International conflict, military rule, and violent authoritarian breakdown

ABSTRACT

Why do some transitions of power from military rule occur violently while others do not? What effect, if any, does the international security environment have on how violent breakdowns of authoritarian rule are? I argue a conflict-prone security environment ameliorates the commitment problem by ensuring an influential role for the military out of power. Therefore, when facing a domestic crisis in a threatening security environment, military leaders are more likely to peacefully cede power rather than wield violent measures to stay in office. Perhaps counter-intuitively, international conflicts thus lead to transitions of power from military rule that minimize violence and human costs. International conflicts do not have this moderating effect on other types of authoritarian rule.

KEYWORDS:

• Military rule
• civil military relations
• international conflict
• authoritarian breakdown

Notes

¹ This feature of military regimes represents a departure from the assumption that political actors in office seek to maintain power above all else.

² The American ambassador in Seoul during the democratization period was of the view that “an opportunistic and terrorist-minded neighbor to the north that might take advantage of civil dislocation in the South” influenced the decision of South Korean generals in a way that moderated the use of force (Lilley 2004: 275).

³ When the higher costs give additional leverage to all sides that can benefit from coordination, and there are incentives for all sides to try to send costly signals in the hope of reaching the coordination option most favorable to them, failure to coordinate can result (Crawford 1982; Levenontoglu and Tarar 2005).

⁴ Substantively, military leaders across all factions could believe that the nation’s security is better served with the military in power when facing a higher probability of conflict. As a result, factions that have policy differences with the one in power – and which can force the regime to give up power through a credible threat to head back to the barracks – may refrain from action against the government within a threatening security environment and even side with the government in a crackdown for the national interest. This would decrease the likelihood of regime transition. On the other hand, prospects of cracking down on resistance against the regime and the polarization that it would bring to the military may favor the faction that prefers a return to the barracks, thereby lowering the chances of a violent transition from military rule.

⁵ Similarly, it is quite different to refrain from repression when not facing a threat to power, and to do so when facing such a threat.

⁶ A recent account has argued that democratic institutions provide this credible assurance to the military. Because democracies rely less on violence as a means to determine who governs they reduce the potential coercive influence of militaries after democratic transitions (Debs 2016). Therefore the military poses less of a threat to the new (democratic) ruler, taking away the incentive for the new regime to punish or weaken the military out of power. As a result, military dictators can expect a safer posttenure fate, and by implication, are more likely to cede power peacefully to democratic rule. Similar to the literature on military violence, however, this argument cannot account for the mode of transition: why some cases of democratization occur violently while others do peacefully. Moreover, it fails to account for variation in violence during transitions to other autocracies for the theory implies that they should be violent because the military is not assured of its postexit safety.

⁷ 94.5% of voters supported the constitutional amendment.

⁸ Author retracted.

⁹ When there is uncertainty about whether casualties occurred during the regime transition or not, the dataset limits casualties to those within two months of the date of regime change.

¹⁰ Cheibub, Gandhi, and Vreeland 2010, available at: https://sites.google.com/site/joseantoniocheibub/datasets/democracy-and-dictatorship-revisited .

¹¹ Due to the coding for rivalries ending in 2001, and the GWF coding ending in 2010, the 13 regimes that came to power between 2001 and 2009 and that were not part of a rivalry before 2001 were deleted from the sample. This is just under 5% of the sample. They were, with the exception of Venezuela 05-, all regimes that emerged in countries that emerged from the break up of the USSR.

¹² About 21% of single party regimes fell with over 25 casualties (6 of 28).

¹³ A regime was considered to be a party to an enduring rivalry if it was part of one for over half of its duration in power.

¹⁴ The two violent collapses of military rule (Thailand 57–73 and Pakistan 58–71) were both hybrid military-personalist regimes. As noted above, pure personalist regimes have a much higher propensity for violent breakdowns under the shadow of conflict, and hence it is not surprising that the only two cases of violent breakdown for military rule were actually hybrid military-personalist regimes.

¹⁵ Applying this model does require transforming the ordinal casualty data into a binary variable by imposing a cutoff at a certain threshold.

¹⁶ Results of robustness checks that run the key models excluding hybrid types from the definition of military regimes are reported in the Appendix A2 and A3. The results are robust to such coding rule changes.

¹⁷ In practice, because over 95% of regimes that have collapsed either were never in a rivalry or were always in a rivalry during their duration in power, the operationalization makes little difference. Running the same models below with the dichotomous version of the rivalry variable, coded as a one if an authoritarian regime was in a rivalry for 50% or more of its duration in power and a zero otherwise, results in similarly significant, though slightly weaker, results (both substantively and statistically) on the key interaction terms. Changing the threshold to 60 or 70 and so on makes no difference. Refer to Appendix A1.

¹⁸ All GDP data is from the Penn World Tables dataset, Version7.0.

¹⁹ An example is the tacit, and at times overt, guarantee the Soviets gave toward Communist Parties in Eastern Europe during the Cold War externally through the Warsaw Pact and internally through intervention when ally communist parties were threatened.

²⁰ A forced removal of a leader and imposition of a new government after an invasion by a foreign force are not only interventions into the lifespan of each regime, they usually lead to resistance against foreign forces that affect casualty levels by a different mechanism than the intraregime politics that are the focus in this paper.

²¹ Among the 199 collapses of authoritarian rule in the dataset, 37 occurred with between 1 and 25 casualties, 24 with between 25 and 1000, and 26 with over 1000 casualties. The rest (112) resulted in no casualties.

²² These results are robust to specifications that define military regimes in different ways. In Appendixes A2 and A3 are the results of Tables 2 and 3 but with personalist-military regimes and indirect military regimes excluded from the military regime variable. Leaving only the pure military regime in the military regime variable lead to stronger results.

²³ From Column 4 of the regression table, a military regime not in a rivalry will have a 5.8% probability in a given regime-year of breaking down with at least a single casualty. Within a rivalry, this probability falls to 0.6%. This is a 90% decrease in the probability of a violent fall. As expected, the effect of the security environment is substantively more pronounced with a higher threshold for defining violence, with the predicted probability in a given regime year increasing nearly 10 times with the shift in the security environment. Due in part to the rareness of such events however, the statistical significance of this difference is weaker at about p = .12.

²⁴ Results available upon request.

²⁵ Noninstitutional means of autocratic breakdown are also more frequent, however, than those with casualties.

²⁶ Again, because personalist regimes are the omitted class of regimes, the coefficient on the rivalry variable is interpreted as its effect on the personalist regimes.

²⁷ Arguably this endogeneity would underestimate the effect of the external threats on the moderation of violence under military rule.

²⁸ Neither lagging the GDP variable to include the previous year’s GDP growth or including the lagged GDP growth with the GDP performance of the year of the breakdown substantively changes the results.