Unconditional Loyalty: The Survival of Minority Autocracies
Unconditional Loyalty: The Survival of Minority Autocracies (DOI: https://doi.org/10.1017/S0003055425000218) was written by Salam Alsaadi in 2025. It was published in the American Political Science Review.
The expectation is that, in ethnically-divided societies, a minority government is less stable than a majority one. But there are notable exceptions:
through the Arab Spring, the governments that resisted rebellion were minority-led autocracies (i.e., Bahrain, Jordan, Syria)
The authors first model autocratic regime survival using data from 1900 to 2015. ("extended version of Geddes, Wright, and Frantz’s (GWF) (2014) dataset as presented in Casey et al. (2020") The data is subset to states with ethnic divisions; this involved referencing classifications in this and other notable datasets. This procedure yields 71 regimes.
If an ethnic out-group represents more than 50% of the population, then the regime is classified as excluding a majority. These include:
Altogether there are now three candidate predictors:
- minority government
- minority government that excludes a majority
- minority government that excludes other minorities
The authors fit a logistic model of regime survival on those predictors using country-year data. They control for GDP per capita, GDP growth, oil production, regime type ("party-based, personalist, military, and monarchical"), and "whether the regime relies on a colonial or foreign state to sustain its rule". They also use fixed effects for country and year.
They find that minority governments that exclude a majority are more likely to survive. To check robustness, this analysis is repeated in other datasets. To address imbalances between control and treatment cohorts, they use coarsened exact matching to pre-process the data. The weighted logistic model reproduces the main model's findings. They also perform a design analysis following Gelman and Carlin 2014; there is negligible Type S error and 7.9% exaggeration factor.
They next model challenges to autocratic regimes using the NAVCO 2.1 data, covering 1945 to 2013.
The outcomes are:
- 'challenge' success
- elite defections
- mobilization for the autocracy
These outcomes are regressed on the identifier for a minority government that excludes a majority. The first two are fit to logistic models, while the third is fit by OLS. They control for GDP per capita, population size, size of the challenge, use of violence, external support (two components: for the regime, or the challenge), duration of the challenge, opposition unity. They also use fixed effects for country and year.
They find that challenges to minority governments that exclude a majority are less likely to succeed. They also find that elite defections are less likely to occur. These two effects are particularly strong and significant. They also find that mobilizations for the autocracy are larger, though this is not as strong an effect. To address imbalances that may affect the third model particularly, they use a generalized synthetic control method; the treated cases are compared instead to synthetic cases sampled from a donor pool of untreated cases. This is done for both minority governments that exclude a majority and those that exclude other minorities. The former is a significant effect, while the latter is not.
The authors examine the case study of Bahrain and fit a theory to the quantitative and qualitative findings.
- Opposition and protest leaders in Bahrain made efforts to cast off labels of a Shiite movement. Despite this, Sunni opposition parties did not support the movement.
- Civilian Sunnis organized in militia, 'popular committees', and 'local security' to attack Shiites.
- Military repression of protests was brutal, but there were no notable Sunni defections.
They argue that being the minority in an ethnically-divided society creates important structures.
- the ruler's coethnics are demobilized
- 'linked fate'
- collective interest over personal interest
- coethnics self-police and enforce by social pressure and sanction
- elite defection is unlikely
- if the majority mobilizes, the ruler's coethnics can be counter-mobilized
- perception of heightened threat for coethnics
- repression is legitimized, and state capacity is enlarged
- elite are homogeneous by the salient social stratifier
- military can be empowered without fear of inviting a coup
Reading Notes
I'm not sure that a logistic regression is the best way to model regime survival. I've briefly read some content about using Weibull and Cox proportional hazards instead.