Contracting on Violence: The Moral Hazard in Authoritarian Repression and Military Intervention in Politics

Introduction

Scholars have examined institutions in dictatorships and how internal dynamics affect their operation.
Military dictatorships are a distinct category, impacting political outcomes.
- They are the most common form of government between 1945 and 1990.
- They have the shortest life span.
- Leaders are less likely to suffer violent removal compared to nonmilitary ones.
- They are less prone to coups.
- Their effects persist even after a country transitions to democracy, affecting authoritarian reversals and impeachments.
Research has focused more on the consequences of military intervention than its causes.
The central question is why the military intervenes in some dictatorships but remains under civilian control in others.
From 1946 to 2002, roughly two of three authoritarian leaders were installed or removed by the military.
The Soviet Union maintained firm control over its armed forces for seven decades.
The article addresses why some democracies break down due to military intervention while others do not.

Main Argument

Military intervention in politics is linked to a moral hazard problem in authoritarian repression.
Dictatorships facing mass, organized, and potentially violent opposition use the military as a repressive agent of last resort.
Domestic repression comes at a price: the military gains a pivotal role and demands influence over policy.
Failure of authoritarian leadership to concede to these demands may lead to military intervention.
Dictatorships that rely on military repression risk exposing themselves to challenges from the regime's repressive apparatus.
The article develops a theoretical model to explain how the likelihood of military intervention depends on the threat posed by those excluded from power.
The military's institutional autonomy and resources play a crucial role.
The observed level of latent conflict and the likelihood of military intervention are interconnected.
The likelihood of military intervention is highest at intermediate levels of mass threats.
- When threats are low, the military lacks the resources for intervention.
- When threats are very high, they are completely deterred.

Theoretical Underpinnings

Military intervention is first increasing and then decreasing in the magnitude of the threat from those excluded from power.
The article uses economic inequality as an empirical proxy for latent conflict between the authoritarian elite and the masses.
Examples of military concessions include:
- Donativa and privileges to praetorian guards and the army in ancient Rome.
- Military-run enterprises in Indonesia.
- Illegal activities in Paraguay.
- Institutional autonomy through budgetary, personnel, and procurement decisions, as well as control over internal security.
The alliance between the government and the military may fail due to limited policy expertise within the military.
Overt intervention is costly for the military because it may fail, leading to imprisonment or execution, and it may highlight political differences within the armed forces.
Information asymmetries and the costliness of military intervention create a temptation for the government to defect from agreements and a reason for the military to suspect defection.
Military interventions may occur even when both the government and the military prefer to avoid them.
When the threat from those excluded from power is low, the military lacks the resources for intervention.
Once threats exceed a threshold, conflict between the government and the military fails with positive probability.
Military intervention is greatest immediately past this threshold.

Empirical Analysis

Empirical analysis supports the theoretical arguments.
Economic inequality proxies for latent conflict between the authoritarian elite and those excluded from power.
The relationship between military intervention and economic inequality is nonmonotonic.
The study uses original data on military intervention, including:
- Participation of the military in the entry and exit of authoritarian leaders.
- 738 leaders from 139 countries between 1946 and 2002.
The military has intervened in the entry of 291 and the exit of 248 authoritarian leaders.
The Gini coefficient and the Theil statistic are used to measure economic inequality.
The Gini coefficient ranges from eighteen (Bulgaria in 1968) to sixty-eight (Sierra Leone in 1989).
The Theil statistic ranges from twenty (Czechoslovakia in 1988) to sixty (South Africa in 1993).
Communist regimes are the most egalitarian, while oil-rich Middle Eastern countries and some South American and sub-Saharan African countries are the most unequal.

Data and Methodology

Missing data on economic inequality is addressed using polynomial intra- and extrapolations of the Gini coefficient and multiple imputation of missing values for the Theil data.
A country-level, random-intercept logistic regression model is used to estimate the relationship between military intervention and economic inequality.
The model accounts for covariate effects and estimation concerns specific to cross-sectional time-series data on dictatorships.
Control variables include:
- Log of gross domestic product (GDP) per capita.
- GDP growth.
- Log of trade openness.
- Fuel and ore exports.
- Cold War era.
- Democratic neighbors.
- Ethnic fractionalization.
- Interstate war.
- Civil war.
- Military leader.
- Time.

Results

The likelihood of military intervention is first increasing and then decreasing in economic inequality.
This relationship is statistically significant across different measures of military intervention and economic inequality.
Including a quadratic term for economic inequality improves the fit of the models.
The estimated effect of economic inequality on military intervention is plotted in Figure 4.
The random effects capture the effect of unobserved or omitted country-level factors.
Mexico ranks near the bottom of the distribution of predicted random effects, while Greece ranks near the top, indicating country-specific factors influencing military intervention.

Formal Model

An authoritarian government faces a threat of magnitude R > 0 from those excluded from power.
The government does not perfectly observe R, but knows it is distributed uniformly on the interval [\underline{R}, \overline{R}], with the expected magnitude \overline{R} = (\underline{R} + \overline{R})/2.
The government endows the military with resources of size r \in [0, \infty].
A military with resources r defeats the threat with probability \phi(R, r) = r/(r + R).
The polity can be in two states of the world, \theta = {A, B}, where state \theta occurs with probability \gamma \in (0, 1), \gamma = Pr(\theta = A).
The government adopts one of two policies p = {pA, pB}.
Ideally, the government would adopt policy p_A, regardless of the state \theta.
The government agrees to adopt policy pA only in state A and policy pB in state B.
The probability \gamma measures the extent to which this policy compromise favors the government.
The military intervenes with probability \beta(r) and typically takes the form of a coup d'état that replaces the civilian government by a military one.
The probability p(r) is an increasing, concave, and differentiable function of the military's resources r, p'(r) > 0, p''(r) < 0, such that p(0) = 0 and \lim_{r \to \infty} p(r) = 1.
The payoffs to the government and the military depend on the policy that the government adopts, the resources spent on repression, the outcome of the coup if the military intervenes, and on whether the mass threat is defeated at the end of the game.

Model Payoffs

If the threat is defeated, the government's payoffs are g - r and 1 - r when the adopted policy is pA and pB, respectively, and the military acquiesces or intervenes but the coup fails; g > 1.
The government's payoff when the military intervenes and the coup succeeds is -r.
The government's worst outcome is when the military intervenes and the coup fails, in which case its payoff is -(r + R).
The military prefers that the government comply with the policy compromise.
The military's payoffs are normalized to 0 and -1 when the government complies and reneges, respectively, and the military acquiesces.

Probability Calculations

The military intervenes with probability \beta^* = \frac{\gamma - \frac{c}{[1 - p(r)]m + c}}{\frac{\gamma (1 - \gamma)}{\kappa_{bB}} \frac{c}{p(r)m + c}}

Key Findings

The likelihood of military intervention is greatest immediately past this threshold because the military's autonomy and resources are large enough that it can successfully intervene, yet because such an intervention may still fail, its mere threat does not deter the government from reneging on a policy compromise.
- As the threat from those excluded from power grows, the government must cede an ever-increasing amount of resources to the military and is thus increasingly tempted to defect from the policy compromise.

Proposition 1

Proposition 1 (Military Intervention in Authoritarian Politics): In a perfect Bayesian equilibrium,
1. if \overline{R} < R_1, r^* = -\overline{R} + \sqrt{\overline{R}(g + \overline{R})}, \alpha^* = 1, \beta^* = 0
2. if R1 < \overline{R} < R2, r^ solves p(r^) = c/m, \alpha^* = 1, \beta^* = 0
3. if \overline{R} > R2, r^* = -\overline{R} + \sqrt{\overline{R}(g' + \overline{R})}, \alpha^* = \frac{\gamma - \frac{\kappa{bA}}{\kappa{bB}} [1 - p(r)]m + c}{\gamma (1 - \gamma) \frac{\kappa{bA}}{\kappa{bB}} p(r)m + c}, \beta^* = 1 - \frac{g - 1}{\kappa{bB} p(r^*) g}

Conclusion

Authoritarian repression entails a fundamental moral hazard problem, explaining military intervention in politics.
Dictators rely on the military to deter challenges when facing mass, organized, and potentially violent opposition.
Military intervention is an indirect, political cost of authoritarian repression.
Dictatorships attempt to minimize vulnerability to challenges from their repressive apparatus through methods like parallel command structures, rotation of commanders, and ethnic or religious selection in recruitment.
Military intervention is more likely in new democracies with inherited military autonomy and resources.
The institutional makeup of dictatorships is endogenous to underlying structural factors like economic inequality or ethnic and religious divisions.
Structural factors determine the form and magnitude of polity-wide conflict and the dictatorship's optimal institutional response.
Soldiers repeatedly intervene in the politics of some countries while remaining under firm civilian control in others due to persistent structural factors.