How do Dictatorships end Civil War?

Table of Contents

Abstract

1. Introduction

2. Previous Literature

3. How Regimes impact Conflict Termination
3.1 Nonpersonalist Civilian Regime
3.2 Personalist Civilian Regime
3.3 Personalist Military Regime
3.4 Nonpersonalist Military Regime

4. Data and Research Design
4.1 Dependent Variable
4.2 Independent Variable
4.3 Control Variables
4.4 Model

5. Results

6. Discussion and Conclusion

References
List of Figures
List of Tables
Variable Coding
Appendix
A. Regime Types
B. Manually changed Regime Types

Abstract

This study investigates whether there is a significant impact of different authoritarian regimes, namely personalist and nonpersonalist civilian autocracies as well as their respective military counterparts, on the termination of civil war. The paper suggests that in civil wars those regimes that function on a civilian basis have an advantage in co-opting their opponents to reach a settlement of the conflict. Furthermore, military leaders are expected to be less reluctant towards the use of force than civilian ones which should make them more likely to win civil wars. The same applies to personalist dictators who are hypothesised to view the state as their own and are inclined to forcefully defend their achievements. Several multinomial logit models are tested for the time period between 1946 and 2010. The average predicted probabilities show that personalist civilian regimes indeed are more likely to reach a settlement than other types. Nonpersonalist civilian dictatorships are correctly hypothesised to least likely win an intra-state war, whereas personalist military leaders do so with a comparatively high probability.

1. Introduction

Research seems to agree on the fact that civil wars rarely end in decisive outcomes – that is victory by one of the warring parties (Fortna, 2009; Licklider, 1995). Rather, intra-state conflicts nowadays are more likely to lead to negotiated settlements than to any other outcome (Human Security Report Project, 2008). But does this relationship hold for civil war in all forms of autocratic regimes? Are certain forms of dictatorships^[1] more likely to end civil wars in a specific outcome of intra-state conflict?

The differentiation among authoritarian regimes has recently been introduced to many fields of research. It has been under investigation how different authoritarian regimes have varying influence on civil war onset (Fjelde, 2010) and international conflict initiation (Peceny & Butler, 2004; Weeks, 2012). In the research about civil war termination, however, the group of authoritarian regimes has continuously been viewed as a homogenous entity (see DeRouen & Sobek, 2004) or has only been differentiated by considering a regime’s level of democracy according to Polity IV (see Lyall & Wilson III, 2009). A possible heterogeneous impact of characteristics of a dictator on the outcome of intra-state conflict has been ignored so far. This paper aims at investigating the influence of an authoritarian regime’s type on the outcome of civil war and distinguishes between four different types of dictatorships which are determined along two different dimensions – personalist versus nonpersonalist and military versus civilian. Furthermore, autocratic regimes that do not fall into any of these categories are assigned to an additional category called other. The dependent variable is threefold: possible outcomes are government win, rebel win and settlement where both parties agreed on not fighting any longer through either signing a peace agreement, reaching a ceasefire or letting the war pass to low activity as tacit agreement.

This paper suggests that in civil wars those regimes that function on a civilian basis have an advantage in co-opting their opponents to reach a settlement of the conflict. A second point that plays an important role is a regime’s position concerning the use of weapons. Civilian leaders are expected to be more reluctant towards the use of force than military ones. Personalist dictators furthermore view the state as their own and are willing to act strongly against rebels in order to defend their own achievements.

To investigate these theories, the text makes use of data from the Uppsala Conflict Data Program (UCDP) Conflict Termination dataset (Kreutz, 2010) classifying civil wars according to the type of conflict termination they have experienced. A second source is the dataset by Geddes, Wright, and Frantz (2014), which differentiates between several kinds of authoritarian regimes. In the first instance, a multinomial logit model is used. A second step adds the calculations of average predicted probabilities using the observed-value approach.

The analysis does not reveal any significant relationship between authoritarian regime type and the form of conflict termination. Only some of the control variables show significant impact on conflict termination. The average predicted probabilities, however, show that a number of the sub-conjectures are met by the data. Personalist civilian regimes indeed seem more likely to negotiate settlement than other types. Party-based autocracies are correctly hypothesised to least likely win a civil war, whereas personalist military leaders fare – as it is expected – comparatively well.

The paper proceeds as follows: the next section introduces existing literature about civil war termination; the subsequent part provides the development of the theoretical argument followed by a description of the data and the conception of a research design; and the final section presents the results.

2. Previous Literature

Civil war has many components that recently received the attention of political science research. With the surge of datasets gathering information about the outbreak of civil war, its dynamics like recruitment of fighters and violence against civilians, duration, termination of conflict, as well as duration of the reached outcome, many new possibilities for researchers have opened. That is clearly visible in the increase of empirically led papers investigating the field of civil war. Concentrating on civil war termination, the following section gives an overview about the literature examining how conflicts end.

One early qualitative piece about conflict termination is delivered by Coser (1961) who distinguishes between absolute and institutionalized conflicts. While the former is a zero-sum game where one of the two warring sides has to be completely defeated for the war to end, the latter is subject to certain regulations facilitating the conclusion of a negotiated settlement or the determination of the victor without fighting until the last man.

Choosing a psychological approach, Bar-Tal, Kruglanski, and Klar (1989) state that for a conflict to be fully terminated, all parties involved have to finish the conflict mentally. If that is not the case, the simple abandonment of combat operations by both parties, the signing of a peace agreement or even a victory by one side does not suffice to actually end the conflict that the authors describe as ‘content of knowledge’. Conflict termination, thus, can only be reached through conflict resolution where a change in the conflict beliefs occurs or through the subordination of previously more important beliefs, so-called conflict dissolution.

Over time, empirically led studies gained considerably in importance. With the substantial increase in international interventions into civil wars by organizations like the United Nations or NATO, which can be seen at the end of the 1980s and beginning of the 1990s (Regan, 1996, 2002), research on how to organize third-party interventions aimed at finding a settlement successfully has seen a surge as well. Mason and Fett (1996) still argue that only a small number of intra-state wars come to an end in a negotiated settlement. Regan (1996) and Walter (1997) describe the situation similarly but start investigating which characteristics and conditions of a war lead to successful intervention and therefore negotiated settlements. DeRouen and Sobek (2004) find interventions from the United Nations to be likely followed by a peaceful resolution of the conflict. Civil wars where the UN intervenes have – all other variables at their means – a 57% chance to end in a treaty, compared to 15% for wars without UN intervention. Walter (1997) comes to analogous conclusions by calculating the correlation of potential explanations for settlement or victory with these two outcomes of civil war. Security guarantees made by intervening third-parties have a significant impact not only in all of the 41 found cases in the period from 1940 to 1990 but also when concentrating on those 17 civil wars that initiated bilateral negotiations before another party stepped in to provide guarantees overlooking the implementation of the terms agreed upon in the settlement. Furthermore, power sharing pacts without those third-party guarantees are insufficient to lead to a peaceful resolution of intra-state conflict. The strength of the security guarantees does not play any role. Stronger guarantees do not seem to lead to statistically better outcomes in terms of civil war settlement (Walter, 2002). Mixed strategies are superior to merely military or economic ones in reaching a conflict’s successful settlement (Regan, 1996). Purely military support from external entities is found to decrease the likelihood of a rebel win (Lyall & Wilson III, 2009). Gent (2008) additionally concludes that negotiated settlements become more likely as well. While Regan (1996) concludes that interventions supporting the state over the rebels raise the chance to stop the fighting peacefully (see as well Svensson, 2007), Balch-Lindsay and Enterline (2000) observe that an even-handed division of interventions by outsiders is supposed to lead to a stalemated conflict. Together with earlier findings suggesting that a ‘mutually hurting stalemate’ makes successful mediation from third parties more likely (Zartman, 1985), this corresponds to what Walter (1997, 2002) has reported. Walter (1997) even tests this relationship empirically and finds support. When negotiations between the conflict parties already took place, military stalemate indeed predicts a successful settlement of the intra-state conflict.

Considering that military stalemate means the prolongation of war (Balch-Lindsay & Enterline, 2000), these findings are consistent with results suggesting that the longer a war endures the more likely it ends in a negotiated settlement (Mason & Fett, 1996) and it is less likely that this conflict experiences a government victory (Cunningham, Gleditsch, & Salehyan, 2009). Moreover, the results would also support the finding that the shorter a war is the more likely it ends in a victory by either of the opponents (Kreutz, 2010). DeRouen and Sobek (2004), however, find that duration decreases not only the probability of a government victory but of a treaty compared to the reference category of ongoing war as well. The second analysis of their paper shows again the initially outlined relation of duration and the probability of a negotiated settlement. When moving from the 25% to the 75% quantile while all other variables are held constant at their means, the probability of a treaty decreases by 9% to 42% compared to all other outcomes. Those rebel groups which challenge the state units militarily appear to shorten the war and thereby increase the probability of favorable outcomes for the rebels. Nevertheless, the likelihood of reaching a settlement remains remarkably small compared to all other outcomes (Cunningham et al., 2009).

Findings about the relationship between the size of the government army and the outcome of civil war are ambiguous. For one thing, settlements are found to be less likely the larger the army is (Mason & Fett, 1996), then again army size seems to make all outcomes more likely to occur except for the base category ongoing war (DeRouen & Sobek, 2004). Moving army size from the first to the third quartile does not have an effect at all on either of the war terminating outcomes’ probabilities. Signing a peace treaty remains at 30% likelihood, the probability of government victory increases by 1% to 64%. Mason, Weingarten, and Fett (1999) report government victory to be more likely with increasing army size. The positive effect on rebel victory is not significant.

The results are more consistent for the effect of a conflict’s type even though some researchers have not found any significant impact of whether the civil war was fought about identical, that means ethnic or religious, or ideological matters (Mason & Fett, 1996; Walter, 1997). In conformity with the findings of DeRouen and Sobek (2004) and Gent (2008), Mason et al. (1999) report that ethnic wars are more likely to see government than rebel victories. In non-ethnic, ideological conflicts, rebel groups fare considerably better but their chance to win a war is reduced by 28 percentage points (DeRouen & Sobek, 2004). The probability of ending a war peacefully with a settlement seems not be affected by the reason the war is fought over. Neither the distinction of ideological and ethnic conflicts reveals a significant difference (DeRouen & Sobek, 2004; Mason et al., 1999), nor does the identity against political-economic division in Licklider’s (1995) paper. Cunningham et al. (2009), on the contrary, report a significantly increased probability of a settlement compared to the base category ongoing war. A second distinction of war type made in the literature aims at differentiating revolutionary wars from conflicts fought over separatist and secessionist issues. However, no statistical relationship could be proved for these variables (Mason et al., 1999; Mason & Fett, 1996; Walter, 1997).

The same applies for the battle death rate. Neither Regan (1996), nor Mason and Fett (1996), nor Mason et al. (1999) can prove a substantial impact of the number of battle-related deaths on the outcome of civil war. While the two former analyses make use of a war’s overall casualty number, the latter paper calculates the variable as ratio of the country’s population. Walter (1997) introduces two different measures of this variable. A war’s ‘magnitude’ is operationalized as number of battle deaths per 1000 inhabitants of the country the war is fought in. Using this operationalisation, the study reveals a positive correlation with settlement of conflict when testing for the universe of all 41 examined cases. This relationship loses its significance when looking only at the subset of cases with negotiations. For ‘intensity’ indicating the number of battle deaths per month, no effect is found in either of the two analyses. Gent (2008), however, finds a significant positive impact of the monthly battle death toll on the probability of government victory and a negative effect on rebel win.

DeRouen and Sobek (2004) also investigate further relations. They thereby find that in states with a well-functioning bureaucracy, rebel victories become less likely than ongoing war, while the variable does not have any significant effect on either of the other outcomes. Wars on the African continent seem to be less prone to government victory, truce or treaty and the unequal distribution of income as measure for grievances appear to lead to fewer truces as well. Forested terrain makes every outcome less likely compared to continuation of the war and mountains increase the probabilities of all outcomes but government victory. Gent (2008) only finds a significant positive effect on the likelihood of negotiated settlements. Regime type does not seem to change much in DeRouen and Sobek’s (2004) study. None of the coefficients are significant. Lyall and Wilson III (2009), however, find democracy to decrease the probability of government victory after 1945. Results from Cunningham et al. (2009), correspond to the general finding. Additionally, they report that democracy lowers the likelihood of a low activity outcome. The first two analyses derive the democracy variable from Polity IV data, whereas in the latter a dichotomous variable is created to investigate the relationship. Democracy is found to increase the likelihood of negotiated settlement elsewhere (Gent, 2008).

The literature on civil war termination has clearly attracted widespread attention during the last two decades. Many researchers contributed with analyses on the outcomes of intra-state conflict looking from several points of view and investigating many potential explanations. Even though the influence of different regime types has been occasionally included by controlling for the level of democracy, particularly the differences among authoritarian regimes have been neglected so far. To close this gap, theoretical arguments of how different types of authoritarian regimes could impact the termination of civil wars are outlined in the following section.

3. How Regimes impact Conflict Termination

Iraq has experienced eight different civil wars between 1946 and 2010. The first two were fought under the aegis of personalist military leaders and were won by the government. The six subsequent conflicts occurred within civilian regimes, five of which were personalistic autocracies. They all ended in a settlement between the warring parties. Did this pattern arise coincidentally or are there indeed observable causalities between the regime type and the outcome of civil war in autocratic regimes? This paper investigates if there is empirical evidence for differences in civil war termination among different types of authoritarian regimes. In the following, the underlying theoretical arguments will be developed for why the four different regime types investigated in the study reach different outcomes in civil wars. Therefor, nonpersonalist civilian regimes are distinguished from personalist civilian autocracies and the respective military forms of these two kinds of dictatorships.

No matter what kind of dictatorship one looks at, its leader’s main goal is to stay in power (Bueno de Mesquita, Smith, Siverson, & Morrow, 2003). That is why the heads of these regimes in the first place should try to avoid civil wars constituting a serious threat to their reign. Sometimes intra-state conflict occurs nonetheless. However, the primary aim of keeping the position steady has not changed. Different kinds of dictatorships may, due to their divergent capabilities and perceptions, try to reach different conflict outcomes (either governmental victory or conflict settlement) to do so.

To elaborate the theory about how diverse autocracies end civil war, this paper will look at (1) what the attitude is towards the use of arms, of both the regime’s leader and the respective domestic audience, and (2), to what degree a regime is able to coopt its adversaries. Furthermore, (3) the self-perception of personalist dictators should play a crucial role in answering rebel activities. The first and third point are expected to have an influence on the likelihood of a regime to win a civil war, the second one should impact the probability of reaching any kind of settlement (official or tacit). Below, they are investigated further.

Only nonpersonalistic regimes have a domestic audience (Weeks, 2008). It is the group of people whose members are able to remove a regime’s leader from office. They gain their position because of the long time they have served the regime and their meritoriousness rather than personal relationships to the incumbent. An elite-constrained leader therefore is accountable to it and tries not to antagonize them by disregarding their opinion. That is why a domestic audience’s attitude toward arms is expected to have an impact on a leader’s behavior in the case of civil war. Personalist autocrats on the other side are not liable to anyone and therefore only their own attitude is to be considered. Second, the capability of using cooptation as a tool of stabilization in autocratic regimes (Frantz & Kendall-Taylor, 2014; Gerschewski, 2013) further hypothesises a varying behavior in the case of civil war and influences this time the likelihood of seeing an agreement between the warring parties. While military regimes due to their selectiveness of membership are disadvantaged in bribing and giving spoils (Fjelde, 2010), civilian dictatorships are not as exclusionary towards broad parts of the society. This indicates that cooptation as a tool to appease rebels is available and thus makes it easier for these kinds of autocracies to find consensus with rebelling forces. The likelihood of settlements should increase. Lastly, personalist leaders tend to create cults of personality (Ezrow & Frantz, 2011) and they pursue a reign of self-gratification and glorification (Decalo, 1985). This suggests that personalist autocrats are to a high extent devoted to what they have achieved and are more inclined to fight for it. Geddes (1999) writes that personalist leaders in particular eradicate political enemies to secure their position. This position is compatible with arguments from Peceny and Beer (2003) who trace back personalists’ belligerent attitude to the absence of decisional restrictions partly applicable to the context of intra-state conflict as well. Unconstrained autocrats therefore are expected to order heavy military action and thereby increase the probability of a government victory.

In the following, each regime type will be analyzed separately on how they fare on the named theoretical aspects and what this indicates with respect to the expected outcome of civil war. For better understanding, this happens with the help of a historical example of each regime type.

3.1 Nonpersonalist Civilian Regime

Party-based (nonpersonalist civilian) regimes like the Soviet Union until its dissolution in 1991 have a domestic audience the leader is accountable to. In the example of the Soviet Union the Politburo of the Communist Party represents this audience. It has an impact on the decisions the leader takes. The domestic audience does not have a military background (Weeks, 2012) and, therefore, does not incite to take military action (Sechser, 2004). Additionally, a party’s leader probably is most reluctant to the use of force compared to other autocratic regimes since he in most cases neither has a military background nor did he have to forcefully fight his way to the leading position. Rather he served the party for a long time in an outstanding manner. The same applies to the officials in the domestic audience and was the case in the Soviet Union as well (Mawdsley, 2004). Of course, even in parties it is not all about seniority and performance, but the element of force is not as commonly used as in other regimes. Weeks (2012) states that civilian domestic audiences view the costs of fighting as high as democracies. Therefore, party-based regimes are expected to use force of arms hesitantly, which subsequently reduces the likelihood of a government victory. They, on the other hand, due to their parties’ institutional framework possess elaborated cooptation capabilities. Their long lasting structure allows them to effectively distribute spoils and make credible commitments for the future (Fjelde, 2010). This mechanism is expected to work even when a civil war has already broken out. Parties thus should very likely experience a settlement.

3.2 Personalist Civilian Regime

Although personalist civilian leaders lack the military education and the allegiance to the army, they nevertheless might have used arms to overthrow the previous regime. Many personalist civilians take power by making use of violent means like revolution or civil war (Weeks, 2012). An example of such a leader is Yoweri Museveni, president of Uganda since 1986. He overthrew the previous regime with his National Resistance Army and took over control (Tripp, 2010). Being a former rebel, the costs of use of force subsequently are perceived as low (Colgan, 2010) and the repeated use of violence becomes more likely (Horowitz & Stam, 2014). Additionally, the regime lacks a domestic audience that due to their civilian roots would mitigate the military drive of the leader. As stated above, personalist leaders have a high inclination for the regime which is expressed by a cult of personality. Museveni in Uganda created such a cult as well (Omach, 2017). All these points together should make personalist civilian autocrats more likely to end civil conflict with a victory of the government. Due to their civilian nature, these regimes are not as exclusionary to broad parts of the society as both personalist and nonpersonalist military regimes are (Fjelde, 2010). This indicates that cooptation as a tool to appease rebels is available. Nevertheless, they cannot rely on parties as a “forum where the dictator can make offers to distribute spoils in exchange for cooperation” (Fjelde, 2010, p. 203). Furthermore, the lack of such an institution does not guarantee a continuation of spoils obtainment in the long run. Rebel group leaders are therefore expected to accept bribery offered by personalist civilian leaders less likely than from party-based regimes. In accordance with the arguments outlined above, dictatorships led by personalistic civilians are expected to experience a negotiated settlement to end an intra-state conflict less likely than party-based regimes but still more likely than military ones.

[...]

^[1] The terms authoritarian regime, autocracy, and dictatorship are used interchangeably.