Coup d'état

From Wikipedia, the free encyclopedia

https://en.wikipedia.org/wiki/Coup_d%27%C3%A9tat

General Napoleon Bonaparte during the Coup of 18 Brumaire in Saint-Cloud, detail of painting by François Bouchot, 1840

A coup d'état, or simply a coup, is an illegal and overt attempt by the military or other government elites to unseat the incumbent leader. A self-coup is when a leader, having come to power through legal means, tries to stay in power through illegal means.

By one estimate, there were 457 coup attempts from 1950 to 2010, half of which were successful. Most coup attempts occurred in the mid-1960s, but there were also large numbers of coup attempts in the mid-1970s and the early 1990s. Coups occurring in the post-Cold War period have been more likely to result in democratic systems than Cold War coups, though coups still mostly perpetuate authoritarianism.

Many factors may lead to the occurrence of a coup, as well as determine the success or failure of a coup. Once a coup is underway, coup success is driven by coup-makers' ability to get elites and the public to believe that their coup attempt will be successful. The number of successful coups has decreased over time. Failed coups in authoritarian systems are likely to strengthen the power of the authoritarian ruler. The cumulative number of coups is a strong predictor of future coups, a phenomenon referred to as the "coup trap".

In what is referred to as "coup-proofing", regimes create structures that make it hard for any small group to seize power. These coup-proofing strategies may include the strategic placing of family, ethnic, and religious groups in the military; and fragmenting of military and security agencies. However, coup-proofing reduces military effectiveness. One reason why authoritarian governments tend to have incompetent militaries is that authoritarian regimes fear that their military will stage a coup or allow a domestic uprising to proceed uninterrupted – as a consequence, authoritarian rulers have incentives to place incompetent loyalists in key positions in the military.

Etymology

The term comes from French coup d'État, literally meaning a 'stroke of state' or 'blow of state'. In French, the word État (French: ) is capitalized when it denotes a sovereign political entity.

Although the concept of a coup d'état has featured in politics since antiquity, the phrase is of relatively recent coinage. It did not appear within an English text before the 19th century except when used in the translation of a French source, there being no simple phrase in English to convey the contextualized idea of a 'knockout blow to the existing administration within a state'.

One early use within text translated from French was in 1785 in a printed translation of a letter from a French merchant, commenting on an arbitrary decree or arrêt issued by the French king restricting the import of British wool. What may be its first published use within a text composed in English is an editor's note in the London Morning Chronicle, January 7, 1802, reporting the arrest by Napoleon in France, of Moreau, Berthier, Masséna, and Bernadotte: "There was a report in circulation yesterday of a sort of coup d'état having taken place in France, in consequence of some formidable conspiracy against the existing government."

In the British press, the phrase came to be used to describe the various murders by Napoleon's alleged secret police, the Gens d'Armes d'Elite, who executed the Duke of Enghien: "the actors in torture, the distributors of the poisoning draughts, and the secret executioners of those unfortunate individuals or families, whom Bonaparte's measures of safety require to remove. In what revolutionary tyrants call grand[s] coups d'état, as butchering, or poisoning, or drowning, en masse, they are exclusively employed."

Related terms

Self coup

A self-coup, also called an autocoup (from the Spanish: autogolpe), or coup from the top, is a form of coup d'état in which a nation's head, having come to power through legal means, tries to stay in power through illegal means. The leader may dissolve or render powerless the national legislature and unlawfully assume extraordinary powers not granted under normal circumstances. Other measures may include annulling the nation's constitution, suspending civil courts, and having the head of government assume dictatorial powers.

Between 1946 and 2022, an estimated 148 self-coup attempts took place, 110 in autocracies and 38 in democracies.

Soft coup

A soft coup, sometimes referred to as a silent coup or a bloodless coup, is an illegal overthrow of a government, but unlike a regular coup d'état it is achieved without the use of force or violence.

Palace coup

A palace coup or palace revolution is a coup in which one faction within the ruling group displaces another faction within a ruling group. Along with popular protests, palace coups are a major threat to dictators. The Harem conspiracy of the 12th century BC was one of the earliest. Palace coups were common in Imperial China. They have also occurred among the Habsburg dynasty in Austria, the Al-Thani dynasty in Qatar, and in Haiti in the 19th to early 20th centuries. The majority of Russian tsars between 1725 and 1801 were either overthrown or usurped power in palace coups.

Putsch

The term Putsch ([pʊtʃ], from Swiss-German 'knock'), denotes the political-military actions of an unsuccessful minority reactionary coup. The term was initially coined for the Züriputsch of 6 September 1839 in Switzerland. It was also used for attempted coups in Weimar Germany, such as the 1920 Kapp Putsch, Küstrin Putsch, and the 1923 Beer Hall Putsch by Adolf Hitler.

During the Night of the Long Knives in 1934, a supposed putsch was the underpinning of a disinformation tactic by Hitler and other Nazi party members. After initiating a purge, the idea of an imminent coup allowed them to falsely claim the killing was justified (as a means to suppress an uprising). Germans still use the term Röhm-Putsch to describe the event, the term given to it by the Nazi regime, despite the unproven implication that the murders were necessary to prevent a reactionary coup. Thus, German authors often use quotation marks or write about the sogenannter Röhm-Putsch ('so-called Röhm Putsch') for emphasis.

The 1961 Algiers Putsch and the 1991 August Putsch also use the term.

Pronunciamiento

Main article: Pronunciamiento

Pronunciamiento ('pronouncement') is a term of Spanish origin for a type of coup d'état. The pronunciamiento is the formal explanation for deposing the regnant government, justifying the installation of the new government that was affected by the golpe de estado. A "barracks revolt" or cuartelazo is also a term for military revolt, from the Spanish term cuartel ('quarter' or 'barracks'). Specific military garrisons are the sparking factor for a larger military revolt against the government.

One author makes a distinction between a coup and a pronunciamiento. In a coup, it is the military, paramilitary, or opposing political faction that deposes the current government and assumes power; whereas, in the pronunciamiento, the military deposes the existing government and installs an ostensibly civilian government.

Other

Other types of actual or attempted unilateral seizures of power are sometimes called "coups with adjectives". The appropriate term can be subjective and carries normative, analytical, and political implications.

• Civil society coup
• Constitutional coup
• Counter-coup, a coup to repeal the result of a previous coup
• Democratic coup
• Electoral coup
• Judicial coup
• Market coup
• Military coup
• Parliamentary coup
• Presidential coup
• Royal coup, in which a monarch dismisses democratically elected leaders and seizes all power; for example the 6 January Dictatorship
• Slow-motion coup
• Slow-moving coup
• Slow-rolling coup

Revolution, rebellion

A revolution or rebellion can have the same outcome as a coup, in that a ruler or government can be replaced by unconstitutional means. However, while a coup is usually made by a small group and planned beforehand, a revolution or rebellion is usually started more spontaneously and by larger groups of uncoordinated people. The distinction is not always clear. Sometimes, a coup is also labelled as a revolution by the coup makers to try to give it a form of democratic legitimacy.

Prevalence and history

Further information: List of coups and coup attempts and List of coups and coup attempts by country

According to Clayton Thyne and Jonathan Powell's coup data set, there were 457 coup attempts from 1950 to 2010, of which 227 (49.7%) were successful and 230 (50.3%) were unsuccessful. They find that coups have "been most common in Africa and the Americas (36.5% and 31.9%, respectively). Asia and the Middle East have experienced 13.1% and 15.8% of total global coups, respectively. Europe has experienced by far the fewest coup attempts: 2.6%." Most coup attempts occurred in the mid-1960s, but there were also large numbers of coup attempts in the mid-1970s and the early 1990s. From 1950 to 2010, a majority of coups failed in the Middle East and Latin America. They had a somewhat higher chance of success in Africa and Asia. Numbers of successful coups have decreased over time.

Outcomes

Successful coups are one method of regime change that thwarts the peaceful transition of power. A 2016 study categorizes four possible outcomes to coups in dictatorships:

• Failed coup
• No regime change, such as when a leader is illegally shuffled out of power without changing the identity of the group in power or the rules for governing
• Replacement of incumbent with another dictatorship
• Ousting of the dictatorship followed by democratization (also called "democratic coups")

The study found that about half of all coups in dictatorships—both during and after the Cold War—install new autocratic regimes. New dictatorships launched by coups engage in higher levels of repression in the year that follows the coup than existed in the year leading to the coup. One-third of coups in dictatorships during the Cold War and 10% of later ones reshuffled the regime leadership. Democracies were installed in the wake of 12% of Cold War coups in dictatorships and 40% of post-Cold War ones.

Coups occurring in the post-Cold War period have been more likely to result in democratic systems than Cold War coups, though coups still mostly perpetuate authoritarianism. Coups that occur during civil wars shorten the war's duration.

Predictors

A 2003 review of the academic literature found that the following factors were associated with coups:

• officers' personal grievances
• military organizational grievances
• military popularity
• military attitudinal cohesiveness
• economic decline
• domestic political crisis
• contagion from other regional coups
• external threat
• participation in war
• collusion with a foreign military power
• military's national security doctrine
• officers' political culture
• noninclusive institutions
• colonial legacy
• economic development
• undiversified exports
• officers' class composition
• military size
• strength of civil society
• regime legitimacy and past coups.

The literature review in a 2016 study includes mentions of ethnic factionalism, supportive foreign governments, leader inexperience, slow growth, commodity price shocks, and poverty.

Coups have been found to appear in environments that are heavily influenced by military powers. Multiple of the above factors are connected to military culture and power dynamics. These factors can be divided into multiple categories, with two of these categories being a threat to military interests and support for military interests. If interests go in either direction, the military will find itself either capitalizing off that power or attempting to gain it back.

Often times, military spending is a indicator of the likelihood of a coup taking place. Nordvik found that about 75% of coups that took place in many different countries rooted from military spending and oil windfalls.

Coup trap

The cumulative number of coups is a strong predictor of future coups. This phenomenon is called the coup trap. A 2014 study of 18 Latin American countries found that the establishment of open political competition helps bring countries out of the "coup trap" and reduces cycles of political instability.

Regime type and polarization

Hybrid regimes are more vulnerable to coups than very authoritarian states or democratic states. A 2021 study found that democratic regimes were not substantially more likely to experience coups. A 2015 study finds that terrorism is strongly associated with re-shuffling coups. A 2016 study finds that there is an ethnic component to coups: "When leaders attempt to build ethnic armies, or dismantle those created by their predecessors, they provoke violent resistance from military officers." Another 2016 study shows that protests increase the risk of coups, presumably because they ease coordination obstacles among coup plotters and make international actors less likely to punish coup leaders. A third 2016 study finds that coups become more likely in the wake of elections in autocracies when the results reveal electoral weakness for the incumbent autocrat. A fourth 2016 study finds that inequality between social classes increases the likelihood of coups. A fifth 2016 study finds no evidence that coups are contagious; one coup in a region does not make other coups in the region likely to follow. One study found that coups are more likely to occur in states with small populations, as there are smaller coordination problems for coup-plotters.

A 2019 study found that when a country's politics is polarized and electoral competition is low, civilian-recruited coups become more likely.

A 2023 study found that civilian elites are more likely to be associated with instigating military coups while civilians embedded in social networks are more likely to be associated with consolidating military coups.

In autocracies, the frequency of coups seems to be affected by the succession rules in place, with monarchies with a fixed succession rule being much less plagued by instability than less institutionalized autocracies.

A 2014 study of 18 Latin American countries in the 20th-century study found the legislative powers of the presidency does not influence coup frequency.

Territorial disputes, internal conflicts, and armed conflicts

A 2017 study found that autocratic leaders whose states were involved in international rivalries over disputed territory were more likely to be overthrown in a coup. The authors of the study provide the following logic for why this is:

Autocratic incumbents invested in spatial rivalries need to strengthen the military in order to compete with a foreign adversary. The imperative of developing a strong army puts dictators in a paradoxical situation: to compete with a rival state, they must empower the very agency—the military—that is most likely to threaten their own survival in office.

However, two 2016 studies found that leaders who were involved in militarized confrontations and conflicts were less likely to face a coup.

A 2019 study found that states that had recently signed civil war peace agreements were much more likely to experience coups, in particular when those agreements contained provisions that jeopardized the interests of the military.

Popular opposition and regional rebellions

Research suggests that protests spur coups, as they help elites within the state apparatus to coordinate coups.

A 2019 study found that regional rebellions made coups by the military more likely.

Effect of the military

A 2018 study found that coup attempts were less likely in states where the militaries derived significant incomes from peacekeeping missions. The study argued that militaries were dissuaded from staging coups because they feared that the UN would no longer enlist the military in peacekeeping missions.

A separate 2018 study found that the presence of military academies were linked to coups. The authors argue that military academies make it easier for military officers to plan coups, as the schools build networks among military officers.

Economy, development, and resource factors

A 2018 study found that "oil price shocks are seen to promote coups in onshore-intensive oil countries, while preventing them in offshore-intensive oil countries". The study argues that states which have onshore oil wealth tend to build up their military to protect the oil, whereas states do not do that for offshore oil wealth.

A 2020 study found that elections had a two-sided impact on coup attempts, depending on the state of the economy. During periods of economic expansion, elections reduced the likelihood of coup attempts, whereas elections during economic crises increased the likelihood of coup attempts.

A 2021 study found that oil wealthy nations see a pronounced risk of coup attempts but these coups are unlikely to succeed.

A 2014 study of 18 Latin American countries in the 20th century study found that coup frequency does not vary with development levels, economic inequality, or the rate of economic growth.

Coup-proofing

In what is referred to as "coup-proofing", regimes create structures that make it hard for any small group to seize power. These coup-proofing strategies may include the strategic placing of family, ethnic, and religious groups in the military; creation of an armed force parallel to the regular military; and development of multiple internal security agencies with overlapping jurisdiction that constantly monitor one another. It may also involve frequent salary hikes and promotions for members of the military, and the deliberate use of diverse bureaucrats. Research shows that some coup-proofing strategies reduce the risk of coups occurring. However, coup-proofing reduces military effectiveness, and limits the rents that an incumbent can extract. One reason why authoritarian governments tend to have incompetent militaries is that authoritarian regimes fear that their military will stage a coup or allow a domestic uprising to proceed uninterrupted – as a consequence, authoritarian rulers have incentives to place incompetent loyalists in key positions in the military.

A 2016 study shows that the implementation of succession rules reduce the occurrence of coup attempts. Succession rules are believed to hamper coordination efforts among coup plotters by assuaging elites who have more to gain by patience than by plotting.

According to political scientists Curtis Bell and Jonathan Powell, coup attempts in neighbouring countries lead to greater coup-proofing and coup-related repression in a region. A 2017 study finds that countries' coup-proofing strategies are heavily influenced by other countries with similar histories. Coup-proofing is more likely in former French colonies.

A 2018 study in the Journal of Peace Research found that leaders who survive coup attempts and respond by purging known and potential rivals are likely to have longer tenures as leaders. A 2019 study in Conflict Management and Peace Science found that personalist dictatorships are more likely to take coup-proofing measures than other authoritarian regimes; the authors argue that this is because "personalists are characterized by weak institutions and narrow support bases, a lack of unifying ideologies and informal links to the ruler".

Impact

Democracy

Research suggests that coups promoting democratization in staunchly authoritarian regimes have become less likely to end in democracy over time, and that the positive influence has strengthened since the end of the Cold War.

A 2014 study found that "coups promote democratization, particularly among states that are least likely to democratize otherwise". The authors argue that coup attempts can have this consequence because leaders of successful coups have incentives to democratize quickly in order to establish political legitimacy and economic growth, while leaders who stay in power after failed coup attempts see it as a sign that they must enact meaningful reforms to remain in power. A 2014 study found that 40% of post-Cold War coups were successful. The authors argue that this may be due to the incentives created by international pressure. A 2016 study found that democracies were installed in 12% of Cold War coups and 40% of the post-Cold War coups. A 2020 study found that coups tended to lead to increases in state repression, not reductions.

According to a 2020 study, "external reactions to coups play important roles in whether coup leaders move toward authoritarianism or democratic governance. When supported by external democratic actors, coup leaders have an incentive to push for elections to retain external support and consolidate domestic legitimacy. When condemned, coup leaders are apt to trend toward authoritarianism to assure their survival."

According to legal scholar Ilya Somin a coup to forcibly overthrow democratic government might be sometimes justified. Commenting on the 2016 Turkish coup d'état attempt, Somin opined,

There should be a strong presumption against forcibly removing a democratic regime. But that presumption might be overcome if the government in question poses a grave threat to human rights, or is likely to destroy democracy itself by shutting down future political competition.

Repression and counter-coups

According to Naunihal Singh, author of Seizing Power: The Strategic Logic of Military Coups (2014), it is "fairly rare" for the prevailing existing government to violently purge the army after a coup has been foiled. If it starts the mass killing of elements of the army, including officers who were not involved in the coup, this may trigger a "counter-coup" by soldiers who are afraid they will be next. To prevent such a desperate counter-coup that may be more successful than the initial attempt, governments usually resort to firing prominent officers and replacing them with loyalists instead.

Some research suggests that increased repression and violence typically follow both successful and unsuccessful coup attempts. However, some tentative analysis by political scientist Jay Ulfelder finds no clear pattern of deterioration in human rights practices in wake of failed coups in post-Cold War era.

Notable counter-coups include the Ottoman countercoup of 1909, the 1960 Laotian counter-coup, the Indonesian mass killings of 1965–66, the 1966 Nigerian counter-coup, the 1967 Greek counter-coup, 1971 Sudanese counter-coup, and the Coup d'état of December Twelfth in South Korea.

A 2017 study finds that the use of state broadcasting by the putschist regime after Mali's 2012 coup did not elevate explicit approval for the regime.

According to a 2019 study, coup attempts lead to a reduction in physical integrity rights.

International response

The international community tends to react adversely to coups by reducing aid and imposing sanctions. A 2015 study finds that "coups against democracies, coups after the Cold War, and coups in states heavily integrated into the international community are all more likely to elicit global reaction." Another 2015 study shows that coups are the strongest predictor for the imposition of democratic sanctions. A third 2015 study finds that Western states react strongest against coups of possible democratic and human rights abuses. A 2016 study shows that the international donor community in the post-Cold War period penalizes coups by reducing foreign aid. The US has been inconsistent in applying aid sanctions against coups both during the Cold War and post-Cold War periods, a likely consequence of its geopolitical interests.

Organizations such as the African Union (AU) and the Organization of American States (OAS) have adopted anti-coup frameworks. Through the threat of sanctions, the organizations actively try to curb coups. A 2016 study finds that the AU has played a meaningful role in reducing African coups.

A 2017 study found that negative international responses, especially from powerful actors, have a significant effect in shortening the duration of regimes created in coups.

According to a 2020 study, coups increase the cost of borrowing and increase the likelihood of sovereign default.

Deep state

From Wikipedia, the free encyclopedia

https://en.wikipedia.org/wiki/Deep_state

A deep state is a type of governance made up of potentially secret and unauthorized networks of power operating independently of a state's political leadership in pursuit of their own agenda and goals. In popular usage, the term carries overwhelmingly negative connotations. The range of possible uses of the term is similar to that for shadow government. The expression state within a state is an older and similar concept. Historically, it designated a well-defined organization which seeks to function independently, whereas the deep state refers more to a hidden organization seeking to manipulate the public state.

Potential sources for deep state organization include rogue elements among organs of state, such as the armed forces or public authorities such as intelligence agencies, police, secret police, administrative agencies, and government bureaucracy. During the presidency of Donald Trump, deep-state rhetoric has been used in the United States to describe the "permanent government" of entrenched career bureaucrats or civil servants acting in accordance with the mandate of their agency and congressional statutes, when seen as in conflict with the incumbent presidential administration. The intent of a deep state can include continuity of the state itself, job security of civil servants, enhanced power and authority, and the pursuit of ideological or programmatic objectives. It can operate in opposition to the agenda of elected officials, by obstructing, resisting, and subverting their policies, conditions and directives.

Forms

Deep state may refer to:

• Plans for an emergency government that takes over in the event of a disaster, see continuity of government
• A government run by:
  • An unelected bureaucracy or branch of the security services – see section Cases
  • A state within a state or deep state – specific examples include: Deep state in Turkey and Deep state in the United States.
• A 'shadow government', a conspiracy theory of a secret government.
• A broad definition of the state on the subject of state liability.

Etymology and historical usage

The modern concept of a deep state is associated with Turkey, a presumed secret network of military officers and their civilian allies trying to preserve the secular order based on the ideas of Mustafa Kemal Atatürk from 1923. There are also opinions that the deep state in Turkey and "Counter-Guerrilla" was established in the Cold War era as a part of Gladio Organization to sway Turkey more into NATO against the threat of the expansion of Soviet communism. A similar concept is the Greek language κράτος ἐν κράτει, (kratos en kratei) that was later adopted into Latin as imperium in imperio or status in statu).

In the seventeenth and eighteenth centuries political debate surrounding the separation of church and state often revolved around the perception that if left unchecked the Church might turn into a kind of State within a State, an illegitimate encroachment of the State's civil power monopoly.

At the beginning of the 20th century, the deep state was also used to refer to government-owned corporations or private companies that seem to operate largely outside of regulatory or governmental control.

Scholarly understanding

See also: State (polity)

Within social science in general and political science specifically, scholars distinguish between positivism ("what is") and normativism ("what should be"). Because political science deals with topics which are inherently political and often controversial, this distinction between "what is" (positive) and "what should be" (normative) is critical because it allows diverse people with different preferred worlds to discuss the causes, workings, and effects of policies and social structures; while readers may disagree on the normative qualities of the deep state (i.e. whether it is good or bad), it is still possible to study the positive qualities (i.e. its origins and effects) without requiring a normative judgement.

In a 1955 article in the Bulletin of the Atomic Scientists, the realist international relations scholar Hans Morgenthau quoted others speaking about a "dual state" existing in the United States: the democratic façade of elected politicians who operate according to the law, and a hidden national security hierarchy and shadow government that operates to monitor and control the former. This has been said to be the origin of the notion of a deep state in the United States.

In the field of political science, the normative pop culture concept of the deep state is studied within the literature on the state. Current literature on the state generally traces a lineage to Bringing the State Back In (1985) and remains an active body of scholarly research as of 2020. Within this literature, the state is understood as both venue (a set of rules under which others act and interact) as well as actor (with its own agenda). An example of a non-conspiratorial version of the 'state as actor' from the empirical scholarly literature would be "doing truth to power" (as a play on speaking truth to power, which is what journalists often aspire to do) as studied by Todd La Porte. Under this dual understanding, the conspiratorial version of the deep state concept would be one version of the 'state as actor' while the non-conspiratorial version would be another version of the 'state as actor.'

The fundamental takeaway from the scholarly literature on the dual nature of the state is that the "state as actor" (deep state) is a functional characteristic of all states which has effects that may be normatively judged as "good" or "bad" in different times, places, and contexts. From a positivist scientific perspective, the state-as-venue, colloquially known as the "deep state", simply "is" and should not be assumed to be "bad" by default.

Intellectual history of concept

See also: New institutionalism

While the state has been one of the longest-studied topics in political science, sociology, and economics, the rise of new institutionalism(s) in the 1970s brought to the forefront the dual nature of the state as both venue (a set of rules under which others act and interact) as well as actor (with its own agenda). This new institutionalism stands in contrast to the immediately prior behavioral revolution which focused on society-centered explanations for political outcomes where the state was primarily or solely seen as an arena where interest groups vied for political power.

State-as-actor versus state-as-venue

See also: State (polity) § State autonomy within institutionalism

The normative pop culture concept of the deep state is distinguished from the classical concept of the state within the scholarly literature on the state by the dual nature of the state as both an actor (which pursues certain ends) and a venue (which structures interaction between actors). In this dyad, the deep state is called the state-as-actor, while the classical concept of the state is called the state-as-venue.

State-as-venue

Main article: State (polity) § State and government

To distinguish the traditional, formal processes of the state from the state-as-actor, the state-as-venue view reflects the state serving as an arena in which actors act. Under this concept, the state is seen as a passive organizational structure within which societal actors (e.g. interest groups, classes) compete for power, influence, and resources.

State-as-actor

The state-as-actor concept subsumes the activities described by the pop culture concept of the deep state by focusing on all forms of state goal formation and pursuit which are independent of external societal actors (e.g. interest groups, classes).

Positivist political science and sociology further break this concept down into state autonomy and state capacity. State autonomy refers to a state's ability to pursue interests insulated from external social and economic influence. State capacity reflects the state's skills, knowledge, tools, equipment, and other resources needed to do their jobs competently. Together, autonomy and capacity are necessary for states to implement all policy including that delegated by political leaders, court decisions, and agency or ministry programmatic as well as the subversive or clandestine ends suggested by the popular usage of the deep-state concept.

Popular understanding

After the 2016 United States presidential election, deep state became much more widely used as a pejorative term with an overwhelmingly negative definition by both the Donald Trump administration and conservative-leaning media outlets.

Deep state in public international law

According to public international law (PIL), and the issue of state liability in particular, the term "deep state" is used to refer to the state in a broad sense. In this way, a state may need to answer for actions not only carried out by state authorities or representatives, but also other officials, regional authorities, publicly owned companies and other private bodies with state influence. This rule appears, among other things, in the ARSIWA report, produced by the International Law Commission, and in the case law of the International Court of Justice.

Cases

Chechnya

According to the journalist Julia Ioffe, the Russian Republic of Chechnya, under leadership of Ramzan Kadyrov, had become a state within a state by 2015.

Egypt

In 2013, author Abdul-Azim Ahmed wrote the deep state was being used to refer to Egyptian military/security networks, particularly the Supreme Council of the Armed Forces after the 2011 Egyptian revolution. They are "non-democratic leaders within a country" whose power is "independent of any political changes that take place". They are "often hidden beneath layers of bureaucracy" and may not be "in complete control at all times" but have "tangible control of key resources (whether human or financial)". He also wrote: "The 'deep state' is beginning to become short hand for the embedded anti-democratic power structures within a government, something very few democracies can claim to be free from."

Israel

In May 2020, an article in Haaretz describes how people meeting Prime Minister Benjamin Netanyahu "have heard lengthy speeches ... that even though he has been elected repeatedly, in reality, the country is controlled by a 'deep state.'"

Italy

The most famous case is Propaganda Due. Propaganda Due (better known as P2) was a Masonic lodge belonging to the Grand Orient of Italy (GOI). It was founded in 1877 with the name of Masonic Propaganda, in the period of its management by the entrepreneur Licio Gelli it assumed deviated forms with respect to the statutes of the Freemasonry and became subversive towards the Italian legal order. The P2 was suspended by the GOI on 26 July 1976; subsequently, the parliamentary commission of inquiry into the P2 Masonic lodge under the presidency of Minister Tina Anselmi concluded the P2 case by denouncing the lodge as a real "criminal organization" and "subversive". It was dissolved with a special law, the n. 17 of 25 January 1982.

Middle East

Robert Worth argues deep state is "just as apt" for networks in many states in the Middle East where governments have colluded with smugglers and jihadis (Syria), jihadi veterans of the Soviet–Afghan War (Yemen), and other criminals working as irregular forces (Egypt and Algeria). In his book From Deep State to Islamic State, he describes a hard core of regimes in Syria, Egypt, and Yemen that staged successful counter-revolutions against the Arab Spring in those countries, comparing them with the Mamluks of Egypt and the Levant 1250–1517 in that they proclaim themselves servants of the putative rulers while actually ruling themselves.

Pakistan

Since independence, the Pakistan armed forces have always had a huge influence in the country's politics. In addition to the decades of direct rule by the military government, the military also has many constraints on the power of the elected prime ministers, and also has been accused of being a deep state.

Soviet Union and post-Soviet Russia

Main article: Silovik

The Soviet secret police have been frequently described by historians as a "state within a state". According to the investigative journalist Yevgenia Albats, most KGB leaders, including Lavrenty Beria, Yuri Andropov, and Vladimir Kryuchkov, always competed for power with the Communist Party and manipulated communist leaders.

According to historian Abdurakhman Avtorkhanov in 1991, "It is not true that the Political Bureau of the Central Committee of the Communist Party is a supreme power. The Political Bureau is only a shadow of the real supreme power that stands behind the chair of every Bureau member ... The real power thinks, acts and dictates for all of us. The name of the power is NKVD–MVD–MGB. The Stalin regime is based not on the Soviets, Party ideals, the power of the Political Bureau or Stalin's personality, but on the organization and the techniques of the Soviet political police where Stalin plays the role of the first policeman." However, he also noted that "To say that NKVD is 'a state within the state' means to belittle the importance of the NKVD because this question allows two forces – a normal state and a supernormal NKVD – whereas the only force is Chekism".

According to former general Ion Mihai Pacepa in 2006, "In the Soviet Union, the KGB was a state within a state. Now former KGB officers are running the state. They have custody of the country's 6,000 nuclear weapons, entrusted to the KGB in the 1950s, and they now also manage the strategic oil industry renationalized by Putin. The KGB successor, rechristened FSB, still has the right to electronically monitor the population, control political groups, search homes and businesses, infiltrate the federal government, create its own front enterprises, investigate cases, and run its own prison system. The Soviet Union had one KGB officer for every 428 citizens. Putin's Russia has one FSB-ist for every 297 citizens."

Turkey

Main article: Deep state in Turkey

According to the Journalist Robert F. Worth, "The expression 'deep state' had originated in Turkey in the 1990s, where the military colluded with drug traffickers and hit men to wage a dirty war against Kurdish insurgents". Professor Ryan Gingeras wrote that the Turkish term derin devlet "colloquially speaking" refers to "'criminal' or 'rogue' element that have somehow muscled their way into power". The journalist Dexter Filkins wrote of a "presumed clandestine network" of Turkish "military officers and their civilian allies" who, for decades, "suppressed and sometimes murdered dissidents, Communists, reporters, Islamists, Christian missionaries, and members of minority groups—anyone thought to pose a threat to the secular order". Journalist Hugh Roberts has described the "shady nexus" between the police and intelligence services, "certain politicians and organised crime", whose members believe they are authorised "to get up to all sorts of unavowable things" because they are "custodians of the higher interests of the nation".

United Kingdom

The Civil Service has been called a deep state by senior politicians. Tony Blair said: "You cannot underestimate how much they believe it's their job to actually run the country and to resist the changes put forward by people they dismiss as 'here today, gone tomorrow' politicians. They genuinely see themselves as the true guardians of the national interest, and think that their job is simply to wear you down and wait you out." The efforts of the Civil Service to frustrate elected politicians is the subject of the popular satirical BBC TV comedy Yes Minister, which originated in the 1980s.

United States

Main article: Deep state in the United States

Since at least 2013, the deep state conspiracy theory has been used to describe "a hybrid association of government elements and parts of top-level industry and finance that is effectively able to govern the United States without reference to the consent of the governed as expressed through the formal political process." Intelligence agencies such as the CIA have been accused by elements of the Donald Trump administration of attempting to thwart its policy goals. Writing for The New York Times, the analyst Issandr El Amani warned against the "growing discord between a president and his bureaucratic rank-and-file", while analysts of the column The Interpreter wrote:

Though the deep state is sometimes discussed as a shadowy conspiracy, it helps to think of it instead as a political conflict between a nation's leader and its governing institutions.

— Amanda Taub and Max Fisher, The Interpreter

According to the political commentator David Gergen, quoted by Time in early 2017, the term has been appropriated by Steve Bannon and Breitbart News, and other supporters of the Trump Administration in order to delegitimize critics of the Trump presidency. In February 2017, deep-state theory was dismissed by authors for The New York Times, as well as The New York Observer. In October 2019 The New York Times gave credence to the general idea by publishing an opinion piece arguing that the deep state in the civil service was created to "battle people like Trump".

Scholars have generally disputed the notion that the U.S. executive branch bureaucracy represents a true deep state as the term is formally understood but have taken a range of views on the role of that bureaucracy in constraining or empowering the U.S. president.

Venezuela

Further information: Cartel of the Suns

The Cartel of the Suns, a group of high-ranking officials within the Bolivarian Government of Venezuela, has been described as "a series of often competing networks buried deep within the Chavista regime". Following the Bolivarian Revolution, the Bolivarian government initially embezzled until there were no more funds to embezzle, which required them to turn to drug trafficking. President Hugo Chávez made partnerships with the Colombian leftist militia Revolutionary Armed Forces of Colombia (FARC) and his successor Nicolás Maduro continued the process, promoting officials to high-ranking positions after they were accused of drug trafficking.

Other alleged cases

• Imperial Japan's Army, Kwantung Army and the post-war ninkyō dantai
• Imperial Brazil's Diretório Monárquico do Brasil and Black Guard
• India's RSS
• Indonesia's Kopassus
• Iran's IRGC
• Morocco's Makhzen
• Pakistan's Intelligence Community: ISI, FIA, NAB, and/or IB
• Serbia's Black Hand
• Serbia and Montenegro's RDB
• Spain's Grupos Antiterroristas de Liberación, "Antiterrorist Liberation Groups"
• Thailand's Military-Monarchy Nexus or Network monarchy
• United Kingdom's City of London Corporation

Ensemble learning

From Wikipedia, the free encyclopedia
https://en.wikipedia.org/wiki/Ensemble_learning

In statistics and machine learning, ensemble methods use multiple learning algorithms to obtain better predictive performance than could be obtained from any of the constituent learning algorithms alone. Unlike a statistical ensemble in statistical mechanics, which is usually infinite, a machine learning ensemble consists of only a concrete finite set of alternative models, but typically allows for much more flexible structure to exist among those alternatives.

Overview

Supervised learning algorithms perform the task of searching through a hypothesis space to find a suitable hypothesis that will make good predictions with a particular problem. Even if the hypothesis space contains hypotheses that are very well-suited for a particular problem, it may be very difficult to find a good one. Ensembles combine multiple hypotheses to form a (hopefully) better hypothesis. The term ensemble is usually reserved for methods that generate multiple hypotheses using the same base learner. The broader term of multiple classifier systems also covers hybridization of hypotheses that are not induced by the same base learner.

Evaluating the prediction of an ensemble typically requires more computation than evaluating the prediction of a single model. In one sense, ensemble learning may be thought of as a way to compensate for poor learning algorithms by performing a lot of extra computation. On the other hand, the alternative is to do a lot more learning on one non-ensemble system. An ensemble system may be more efficient at improving overall accuracy for the same increase in compute, storage, or communication resources by using that increase on two or more methods, than would have been improved by increasing resource use for a single method. Fast algorithms such as decision trees are commonly used in ensemble methods (for example, random forests), although slower algorithms can benefit from ensemble techniques as well.

By analogy, ensemble techniques have been used also in unsupervised learning scenarios, for example in consensus clustering or in anomaly detection.

Ensemble theory

Empirically, ensembles tend to yield better results when there is a significant diversity among the models. Many ensemble methods, therefore, seek to promote diversity among the models they combine. Although perhaps non-intuitive, more random algorithms (like random decision trees) can be used to produce a stronger ensemble than very deliberate algorithms (like entropy-reducing decision trees). Using a variety of strong learning algorithms, however, has been shown to be more effective than using techniques that attempt to dumb-down the models in order to promote diversity. It is possible to increase diversity in the training stage of the model using correlation for regression tasks or using information measures such as cross entropy for classification tasks.

Theoretically, one can justify the diversity concept because the lower bound of the error rate of an ensemble system can be decomposed into accuracy, diversity, and the other term. 

Ensemble size

While the number of component classifiers of an ensemble has a great impact on the accuracy of prediction, there is a limited number of studies addressing this problem. A priori determining of ensemble size and the volume and velocity of big data streams make this even more crucial for online ensemble classifiers. Mostly statistical tests were used for determining the proper number of components. More recently, a theoretical framework suggested that there is an ideal number of component classifiers for an ensemble such that having more or less than this number of classifiers would deteriorate the accuracy. It is called "the law of diminishing returns in ensemble construction." Their theoretical framework shows that using the same number of independent component classifiers as class labels gives the highest accuracy.

Common types of ensembles

Bayes optimal classifier

Main article: Bayes classifier

The Bayes optimal classifier is a classification technique. It is an ensemble of all the hypotheses in the hypothesis space. On average, no other ensemble can outperform it. The Naive Bayes classifier is a version of this that assumes that the data is conditionally independent on the class and makes the computation more feasible. Each hypothesis is given a vote proportional to the likelihood that the training dataset would be sampled from a system if that hypothesis were true. To facilitate training data of finite size, the vote of each hypothesis is also multiplied by the prior probability of that hypothesis. The Bayes optimal classifier can be expressed with the following equation:

${\displaystyle y=\underset{c_j\in C}{\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{m}\mathrm{a}\mathrm{x}}\sum _{h_i\in H}P(c_j|h_i)P(T|h_i)P(h_i)}$

where ${\displaystyle y}$ is the predicted class, ${\displaystyle C}$ is the set of all possible classes, ${\displaystyle H}$ is the hypothesis space, ${\displaystyle P}$ refers to a probability, and ${\displaystyle T}$ is the training data. As an ensemble, the Bayes optimal classifier represents a hypothesis that is not necessarily in ${\displaystyle H}$. The hypothesis represented by the Bayes optimal classifier, however, is the optimal hypothesis in ensemble space (the space of all possible ensembles consisting only of hypotheses in ${\displaystyle H}$).

This formula can be restated using Bayes' theorem, which says that the posterior is proportional to the likelihood times the prior:

${\displaystyle P(h_i|T)\propto P(T|h_i)P(h_i)}$

hence,

${\displaystyle y=\underset{c_j\in C}{\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{m}\mathrm{a}\mathrm{x}}\sum _{h_i\in H}P(c_j|h_i)P(h_i|T)}$

Bootstrap aggregating (bagging)

Main article: Bootstrap aggregating

Three datasets bootstrapped from an original set. Example A occurs twice in set 1 because these are chosen with replacement.

Bootstrap aggregation (bagging) involves training an ensemble on bootstrapped data sets. A bootstrapped set is created by selecting from original training data set with replacement. Thus, a bootstrap set may contain a given example zero, one, or multiple times. Ensemble members can also have limits on the features (e.g., nodes of a decision tree), to encourage exploring of diverse features. The variance of local information in the bootstrap sets and feature considerations promote diversity in the ensemble, and can strengthen the ensemble. To reduce overfitting, a member can be validated using the out-of-bag set (the examples that are not in its bootstrap set).

Inference is done by voting of predictions of ensemble members, called aggregation. It is illustrated below with an ensemble of four decision trees. The query example is classified by each tree. Because three of the four predict the positive class, the ensemble's overall classification is positive. Random forests like the one shown are a common application of bagging.

Boosting

Main article: Boosting (meta-algorithm)

Boosting involves training successively models by emphasizing training data mis-classified by previously learned models. Initially, all data (D1) has equal weight and is used to learn a base model M1. The examples mis-classified by M1 are assigned a weight greater than correctly classified examples. This boosted data (D2) is used to train a second base model M2, and so on. Inference is done by voting.

In some cases, boosting has yielded better accuracy than bagging, but tends to over-fit more. The most common implementation of boosting is Adaboost, but some newer algorithms are reported to achieve better results.

Bayesian model averaging

Bayesian model averaging (BMA) makes predictions by averaging the predictions of models weighted by their posterior probabilities given the data. BMA is known to generally give better answers than a single model, obtained, e.g., via stepwise regression, especially where very different models have nearly identical performance in the training set but may otherwise perform quite differently.

The question with any use of Bayes' theorem is the prior, i.e., the probability (perhaps subjective) that each model is the best to use for a given purpose. Conceptually, BMA can be used with any prior. R packages ensembleBMA and BMA use the prior implied by the Bayesian information criterion, (BIC), following Raftery (1995). R package BAS supports the use of the priors implied by Akaike information criterion (AIC) and other criteria over the alternative models as well as priors over the coefficients.

The difference between BIC and AIC is the strength of preference for parsimony. BIC's penalty for model complexity is ${\displaystyle \ln (n)k}$ , while AIC's is ${\displaystyle 2k}$. Large-sample asymptotic theory establishes that if there is a best model, then with increasing sample sizes, BIC is strongly consistent, i.e., will almost certainly find it, while AIC may not, because AIC may continue to place excessive posterior probability on models that are more complicated than they need to be. On the other hand, AIC and AICc are asymptotically “efficient” (i.e., minimum mean square prediction error), while BIC is not .

Haussler et al. (1994) showed that when BMA is used for classification, its expected error is at most twice the expected error of the Bayes optimal classifier. Burnham and Anderson (1998, 2002) contributed greatly to introducing a wider audience to the basic ideas of Bayesian model averaging and popularizing the methodology. The availability of software, including other free open-source packages for R beyond those mentioned above, helped make the methods accessible to a wider audience.

Bayesian model combination

Bayesian model combination (BMC) is an algorithmic correction to Bayesian model averaging (BMA). Instead of sampling each model in the ensemble individually, it samples from the space of possible ensembles (with model weights drawn randomly from a Dirichlet distribution having uniform parameters). This modification overcomes the tendency of BMA to converge toward giving all the weight to a single model. Although BMC is somewhat more computationally expensive than BMA, it tends to yield dramatically better results. BMC has been shown to be better on average (with statistical significance) than BMA and bagging.

Use of Bayes' law to compute model weights requires computing the probability of the data given each model. Typically, none of the models in the ensemble are exactly the distribution from which the training data were generated, so all of them correctly receive a value close to zero for this term. This would work well if the ensemble were big enough to sample the entire model-space, but this is rarely possible. Consequently, each pattern in the training data will cause the ensemble weight to shift toward the model in the ensemble that is closest to the distribution of the training data. It essentially reduces to an unnecessarily complex method for doing model selection.

The possible weightings for an ensemble can be visualized as lying on a simplex. At each vertex of the simplex, all of the weight is given to a single model in the ensemble. BMA converges toward the vertex that is closest to the distribution of the training data. By contrast, BMC converges toward the point where this distribution projects onto the simplex. In other words, instead of selecting the one model that is closest to the generating distribution, it seeks the combination of models that is closest to the generating distribution.

The results from BMA can often be approximated by using cross-validation to select the best model from a bucket of models. Likewise, the results from BMC may be approximated by using cross-validation to select the best ensemble combination from a random sampling of possible weightings.

Bucket of models

A "bucket of models" is an ensemble technique in which a model selection algorithm is used to choose the best model for each problem. When tested with only one problem, a bucket of models can produce no better results than the best model in the set, but when evaluated across many problems, it will typically produce much better results, on average, than any model in the set.

The most common approach used for model-selection is cross-validation selection (sometimes called a "bake-off contest"). It is described with the following pseudo-code:

For each model m in the bucket:
    Do c times: (where 'c' is some constant)
        Randomly divide the training dataset into two sets: A and B
        Train m with A
        Test m with B
Select the model that obtains the highest average score

Cross-Validation Selection can be summed up as: "try them all with the training set, and pick the one that works best".^[30]

Gating is a generalization of Cross-Validation Selection. It involves training another learning model to decide which of the models in the bucket is best-suited to solve the problem. Often, a perceptron is used for the gating model. It can be used to pick the "best" model, or it can be used to give a linear weight to the predictions from each model in the bucket.

When a bucket of models is used with a large set of problems, it may be desirable to avoid training some of the models that take a long time to train. Landmark learning is a meta-learning approach that seeks to solve this problem. It involves training only the fast (but imprecise) algorithms in the bucket, and then using the performance of these algorithms to help determine which slow (but accurate) algorithm is most likely to do best.

Stacking

Stacking (sometimes called stacked generalization) involves training a model to combine the predictions of several other learning algorithms. First, all of the other algorithms are trained using the available data, then a combiner algorithm (final estimator) is trained to make a final prediction using all the predictions of the other algorithms (base estimators) as additional inputs or using cross-validated predictions from the base estimators which can prevent overfitting. If an arbitrary combiner algorithm is used, then stacking can theoretically represent any of the ensemble techniques described in this article, although, in practice, a logistic regression model is often used as the combiner.

Stacking typically yields performance better than any single one of the trained models. It has been successfully used on both supervised learning tasks (regression, classification and distance learning ) and unsupervised learning (density estimation). It has also been used to estimate bagging's error rate. It has been reported to out-perform Bayesian model-averaging. The two top-performers in the Netflix competition utilized blending, which may be considered a form of stacking.

Voting

Voting is another form of ensembling. See e.g. Weighted majority algorithm (machine learning).

Implementations in statistics packages

• R: at least three packages offer Bayesian model averaging tools, including the BMS (an acronym for Bayesian Model Selection) package, the BAS (an acronym for Bayesian Adaptive Sampling) package, and the BMA package.
• Python: scikit-learn, a package for machine learning in Python offers packages for ensemble learning including packages for bagging, voting and averaging methods.
• MATLAB: classification ensembles are implemented in Statistics and Machine Learning Toolbox.

Ensemble learning applications

In recent years, due to growing computational power, which allows for training in large ensemble learning in a reasonable time frame, the number of ensemble learning applications has grown increasingly. Some of the applications of ensemble classifiers include:

Remote sensing

Main article: Remote sensing

Land cover mapping

Land cover mapping is one of the major applications of Earth observation satellite sensors, using remote sensing and geospatial data, to identify the materials and objects which are located on the surface of target areas. Generally, the classes of target materials include roads, buildings, rivers, lakes, and vegetation. Some different ensemble learning approaches based on artificial neural networks, kernel principal component analysis (KPCA), decision trees with boosting, random forest and automatic design of multiple classifier systems, are proposed to efficiently identify land cover objects.

Change detection

Change detection is an image analysis problem, consisting of the identification of places where the land cover has changed over time. Change detection is widely used in fields such as urban growth, forest and vegetation dynamics, land use and disaster monitoring. The earliest applications of ensemble classifiers in change detection are designed with the majority voting, Bayesian model averaging, and the maximum posterior probability. Given the growth of satellite data over time, the past decade sees more use of time series methods for continuous change detection from image stacks. One example is a Bayesian ensemble changepoint detection method called BEAST, with the software available as a package Rbeast in R, Python, and Matlab.

Computer security

Distributed denial of service

Distributed denial of service is one of the most threatening cyber-attacks that may happen to an internet service provider. By combining the output of single classifiers, ensemble classifiers reduce the total error of detecting and discriminating such attacks from legitimate flash crowds.

Malware Detection

Classification of malware codes such as computer viruses, computer worms, trojans, ransomware and spywares with the usage of machine learning techniques, is inspired by the document categorization problem. Ensemble learning systems have shown a proper efficacy in this area.

Intrusion detection

An intrusion detection system monitors computer network or computer systems to identify intruder codes like an anomaly detection process. Ensemble learning successfully aids such monitoring systems to reduce their total error.

Face recognition

Main article: Face recognition

Face recognition, which recently has become one of the most popular research areas of pattern recognition, copes with identification or verification of a person by their digital images.

Hierarchical ensembles based on Gabor Fisher classifier and independent component analysis preprocessing techniques are some of the earliest ensembles employed in this field.

Emotion recognition

Main article: Emotion recognition

While speech recognition is mainly based on deep learning because most of the industry players in this field like Google, Microsoft and IBM reveal that the core technology of their speech recognition is based on this approach, speech-based emotion recognition can also have a satisfactory performance with ensemble learning.

It is also being successfully used in facial emotion recognition.

Fraud detection

Main article: Fraud detection

Fraud detection deals with the identification of bank fraud, such as money laundering, credit card fraud and telecommunication fraud, which have vast domains of research and applications of machine learning. Because ensemble learning improves the robustness of the normal behavior modelling, it has been proposed as an efficient technique to detect such fraudulent cases and activities in banking and credit card systems.

Financial decision-making

The accuracy of prediction of business failure is a very crucial issue in financial decision-making. Therefore, different ensemble classifiers are proposed to predict financial crises and financial distress. Also, in the trade-based manipulation problem, where traders attempt to manipulate stock prices by buying and selling activities, ensemble classifiers are required to analyze the changes in the stock market data and detect suspicious symptom of stock price manipulation.

Medicine

Ensemble classifiers have been successfully applied in neuroscience, proteomics and medical diagnosis like in neuro-cognitive disorder (i.e. Alzheimer or myotonic dystrophy) detection based on MRI datasets, and cervical cytology classification.