, when inequality exists alongside a stagnant economy the likelihood of coups increases. When faced with inequality, the dissatisfied masses are likely to demand change and more resources or democratization from the government in order to bring change

12 Cavdar 12 themselves. The more unequal the society is and the more they revolt, while their success depends on structural factors more than their will to protest. The government is likely to grant the masses more resources or redistribute wealth to appease them, since the executive wants to remain in power. The government will remain in power if they are able to appease the masses while also keeping the elites satisfied. This is especially difficult in societies with very high income inequality because the rich elite tend to have a lot of power. Additionally, the governments of unequal states may have less coup-proofing and form strong militaries to crush uprisings, which would make the military strong enough to carry out the coup and make it easier for them to coordinate (Houle 2016). For instance, during the 2011 Arab uprisings, the military's willingness to use force on the people was determinant of whether or not the uprisings would succeed (Bellin 2012). When a government s legitimacy is wavering as civil unrest erupts, the elite may see this as an opportunity (Powell 2012; Johnson 2016). The elites, who want to maintain their power and wealth, will either stage a coup preemptively, to crush any potential liberalization or revolution, or as a reaction to a liberal reform, change in the status quo, or a shift towards the left. They are more likely to attempt a coup when there is higher inequality, more civil unrest, declining government legitimacy, and less coup proofing because it decreases the risks of failing and being purged. While the existing system is waning, the masses will be more likely to support the coup if the elite make promises, even if the promises are empty and false, in order to attain legitimacy and keep the masses support on their side for after the coup. In short, this theory is based on groups and group level decisions, not on individuals. When the elite class stages a coup, they tend to be

13 Cavdar 13 against a policy change or potential policy change, the masses tend to be unsatisfied with the status quo (and inequality), and the current existing government tends to be falling from favor. However, the situation differs based on governments. In non-democratic states, elites will be more likely to stage coup attempts to defend their own interests, to stop the government from giving the people more resources for appeasement, or out of fear that the masses will start a revolution to democratize the state. Any shift from the status quo towards higher taxes, redistribution of wealth, and resources for the masses will anger the wealthy elite group. Elites prefer non-democracies because they get more resources and power. Rich elites oppose democratization because if the people start making the policies, the median voter the voter in the middle of the political spectrum whose preference determines what is selected in a majority voting system will be poor and want higher taxes on the rich to redistribute their wealth (Robinson and Acemoglu 2006). In some democracies, especially the ones with high inequality, the elites may have captured the democracy, such that even if they cannot change who is elected officially, they can have disproportionate amounts of influence in practice via lobbying, investment, and local control (Acemoglu and Robinson 2013). Thus, the elite are still able to control the political process, like in non-democratic states. If the policy begins changing by moving away from the status quo towards the left or unrest and demand for change increases by the masses, then the elite may initiate a coup in reaction or preemptively. In real, established democratic states, policies will reflect the median voter's choices. If there is high inequality and the median voter is poor, there will be policies to tax the rich

14 Cavdar 14 and distribute their wealth. Yet, the very poor might not be fully satisfied due to the inequality, especially if the economy is not doing well. Although not as common, mass political expressions, which are a sign of political instability and government illegitimacy, can be a window of opportunity for plotters to attempt the coup. This can give the elite an incentive to attempt a coup in order to seize more political power and keep their wealth after continuous marginalization of their preferences. This type of coup in well established democracies is more rare than non-democracies and unstable democracies due to the higher risks of failing. Logically speaking, their governments tend to have more legitimacy and more coup-proofing, since uprisings are not an eminent threat. In democracies, there is a clear, legal means to gaining power, so the government tends to be legitimate and fewer uprisings take place. Even if the government becomes illegitimate, there is a certain, established way of setting up a new government, so the people would be less likely to support a coup. When there are succession rules, coup attempts are less likely because the plotters would rather risk less by waiting it out (Frantz and Stein 2016). There is a definite punishment if the plotters are caught, so they are less likely to initiate a coup unless they are certain they will succeed (Hiroi and Omori 2013). Domestically, the military s leaders tend to be weaker because citizens would have less incentive or interest to increase the size of the military or its budget since the voters would want the money for themselves. The military cannot easily coordinate among themselves when planning the coup due to coup proofing and division within the military. Countries such as the United States have branches within the government and the military

15 Cavdar 15 so that the different sections limit each other s power and authority, making it difficult to coordinate. Nevertheless, coups can still happen in democracies they are just less likely. Democracies with high levels of inequality still tend to have coups more than democracies with low levels of inequality. My hypotheses are not true in more equal countries. The more equal the people are, the less likely they are to demand redistribution or revolt for economic reasons. Even if they do, the government would be less likely to appease them because the cost would be too high. Additionally, wealthier people would have less to lose with redistribution and democratization. Their taxes would not be that much higher than the rest of the people. Thus, they would be less likely to initiate a coup attempt. Finally, there would be more coup proofing and weaker military because the people or uprisings would not be an immediate threat. V. Hypotheses hypotheses: Based on the theoretical arguments above, I have come up with the following nine H1: Inequality will increase the occurrences of coups. My main hypothesis is that inequality will increase the probability of coup occurrences. When a country has high levels of inequality, the government in place may be rendered illegitimate and inadequate, and further civil uprisings. Inequality can give the coup plotters both an opening to attempt their coup and a motivation. Countries with high inequality have a powerful, wealthy elite class that will feel threatened if the larger and

16 Cavdar 16 much poorer lower classes begin acts of mass unrest. Out of fear that the masses may seize power, the elite may attempt to overthrow the already weakening government. H2: Relationship between inequality and coup occurrences is an inverted-u shape. Countries with mid-range level of inequality will have a higher probability of experiencing a coup than countries with extremely equal or extremely unequal income distribution. If there is extreme inequality, the masses interests will not pose a serious threat for the elite because the elite will know their power is secure. If the elite are really rich and the masses are really poor, it is easier to pay them off. In very unequal societies, because the very large number of poor people could potentially lead to an uprising or revolution, the military tends to be strong enough to prevent such events. This also means the military will be strong enough to have the capacity to overthrow the government if necessary. The government will be too reliant on the elite and their wealth to anger them with policies that might go against their interests. H3: Civil unrest will increase the occurrences of coups. My theory that inequality increases the chances of coups is dependent on my second hypothesis that civil unrest will also increase the probability of coups. This is due to the fact that inequality without demonstration of dissatisfaction from the people might not be enough to actually increase the chances of a coup. The sitting leader must be worried enough to consider giving the poor masses more resources or power, while the elite must be threatened enough to react through a coup. In some states, the government might be able to suppress the uprisings and no change will come about. But large amounts of people publicly demonstrating their frustration with the regime might even lead to a revolution or

17 Cavdar 17 change in government. When there is civil unrest in unequal countries, coups are more likely. The interaction between inequality and civil unrest is my fourth hypothesis. H4: Inequality and civil unrest will increase the occurrences of coups. I anticipate that the relationship between inequality and occurrences of coups might vary depending on whether there is civil unrest or not. The interaction effect will show that inequality has a larger impact on coups when there is civil unrest, while the impact is not as salient when there is no civil unrest. The effect of inequality on the probability of coups depends on the amount of mass unrest. H5: Economic growth will decrease the occurrences coups. I predict that economic growth will decrease the probability of coups. Economic stagnation is another indication and cause of mass dissatisfaction and the government s unpopularity and illegitimacy. By causing instability and uncertainty, the slowdown in the economy can encourage the elites to attempt the coup and further scare the sitting leader to cede to the masses. Additionally, the coup plotters could promise the people better economic prospects to gain their support over the current regime. Finally, Londregan and Poole s study of 121 countries between 1950 and 1982 demonstrates that the probability of a coup is influenced by economic well being. Thus, when there is economic growth in an unequal country, the probability a government will be overthrown by a coup decreases. H6: Democracies will have fewer coups than non-democracies. Democracies will have a lower chance of coups than non-democracies. Elites in non-democracies have a lot more to lose if they allow a shift in the status quo or democratization, while coups in democracies are significantly more risky for the elite. In

18 Cavdar 18 non-democratic states, elites will be more likely to defend their interests and prevent their wealth and power from going into the hands of the masses. The elite fear a revolution or liberalization because it would mean they would face higher taxes and have less power and autonomy over the state. In democratic states, the governments tend to have more legitimacy and more coup proofing in place. Meaning, the elites and military cannot easily coordinate and if they are able to, they are less likely to succeed. If the coup fails they will be punished. Furthermore, the people are less willing to support the new regime because there is a clear, legal means to gaining power. Even if there are high levels of inequality, non-democracies are less likely to experience coups than democracies. The interaction between inequality and level of democracy is my eighth hypothesis. H7: Relationship between democracy and coup occurrences is an inverted-u shape. Countries that are fully democratic and fully autocratic will have a lower chance of encountering a coup than hybrid or anocratic countries. If a country is very autocratic with an established government, it is difficult to topple it and failure to do so will lead to the punishment of the plotters. Hiroi and Omori argue that intermediate regimes are the most vulnerable because of their lack of legal means to gaining power and the elite s ease of coordination. H8: Inequality and democracy will decrease the occurrences of coups. The relationship between inequality and occurrences of coups might vary depending on the country s level of democracy. I believe the interaction effect will depict that inequality has a smaller impact on coups when a country is more democratic, while the

19 Cavdar 19 impact is greater when the country is more autocratic. The effect of inequality on the probability of coups depends on how democratic a country is. H9: Democracies will have fewer successful coups than non-democracies. Mostly for the same reasons democracies have a less chance of coup occurrences than non-democracies, successful coups must be less likely in democratic countries. Democracies are more likely to have coup proofing and weaker militaries than non-democracies, which have to keep a strong military in place to keep the state in order. The military and elites organizational and coordination problems will make carrying out a coup very difficult, so chances of failure will be higher. VI. Data I carry out my research by running regression analyses on merged combinations of existing datasets on the statistical data analysis software R. The unit of analysis is country-year. The dependent variable is coup occurrences. The main independent variable is income inequality. Other independent variables include civil unrest, economic growth, and level of democracy. Population, GDP per capita, trade as a percentage of GDP, and regime durability are the controls. A summary of the datasets and the descriptive statistics can be found in the appendix. The datasets used to run the tests come from five sources. For the dependent variable, I use Powell and Thyne s Global Instances of Coups data, which covers all coups from 1950 to the present. It is arguably the most recent and comprehensive dataset on coups today. Since 1950, there have been 475 coups in total, spread across 96 countries. Coded from a range of sources, it includes both coup attempts and successful coups. For my

20 Cavdar 20 study, I have made the coup dependent variable take the form of a dummy for whether a coup occurred or not, a value of total number of coup attempts and successful coups, and a value of just the total number of successful coups in a given country and in a given year. No country has had more than four coup attempts and three successful coups in a given year. It should be noted that because the data only has entries for coup occurrences, I change the empty country-year entries to zeros, assuming no coup attempts occurred. Despite how rigorous Powell and Thyne s study is, this is still a speculation that no coups occurred in those cases. The source for the main independent variable is from Solt s Standardized World Income Inequality Database (SWIID), which consists of 176 countries from 1960 to the present. Income inequality refers to the cross-sectional distribution of income at every year. Economic inequality can be caused by a variety of factors, including the labor market, liberalism, globalization, education, taxation policies, ethnic discriminations, gender discriminations, economic development, and others. The Gini index measures and depicts the distribution and inequality of income in a country by using a Lorenz curve, which plots the cumulative percentages of total income against the total population, to portray how much it deviates from perfect equality (Gini 1912, Lorenz 1905). Gini ranges from 0 to 1, where 0 is perfectly equal and 1 is completely unequal. Solt uses a missing-data algorithm that puts together data on Gini coefficients, both market/gross (wage) and net (after tax) income inequality, from a variety of sources to increase the coverage and comparability. SWIID also measures absolute and relative redistribution. Absolute redistribution is the difference between the market-income and net-income Gini indices, while relative is the

21 Cavdar 21 difference divided by the market income Gini and multiplied by 100 or the percentage by which market income inequality is reduced. For the purposes of my study, I use the net gini rather than the market gini because the income inequality before taxes does not accurately depict the effect of the government and its policies (Solt 2016). There is a total of 4,087 entries, with an average net gini of It ranges from 14.1 (Slovakia in 1990) and 67.2 (Namibia in 1993). For the civil unrest independent variable, I work with the University of Illinois Cline Center s Social, Political, Economic Event data (SPEED) for civil unrest, which consists of 165 countries from 1946 to the present. It is divided into categories such as political expression and politically motivated attacks. The Cline Center for Democracy at the University of Illinois defines political expression as the public articulation, by non-governmental actors (or government officials acting in their private capacity), of threatening or unwelcome political messages. The political message must be threatening to societal elites and/or prevailing societal equilibria. These include statements, actions, and symbolic expressions, which could be verbal, written, digital, broadcasted, movies or documentaries, signs, non-violent mass demonstrations, political associations, passive resistance, prayers, boycotts, disrespect to national symbols, and self-inflicted harm. Politically motivated attacks are physical acts, perpetrated by humans for political reasons, which are intended to damage the person or property of others. The initiator must have political reasons for initiating the act (Hayes and Nardulli 2011). This can be divided into five categories: 1. spontaneous attacks (riots or brawls),

22 Cavdar extraordinary attacks (assassinations, suicide attacks, kidnappings, hostage taking, or executions), 3. other politically motivated attacks (personal attacks or property attacks), 4. organized mass attacks (border incidents, blockade, or sieges), and 5. unrealized political attacks involving unsuccessful efforts to damage others for political reasons (attempts or conspiracies). I compile all expressions and politically motivated attacks into a single variable as unrest events total. Then, for clarity, each entry is logged. There is a total of 9,271 event inputs, ranging from 0 to 230 events with an average of two events. After the log is taken, the range becomes 0 to 5.4 and the average becomes 0.5. It should be noted that, similar to the coup data, because the data only has entries for civil unrest events, I change the empty country-year entries to zeroes for no unrest. The SPEED study is very thorough, but I am still making an assumption that no civil unrest events happened in those cases. To measure economic growth, I use the World Bank s GDP per capita growth data ranging from 1960 to 2015 and including 264 countries or regions. I only use the countries. This is the annual percentage growth rate of gross domestic product divided by population, based on constant local currency. There are 8,621 entries, with an average of 2.1 percent and range between -65 percent and percent. My study also retrieves the GDP per capita, total population, and trade controls from the World Bank. They all cover 1960 to 2015 across 264 countries or regions. Population is the total number of residents as a mid-year estimate. Gross domestic product (GDP) is the yearly sum of goods produced and services provided in a country. It is the sum of all consumer spending, government

23 Cavdar 23 spending, investment, and net exports (difference between total exports and total imports). GDP per capita is GDP, in current US dollars, divided by the midyear population. Both population and GDP are logged. Population has 11,997 observations, while GDP has 8,908 observations. Trade is the sum of exports and imports of goods and services as a percentage of GDP. It is a proxy for globalization and trade openness. There are 8,133 observations, with a range of 0 to and mean of Finally, I use Gurr and Marshall s Polity IV data from 1946 to 2013, which includes 167 countries, to measure democracy. It is the level of democracy of states based on their elections, competitiveness, openness, political participation, and checks on executive authority. The score ranges from -10 to 10, with -10 to -6 being autocracies and -5 to 5 being anocracies, and 6 to 10 being democracies. There are 8,687 inputs, with an average polity score of 1. The dataset also includes a regime durability variable, which I use as a control. It is the number of years since the most recent regime change (defined by a three point change in POLITY score over a period of three years or less) or the end of transition period defined by the lack of stable political institutions (denoted by a standardized authority score) (Marshall, Gurr, & Jaggers 2016). With 8,764 observations, it ranges from 0 to 206 with an average of 22. VII. Empirical Design I empirically test my predictions by running cross sectional time series, or panel, analyses with Probit and Ordinary Least Squares (OLS) regressions for each of my hypotheses. The dependent coup dummy variable is applied to the probit models because it is binary, while the total number of coup variable is applied to the OLS model tests. I run

24 Cavdar 24 basic regressions and regressions with controls for both probit and OLS models. I do a robust test with the OLS models by clustering the standard errors at the country level. Finally, I repeat all steps with country and year fixed effects to absorb all unobserved heterogeneity that does not change across countries over time to make sure the variance is not driven by a specific predictor in a country and year. This focuses on the intra-country and intra-time variations, or variation within each country and within each year, rather than inter-country and inter-time variation in order to control for potential omitted variable bias. A full list of all the models can be found in the appendix. Below are the main templates: Model 1: Basic Dummy Probit Regression Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + ε Models 2-6: Dummy Probit Regression with Controls Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(durability) + β6log(population) + β7log(gdp per capita) + β8(trade) + + ε Model 7: Basic Total Coup OLS Regression Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + ε Models 8-12: Total Coup OLS Regression with Controls Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(durability) + β6log(population) + β7log(gdp per capita) + β8 (Trade) + + ε Model 13: Basic Successful Coups Dummy Probit Regression Successful Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity)+ε Models 14-16: Successful Coup Dummy Probit Regression with Controls Successful Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(durability) + β6log(population) + β7log(gdp per capita) + β8(trade) + ε

25 Cavdar 25 Model 17: Basic Total Successful Coup OLS Regression Successful Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + ε Models 18-20: Total Successful Coup OLS Regression with Controls Successful Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(durability) + β6log(population) + β7log(gdp per capita) + β8 (Trade) + ε Models : All models with Country and Year Fixed Effects Coup it = β0 + β1(gini) it + β2log(unrest) it + β3(gdp growth) it + β4(polity) it + + α(country) i + δ(year) t + ε it VIII. Results and Analysis The theory that inequality increases the probability of coups is empirically supported by the data and results. My results indicate that income inequality, during periods of civil unrest and economic decline, creates an incentive and motivation for the state s elites to attempt a coup in order to secure their interests and resources. Additionally, coups are less likely to occur in democratic countries. Thus, non-democratic countries with high income inequality are more vulnerable to the occurrences of coup d états than more democratic and less unequal countries. Overall, almost all my predictions are significantly supported. All model regression tables can be found in the appendix. The results based on each of my hypotheses are detailed below: H1: Inequality will increase the occurrences of coups. My main hypothesis is supported: as inequality increases, the probability of a coup attempt also increases. The regression coefficient is not extremely strong -- it is never more than and the significance varies based on the model. The dummy probit regression

26 Cavdar 26 is very strongly significant, with a p-value < 0.01, while the OLS regression is strongly significant with a p-value < Inequality s impact on whether a coup occurs or not is stronger and more significant than its impact on how many coups occurs. In both models, the control variables lessen the strength and significance of the correlation. In fact, the significance disappears in the OLS model when the controls are in place. Gini and coups appear to be correlated, but it is mostly caused by other variables, such as GDP, population, regime durability, and trade. The OLS regressions with clustered standard errors using the total number of coup dependent variable indicates that for every one-unit increase in gini, occurrence of coups are expected to increase by 0.1% or 0.02% with controls while holding other variables in the model constant. The positive relation remains the same but the correlation becomes extremely weak and the significance completely disappears with country and year fixed effects. This is probably due to gini levels having little within-country and within-year variations, so the fixed effects models cannot assess the effect of inequality on coups. Thus, the fixed effect model only captures and magnifies the few countries that have had a change in inequality. In terms of my hypothesis, these results support the idea that when a country has high levels of inequality, the current government may be established illegitimate and expedite civil uprisings. Inequality can give the coup plotters an opportunity and a motivation to overthrow the leader. Countries with high levels of inequality usually have an influential elite class that will feel threatened if the larger lower classes begin uprising. Elites might overthrow the weakening government to prevent them from giving the people power or wealth.

27 Figure 2: Relationship between Gini (net) and Coup (dummy) Data Cavdar 27 H2: Relationship between inequality and coup occurrences is an inverted-u shape. To test the inverted-u shape of the relationship between inequality and coups, I plotted gini with gini-squared along with the rest of the variables and controls. The positive sign of the gini variable coefficient and the negative sign of the gini-squared variable coefficient support my prediction. It is strongly significant for both the probit regression (p-value < 0.01) and the OLS regression (p-value < 0.05). Based on the OLS regression, moving from the 25th percentile to the 75th percentile of income inequality, the probability of a coup will increase by 6.4 percentage points. Similar to the regressions without the gini-squared, the fixed effect models have the same signs but are no longer significant because inequality does not change within a country. The inverted U-shape is certainly present in Figure 2. It portrays the net gini and coup data's relationship, rather than my

28 Cavdar 28 regression models. Because it is the nonparametric depiction of what the data tells us, it shows that my results are not driven by model specifications. Because graphs with binary dependent variables are not so clear, I use loess smoothing method that focus on the deterministic part of the variation in the data. The parabolic shape of the regression means that most of the action happens in the middle ranges of the gini score. It can be concluded that just enough inequality is needed to make people angry, irritate the government, and threaten the elites. If there is extreme inequality, the rich elite will not be worried about or bothered by the masses because they know they are secure. The government would rely too much on the elites wealth and the military s strength to do anything against them. Plus, it would be easy to appease or pay off many very poor people. However, as seen in figure 3, most of the gini data is in the range between 25 and 50, with a mean of Realistically speaking, no country is perfectly unequal or equal. Figure 3: Gini Density Plot

29 Cavdar 29 These trends can be noticed in the three countries I previously mentioned: Egypt, Thailand, and Turkey. These countries are all anocracies and have had a coup or coup attempt in the last 5 years. Figure 4 depicts the gini of these countries since the 1960s to around 2010 with points referencing the occurrence of a coup (or vertical lines when the exact gini is not available). Turkey s previous coup before the 2016 attempt occurred when inequality was very high. It has steadily declined since then, and so have the number of coups. In Thailand, coups happened at points where the gini is increasing, never when it is declining. No coups occurred in Egypt despite the high levels of inequality around the 1970s and 2000s, which may be a representation of the parabolic shape. Even if inequality was not the direct cause of these coups, they probably had an impact. Figure 4: Inequality & Coups of Egypt, Thailand, & Turkey Across Time

30 Cavdar 30 H3: Civil unrest will increase the occurrences of coups. As civil unrest increases, the occurrences of coups are more likely. In all the regressions, the estimates of the regression coefficient on the unrest variable are positive and significant with p-values < The OLS regressions coefficients are weaker compared to the probit regressions. The robust OLS regression suggests that for every one-unit increase in (logged) unrest, occurrence of coups are expected to increase by 2.1% while holding other variables in the model constant. Mass unrest is a sign of illegitimacy and an opportunity for the elites to stage a coup when they have the motive to topple the current regime. Unrest can also be influenced by inequality. It should be noted that because the unrest events are logged, even though the range is 0 to 5.4, mostly all data are between 0 and 1.5 with a mean of 0.5. The density plot can be found in the appendix. There is a clear upward trend between the civil unrest and the coup datasets, as depicted in figure 5 using the loess smoothing method. Figure 5: Relationship between Unrest (logged) and Coup (dummy) Data

31 Cavdar 31 H4: Inequality and civil unrest will increase the occurrences of coups. The interaction between inequality and civil unrest has mixed results based on which model is used. My prediction was that when a country has unrest, the effect of inequality on the probability of coups would be larger. My hypothesis is supported when I run OLS regressions for both the coup dummy and total number of coups. When a probit regression is run, the coefficient comes out negative. However, almost none of the coefficients are significant. The only significant results are for the OLS regressions without robust standard errors, which has a p-value < 0.1, and the OLS regression with the dummy coup variable, which has a p-values < Because the results that are significant both have a positive coefficient, my hypothesis is supported. The fixed effects are, again, not significant because inequality does not vary much within a country and year. Figure 6: Relationship between Gini and Coup without Unrest and with Unrest Figure 6 depicts the relationship between net gini and coups when no civil unrest events happen and when it does. When events happen, the effect of gini on coups is more.

32 Cavdar 32 This is because the people are already unsatisfied and are publicly demonstrating it. Inequality can further this dissatisfaction and fuel the unrest to the point where the government must appease the poor masses. If there is no or little unrest, or if the state can successfully repress them, the government or elites will not be threatened. The non-binary marginal effect graph is in the appendix. H5: Economic growth will decrease the occurrences coups. According to the empirical tests, economic growth decreases the occurrences of coups. The coefficients are weaker for the OLS regressions than the probit regressions, but all models results are negative and strongly significant with p-values < The robust OLS regressions show that for every one-unit increase in economic growth, occurrence of coups are expected to decrease by 0.4% while holding other variables in the model constant. It should be mentioned that most of the data is between -10% and 10%, with a mean of 2.1%. The density plot can be found in the appendix. The relationship between the GDP growth data and the coup data is negative, as shown in figure 7. Economic decline is an indication of and cause for the public's discontent and the government s illegitimacy. The instability and uncertainty can cause the government to concede to the poor people, while encouraging the elites to attempt a coup to prevent it. Additionally, the elite can use promises of better economic conditions as a tool to garner support from the masses.

33 Figure 7: Relationship between GDP Growth and Coup (dummy) Data Cavdar 33 H6: Democracies will have fewer coups than non-democracies. Democracies decrease the probability of coup occurrences. All models results are negative and very significant, with p-values < As control variables are added, the coefficients get weaker. The OLS regressions with clustered standard errors presents that for every one-unit increase in polity, occurrence of coups are expected to decrease by 0.2% while holding other variables constant. When a country has a higher level of democracy, coups are a lot more risky for the elites, so they do not attempt it in case they fail and get purged afterwards. Democracies tend to have more coup proofing, weaker militaries, and more government legitimacy. Clear rules and means to gaining power means that the people will probably be less likely to support a new regime that took power through illegal means. Figure 8 presents the polity IV data s relationship with the coup dataset, and not my

34 Cavdar 34 exact regression models with the other variables and controls, so the apparent correlation might be misleading. The shape is discussed further in the seventh hypothesis. Figure 8: Relationship between Polity and Coup (dummy) Data H7: Relationship between democracy and coup occurrences is an inverted-u shape. Like the gini-coup regression in the second hypothesis, I plotted polity with polity-squared along with the rest of the variables and controls to test for the parabola shape of the relationship between democracy and coups. Although the graph of the relation between level of democracy and coups seem to support my prediction, it is misleading because it does not factor in my other variables. The data looks like an inverted U-shape, but the coefficients state otherwise. My assumption is false because both the sign of the polity variable coefficient and the polity-squared variable coefficient are negatives. The results of the simple probit regression and OLS regression with fixed effects have statistically significant results. This could be impacted by the data, because most of the

35 Cavdar 35 observations are concentrated at the two ends of the polity spectrum. In figure 9, the polity density plot is depicted over the polity and gini relationship graph from figure 8 (the original polity density plot can be found in the appendix). Coups in democracies are risky and less likely to occur. In non-democratic countries, the elites, because their wealth and power at stake, are more likely to attempt a coup. Plus, in autocracies the elite are stronger and wealthier. Figure 9: Polity Density Plot with Polity-Gini Graph H8: Inequality and democracy will decrease the occurrences of coups. My hypothesis is that the more democratic a country is, the less the effect of inequality would be on the probability of coups. My assumption on the interaction between inequality and democracy is not supported by the results. Not only are the results mixed for each model -- the coefficients are negative for the OLS regressions and positive for the probit regressions -- but none of them are statistically significant. Although insignificant,

36 Cavdar 36 figure 10 depicts the relationship between net gini and coups for different levels of democracy. If significant, it would mean that income inequality has a less impact on the likelihood of coups in democracies. In democracies, even if people are poor, they at least seem to have a democratic say in policies. Meaning, the elites in democracies do not have a complete monopoly on political influence. They are less likely to attempt overthrowing the government because they do not fear losing their power along with their resources, like they would in autocratic countries. Figure 10: Relationship between Gini and Coups for Different Levels of Democracy H9: Democracies will have fewer successful coups than non-democracies. I ran regressions with successful coups and attempted coups as separate dependent variables to compare their polity coefficients. All are negative, meaning the more democratic a country is, the less likely it is to have successful and attempted coups and less coups over all. The correlation between polity and successful coups is highly significant for all regressions. Both probit and OLS regressions, regressions with basic variables and

37 Cavdar 37 regressions with controls, and all regressions with fixed effects all have p-values < The regression between polity and attempted coups are not as strongly significant, nor as strongly correlated as successful coups. The robust OLS regressions indicate that for every one-unit increase in polity, occurrence of successful coups are expected to decrease by 0.2%, while the number of attempted coups is expected to decrease by 0.01%. The more democratic a country is, the more its chances of having a successful coup drops, which is more than its chances of having an attempted coup. Figure 11 depicts the density plot of coups for each polity level. Successful coups (in blue) tend to happen at more autocratic levels, and less at democratic levels. Additionally, more failed coup attempts (in red) have happened at the democratic end than successful coups. Figure 11: Polity Density Plot for Coups The reasoning behind this trend is essentially the same as why democracies have fewer coups overall: coup proofing and weak militaries. If there is coup-proofing in the system and if the military is weak, there will be organizational and coordinating problems.

38 Cavdar 38 The coup will not be easily planned or carried out, decreasing the chances of it being successful. In democracies, gaining power illegally, rather than democratically, will be frowned upon and not accepted by the masses. Moreover, the rich elites will not have as much power and influence in democracies as they do in autocracies. Table 1: Overall Results

39 Cavdar 39 Table 2: Overall Results with Fixed Effects IX. Conclusion Overall, the results of my tests empirically support my theoretical arguments on how income inequality impacts the likelihood of coup d états. My study on the global occurrences of coups since the 1960s exhibits that elites are more likely to instigate a coup attempt in countries with higher levels of income inequality in order to secure their own interests and resources. When there is civil unrest and economic decline, there is a positive correlation between a country s income inequality and the occurrences of coups, especially in non-democracies. Democratic countries have less coup occurrences, but when they do,

40 Cavdar 40 they tend to be less successful than non-democratic countries due to coup-proofing and the risks involved. Therefore, non-democratic countries with high levels of income inequality, civil unrest, and economic stagnation are more liable to coup attempts than countries that are more democratic with lower levels of income inequality, less civil unrest, and economic growth. However, there are a few obvious pitfalls to my research. My paper assumes that no coup attempts have happened that Powell and Thyne have not indicated. Although their data is a comprehensive articulation and improvement of fourteen difference coup datasets and is regularly updated, it is possible that they may have missed a few instances. The same case goes for the civil unrest event dataset. Despite how rigorously prepared, it might not be all encompassing and completely thorough. Reporting bias is possible and very likely. Most of the data and studies rely on Western news outlets. News in unknown, smaller and less developed countries might not be as well reported as news in large, developed countries. Finally, I may have missed significant control variables in my theory that needed to be accounted for. For instance, including variables that measure governmental legitimacy and instability, military strength, and coup proofing is one way this paper can be advanced. Moreover, the study can be furthered by researching income inequality s impact on the duration and outcome of new regimes after successful coups. Nevertheless, this paper has important implications for the study of inequality and coups. My results add a cross-national and recent inquiry to the existing literature. Coups are not a new phenomenon elites around the world have been overtly overthrowing existing regimes for a long time. Yet, coup attempts still remain very unpredictable and

41 Cavdar 41 difficult to explain. Income inequality is one factor that can explain the circumstances that can lead to coups. Examining the role of inequality can reveal why some nations are more vulnerable than others. Inequality also illustrates why some countries elites are more prone to turn to coups while others are not. A greater focus on the role of inequality in political instability has the potential to decrease uncertainty domestically, especially in developing countries, and globally, in terms of foreign policies and international affairs. Pushing for further income equality can potentially stabilize governments, whether democratic or autocratic.

42 Cavdar 42 References Acemoglu, Daron, and James A. Robinson (2001). "A Theory of Political Transitions." American Economic Review 91(4): Acemoglu, Daron, and James A. Robinson (2006). Economic Origins of Dictatorship and Democracy. Cambridge: Cambridge UP. Acemoglu, Daron, and James A. Robinson (2013). Why Nations Fail: The Origins of Power, Prosperity, and Poverty. New York, NY: Crown Business. Alesina, Alberto, and Roberto Perotti (1996). Income Distribution, Political Instability, and Investment. European Economic Review 40(6): Belkin, Aaron, and Evan Schofer (2003). "Toward a Structural Understanding of Coup Risk." Journal of Conflict Resolution 47(5): Bell, Curtis (2016). "Coup D'etat and Democracy." Comparative Political Studies 49(9): Bellin, Eva (2012). Reconsidering the Robustness of Authoritarianism in the Middle East: Lessons from the Arab Spring. Comparative Politics 44(2): Conrad, Justin, Hong-Cheol Kim, and Mark Souva (2013). "Narrow Interests and Military Resource Allocation in Autocratic Regimes." Journal of Peace Research 50(6): Dunne, Michele, and Scott Williamson (2014). Egypt s Unprecedented Instability by the Numbers. Carnegie Endowment For International Peace, March 24, Escriba-Folch, Abel (2013). "Repression, Political Threats, and Survival under Autocracy." International Political Science Review 34(5): Fadel, Leila (2016). Egypt and Turkey: Why A Coup Succeeded In One and Failed In The Other. NPR Parallels, July 27, Fahey, Mark (2016). What We Can Learn From 70 Years of Coups. CNBC, July 21, Frantz, Erica, and Elizabeth A. Stein (2016). "Countering Coups: Leadership Succession Rules in Dictatorships." Comparative Political Studies (2016): Gini, Corrado (1912). Variabilità e mutabilità.

43 Cavdar 43 Hayes, Matthew, and Peter F. Nardulli (2011). SPEED s Societal Stability Protocol and the Study of Civil Unrest: An Overview and Comparison with Other Event Data Projects. Cline Center for Democracy, University of Illinois at Urbana-Champaign. Hiroi, Taeko, and Sawa Omori (2013). "Causes and Triggers of Coups D'etat: An Event History Analysis." Politics & Policy 41(1): Hiroi, Taeko, and Sawa Omori (2015). "Policy Change and Coups: The Role of Income Inequality and Asset Specificity." International Political Science Review 36(4): Houle, Christian (2016). "Why Class Inequality Breeds Coups but Not Civil Wars." Journal of Peace Research 53(5): Johnson, Jaclyn, and Clayton L. Thyne (2016). "Squeaky Wheels and Troop Loyalty." Journal of Conflict Resolution : Lagendijk, Joost. Good Coups, Bad Coups and Double Standards. Huffington Post. Locke, John (1689). Second Treatise on Civil Government. Indianapolis: Hackett, Londregan, John B., and Keith T. Poole (1990). "Poverty, the Coup Trap, and the Seizure of Executive Power." World Politics 42(2): Lorenz, Max O. (1905). "Methods of Measuring the Concentration of Wealth." Publications of the American Statistical Association. Publications of the American Statistical Association Vol. 9, No. 70 9(70): Marshall, Monty G., Ted R. Gurr, and Keith Jaggers (2016). Polity IV Project: Dataset Users Manual v2015. Center for Systemic Peace. Marshall, Monty G., Ted R. Gurr, and Keith Jaggers (2016). Polity IV Project: Political Regime Characteristics and Transitions, Mills, C. Wright (2000). The Power Elite. New York: Oxford UP, Powell, Jonathan M. (2012). Determinants of the Attempting and Outcome of Coups D etat. Journal of Conflict Resolution 56(6): Powell, Jonathan M. (2014). "An Assessment of the Democratic Coup Theory." African Security Review 23(3): Powell, Jonathan M., and Clayton L. Thyne (2011). Global Instances of Coups from 1950 to 2010: A New Dataset. Journal of Peace Research 48(2):

44 Cavdar 44 Quinlivan, James, T. (1999). Coup-proofing: Its Practice and Consequences in the Middle East. International Security 24(2): Solt, Frederick (2016). The Standardized World Income Inequality Database. Social Science Quarterly 97(5): SWIID Version 5.1, July Svolik, Milan (2012). "Contracting on Violence: The Moral Hazard in Authoritarian Repression and Military Intervention in Politics." Journal of Conflict Resolution 57(5): Svolik, Milan (2012). The Politics of Authoritarian Rule. Cambridge: Cambridge UP. Thyne, Clayton L., and Jonathan M. Powell (2014). "Coup D'etat or Coup D'Autocracy? How Coups Impact Democratization, " Foreign Policy Analysis 0: Torchia, Christopher (2016). Why Does Turkey Have a Long History of Coups? US News, July 16, World Development Indicators. World Bank.

45 Cavdar 45 Appendix I. Summary of Datasets Variable Data Source Years Countries Coups Global Instances of Coups Powell & Thyne countries with coups Inequality Standardized World Income Inequality Database Solt countries Civil Unrest Social, Political, Economic Events Database University of Illinois Cline Center countries Economic growth, GDP per capita, Population, Trade GDP growth, GDP per capita, Population, Trade World Bank countries Democracy & Durability Polity IV Gurr & Marshall countries II. Descriptive Statistics N Mean St. Dev. Min Max Coup (total) 8, Coup (success) 8, Coup (attempt) 8, Gini (net) 4, Unrest (logged) 9, Unrest (total) 9, Political Expression 9,

46 Cavdar 46 Political Attack 9, Mass Demonstration 9, Riots 9, GDP Growth 8, Polity 8, Durability 8, Trade (% of GDP) 8, Population (logged) 11, GDP (logged) 8, III. Complete List of Models Model 1: Basic Dummy Probit Regression Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + ε Model 2: Dummy Probit Regression with Controls Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(durability) + β6log(population) + β7log(gdp per capita) + β8(trade) + ε Model 3: Dummy Probit Regression with Controls & Gini Squared Coup (dummy) = β0 + β1(gini) + β2(gini) 2 + β3log(event) + β4(gdp growth) + β5(polity) + β6(durability) + β7log(population) + β8log(gdp per capita) + β9(trade) + ε Model 4: Dummy Probit Regression with Controls & Polity Squared Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(polity) 2 + β6(durability) + β7log(population) + β8log(gdp per capita) + β9 (Trade) + ε Model 5: Dummy Probit Regression with Controls & Interaction between Gini & Unrest Coup (dummy) = β0 + β1(gini) + β2log(event) + β3[(gini) * Log(Event)] + β4(gdp growth) + β5(polity) + β6(durability) + β7log(population) + β8log(gdp per capita) + β9(trade) + ε Model 6: Dummy Probit Regression with Controls & Interaction between Gini & Polity

47 Cavdar 47 Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5[(gini) * (Polity)] + β6(durability) + β7log(population) + β8log(gdp per capita) + β9(trade) + ε Model 7: Basic Total Coup OLS Regression Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + ε Model 8: Total Coup OLS Regression with Controls Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(durability) + β6log(population) + β7log(gdp per capita) + β8 (Trade) + ε Model 9: Total Coup OLS Regression with Controls & Gini Squared Coup (total) = β0 + β1(gini) + β2(gini) 2 + β3log(event) + β4(gdp growth) + β5(polity) + β6(durability) + β7log(population) + β8log(gdp per capita) + β9(trade) + ε Model 10: Total Coup OLS Regression with Controls & Polity Squared Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(polity) 2 + β6(durability) + β7log(population) + β8log(gdp per capita) + β9 (Trade) + ε Model 11: Total Coup OLS Regression with Controls & Interaction between Gini & Unrest Coup (total) = β0 + β1(gini) + β2log(event) + β3[(gini) * Log(Event)] + β4(gdp growth) + β5(polity) + β6(durability) + β7log(population) + β8log(gdp per capita) + β9(trade) + ε Model 12: Total Coup OLS Regression with Controls & Interaction between Gini & Polity Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5[(gini) * (Polity)] + β6(durability) + β7log(population) + β8log(gdp per capita) + β9(trade) + ε Model 13.1 & 13.2: Basic Successful/Attempted Coups Dummy Probit Regression Successful/Attempted Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + ε Model 14.1 & 14.2: Successful/Attempted Coup Dummy Probit Regression with Controls Successful/Attempted Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(durability) + β6log(population) + β7log(gdp per capita) + β8(trade) + ε Model 15.1 & 15.2: Successful/Attempted Coup Dummy Probit Regression with Controls & Gini Squared Successful/Attempted Coup (dummy) = β0 + β1(gini) + β2(gini) 2 + β3log(event) + β4(gdp growth) + β5(polity) + β6(durability) + β7log(population)

48 Cavdar 48 + β8log(gdp per capita) + β9(trade) + ε Model 16.1 & 16.2: Successful/Attempted Coup Dummy Probit Regression with Controls & Polity Squared Successful/Attempted Coup (dummy) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(polity) 2 + β6(durability) + β7log(population) + β8log(gdp per capita) + β9 (Trade) + ε Model 17.1 & 17.2: Basic Total Successful/Attempted Coup OLS Regression Successful/Attempted Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + ε Model 18.1 & 18.2: Total Successful/Attempted Coup OLS Regression with Controls Successful/Attempted Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(durability) + β6log(population) + β7log(gdp per capita) + β8 (Trade) + ε Model 19.1 & 19.2: Total Successful/Attempted Coup OLS Regression Controls & Gini Squared Successful/Attempted Coup (total) = β0 + β1(gini) + β2(gini) 2 + β3log(event) + β4(gdp growth) + β5(polity) + β6(durability) + β7log(population) + β8log(gdp per capita) + β9(trade) + ε Model 20.1 & 20.2: Total Successful/Attempted Coup OLS Regression Controls & Polity Square Successful/Attempted Coup (total) = β0 + β1(gini) + β2log(event) + β3(gdp growth) + β4(polity) + β5(polity) 2 + β6(durability) + β7log(population) + β8log(gdp per capita) + β9 (Trade) + ε Model 21: Basic Dummy Probit Regression Country & Year Fixed Effects Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + α(country) i + δ(year) t + ε Model 22: Dummy Probit Regression with Controls Country and Year Fixed Effects Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5(durability) it + β6log(population) it + β7log(gdp per capita) it + β8(trade) it + α(country) i + δ(year) t + ε Model 23: Dummy Probit Regression with Controls & Gini Squared Country & Year Fixed Effects Coup (dummy) it = β0 + β1(gini) it + β2(gini) 2 it + β3log(event) it + β4(gdp growth) it + β5(polity) it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9(trade) it + α(country) i + δ(year) t + α(country) i + δ(year) t + ε

49 Cavdar 49 Model 24: Dummy Probit Regression with Controls & Polity Squared Country & Year Fixed Effects Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5(polity) 2 it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9 (Trade) it + α(country) i + δ(year) t + ε Model 25: Dummy Probit Regression with Controls & Interaction between Gini & Unrest Country & Year Fixed Effects Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3[(gini) * Log(Event)] it + β4(gdp growth) it + β5(polity) it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9(trade) it + α(country) i + δ(year) t + ε Model 26: Dummy Probit Regression with Controls & Interaction between Gini & Polity Country & Year Fixed Effects Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5[(gini) * (Polity)] it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9(trade) it + α(country) i + δ(year) t + ε Model 27: Basic Total Coup OLS Regression Country & Year Fixed Effects Coup (total) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + α(country) i + δ(year) t + ε Model 28: Total Coup OLS Regression with Controls Country & Year Fixed Effects Coup (total) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5(durability) it + β6log(population) it + β7log(gdp per capita) it + β8 (Trade) it + α(country) i + δ(year) t + ε Model 29: Total Coup OLS Regression with Controls & Gini Squared Country & Year Fixed Effects Coup (total) it = β0 + β1(gini) it + β2(gini) 2 it + β3log(event) it + β4(gdp growth) it + β5(polity) it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9(trade) it + α(country) i + δ(year) t + ε Model 30: Total Coup OLS Regression with Controls & Polity Squared Country & Year Fixed Effects Coup (total) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5(polity) 2 it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9 (Trade) it + α(country) i + δ(year) t + ε Model 31: Total Coup OLS Regression with Controls & Interaction between Gini & Unrest Country & Year Fixed Effects Coup (total) it = β0 + β1(gini) it + β2log(event) it + β3[(gini) * Log(Event)] it + β4(gdp growth) it + β5(polity) it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9(trade) it + α(country) i + δ(year) t + ε

50 Cavdar 50 Model 32: Total Coup OLS Regression with Controls & Interaction between Gini & Polity Country & Year Fixed Effects Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5[(gini) * (Polity)] it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9(trade) it + α(country) i + δ(year) t + ε Model 33.1 & 33.2: Basic Successful/Attempted Coups Dummy Probit Regression Country & Year Fixed Effects Successful/Attempted Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) + α(country) i + δ(year) t + ε Model 34.1 & 34.2: Successful/Attempted Coup Dummy Probit Regression with Controls Country & Year Fixed Effects Successful/Attempted Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5(durability) it + β6log(population) it + β7log(gdp per capita) it + β8(trade) it + α(country) i + δ(year) t + ε Model 35.1 & 35.2: Successful/Attempted Coup Dummy Probit Regression with Controls & Gini Squared Country & Year Fixed Effects Successful/Attempted Coup (dummy) it = β0 + β1(gini) it + β2(gini) 2 it + β3log(event) it + β4(gdp growth) it + β5(polity) it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9(trade) it + α(country) i + δ(year) t + ε Model 36.1 & 36.2: Successful/Attempted Coup Dummy Probit Regression with Controls & Polity Squared Country & Year Fixed Effects Successful/Attempted Coup (dummy) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5(polity) 2 it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9 (Trade) it + α(country) i + δ(year) t + ε Model 37.1 & 37.2: Basic Total Successful/Attempted Coup OLS Regression Country & Year Fixed Effects Successful/Attempted Coup (total) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + α(country) i + δ(year) t + ε Model 38.1 & 38.2: Total Successful/Attempted Coup OLS Regression with Controls Country & Year Fixed Effects Successful/Attempted Coup (total) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5(durability) it + β6log(population) it + β7log(gdp per capita) it + β8 (Trade) it + α(country) i + δ(year) t + ε Model 39.1 & 39.2: Total Successful/Attempted Coup OLS Regression Controls & Gini Squared Country & Year Fixed Effects

51 Cavdar 51 Successful/Attempted Coup (total) it = β0 + β1(gini) it + β2(gini) 2 it + β3log(event) it + β4(gdp growth) it + β5(polity) it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9(trade) it + α(country) i + δ(year) t + ε Model 40.1 & 40.2: Total Successful/Attempted Coup OLS Regression Controls & Polity Square Country & Year Fixed Effects Successful/Attempted Coup (total) it = β0 + β1(gini) it + β2log(event) it + β3(gdp growth) it + β4(polity) it + β5(polity) 2 it + β6(durability) it + β7log(population) it + β8log(gdp per capita) it + β9 (Trade) it + α(country) i + δ(year) t + ε IV. Regression Result Tables

52 Cavdar 52

53 Cavdar 53

54 Cavdar 54

55 Cavdar 55

56 Cavdar 56

57 Cavdar 57

58 Cavdar 58

59 Cavdar 59

60 Cavdar 60 V. Non-Binary Marginal Effects Graph VI. Density Plots of Data

61 Cavdar 61