Military Interventions As Omitted Variable Of Inversed Democratic Peace: An Empirical Evidence

The paper examines the relationship between military interventions and democratisation processes which took place in targeted states. While many researchers try to identify relationship between the regime type and countries’ war proneness, the authors of this paper put these two variables in a reversed order. To test this so-called “inversed democratic peace” thesis based on an argument that an ongoing war is likely to lead to democratisation, we focus our analysis on the US interventions in Afghanistan, Iraq, Libya, and FR Yugoslavia (Kosovo). We deploy three variables: 1) Foreign policy similarity, to determine whether the intervening actor (USA) had similar or different foreign policy goals at the beginning of interventions; 2) Political regime similarity, to indicate whether there were any deviations in the quality of political regime between the intervening state and the target country, as indicated by the democratic peace postulates; 3) military interventions (independent variable). Foreign policy score includes S score dataset developed by Curtis S. Signorino and Jeffrey M. Ritter (1999), while for the political regime quality, the authors deploy Polity IV data. Statistical analysis including Pearsonʼs correlation, logistic regression and descriptive statistics, will be presented for specific dyad level in three specifically designated models. The authors conclude that it is more likely that military interventions affect further democratisation of the targeted post-conflict societies, if observed in a short term rather than in longitudinal domain, while the foreign policy similarity (with the United States) positively correlates in cases with more successful democratisation process.


Putting Military Interventions into the Democratic Peace Matrix
Level of violence among national states within the international system is one of essential topics in IR research. Often being inducted by various factors, interstate violence has been identified and connected with various causes, such as the structure of the system (Levy 1988), national power (Morgenthau 1978), natural resources (Ross 2004), whereas differentiations within the political regime type go to the category of more specific factors (Doyle 1983;Maoz & Russet 1993;Mansfield & Snyder 1996). Nuances in political regime types formed a basis upon which liberal provenance wave of studies eroded in the late 1980s, and cast an argument that democracies were perceived as more stable societies -both in domestic and international level -than authoritarian regimes. Initially, the global gradual democratisation has triggered a more significant academic niche centred on several questions: what factors make countries to ally with each other; when and how countries decide to wage wars; how to determine country's war proneness; how violent is the international system; and, when do the countries consider their decision to intervene unilaterally. This entire set of questions has formulated the basics for establishment of the democratic peace theory (DPT). 1 Its postulates were widely accepted by the scholars ever since the idea was introduced in 1983 (Doyle 1983a;Doyle 1983b). During the last decade of the 20 th century, the linkage between war and "internal variable" -political regime type, was being investigated mostly through quantitative research studies (Maoz & Russet 1993;Meernik 1996;Peceny 1999). Among the divergent implications of the DPT idea, one of the most vivid outcomes is the vast number of available datasets, indexing data on two key variables: level of democracy/autocracy and the level of violence within the international system. 2 However, even though the literature has been enriched with significant research results, not many scholars reference the inversed causal mechanism of these two key variables through the lenses of the DPT. Knowledge on the impact of external military interventions on political regime change is still inconsistent and is initially known as inversed democratic peace. It is argued that inverse nature of democratic peace sheds the lights to an inversed relation among variables according to which coercive force, precisely military interventions, influences peace and the level of democracy within the target state. As concomitant phenomena within the IR, military interventions drew significant academic attention. Peace Research Institute Frankfurt -PRIF's data show that more than 1100 cases of the unilateral use of force were registered between 1946 and 2005 in regards to violation of the air/ terrestrial space of a sovereign country (PRIF 2019). Patricia Sullivan and Michael Koch report on 139 cases of military interventions conducted by powerful states in the same period (Sullivan & Koch 2009). 3 Not only their frequency, but the effects they leave 1) Terms that are being consistently used to mark the DPT are usually titled as "democratic peace", "liberal peace", "democratic peace idea", and "interdemocratic peace". 2) For the comprehensive analysis over the datasets utility in IR research see: Vanhanen, T. (2000)  We will examine whether there are any causal relation between US military interventions and democratization processes which took place in a selected group of states -Afghanistan, Iraq, Libya, and FR Yugoslavia/Serbia. The concept of inversed democratic peace casts an argument according to which a unilateral coercive use of force would leads to further democratization of a targeted state. To inspect the specifics of military interventions, we shall deploy a multi-case study research design, aiming to determine political regime changes as a consequence of military interventions. We chose the selected cases because they depict in the clearest way possible how the pragmatic misuse of DPT forms a ground for legitimisation of military interventions. After the brief introduction of the DPT's key arguments and development of the theoretical framework, we argue an innovative way of military interventions research -the inversed democratic peace. In line with this argumentation, the central part of our article focuses on specific democratisation effects left by the conducted military interventions, and the investigation of similarity in foreign policy decisions between the intervening state and four analysed target countries.

Theorising Inversed Democratic Peace
Without any doubt, democratic peace theory belongs to one of the most discussed theories within the IR discipline. Ever growing number of theoretical postulates brought saturated empirical evidences on the research agenda. Ever since the DPT was pitched into academic discourse, the key debate was branched out into several main versions. Having great research and explanatory potential, the DPT serves as a fruitful analytical tool to both bilateral and systemic level of violence, induced by the internal political regime characteristics. The claim that it is less likely for two countries to get involved into an interstate conflict was initial basis upon which dyadic variant of DPT was conceptualised (Doyle 1983). Dyadic variant casts its assumption according to which dyad (pair of two Security Council members, against both state and non-state targets between 1946 and 2003. countries) is less war prone as the level of democracy rises. Within the dyadic variant, two sub-branches are developed. The first group of authors advocates that any conflict among democracies is impossible to occur (Hermann & Kegley 1998;Mansfield & Snyder 2005), while the second group argues that there were some examples of such conflicts, but they are statistically marginal (Maoz & Russet 1993;Bremer 1993;Dixon 1994;Peceny 1999). According to the central hypothesis, dyads of democratic states "will not be involved in serious military conflicts in their relations" (Bremer 1993). Zeev Maoz and Bruce Russet, who supported the ideas of Bremer, corroborated findings by Simon, Gartzke, and Weart who set forth hypothesis with a positive prognostic prefix, stating that "two democratic states are more likely to form an alliance" or "the alliances that have been formed by democratic states last longer" (Simon & Gartzke 1996: 632). Another massive wave of studies favoured the normative monadic assumption, which stipulates that democratic state externalises its internal norms and values through its foreign policy action, while in a dyadic case, the potential conflict of democratic and non-democratic norms would be resolved in favour of undemocratic ones (Rummel 1998;Russet 2005).
Some considerations whether the global lasting peace could be built by spreading the liberal model of democracy with use of violent methods are still vivid (Anderson 2018;Ray & Dafoe 2018;Bell & Kuek 2018). This issue seems to question the very ethical substance of democratic peace as it "presupposes a morally ambivalent stance in the form of simultaneous advocating the idea of international peace as well as easy legitimisation of the pragmatic use of violence" (Korać & Stekić 2018: 112). Such an internal ethical DPT discrepancy paves the path towards theoretical deepening of the initial postulates.
In the previous section, we have covered some of the key debates and concerns on the DPT development and empirical research. All of these debates were being improved on the one-way (cor)relation among the two key variables. This is why inversed democratic peace can add epistemic value to the theoretical puzzle of the DPT. The central argument should be understood in a way that DPT advancement could be facilitated via inversing the causal mechanism between interventions and democratisation processes SERBIAN POLITICAL THOUGHT occurring right after the intervention took place. Much has been achieved in understanding how the political regime influences war proneness, but there are not many studies investigating this issue from the opposite view. This study intends to contribute in filling this gap.

Methodology
The methodological debates about the vital role of statistics in IR research -and especially about the "big data" usage -raise the question of analytical level of a particular concept -whether it is democracy, development, stability, or peace. Michael Coppedge states that political science concepts, as represented in IR, are extremely sensitive to quantitative analysis, and therefore advises design proportionality in research. These concepts cannot be reduced to a single indicator without losing some important part of their meaning; therefore, development is a term which includes more determinants than average wealth; stability is more than the absence of a coup attempt in a particular country, and democracy is more complex term than the mere holding of regular elections (Coppedge 1999). All of these concepts are often multidimensional because no aspect of the concept can be reduced to another (Coppedge 1999). For example, the right to vote and the competitiveness of elections are two aspects of modern democracy, but knowing one does not allow us to predict the other reliably (Coppedge 1999: 468).
Empirical part of our research is based on three main variables: military interventions (independent variable), the foreign policy similarity index -S Score (control variable), and the type of political regime as a dependent variable. Even though often discussed within the literature, military intervention is not much conceptually contested notion. Clear distinction from the humanitarian intervention, "new" types of wars such as proxy and hybrid wars, and other notions, has provided nuanced approach towards the definition of military interventions. The authors of the Major Powers Military Intervention  dataset, Patricia L. Sullivan and Michael T. Koch, designate military intervention as the official use of armed force involving the engagement of at least 500 members of the regular military force (of any kind and type) for the sake of attaining current political goals, manifested through activity directed at an external enemy (Sullivan & Koch 2009: 709). The first criterion is the use of armed force, which signifies the active use of force by members of the armed forces or the willingness to use it in the face of resistance/attack, while the official criterion relates to authorisation by legitimate political leaders has also to be met in the same time. Military intervention has to enable the accomplishment of concrete political objective, which excludes from the analytical framework the routine and tactical movements of the armed forces for joint military exercises (2009: 709). An external enemy within the context of this definition may be either a state or a non-state actor, such as terrorist organisations in the territory of another state. The political goal of intervention is operationalised as a concrete value (territory, political power, or resources) that is the subject of consideration of the intervening state's political leaders (Sullivan & Koch 2009). The political objective is the crucial element of the definition. According to Sullivan & Koch, it could be specified as maintaining foreign regime authority or its removal, policy change, acquiring or defending territory, maintaining empire, or providing social protection and order (2009: 711).
Foreign policy similarity index, the control variable in this research, is used to determine to what extent does the foreign policy of the intervening actor (state/coalition/alliance) match the goals of targeted state at the beginning of the intervention. Though academic endeavours recognise the importance of quantifying and measuring consensuses of foreign policy actions (Häge 2011;D'Orazio 2013;Häge & Fug 2016;Strüver 2016), most of efforts to operationalise indicators have largely remained partial. The most successful attempt to systematically determine similarity between two states' foreign policies was done in 1999 by Curtis S. Signorino and Jeffrey Ritter. They introduced a complex quantitative measurement they called S Score (Signorino & Ritter 1999). This measure quantifies the level of alignment between any possible pair of states within the system on a scale between 0 and 1 (Signorino & Ritter 1999: 117).
We chose to exploit S Score in our research as it includes these two key variables observed in an inversed causal order, unlike SERBIAN POLITICAL THOUGHT the classical DPT perspective. We deploy a time-series, cross-section research design to examine whether military interventions had positive effects in terms of democratisation of target countries. It is not always the case that single state acts as an intervener; on the contrary, in most cases the military alliance, or a group of states were subjects of interventions (Simon & Gartzke 1996). Though the alliances are acknowledged with status of dyad actor, our research takes sovereign country (member of an international system at a given time) as a unit of analysis. Similarity of foreign policy positions quantitatively expressed between 0 and 1.
S Score (Signorino & Ritter 1999) Source: Authors Some studies conducted at the very end of the 20th century emphasised a strong correlation between military intervention and democratisation process (Hermann & Kegley 1998;Peceny 1999). On the contrary, Mernik reported the lack of existence of correlation between the two variables (Mernik 1996) being among the first scholars who included non-US interventions as a control variable. This provided him further data validity for comparison as he conducted a probit analysis on 159 cases in which the US military interventions took place versus 27 cases in which the US was not the intervening state. Eventually, Mernik found that majority of selected cases did not experience further democratisation after the interventions took place (Mernik 1996: 400). Based on the previous findings, and with the aim to further develop the inversed DPT argument, the central hypothesis of our study states that H: ...US military interventions in FR Yugoslavia, Afghanistan, Iraq, and Libya had positive democratising effects.
To narrow the research focus, we will test two more auxiliary hypotheses. In some studies, the short term effects of the interventions had been measured (Bremer 1993;Meernik 1996). To compare their findings with the ones in selected cases, the first auxiliary hypothesis claims that

H1: …military interventions influence democratisation more effectively in short term (Model I), rather than in longitudinal domains (Models II and III).
The second argument is that control variable (S Score) significantly influences democratisation process. It advocates the positive correlation between the democratisation and S Score measurement before, and after the intervention took place, and therefore presumes that H2: …the more democratic the target state become, the more similar foreign policy indexes with the United States are.
Unlike almost univocal stance on what nurtures the definition of military interventions, the determinants of what constitutes democracy in contemporary scientific literature is contested. In DPT research, democracy is perceived not in an idealistic way, but in more operationalised level through its key sub-variables and their interrelations (Marshall, Gurr & Jaggers 2019). Initial division of political regime type was set to distinct dichotomous autocracy-democracy reality within the concrete society at a given time. Most scholars agree that the temporal unit that describes the level SERBIAN POLITICAL THOUGHT of democracy (or political regime in general) is one calendar year per case. But what does not constitute high academic consensus is the issue of adequate democracy measurement. Quantification of political regimes in IR is a tremendously complex issue and poses a threat for inconsistent measurements and findings.
The DPT sheds light on the exact terms under which the empirical evidence, variables, and indicators should be measured. In order to gain insights into the quality of the political regimes of respective states, authors shall deploy data developed within the Polity IV project. 4 It indexes basic features of the political regimes and classifies them into three categories: democratic, anocratic, and autocratic (Marshall M. et al. 2019). By quantifying key variables of the political regime, Polity IV presents findings on a twenty one scale (between -10 and +10), with all regimes having a score between -10 and -6 denoted as autocratic, anocratic between -5 and +5, while democratic ones score between 6 and 10 (Marshall M. et al. 2019). The Polity IV project covers all states and political-territorial entities with population of over 500.000, with the time span beginning in 1816. Usual temporal domain in similar studies compounds post-World War II period underlines the Cold War legacy. Most critiques regarding temporal problems are addressing the narrow scope of both years taken into analysis and cases. At the very beginning of the 21 st century the structure of international system significantly changed, and so the types and characteristics of the military interventions did change as well. Empirical studies often opt out from analysing case studies. Instead, most of them are focusing on the big data (large N studies) aiming to prove consistence within the vast number of cases which to some extent could not really be statistically compared. This study addresses several selected cases and includes Afghanistan, Iraq, Libya, and the then FR Yugoslavia (Kosovo). All of these cases are being put within the one decade temporal domain subjected to more precise analytical discourse. As this research focuses on four preselected cases, the spatial domain is limited to the territories of the countries subjected to the US-led military interventions. Studies were insofar driven inconsistently in terms of their research design. We will use descriptive statistics to demonstrate the key 4) Annual Polity scores have been plotted for each of the 167 countries currently covered by the Polity IV data series for the period 1946-2018.

Statistical models
Available datasets and other statistical tools on disposal to academics have had improved and advanced statistical studies of DPT. Comprehensive datasets in which every single case is statistically aligned to follow the occurrences on a basis of one calendar year as an analytical unit, alleviates research endeavours. Our study will follow case per year logic to determine whether there were any changes in political regime type at a given time.
To test the robustness of the linkage between the two variables, each case will be subjected to statistical analysis in three models. The first model involves the observation of not only the year in which the intervention occurred but also three years prior to it, as well as three years after the intervention (3-1-3 constellation). A seven-year period will provide insight into the short term impact of interventions towards the democratisation process (Model I). The second analytical model will compound comprehensive period from World War II (1945) up until 2014 for which the data exist. We will apply longitudinal approach with the aim to observe potential deviations within the long term impact of interventions, and the prior stability of the political regimes (Model II). It strives to determine whether the democratisation represents the result of long term endeavours regardless of the nature of their origin (internal or external). All of similar arguments and models were tested in similar studies (Bremer 1993;Mernik 1996). The third analytical model will include statistical observation of PRC only after the intervention has occurred. This model will involve the year in which intervention took place (Model III). All of the analytical parts will include both political regime change and foreign policy similarity index variables.

Results
To demonstrate the general trends of military interventionism within the international system, we present statistical findings for the selected cases based on the exported data from available SERBIAN POLITICAL THOUGHT datasets. We initially present prevalence of all the military interventions occurred after the World War II until today. US willingness to spread democracy wherever possible pledged a motion for a pragmatic misuse of the democratisation concept. Both during the Cold War as well as in the post-Cold War era, the United States has been among the prophets of democratisation spread across the globe, brining what Samuel Huntington called "the third wave of democratization" (Huntington 1991). It is believed that democracy promotion became the official US foreign policy constituent in 1984 when the National Endowment for Democracy was established. The importance of the US promotion of democracy lays behind the concept of the "US excellence" and the willingness of the USA to promote the spread of democracy. Although it is generally acknowledged that frequency of military interventions recently increased, statistics show the opposite. Ever since they are being systematically registered from the 1945, polynominal trend line corroborates their gradual decrease. Explanations for this finding are heterogeneous, stating that military interventions occurring in the last two decades are more intense, and are being placed through viral media so they become more receptive for an average citizen. Another reason could be distinction from the classic conflicts which occur on a daily basis, most of them being classified as interstate rather than unilateral coercive use of force. Source: Compiled data from S Score (Signorino & Ritter 1999) Another longitudinal parameter includes control variable of this research (S Score). To demonstrate the similarity of foreign policy orientations with the United States (respectively) as a main intervener, we present compiled and extracted data from Signorino & Ritter dataset for the selected cases, per annum (Graph 2). The first trend indicates that right after the World War II S Score was the highest in each compared country's record. The second trend shows that Libya was the only country that had negative S Score with the United States in the moment of intervention (-0.28). All others, FR Yugoslavia (0.28), Iraq (0.01), and Afghanistan (0.26) measured positive index of foreign policy alignment with the United States (Signorino & Ritter 1999). This research applies three measuring models to the selected cases. The first, Model I (3-1-3 constellation) indicates the values of political regime quality three years before and three years after the intervention, including the year of intervention (Graph 3). In a seven-year period, FR Yugoslavia (later Serbia) experienced the biggest political regime change, managing to transform itself from autocracy to democracy (from -6 to +6 Polity score). And while Iraq and Afghanistan both transformed from autocracy to closed anocracy (from -9 to -2 in Iraq, and -7 to -2 in Afghan case), Libya was the only stagnating country in terms of democratisation (remained in -6). All the four countries managed to increase their Polity values in the year of intervention, excluding Libyan stagnation.
The Model II intends to determine whether democratisation is caused as a result of long term impact of military interventions regardless of the nature of their origin (internal or external). To test this Model, we compiled data for four selected cases with the period between the 1946 and the respective year in which intervention occurred. Graph 4 demonstrates the longitudinal changes in political regimes type. From 1946, up until the year of intervention, Afghanistan and FR Yugoslavia measured general rise of democracy level, except Iraqi declining, and again, stagnating Libya. The polynominal trend lines indicate that in all cases, except Libya, political regime have been changing their types several times prior to the conducted intervention/s. The final Model demanded inclusion of control variable (S Score). We correlated foreign policy similarity index with the changes of political regime types for all the three models. Table 2 shows that in Model I the strongest correlation between the democracy rise and S Score was measured in Yugoslav case (r = .79, p < 0.01), while the moderately negative correlation was registered only in case of Afghanistan (r = -.64, p < 0.01). Model II demonstrated very strong negative correlations in Afghanistan (r = -.75, p < 0.01) and Yugoslav cases (r = -.69, p < 0.01). Unlike in Model 1, findings for Iraq in Model II have 5) Interruptions shown in the graphs are caused by the lack of data for the given year/s. According to the Polity IV Codebook, these interruptions are classified as: interruption in regime functioning, interregnum in governance, and transition periods within the political regime change.  This section presented the key findings and results for the three proposed models of researching the inversed DPT. Results indicate the justification of all the models, as presented correlations highly vary from model to model. Robustness of the applied mod-els is also demonstrated as each of them corresponds to different temporal domain of the selected cases.

Are Military Interventions an Omitted Variable in Democratic Peace Theory?
In our non-mainstream research design based on the inversed democratic peace, we employed three main variables: military interventions (independent variable), the foreign policy similarity index -S Score (control variable), and the type of political regime as a dependent variable. The results of our research indicate several groups of findings which might foster further academic debate.
First, we demonstrated that proposed descriptive statistical models might be valuable tool for further research since they indicate significant statistical variations regarding the temporal domains in which the key deviations are observed. For instance, Model I results (3-1-3 constellation) demonstrated that the highest rise of the political regime quality was measured in case of FR Yugoslavia (succeeded by Serbia), with impressive 16 pointschange on Polity scale. It was followed by Iraq and Afghanistan, both managing to decrease their autocracy levels for 7 points, while Libya was in stagnating status (Graph 3). The second model on other hand shows that political regime quality varied over the decades with ups and downs in terms of Polity score. Besides low number of statistical items (only 10-15 years for analysis), Model III was interrupted as in some cases Polity wasn't able to measure the scale of democracy/autocracy level. It is why its usage is still not fully valid in statistical studies.
The second group of important findings relates to variations in relations among the inspected variables. Before discussing the general hypothesis findings, arguments and findings of the auxiliary ones shall be firstly presented. The first auxiliary hypothesis argued that the impact of military interventions is more statistically significant when observed in the short term, rather than in longitudinal domain. This research shows that in all cases except Libya, this hypothesis was corroborated. However, when observed within the Model II, this hypothesis could not be fully confirmed, as the temporal domain of several decades implies many factors except POLITICAL THOUGHT coercive use of force. This is why H1 is only partially corroborated based on our findings. The second auxiliary hypothesis brought into correlation the S Score and democratisation process. Correlations between these two variables unequivocally confirm this hypothesis, as it shows in the most obvious way the variations between the two models. The Table 2 demonstrated that in Yugoslav and Iraqi cases, not only these correlations have varied, but their nature has changed from negative in Model I to positive (Iraq), and vice versa in Yugoslav case. Basically what this hypothesis argues is that the rise of democracy correlates with the similarity of targeted country with the United States at a given year. On a short term, Libya and FR Yugoslavia corroborated this claim, while in Model II only Iraq was the one which S Score positively correlated with its democratisation ( Table 2).
The departing general assumption of our research claimed that the US military interventions left positive democratising effects to the target countries. As it is discussed earlier, this claim could be partially accepted, as it was corroborated only in individual cases. Besides, social and political occurrences could also have significant impact to democratisation of a society. Furthermore, simple correlations among variables do not imply causality, nor do they leave a direct link between them. Statistical findings could serve as a complementary addition in understanding depth of the research phenomenon.
In their earlier endeavours to inspect military interventions, Herman and Kegley introduced the pivotal regime change model in which they explicitly marked the Iraqi case. They argue that the power vacuum emerged after the end of the Cold War, with the United States as the only remained superpower, must have been filled in with the promotion of democratic values, which in the most cases needed to be imposed by the coercive use of force (Hermann & Kegley 2008: 24). Even though military intervention in Iraq did not induce democratisation process, it triggered democratisation issue within academic debate, thus it was acknowledged as a "pivot among democratising societies" (2008: 25).
The findings of this study directly tackle classic DPT theoretical matrix. The authors of liberal provenance report theoretical inconsistence which consequently leads to misinterpretation of