Cold Warriors
U.S. and Soviet interventions in Internal Wars
Submission date: 22-06-2018 Supervisor: Abbey Steele Second reader: Sijeong Lim
By Jasper Koops 10090770
Abstract
The Cold War often seems like a history of interventions. The period of decolonization that coincided with the period to the establishment of a large number of weakened states that quickly came under immense pressure from both factions. The civil wars that many of these nation succumbed to provided both the United States and the Soviet Union with an opportunity to intervene and expand their influence. Yet the champions of liberty and equality often found themselves supporting regimes who espousing anything but these values. What factors compelled the great powers of the Cold War to intervene? Why did, or didn't, these interventions take place? This thesis consists of a large N
qualitative analysis to discover which factors influenced the decision to intervene.
Abstract 1
Keywords: Civil War, Cold War, Intervention, United States, Soviet Union, 1
Introduction 3
Literature Review 6
Theory 12
Hypotheses 13
Research Design 14
Cases: Civil Wars 14
Dependent Variable: Intervention 14
Explanatory Variables 15
Analysis 18
Wars, civil wars and interventions 18
'Global security interests' and the overall balance of power 22 OLS Analysis: The U.S. its share in global civil war interventions (%) 24 OLS Analysis: The Soviet Union its share in global civil war interventions (%) 25
Logit Analysis: Soviet Direct Interventions - Balance of Power 26 Logit Analysis: Soviet Indirect Interventions - Balance of Power 27 Logit Analysis: US Direct Interventions - Balance of Power 29 Logit Analysis: U.S. Direct Interventions - Balance of Power 30 Logit Analysis: US support - characteristics of recipient nations 32 Domnestion considerations and their influence on interventions 33
Internal politics: United States 33
OLS analysis: Direct Military Intervention, by Party 34 OLS analysis: Indirect Military Intervention, by Party 35
US Presidents 36
Logit Analysis: Military Intervention, by President 38
Internal politics: United States: 39
Logit Analysis: Military Intervention, by Soviet Leader 41
Combined models 43
Logit Analysis: US Military Intervention (any) , combined model 43 Logit Analysis: Soviet Military Intervention (any) , combined model 45
Predicting Interventions 47
Conclusion 49
Bibliography 51
Introduction
The Cold War was defined not just by the battles that were fought, but also by the battles that weren't. When the two great powers of the postwar world began their ideological struggle, their attention - and thus large parts of their military - was directed at Europe. It was the logical conclusion of the conferences of
Moscow, Yalta and Potsdam where a weakened Europe was divided among the victors. Both the Soviet Union and the United States deployed great armies in Europe, conducted countless studies and ran numerous wargames preparing for a conventional war that would never materialize. Though the nuclear
capabilities of both powers helped dissuade them from open warfare, it did not stop them from engaging in a series of proxy wars. However, most of these
conflicts would not be fought in Europe, where both the United States and Soviet Union were firmly entrenched - but in the ‘third world’. The Cold War would play out in the world of not just the Potsdam conference, but the 1955 Bandung Conference as well.
The Cold War coincided with decline of the old imperial powers and a period of rapid decolonization. Not all of the new states that emerged from this period came about peacefully nor were all of them stable. Both communism and liberalism proved to be an enticing ideology for would be revolutionaries and the new states that they founded - and the third world thus became part of the larger ideological struggle of the Cold War.
The Civil Wars that broke out during this period provided outside powers with an opportunity to exert their influence and 'flip' a country into there sphere. Interventions quickly became a valuable tool in the foreign policy of the Cold War factions. The history of the Cold War can be told as a history of
interventions and historians tend to focus on them - from the Hungarian Uprising as a symbol of the 'Iron Curtain' to the American retrenchment
following the Vietnam War and the intervention in Afghanistan as the symbol of the Soviet demise. However these interventions seem to be as inconsistent as
they are pervasive, and here too the question seems to center not just on the interventions that did occur, but also on those that did not. Communism was a staunchly anti-imperial ideology, but the Soviet Union intervened in only a select number of wars of independence and ended up being outpaced by some of its communist allies. The United States too presented themselves as a staunch opponent of colonialism, but would also intervene on behalf of colonial powers. US also styled themselves as the champions of liberty fighting against the 'Evil Empire' of the Soviet Union, yet they would topple democratic governments and prop up authoritarian dictators.
Scholars writing about interventions during the Cold War have suggested various factors that influenced the decision to intervene in civil wars, but these vary widely among authors. Studies focused either on individual actors, like the United States (Yoon, 2008; Butler, 2003) or Soviet Union (Kaw, 1989) or on all possible interveners (Findley & Teo, 2006). The impact of colonialism or power relations either were or were not taken into account. Some discounted the
importance of ideology (Bender, 1987) whereas others attached great importance to it (Butler, 2003) or argued that pragmatism was that factor that channeled ideological zeal. (Gent, 2008). Interventions were either influenced by the
worldview of individual leaders (Westad, 2007; Grow, 2008; Metz, 1984; Findley & Teo, 2006) or the result of larger processes outside the control of individuals (Gent, 2008; Kegley & Hermann, 1996; Lockyer, 2011). Studies have been conducted that included variables from a wide range of possible explanations (Yoon, 2008), moved the focus from the wars to the actors that did or did not intervene (Findley & Teo, 2006) or employed machine learning decision tree models to mimic the decision making process of the leaders involved (Kaw, 1989).
This study aims to provide a expansive analysis of Cold War interventions in civil wars, combining many of the factors proposed above into a single large N analysis. The question that this thesis seeks to answer is: What factors
influenced Cold War military interventions in civil wars?
The analysis combines multiple datasets, providing a broad range of variables on both the countries that the civil wars were fought in (e.g. location, objective of the war being fought, relative power, political leanings, recent
history of revolutions and possible colonial history) and the nations that
intervened (e.g. current leaders, type of intervention, objective of intervention, whether they had intervened in that nation before). The study will focus on the interventions of both the primary actors in the Cold War (United States and Soviet Union) but also on the former colonial powers and third world communist states that had a history of anti colonial interventions (Cuba, China). These groups form the most prolific third party interveners in civil wars. Many scholars (Westad, 2007; Hironaka, 2008; Metz 1984) argue that decolonisation had a great influence on the progression of the Cold War and its interventions, with some going as far as viewing the interventions as extensions of colonial warfare ( Akindele, 1985). Both the (former) colonial powers and 'third world' communist powers such as Cuba provide interesting alternatives to the United States and Soviet Union.
European attempts to retain / influence their (former) colonial empires were at first opposed by the United States, before their struggle was co opted into the Cold War and they started receiving support. 'Third World' communist nations like Cuba would at times surpass the Soviet Union in armed
interventions and their policy was - more than the Soviet Union - dictated by the anti imperial tenants of the communist 'eternal revolution'. Both groups thus operated both within the traditional Cold War framework and a (anti) colonial struggle.
This thesis is organized in into three chapters. The first will be a
literature review and provide a 'lay of the land' on scholarly discussion on civil wars and interventions during the cold war. The second chapter contains a both a detailed exploration of my hypothesis and the data that is used to research them. The third chapter will contain the actual analysis. The analysis will be done primarily by using logistic regressions. But following Marita Kaw (1989) her example, I will also use machine learning models to train classifiers and 'predict' interventions. I will expand on her research by using a larger dataset, and - exploiting the computational advantage that Moore's law has granted me - using more advanced 'random forest' and 'Multi-layer Perceptron' models as well.
The thesis will conclude by providing both a summary of my findings and suggestions for possible future research.
Literature Review
Foreign intervention might not bring victory, but it might prevent defeat -
though at the cost of extending the conflict and raising the level of violence. This is a argument proposed by multiple scholars. In 'Neverending Wars: The
International Community, Weak States and the Perpetuation of War' Ann Hironaka argues that the reason that civil wars are lasting longer is
proliferation of weak and unstable states following decolonisation. Because of their weakness these states are much more likely to experience civil wars - and when they do there exists a relative parity between rebels and government. Their weakness also makes them unable to prevent outside intervention, at times their governments might even welcome it.
The influence of outside intervention also extends to the way the wars are waged. Adam Lockyer argues that the balance of power among (and the overall material capabilities of) the belligerents alters the way a war is waged. Outside intervention alters these variables and can change a conflict from low intensity 'irregular warfare' to high intensity guerilla or even conventional warfare. (Lockyer, 2011, p2341-2342) A similar argument is espoused by Stathis Kalyvas and Laia Balcells. They note that it was not just material support that changed the nature of warfare, but that foreign intervention during the Cold War often included a ideological component as well. (Kalyvas & Balcells, p418-419)
The writings of Zendong and Guevarra inspired revolutionaries
worldwide, helping not just to 'arouse passionate ideological commitments' but also 'shape perceptions about the feasibility of radical change via the path of armed struggle' - they began to believe that David could beat Mcnamara. (Kalyvas & Balcells, p420) Combined, these factors helped to bring just that. Where before 1945 states regularly defeated irregular armies, they would now be defeated, or at least fought to a draw. Kalyvis and Balcells also note that as Soviet support propped up revolutionaries, the United States would soon swoop
in to support the government. Thus these irregular wars - a type of warfare that by itself tended to last a long time - were artificially extended by outside support that granted neither side victory but - as long as support was forthcoming - helped postpone defeat indefinitely. (Kalyvas & Balcells, p420-421) The Cold War has a prominent role in their argument as the intervention on behalf of one party was often followed by an counter-intervention on behalf of its opponent. As a consequence the pattern of foreign interventions during the Cold War had the effect of raising the capacity of states worldwide' leading to increased conflict around the globe. (Kalyvas & Balcells, p418)
Many scholars writing about Cold War interventions emphasize the effect that decolonization had on the proliferation of civil wars. Both Hironake and Kalyvas attribute the existence of large number of weak states to this process. Andreas Wimmer argues that not just decolonization, but the process of nation state creation itself tends to lead to ‘waves of war’. These new nations often came about after wars of independence but - in the case of former colonies - the nation lacked the institutional strength to build an inclusive nation state, instead
resorting to building a nation around the ethnic community of the new elite. This institutionalized exclusion in turn leading to a new 'wave' of civil wars.
(Wimmer, 213)
Odd Arne Westad argues that the Cold War was fought on the battlefield of a decolonizing third world - and that both opponents did so through
interventions. Both the liberalism espoused by the United States and the 'eternal revolution' of the Soviet Union were staunchly anti imperial ideologies that were favored by different groups of revolutionaries. Though both superpowers initially supported the attempts of these groups to gain independence, the Cold War ensured that this support quickly came to depend on the exact ideology of these groups. Initially the 'nativist' factions seemed to be the most successful, with Westad recalling how Nasser told a US diplomat that 'The match is between two teams, Communism and nationalism. And if you insist on playing, you are going to spoil the game for others." (Westad, 2007, p124) But the United States
certainly did insist on playing, and Westad notes that Cold War considerations tended to supercede ideological purity, a sentiment that was best epitomized by a
member of the US state department who stated that 'Batista may be a son of a bitch, but at least he is our son of a bitch'. (Westad, 2007, p171) The fear of communism that drove the United States to prop up dictators, at times even support imperial powers helped alienate large parts of the third world from them. Westad argues that the importance of the third world was such that both powers used it as a barometer for their relative success. It was the resurgence of communism during the 70's that helped convince the United States that they were 'losing' the cold war. The defeat in Afghanistan was not just a military defeat but an ideological disaster.
Westad also writes extensively about the prominent role that the
'independent branches of communism' - as adopted by China and Cuba - played in this theater. With the latter giving the Soviet Union a run for their money when it comes to anti imperialist interventions and attributing the rift with the former in part because of the Soviets alleged inactivity in the third world.
(Westad, 2007, p176-177)
Westad is not alone in emphasising the importance of the third world. Cheikh Anta Babou attributes the relative lack of success of the Soviet Union compared to China and Cuba to the legacy of Stalin his focus on Europe. With Akindele voicing similar arguments. (Babou, 2010; Akindele, 1985) Mary Dudziak linkes the initial weak position of the United States to its domestic racial tensions. (Dudziak, 2011) Steven Metz argues that the United States were well aware of the importance of this theatre, quoting Kennedy: ‘In the struggle with the Russians, Africa is the most critical area in the world’ (Metz, 1984, p518)
If a relative consensus about the both the importance of the third world and the effects of intervention exists, the same cannot be said about what factors might compel a third party to intervene in a civil war. Most of the possible
factors fall in what Huntington (1987) describes as the 'three types of interests' These are 'concrete interests' (access to resources and security concerns), 'general interests' (Human rights, democracy, promotion of the market economy in the case of the United States and promotion of communism and the opposition of imperialism in the case of the Soviet Union) and 'global security interests'; which is derived from the zero-sum bipolar competition that they perceived the
Cold War to be (Thus any move from the other had to be opposed out of pure self interest). (Huntington 1987, 39)
Research on the motivations of the United States is the most
comprehensive. Michael Butler sought to determine if US interventions followed the tenants of the 'Just war theory' and concluded that they did, but that further research was required to determine if ‘just war rhetoric is simply being used as a purposefully opaque shield for the pursuit of realpolitik’ . (Butler, 2003, p245) Stephen Gent would argue that in favor of the latter, as he concludes that decisions to intervene were based for a large part by pragmatic concerns on where they could make the most difference with a minimal investment. (Gent, 2008) His analysis focuses most on relative power, both between combatants and between combatants and potential interveners - thus echoing the observation from Kalyvas that the effects of intervention are most noticeable in conflicts were the combatants were the weakest. Charles Kegley & Margaret Hermann
emphasize the importance of ideology, arguing that intervention serves as a tool
'to use in maintaining, restoring and controlling the status quo in the community of states with which their political systems are most associated' (Kegley &
Hermann, 1996, p318)
A number of scholars examine the effect of individual leaders on the policies of intervention of their nations. Focussing on the interventions in Latin America, Michael Grow (2008) argues that rather than the reality that existed within the individual nations only mattered as how it was interpreted by the various presidents. Providing a constructivist interpretation he remarks that where one president saw a moderate leaders, another saw a communist.
Elizabeth Saunders follows this argument as well, she asserts that the factors that influenced a president to intervene, differ per president. Whereas Kennedy was more inclined to take a states domestic policies into account, Eisenhower focussed much more on the international balance of power between the U.S. and Soviet Union (Saunders, 53, 94). Other scholars provide a similar argument from the Soviet perspective. Westad (2007) notes how Stalin's lack of interest in third world was something that following leaders had to amend and gave the United States a head start. Babou (2010) emphasizes Stalin's missed
opportunity and how it took the Soviet Union until the 70's to get back on track. Both he and Westad (2007) demonstrate how this created an opportunity for China and Cuba to exert themself themselves as anti imperialist powers.
The focus on the leaders is a part of the ‘actor centric approach’ that Findley and Teo (2006) promote as the proper tool to analyze foreign
interventions. Though they recognize that the characteristics of the individual civil wars matter they argue that this ‘phenomenon-centric’ approach ignores important dynamics such as the overall geo-political environment in which these interventions take place. In their analysis they demonstrate how 'the sequence of decision by other third parties' had a statistically significant influence on
whether intervention took place. Giving an Cold War example they note that the US had little interest in Afghanistan, until the Soviet Union decided to
intervene. They conclude that a mix of a 'phenomenon-centric' and a
'actor-centric' approach would provide a more complete overview of the variables that influence interventions into civil wars. (Findley & Teo, 828- 829, 836)
Mi Yung Yoon (1997) provides an example of such a comprehensive analysis. A quantitative analysis of U.S. interventions in third world civil wars, Yoon takes into account both the 'phenomenon-centric' and a 'actor-centric' variables such as t location of the war, the presence of support from either the Soviet Union or its allies, the level of foreign investment the nation received from the US (to measure its relative importance). Yoon also included various political variables, such as whether or not the United States are in a election year,
whether or Reagan (but only Reagan) is president and the years since (or leading up to) the Vietnam War to measure the presence of the ‘Vietnam syndrome’. (Yoon 1997, 593). The diverse set of variables used in this study allows Yoon to make various inferences on how the various factors interact with each other. For example, Yoon investigates not just if US interventions were influenced by the 'Vietnam syndrome' but also if Reagan helped mitigate it. (Which his analysis show that it did) (Yoon 1997, 597).
Yoon concludes that his approach and focus on 'strategic, economic and domestic factors' was effective in predicting military interventions (either in the form of troops or indirect support) - though less successful in non military
interventions. He also provides possible future extension of the study with the inclusion of variables that measure the cause / goal of the civil war as he suspects that the type of internal war is associated with U.S. intervention. (Yoon 1997, 580, 597)
The previous authors stood out because of the various possible
explanations they provided. Marita Kaw (1989) stands out not just because she provides a quantitative analysis on Soviet Intervention but because of her
method. Her research uses a ‘decision tree classifier’ machine learning algorithm to model the decision making process of the Soviet leadership, which reaches a accuracy of 88%. She concludes that Soviet interventions are primarily defensive in nature and serve to protect the regimes of client states. U.S. involvement does not deter the Soviets from intervening, but it will cause them to limit the scope of their intervention. If the number of Soviet interventions rose over time, it did so because the number of friendly regimes did as well. (Kaw 1989, 405-413, 425)
Theory
What factors influenced Cold War military interventions in civil wars?
The studies above provided a wide range range of possible answers and ran their analysis against different datasets. It is very plausible that the internal politics of the great powers influenced their decision to intervene - just like an analysis tuned towards more ‘realpolitik’ indicators, or focussing more on ideology will return interesting results.
Previous research covered most of the hypothesis of this thesis- my analysis seeks to distinguish itself not by filling a single 'gap' but by combining previous research all into one a single analysis. I will combine their datasets and variables in order to test multiple theories at once. Thus adopting the 'mixed approach' Findley & Teo advocated, but with more data and variables at my disposal. My research will contain the political variables of Yoon and Saunders, the focus on the dynamics of the war as seen in Kalyvas, the extensive domestic variables of Wimmer & Min and Regan and - as in Findley & Teo - will focus on the actions of the other nations in the system as well. Since most of the variables apply to all nations, I can also use them to not just ascertain the presence but also explore the characteristics of 'other' third party interveners in conflicts were multiple parties intervened.
Previous research focussed not just on the variables, but on the methods used to analyze them as well. Complementing the use (logit) regressions with simulations and various classification algorithms. (Yoon, 1997; Kaw 1989; Fearon, 2004) The relatively large dataset used for this thesis - combined with computational advancements - allow me to use a number of advanced
classification algorithms and test whether the results from the analysis allow for an accurate prediction of civil war interventions during the Cold War.
The various hypotheses have been adapted from Huntington his 'three types of interest thesis', which subdivided U.S. and Soviet interests in the realms of 'concrete interests' (access to resources and security concerns), 'general
interests' (Human rights, democracy, promotion of the market economy in the case of the United States and promotion of communism and the opposition of imperialism in the case of the Soviet Union) and 'global security interests'. (Huntington 1987, 39)
Hypotheses
H1a - Interventions were dictated by 'Global Security interests' H1b – US Interventions were ‘offensive’
H1c: - Soviet Interventions were ‘defensive’
H2a - In U.S. Interventions, 'realpolitik' superseded ideological considerations H2b - Soviet Interventions were more a struggle for the third world than a struggle against the United States
H3a - The internal politics of the United States and the Soviet Union influence their decision to intervene
H3b - Republicans were more prone to intervene
H3d - Compared to Stalin, other leaders were more prone to intervene
Research Design
This study will be a large N quantitative analysis conducted on a dataset of interventions in civil wars during the cold war. This dataset is itself a
combination of various other datasets. The goals to get as many variables on not just the civil war itself, but also the countries involved and the various countries that are of interest to this study (even it they stay out of that particular conflict).
Cases: Civil Wars
I use Fearons definition for civil war, which describes them as a;
'Fight between between a central government and a organized non-state group who seeks to either take control of a government, take power in a region or use violence to change government policies. The Conflict killed at least 1000 over its course. (Fearon, 2004, p278)
The actual data on these wars comes from the dataset from Wimmer and Min. Their dataset is based on the dataset from the Correlates of War project, but was expanded to include pre-imperial kingdoms and entities. (Wimmer & Min 2006) The authors made a distinction between civil wars (using a definition similar to the one Fearon provided) and wars for independence, defining those as a; 'Fight against the political center with the aims to establish a separate state' (Wimmer and Min, 2006, p403)
I re-coded both types of wars as civil wars for my dataset, as both wars imply some sort of rebellion against a centralized state and wars for
independence were a type of conflict which is closely tied to decolonization; a process that is part of my analysis.
Dependent Variable: Intervention
For this study I use Yoon his definition for intervention, which is the: ‘Entry into a an internal war situation to support or oppose one of the factions. (Yoon, 1997, p585)
This encompasses different levels of intervention which I divide into two categories.
1. Direct intervention as defined by Pearson and Baumann (1993, p1): The movement of regular troops or forces (airborne, seaborne, shelling, etc.) of one country inside another, in the context of some political issue or dispute. 2. Indirect intervention, as defined by Yoon: Arms supply, economic aid or
deployment of military advisors without participation in actual fighting. (1997, p585)
The direct interventions are taken from the 'International military Intervention' dataset from Frederic Pearson and Robert Baumann (1993), this dataset
provides not just the intervention itself, but also provides various other variables such as whether the intervener has committed any previous interventions in the country and the 'direction' of the intervention, which I have recoded into
'government_support' and 'rebel_support' dummy variables.
Data for indirect interventions is pulled from two different datasets. The 'External Support' dataset from the Uppsala Conflict Data program is the most comprehensive, but limited by the fact that it only contains data from 1975 onwards. It accounts for many different kinds of interventions (e.g. economic, logistic, intelligence) which I've which is transcoded into a 'non-direct_support' dummie of the main dataset. The pre-1975 data is supplied by the 'International Crisis behavior' dataset. This dataset uses individual crisis as their cases, which I had to group by their respective wars. (Breacher and Wilkenfield, 1997)
Explanatory Variables
For this study I am using a wide range of explanatory variables, ranging from the characteristics of the wars themselves to the domestic conditions in the intervening countries and the overall balance of power. The ‘From Empire to Nation-State’ dataset from Andreas Wimmer and Brian Min proved to be a valuable starting point for this study. It is a country-year dataset that not only maps various economic, political, societal and geographical variables, but also lists if there is a war occurring and - if so - provides additional variables on the
war. The wars themselves come from their own dataset, which limits my effort here to filtering out all inter-state wars. Since this dataset was used for a study that placed heavy emphasis on decolonization, it includes various variables related to that process as well. (Wimmer & Min 2006).
I then combined their dataset with the 'National Material Capabilities' dataset from the Correlates of war project. Thereby adding additional variables such as the total iron and steel production, energy consumption and military personal and spending for individual nations. It also provides a 'Composite Index of National Capability' (CINC) which is based on share of the worldwide sum a country has for six of these variables (Total population, urban population, iron / steel production, energy consumption, military personnel, and military
expenditure) - thus providing a good indication of the balance of power. (Wimmer & Min, 2006; Singer, 1987, p117-131).
Measuring the effects of the domestic politics of the Cold War superpowers required adding some additional variables. To analyse the effects of individual leaders I added dummy variables for each Soviet Leader, U.S. president and the political party in charge of the United States that year. I also created dummy variable to indicate if the current year was an election year and a numerical variable that indicated the number of years leading up to / since the Vietnam war.
Having filtered out all interstate wars from Wimmer's dataset the interventions where then merged into this dataset, using the wars as foreign keys - thus ensuring that only interventions in civil wars (and wars of
independence) made it into the main dataset.
The end result is an extremely comprehensive dataset tracking more than 300 separate variables, tracking economic, political, military, industrial,
ideological geographic societal factors of the both countries in the civil war and the countries that are (or aren't) intervening in the conflict. (Full list included in the appendages). During the analysis the dataset is cloned into 2 distinct
datasets:
2. The same dataset, grouped by year (excluding the variables where this operation was impossible)
The later being used for used for analysis where the sum rather than individual interventions is important - such as graphs demonstrating the varying rates of intervention between leaders / political parties.
The merging and slicing of the various datasets is conducted with a python script using the pandas library - which allows it to mimic all of the functions of the R programming language, but with the added benefit of including all python's superior data wrangling abilities. The analysis is done with the 'statsmodels' library and the machine learning algorithms come from the 'scikit-learn' library. The scripts are all stored as jupyter notebooks which allows for quick and easy replication / editing and they, together with the source files, filtered files and the master files all have been added to a git repository which can be accessed on the following url:
Analysis
Wars, civil wars and interventions
When graphing the wars and interventions contained in the dataset the expected patterns emerge: The number of civil wars is growing and their duration is
increasing. The period of the Cold War really was a period of civil wars (Fearon, 1994; Goldsmith & He, 2008) When plotting the 10 year moving average of the number of starting wars the 'waves of war' as described in Wimmer (2012) start to emerge. As per his argument the initial wave of wars of independence is
followed by a number of interstate wars as the new nationstates attempt to incorporate foreign diaspora in their nation.
Fearon his thesis of the lengthening duration of civil wars (as
demonstrated above) complements Wimmer his focus on ethnicity. According to Fearon the increased duration of wars can be attributed not to wars of
independence or coups (both of which, ideologically motivated or not, on average end relatively quickly) but to 'sons of the soil' wars, that arise when a ethnic minority is displaced by the state supported migrants from the ethnic majority. (1994, p275) Both scholars thus tell a story of colonization and civil wars that operates outside the traditional Cold War narrative of ideological struggle. This struggle, and the interventions that accompanied it, is visible in the dataset as well.
The United States conducted far more direct interventions than the Soviet Union.
When it comes to indirect interventions the United States still take the lead, though with a much less profound difference. When the indirect support is split into the various types it becomes clear that the United States in general provide more material support than their Soviet counterparts - who excel in training but cannot match the United States in the other categories. Since both powers often offer more than a single type of support at once, the sum of their support greater than the previous comparison. (Which used the 'indirect_support' dummy.)
Various authors describe Ronald Reagan as the president who reignited the Cold War in an attempt to beat the Soviet Union and the graph above seems to reflect their argument. (Cox, 1990; Powaski, 2017; Grow, 2010). When considering the sum of direct and indirect interventions the Soviet Union and the United States go toe to toe for most of the Cold War, until the onset of the 80's. A period that
coincides with both the Reagan presidency and the Soviet intervention in Afghanistan and this thesis will show that he had a positive influence on both the number of interventions as well as the likelyhood to intervene.
As shown in the graph above, the US engaged in far more direct interventions than their Soviet counterparts. Any effects from the Soviet Afghan war is no longer discernible, though the U.S. 'Reagan era' increase is still noticeable.
The lack of Soviet interventions in the initial phase of the Cold War is explained by some scholars by the lack of interest of Stalin, only picking up in the
Khrushchev years. A regression analysis later in this thesis will demonstrate
'Global security interests' and the overall balance of power
According to Huntington on the 'global security interests' were part of a zero sum competition that in its very nature necessitated a countermove to each move the opponent made. If the war was to be decided, it would be in this arena.(Huntington, 1987, 39).
When graphing various variables concerning the balance of power the progression of the Cold War becomes easily discernible.
The graph tracking the military expenditures of both powers shows an arms race that began gradually, escalated in the 1980s and ended in the sudden collapse in the Soviet Union. As shown in the next graph the Soviet Union and United States matched their military expenditures in lockstep.
All these expenses needed to be paid for, but the Soviet Union could not keep up economically, a gap that only grew over the years. The Soviet CINC score was also inferior for a large portion of the conflict and did not vary much. During the latter part of the Cold War the Soviet Union began to stagnate. The graphs all reflect this process, and the dataset seems to reflect the various factors of the Cold War well.
The intervention scatter plots demonstrate that both the U.S. and Soviet Union behavior was very similar. When comparing the sum of interventions, by type, the regression lines confirm that both powers often intervened in the same civil wars. When considering the type of intervention it becomes clear that the Soviet Union conducted direct interventions less frequently.
The regression analysis uses both nations their share in the total
worldwide civil war interventions (that year). Both tables provide the results for both 'Direct' and 'Indirect' interventions as explanatory variables. Their
opponents share in interventions (direct and indirect) are used and their opponents CINC score, which is also a relative global percentage are used as explanatory variables.
OLS Analysis: The U.S. its share in global civil war interventions (%)
Dep variable Model 1: Share U.S. Direct U.S. Indirect Soviet Direct Share 0.669 **
(0.389)
0.215 (0.379 ) Soviet Indirect Share 0.222 ***
(0.089) 0.879*** (0.087) US Cinc Score -0.016 (0.249) -0.494 *** (0.243) const -0.0002 (0.061) 0.206 (0.059)*** N 46 46 Adj R-Squared 0.191 0.72 *p= .10, **p <.10, ***p<0.5
Looking at direct US interventions, any increase by the Soviet Union is matched by the United States. Indirect support from the Soviet Union is met with an increase as well. This could point towards a tendency from the United States to act 'offensive' with their interventions (Westad, 2007). As the relative amount of Soviet indirect increases, so will the strength of rebel groups that this aid tends
to support, a growing strength that the United States is willing to meet with direct intervention.
These results seem to support hypothesis H1b (U.S. interventions were ‘offensive’) though further analysis is required.
OLS Analysis: The Soviet Union its share in global civil war interventions (%)
Dep variable Share Soviet Direct Soviet Indirect U.S. Direct Share 0.105
(0.068)
-0.092 (0.162) U.S. Indirect Share 0.044
(0.040) 0.824 *** (0.096) SU Cinc Score -0.218 (0.422) 0.156 (1.01) const 0.0467 (0.074) -0.026 0.177 N 46 46 Adj R-Squared 0.109 0.69 *p= .10, **p <.10, ***p<0.5
Direct Soviet intervention is not generated in response to a growing U.S. share, whereas indirect support is met with indirect support. This seems to confirm the hypothesis H1c (Soviet . interventions were 'Defensive') in that the Soviet Union seems to be avoiding direct confrontation.
The following Logit models provide a mode precise analysis on both the US and Soviet their decisions on individual cases. Both direct and indirect
interventions are used as separate dependent variables. The other variables consist of indicators of relative economic and industrial power, the opportunity to 'flip' or 'save' a member of their (opponents) spheres of influences and two
dummies to control for both the presence of a war of independence and if the nation is a former colony.
Logit Analysis: Soviet Direct Interventions - Balance of Power
Model 1 Model 2 Model 3 Model 4
U.S. Direct - - -1.397 (2.421) - U.S. Indirect 2.887 *** (0.772) 3.051 *** (1.193) - 2.9647 *** (0.824) U.S. Share 7.399 *** (3.642) 5.551 (5.131) 6.798 (6.138) 6.8125 ** (3.834) U.S. Share (indirect) 1.478 2.652 -4.758 (3.761) -4.911 (3.955) 1.3507 (2.657) U.S. CINC Score -1.113 (15.625) 48.279 ** (27.336) 67.698 *** (28.09) -2.2952 (15.799) Soviet CINC Score -21.215 *** (10.23) -68.375 *** (23.388) -70.292 *** (24.927) -20.5148 *** (10.288) CINC -282.296 (212.46) -397.058 (454.48) -571.557 (460.470) -252.5567 (220.111) gdppc - 0.001 *** (0) -0.001 (0) -0.001 ** (0) -0.001 *** (0.000) Colonial_dumm y - - - 0.6438 (1.472) Non-aligned leaning West - -15.365 (9475.362) -28.651 (8.14e+06) - Non-aligned leaning East - 4.895 *** (1.359) 5.817 *** (1.448) -
Non-aligned - 6.5496 *** (1.808) -14.375 (2.08e+04) - War of Independence - -0.737 (1.029) -1.045 (1.009) -0.3464 ( 0.861) N 258 258 252 293 Pseudo R-Squared 0.328 0.61 0.49 0.33 *p= .10 **p <.10 ***p<0.5
In this analysis to the Soviet Union seems to persists in their policy of not meeting direct interventions with direct interventions. Indirect interventions from the U.S. act as incentive for Soviet intervention. Neither the presence of a war of independence nor if the target is a former colony was statistically
significant. The Soviets do not seem inclined to intervene in 'West Leaning' nations, but will rush to the defence of 'East leaning' nations.
Logit Analysis: Soviet Indirect Interventions - Balance of Power
Model 1 Model 2 Model 3 Model 4
U.S. Direct - - 0.7975 (0.575) 1.2565 *** (0.309) U.S. Indirect 1.0982 *** ( 0.297) 1.1637 *** (0.314) - 2.1818 ( 2.028) U.S. Share 2.9495 ( 1.955) 1.9250 ( 2.097) 1.8159 (2.395) -0.0652 1.249 U.S. Share (indirect) -0.1170 (1.221) -0.8237 (1.309) 0.448 (1.452) 17.787*** ( 5.992) U.S. CINC Score 18.7512 *** ( 5.74) 19.5134*** ( 6.151) 7.015 (8.27) -23.697 *** (4.511)
Soviet CINC Score -23.2981 *** (4.314) -25.3353*** (4.799) -16.201 *** (5.714) -3.3603 (11.320) CINC -4.9414 (11.127) -2.8918 (11.849) -9.3689 (18.024) -0.0001 (0.000) gdppc -0.0002 0.000 -5.372e-05 (0.000) 3.274e-05 (0) 1.5911 (0.619) Colonial_dumm y - - - 1.5814 *** ( 0.62) Non-aligned leaning West - 0.8434 (0.709) 0.351 (0.730) Non-aligned leaning East - 0.5797 (0.424) 0.963 *** (0.419) - Non-aligned - 3.3371 *** ( 1.105) 3.194 *** (1.11) - War of Independence - 0.2175 (0.358) 0.237 (0.385) 0.3013 ( 0.348) N 293 293 272 293 Pseudo R-Squared 0.164 0.21 0.14 0.18 *p= .10 **p <.10 ***p<0.5
The analysis of indirect Soviet interventions shows a number of interesting deviations. In the case of a war of former colony the Soviets are prepare to provide indirect support, these likely involve attempts to retake these colonies, or reflect that many of these former colonies became ideological battlegrounds, as was the case in Vietnam. Non aligned nations appear to be tempting targets, which is one of the few 'offensive' actions contained in this analysis. Based on the results of the analysis H1c seems very plausible.
Logit Analysis: US Direct Interventions - Balance of Power
Model 1 Model 2 Model 3
Soviet. Direct - - 0.0614 (1.24) Soviet Indirect 0.8644 *** (0.419) 0.6842 (0.534) - Soviet Share -8.3535 (6.101) 0.4527 (9.214) 8.2777 (11.689) Soviet Share (indirect) 6.7611 *** (2.095) 5.4637 *** (2.635) 6.6516 *** (3.869) U.S. CINC Score 1.9337
(9.183)
-8.3802 (12.397)
-37.2350 *** (21.436)
Soviet CINC Score -19.8517 *** (6.673) -17.2267 ** (8.916) -0.6244 (13.482) CINC 24.3073 (15.495) 103.0596 *** (31.553) 85.2556 *** (27.155) gdppc -0.0004 *** (0.000) -0.0002 (0.000) 5.189e-05 (0.000) Non-aligned leaning West - 3.5862 *** (0.768) 3.9262 *** (0.808) Non-aligned leaning East - 0.2622 (0.849) 0.9471 (0.993) Non-aligned - 4.5908 *** (1.331) 3.5383 *** (1.671) War of Independence - -3.6931 *** (1.546) -1.9320 (1.292) N 272 272 252 Pseudo R-Squared 0.189 0.421 0.253 *p= .10 **p <.10 ***p<0.5
When it comes to direct support, the United States are not deterred by the presence of Soviet support. In Fact, it seems to increase the chances of it happening, as seen in model 1. The 'Soviet (indirect) Share' variable, which is more abstract and reflects the balance of power in the world, also has a
significant positive effect. Growing support for communist groups is the strongest indicator that the United States will act more forcefully and move against it.
Logit Analysis: U.S. Direct Interventions - Balance of Power
Model 1 Model 2 Model 3
Soviet. Direct - - 4.4520*** (1.396) Soviet Indirect 1.1273 *** (0.304) 1.1566 *** (0.326) - Soviet Share -8.9787 *** (4.566) -9.1810 ** (4.723) -14.5447 *** (5.671) Soviet Share (indirect) 2.5491 (1.664) 2.7417 (1.777) -0.3642 (2.751)
U.S. CINC Score 10.0822 (6.682)
11.7636 ** (6.936)
(9.4468 (10.788) Soviet CINC Score -9.8177 ***
(4.663) -11.5705 *** (4.997) -5.4614 (6.431) CINC 9.4238 (13.114) 22.1631 (15.423) 22.2543 (15.278) gdppc -0.0003 *** (0.000) -0.0004 *** (0.000) -0.0002 (0.000) Non-aligned leaning West - -0.6775 (0.736) -0.7294 (0.821)
Non-aligned leaning East - 1.3631 *** (0.448) 0.6556 (0.496) Non-aligned - -0.2793 (0.709) -1.5265 (1.197) War of Independence - -0.6601 *** (0.365) -0.2602 (0.378) N 293 292 258 Pseudo R-Squared 0.167 0.20 0.152 *p= .10 **p <.10 ***p<0.5
The presence of any form of Soviet intervention increases the change of U.S indirect intervention. An increase of worldwide Soviet direct interventions seems to dapen that effect, but a possible explanation could be that resources are
transferred to indirect interventions in those conflicts, especially considering the reponsis to this variable in the previous analysis. It is also highly noteworthy that the United States are more affected by the change to 'flip' an potential Soviet ally than to defend their sphere, as 'non-aligned leaning West' is not statistically significant, whilst its east-leaning counterpart is.
To get a better sense of the U.S. its motivations an additional analysis is run against all cases where the United States intervened in a civil war, with the 'objective: support government' as the dependent variable. Various other
variables are used to gauge the characteristics of the governments that receive support - with and without the presence of any Soviet interference.
Logit Analysis: US support - characteristics of recipient nations
Model 1 Model 2 Model 3 Model 4
Democracy -1.2238 *** (0.509) -0.9764 ** (0.520) -1.6504 *** (0.636) -1.9463 *** (0.619) Autocracy -2.0455*** (0.274) -1.3866 *** (0.328) -1.7428 *** (0.381) -2.3733 *** (0.324) Anocracy -2.4423 *** (0.521) -2.0373 *** (0.534) -2.4696 *** ( 0.592) -2.8385 *** (0.577) Soviet supported Rebels - -1.1804 *** (0.383) -1.0769 *** (0.390) - Recent Coup - - 0.2178 (0.464) 0.0416 (0.459) Reagan Presidency - - 0.8674 ** (0.451) 1.1353 *** (0.444) N 225 225 225 225 Pseudo R-Squared 0.01 0.05 0.07 0.03 *p= .10 **p <.10 ***p<0.5
The United STates normally seem to shy away from supporting autocracies, but the presence of Soviet supported rebels makes authoritarian governments much more palpable. This effect is increased even further in the case of a Reagan presidency (who, according to the model already has less scruples for supporting non-democratic regimes.) The preference of acting against the Soviet interests, instead of supporting democracy, further supports both H1a (Interventions were dictated by 'Global Security interests) and H1b.
Domnestion considerations and their influence on interventions
The time spent in office varies per leader, with some presidents finishing only 1 term (and others none at all). This difference becomes even more jarring when considering the Soviet leaders - with Brezhnev seemingly refusing to die and his successors refusing to live. The analysis accounts for this fact by providing 'normalized' scores, which divide the values by the number of years in office.Internal politics: United States
A direct intervention was the most dramatic step a nation could take in a foreign civil war, and the Republicans certainly didn't shy away from it. Both their sum and normalized show that Republican Presidents conducted twice the numbers of interventions than their Democrat counterparts.
The difference becomes less profound when measuring indirect
interventions, with the difference in the yearly average interventions having shrunk to 1. Hypothesis H3B suddenly seems very plausible.
However, these interventions are also a question of opportunity: before any president can intervene there needs to be a war to intervene in. The following linear regression analysis accounts for this by using a 'number of Civil Wars' control variable, which is the summed total of all occuring Civil Wars that year. Model 1 also uses an 'number of Soviet Interventions' control variable as
interventions could be triggered not just by civil wars, but by civil wars the Soviet Union has intervened in.
OLS analysis: Direct Military Intervention, by Party
Model 1 Model 2 Democrats 0.200 (0.211) 0.0204 (0.198) Republicans 0.339 (0.262) 0.3529 (0.271)
Soviet Direct Interventions (Sum)
0.798*** (0,395)
-
Number of Civil Wars (Sum) -0.009 (0.021) 0.0177 (0.017) Constant 0.539 ** (0,298) 0.3733 (0.297) Adj R Squared 0.08 0.02 N 46 46 *p= .10, **p <.10, ***p<0.5
The analysis shows no effect for the political party variables, with only the number of Soviet Interventions being significant, as is to be expected.
OLS analysis: Indirect Military Intervention, by Party
Model 1 Model 2 Democrats 1.043*** (0.358) 0.7023 *** (0.340) Republicans 0.279 (0.446) 0.3054 (0.467) Soviet Direct Interventions
(Sum)
1.509*** (0,671)
-
Number of Civil Wars (Sum) -0.076*** (0.036) 0.1264 *** (0.030) Constant 1.322 *** (0,508) 1.0077 ** (0.510) Adj R Squared 0.34 0.27 N 46 46 *p= .10, **p <.10, ***p<0.5
When analyzing indirect interventions the situation becomes very different. Both control variables are significant (reinforcing the general trend of number of wars and Soviet interventions being highly influential on U.S. interventions), but so is the 'Democrat' dummy variable. Where's the graphs seemed to suggest that Republicans were more prone to intervene than democrats - with a smaller difference for indirect interventions - the results of the analysis state the
opposite. There is no significant influence on direct interventions, and in indirect interventions the Democrats prolific interveners.
The analysis shows now support for hypothesis H3B, (Republicans are more likely to intervene), whilst providing tentative support for hypothesis H3A. (The internal politics of the United States and the Soviet Union influence their decision to intervene)
The next analysis will focus on individual presidents and other political variables that will help further test this hypothesis.
US Presidents
Ronald Reagan, features prominently in the graph above, but when looking at normalized values it seems that it's Nixon who is the ultimate 'cold warrior'. The graphs for both Kennedy and Johnson seeom to explain part of the results of previous analysis, as both democrats had quite a number of interventions to their name.
The graphs for indirect interventions show that it may not necessarily be
interventions that presidents shy away from, but direct military involvement - as presidents who were underrepresented in the previous graph are shown to be quite willing to intervene indirectly. This possible phenomenon will be looked at more closely in the next analysis. This logit analysis contains not just dummy variables for each president, but also a number of other political variables. Election years are taken into account, as they may prompt a president to 'look tough' or instead prevent intervention as these could prove quite unpopular. On particularly unpopular intervention has gotten its own dummy variable. The 'Vietnam_dummy' is coded '1' for all years past the end of direct U.S.
involvement in the Vietnam war (1975) to analyse if the 'Vietnam syndrome' (Yoon, 1997, p589-590) is statistically significant in my analysis. The 'US_gdppc'
is also taken into account, though it has a strong political impact as well, as one unincluded president will attest. Truman is used as the 'base' for this analysis.
Logit Analysis: Military Intervention, by President
Model 1 Model 2 Eisenhower -0.429 (0.400) 0.291 (0.504) Kennedy 0.602 (0.648) 1.512 *** (0.757) Johnson 0.642 (0.433) 1.626 *** (0.603) Nixon 0.931 ** (0.476) 1.999*** (0.660) Ford 0.841 (0.783) 2.071 *** (0.947) Carter 1.524 (1.012) 2.761 *** (1.148) Reagan 2.436 *** (1.099) 3.024 *** (1.145) Bush 2.505 *** (1.162) 3.409 *** (1.240) US_gdppc - -0.002 *** (7.59e-05) Vietnam -0.6344 (0.856) -0.760 (0.863) Soviet Intervention 2.18 *** (0.227) 2.245 *** (0.229)
Presidential election year 0.165 (0.255) 0.112 (0.257) n_civil_wars -0.071 *** (0.017) 0.01 (0.037) N 562 Pseudo-R squared 0.19 0.20 *p= .10, **p <.10, ***p<0.5
Both models show that individual presidents influenced the decision to intervene to varying degrees. When taking the 'gdppc' into account almost all all presidents had a statistically relevant impact on the decision to intervene. Both the
'Vietnam'' and the 'Presidential election year' dummy variables did not score as significant. This model is flawed in the sense that it ignores virtually all factors expect domestic politics, whereas interventions were very much influenced by outside factors. At the end of the thesis I will demonstrate a model that includes variables from all chapters. The results of the analysis above do seem to
reinforce hypothesis H3a, but an analysis of the Soviet side is required before it can be confirmed.
Internal politics: United States:
The Soviet side is interesting in the sense that some of its leaders were in office for quite a time and others served for only a few months. This means that an individual leader could determine the Soviet policy for decades, or barely at all. With the Soviet side the 'normalized' scores thus become much more important, as Andropov his contribution is impossible to assess otherwise.
Khrushchev his interventions were not coded as direct in the source material, hence his disappearance from the graph. When normalized, both Andropov and
Chernenko seem more intervention prone than Brezhnev, who in much of the literature is regarded as quite passive (Westad, 2007; Cox, 1990). Stalin (who is normalized only for the years that he appears in the dataset) reigned during the
critical, early years of decolonization and barely intervened, but he also had to
rebuild a wartorn country and focus on the critical events in Europe.
Brezhnev his huge spike of indirect interventions is as much to his longevity as to his assertiveness. As the graph for the normalized values does show that Gorbachev, generally regarded as one of the men who ended the Cold War, was one of the most active Soviet leaders when it came to intervening in the wars primary battlefields.
The logit analysis of the Soviet leaders does away with the 'presidential election' and 'Vietnam' dummy variables as they have no equivalent in the Soviet context. Afghanistan might have been the 'Vietnam' of the Soviet Union, but since the Soviet retreat was so close to the end of the Soviet Union the variable would hold no value in this analysis ('Gorbachev' would be synonymous with it)
Logit Analysis: Military Intervention, by Soviet Leader
Khrushchev -0.582 (0.386) -0.01 (0.548) Brezhnev 0.728 ** (0.418) 1.799 *** (0.852) Andropov 0.461 (0.760) 1.303 (0.964) Chernenko 0.764 (0.952) 1.378 (1.05) Gorbachev 1.645 *** (0.744) 2.259 *** (0.870)
Soviet gdppc - -0.0003 (0) US Intervention 2.308 *** (0.224) 2.332 *** (0.226) n_civil_wars -0.0953 *** (0.017) -0.05 (0.035) N 562 Pseudo-R squared 0.248 0.251
As with the U.S. model GDPPC is statistically relevant, though in the Soviet case it does not change the number of Statistically relevant leaders. What's noteworthy is not that, compared to Stalin, Gorbachev and Brezhnev are more likely to intervene but that Khrushchev wasn't.
Combined models
Logit Analysis: US Military Intervention (any) , combined model
Model 1 Model 2
Soviet Intervention (any) 2.0500 *** 0.274
2.0118 *** 0.267 Soviet Direct Share -3.5245
4.058
-1.4898 3.781
Soviet Indirect Share -0.2237 1.232
-0.5736 0.722
Soviet CINC score -2.5875 5.777
-
U.S. CINC score -0.7862 3.526
-
Non_aligned leaning west 0.5144 0.551
-0.0083 0.451
Non_aligned leaning east 0.9663 0.718
1.2832 *** 0.446 Intervention from (former)
Colonial Power
-1.0458 ** 0.566
-
Third World Communist Intervention 0.5785 0.753 - War of Independence -0.7504 *** 0.304 - CINC score 9.7782 7.073 5.7130 6.476 gdppc -0.0003 *** 0.000 -0.0003 *** 0.000 Democracy -0.5176 0.355 -0.6097 ** 0.346
Autocracy -0.2535 0.284 -0.4327 0.272 Recent Coup 0.4784 0.305 0.5026 ** 0.295 Ronald Reagan President 0.0271
0.374 - Republican president (Compared to Democrat) - -0.0203 0.312 US GDPPC 3.289e-05 5.13e-05 -8.442e-06 2.2e-05 Vietnam Dummy -0.1989 0.640 -
Presidential election year -0.0513 0.271 -0.0837 0.259 Pseudo R Squared 0.183 0.1618 N 426 426 *p= .10, **p <.10 ***p<0.5
The findings of the combined model support both H1a and H1b. U.S behavior is influenced greatly by any presence of the Soviet Union, which the U.S. seem to seek out. Like in previous models, the 'Non_aligned_leaning_East' variable has a strong positive association with interventions.
Unlike the last time it was tested, the H2 hypothesis was not supported in this analysis. Democracy remains a stable deterrent irregardless of the presence of variables measuring Soviet intervention have been removed. (In which case it actually drops to 0.56).
Logit Analysis: Soviet Military Intervention (any) , combined model Model 1 Model 2 US Intervention (any) 1.4929 *** (0.589) 1.8928 *** (0.266) US Direct Share -1.2201 (2.240) -1.1273 (1.555) US Indirect Share 2.7210 *** (1.249) 1.5817 ** (0.890)
Soviet CINC score -1.0069 (6.435)
-4.9833 ( 6.116) U.S. CINC score -1.3987
(5.965)
-4.1207 (3.488)
Non_aligned leaning west -0.7716 (0.751)
-0.7099 (0.630) Non_aligned leaning east 1.4406 ***
(0.605)
0.1957 (0.448) Intervention from Colonial
Power
0.8498 (0.551)
-
Third World Communist Intervention -1.4645 *** (0.665) - (Former) Colony_dummy - 1.3970 *** (0.570) War of Independence 0.3081 (0.393) -0.1071 (0.314) CINC score -22.3984 (16.566) 3.9621 (6.489) gdppc 0.0002 (0.000) -0.0003 *** (0.000) Democracy -0.1886 (0.528) 0.2063 (0.408)
Autocracy 0.2052 (0.384) 0.3921 (0.306) Recent Coup -0.4126 (0.425) -0.3290 (0.336) SU GDPPC -0.0003 *** (0.000) -0.0002 *** (9.43e-05) Pseudo R Squared 0.113 0.1807 N 279 426 *p= .10, **p <.10 ***p<0.5
H2b receives some support with the 'Colony' dummy testing as statistically significant. H1c seems to be supported by the combined model - their opposition to U.S. interventions seems to consist mainly of indirect interventions. The Soviet will defend their sphere of influence, as 'East Aligned' again gets a
significant result. It's interesting to see that 'other' communist support acts as a deterrent. It could be that the Soviets are satisfied to let others 'cut their meat'. It could also be a reflection of the larger intra-communist world struggle - future research could expand on this.
Predicting Interventions
The variables from both combined models are fitted to various Machine Learning algorithms, to determine its effectiveness in predicting interventions.
To prevent overfitting, the data is split into a 'training' test and a 'testing set. The exact cases are randomly selected, with 75% going to the 'training' dataset and the remainder being used for testing.
The algorithm is than trained against the training data, but scored against the 'testing' data. The a score signifying the percentage of correct predictions is added to a list. As the initial split is made at random, the results will inevitably vary per run. To mitigate this the whole process is repeated 500 times. The average of the resulting list of 500 unique predictions is then used as the value for the table below.
Classifiers: Predicting Soviet Interventions
Algorithm Score (Model 1) Score (Model 2)
Logistic model 70.98% 79.59%
Decision Tree Classifier 78.33% 79.95%
Random Forest Classifier 79.78% 81.92%
Support Vector Machine Classifier
70.12% 70.60%
Multi Layer Perceptron Classifier
Classifiers: Predicting U.S. Interventions
Algorithm Score (Model 1) Score (Model 2)
Logistic model 76.53% 75.19%
Decision Tree Classifier 76.66% 76.63%
Random Forest Classifier 81.21% 80.56%
Support Vector Machine Classifier
59.7% 60.56%
Multi Layer Perceptron Classifier
61.12% 59.83%
Conclusion
What factors influenced Cold War military interventions in civil wars? The answer seems to lie somewhere in the middle between Huntington and Findley & Teo. As the analysis have shown 'Global security interests' seem to have been an important motivator for U.S. interventions.
H1b (US Interventions were ‘offensive’) has been confirmed over repeated analysis, containing different variables. The U.S. repeatedly sought out members of the Soviet sphere of influence, and the presence of often Soviet support acted as an encouragement.
Indicators for H1c: - (Soviet Interventions were ‘defensive’) have been confirmed multiple times as well - the most 'offensive' action regarding other nations seems to be the repeated positive correlation between Soviet
Interventions and 'non-aligned' governments. This could be indicative of an attempts to 'flip' them, or an indicator for anti-imperial actions.
In the realm of what Huntington termed 'general interests' the zeal that the U.S at times displayed in their pursuit of the Soviet Union seem to confirm H2a (In U.S. Interventions, 'realpolitik' superseded ideological considerations ). When considering the presence of rebels with Soviet support, the disdain for non-democratic governments that the U.S has by default virtually
H2b (Soviet Interventions were more a struggle for the third world than a struggle against the United States) demands further research. Some indicators came out in support, especially their preference for (former) colonial nations and
non-aligned nations. But 'war of national independence' dummy variables came out inconclusive. It is clear that Soviet policy did operate according to different priorities, but other than their defensive character this study has been unable to make as significant an estimation as to what they might be as it has done with the U.S.
'Concrete interests', as indicated by the CINC, and GDPPC values of the target nations did not get any conclusive results. Future studies could include more geographic indicators account for more abstract considerations such as a countries desirable strategic position.
Thus my analysis seems to confirm most of the existing literature. The models created for this study also proved to be relatively accurate predictors for both Soviet and U.S. interventions.
Bibliography
Akindele, R. A. "Africa and the Great Powers, with Particular Reference to the United States, the Soviet Union and China." Africa Spectrum (1985): 125-151.
Babou, C.A., 2010. Decolonization or national liberation: Debating the end of British colonial rule in Africa. The ANNALS of the American Academy of
Political and Social Science, 632(1), pp.41-54.
Bender, G.J., 1987. The Eagle and the Bear in Angola. The ANNALS of the American Academy of Political and Social Science, 489 (1), 123–132. Breacher, M. and Wilkenfeld, J., 1997. A Study of Crisis.
Butler, M.J., 2005. Elephants of a Feather? The Role of ‘Justice’ in Canadian and American Cold War Military Interventions. Canadian Journal of Political Science/Revue canadienne de science politique, 38 (01).
Butler, M.J., 2003. U.S. Military Intervention in Crisis, 1945-1994. Journal of Conflict Resolution, 47 (2), 226–248.
Cox, M., 1990. Whatever happened to the ‘Second’ Cold War? Soviet—American relations: 1980–1988. Review of International Studies, 16 (02), 155.
Dallin, A. and Lapidus, G.W., 1987. Reagan and the Russians: American policy toward the Soviet Union. publisher not identified.
democracy. Princeton University Press.
Fearon, J.D., 2004. Why Do Some Civil Wars Last So Much Longer than Others? Journal of Peace Research, 41 (3), 275–301.
Findley, M.G. and Teo, T.K., 2006. Rethinking Third-Party Interventions into Civil Wars: An Actor-Centric Approach. The Journal of Politics, 68 (4), 828–837.
Gent, S.E., 2008. Going in When it Counts: Military Intervention and the Outcome of Civil Conflicts*. International Studies Quarterly, 52 (4), 713–735.
Grow, M., 2008. US presidents and Latin American interventions pursuing regime change in the Cold War. University Press of Kansas.
Hasan, Z., 1984. New Thrust of Imperialism in South Asia. Social Scientist, 12 (11), 34.
Hironaka, A., 2008. Neverending wars: the international community, weak states, and the perpetuation of civil war. Harvard University Press.
Huntington, S.P., 2000. Patterns of Response. Culture and Politics, 348–357. Kalyvas, S.N. and Balcells, L., 2010. International System and Technologies of
Rebellion: How the End of the Cold War Shaped Internal Conflict. American Political Science Review, 104 (03), 415–429.
Kaw, M., 1989. Predicting Soviet Military Intervention. Journal of Conflict Resolution, 33 (3), 402–429.
Kegley, C.W. and Hermann, M.G., 1996. How Democracies Use Intervention: A Neglected Dimension in Studies of the Democratic Peace. Journal of Peace Research, 33 (3), 309–322.
Lockyer, A., 2011. Foreign intervention and warfare in civil wars. Review of International Studies, 37 (05), 2337–2364.
Metz, S., 1984. American Attitudes Toward Decolonization in Africa. Political Science Quarterly, 99 (3), 515.
Powaski, R.E., 2017. Ronald Reagan, George Shultz, and Caspar Weinberger: Winding Down the Cold War, 1984–1988. In American Presidential
Statecraft (pp. 175-223). Palgrave Macmillan, Cham.
Pearson, F.S. and Baumann, R.A., 1993. International Military Intervention, 1946-1988. ICPSR Data Holdings.
Regan, P.M. and Meachum, M.S., 2013. Data on interventions during periods of political instability. Journal of Peace Research, 51 (1), 127–135.
Saunders, E.N., 2014. Leaders at war - how presidents shape military interventions. Cornell University Press.
Singer, J.D., 1988. Reconstructing the correlates of war dataset on material capabilities of states, 1816–1985. International Interactions, 14 (2), 115–132.
Westad, O.A., 2007. The global Cold War: third world interventions and the making of our times. Cambridge University Press.
Wimmer, A. and Min, B., 2006. From Empire to Nation-State: Explaining Wars in the Modern World, 1816–2001. American Sociological Review, 71 (6), 867–897.
Wimmer, A. and Min, B., 2009. The Location and Purpose of Wars Around the World: A New Global Dataset, 1816–2001. International Interactions, 35 (4), 390–417.