Regime spoiler or regime pawn: the military and distributional conflict in non-democracies

I consider a model in which an autocrat can be removed from power either through a military coup or a revolution by the citizens. In the event of a revolt by the citizens, the military may choose to support the autocrat to crush the revolt or play a passive role. The autocrat determines the distribution of the country's wealth among himself, the military, and the citizens. I find that, under certain conditions, there exists a unique Markov perfect equilibrium in which there are no coups, the citizens revolt in each period, and the military fights on behalf of the autocrat. Under a different set of conditions, there is another Markov perfect equilibrium in which there are no coups, the citizens always revolt, but the military does not fight the revolt. However, peace (no revolts) is also an equilibrium of the model. The model is consistent with the persistence of social unrest or civil wars in certain countries and the different roles played by the military in different countries. Surprisingly, I find that if the citizens' outside option (i.e., payoff in a democracy) improves, this is likely to make them worse off. Furthermore, an increase in natural resources is likely to make the citizens worse off because it reduces the probability of a transition to democracy or the prospect of good governance in autocracy. I discuss other implications of the model and relate it to real-world events.