In this 2001 AER, Acemoglu and Robinson propose a simple theoretical model of coups and revolutions meant to explain the economic aspects of political transition and consolidation. The basic tradeoff is straightforward: in a democracy, the proletariat must set taxes such that they do not encourage an elite to overthrow the democratic leader in order to escape onerous taxes, and in a dictatorship, the elite must provide sufficient transfers to the proletariat that they do not begin a revolution. Coups are uncommon because a coup leads to a one-time downward shock to the economy. Revolutions are uncommon because they in turn lead to temporary lower output, and because the elite can coopt revolutions by extending the democratic franchise, which de facto allows commitment to higher transfers to the poor in the future by shifting the median voter.
More extensively, the model is as follows. There are two classes of fixed size, an elite and a proletariat. The elite are in the minority, and at time zero, the elite control political power. The level of income is stochastic, with two states, an uncommon low level (recession) and a common high level (“normal times”). A stock of assets, with higher levels held by members of the elite, provides the stochastic income. Taxation with a convex deadweight loss can be imposed on income, with revenue transferred equally to all members of society; since their income is lower, the proletariat prefer higher tax and transfers than the elite, who prefer none. In any period, the poor may begin a permanent revolution, seizing an exogenous portion of the elite’s assets forever, at the cost of destroying an exogenous amount of output in the first period of the revolution. The elite can voluntarily extend democracy in order to prevent revolutions; in that case, the median voter rather than the elite will set the tax rate. If the current state is democracy, the elite can mount a coup, at the cost of destroying an exogenous portion of the economy’s output in the period of the coup. The equilibrium concept is Markov perfection, where each agent chooses a (possibly mixed) strategy in each of six states: dictatorship, democracy and revolutionary government under the recessionary and normal governments. The action space includes whether to mount coups or revolutions, whether to extend the democratic franchise, and what tax rate to set. Agents are not myopic: they solve to maximize their total future welfare, conditional on equilibrium Markov perfect actions by both players. There is no coordination problem: every member of the proletariat agrees on the optimal action, as does every member of the elite. Assumptions on parameters are such that coups will not take place during normal times, and that no transfers will be offered to the poor in normal times.
Since there is no threat of coup in normal times, a democratic government (or a revolutionary government) will choose taxes and transfers in order to maximize the income of the proletariat. Coups are more likely when recessions are severe, since the percentage loss in income due to a coup is lower, yet the gains from not having to offer transfers in the future are unchanged. Coups are also more likely in unequal societies – where inequality is defined in terms of assets, though this maps directly into pre-tax income – because the optimal tax rate in democracy is higher. In particularly unequal societies with particularly strong recessions, democratic governments cannot set their optimal tax rate: rather they lower taxes during recessions because they are worried about a coup occurring. When the level of inequality gets very high, the democratic government cannot prevent coups during recessions even with zero taxes.
If the economy is in dictatorship, the elite may wish to give transfers to the poor during recessions to prevent a revolution which could expropriate the elite income forever. The elite always want to prevent coups, so they will do this whenever the economy is in recession. However, for some parameter values, even offering the transfer-maximizing tax rate to the poor is not enough to prevent a recession. In that case, the democratic franchise will be extended; this allows the elite to commit to higher taxes in nonrecessionary future periods by shifting the median voter, although coups may occur during future recessions, and the proletariat take that into account when deciding to revolt in the face of an offer for democracy. In particular, when recessions are relatively common, it is easier for the elite to avoid extending democracy because the elite can now credibly commit to sharing wealth more often as they will always do so in recessionary periods due to the threat of revolution.
There are then, depending on four parameters, four paths for society: the dictatorship remains forever, society democratizes in the first recession and thereafter sets transfers optimally from the perspective of the poor, society democratizes but sets taxes lower than the level optimal for the poor in order to preemptively prevent coups, or society oscillates between democratic and nondemocratic government. The last two cases are more likely in more unequal societies with more severe recessions.
A final section of the paper discusses how asset redistribution may be used to consolidate democracy or dictatorship. Redistribution is irreversible and causes a negative shock to the total level of assets. Democracies can redistribute assets in the first period they come to power. In particularly unequal societies, the democratic government will redistribute in order to prevent future coups and thus consolidate democracy, since coups are mounted more often in more unequal society. This redistribution may actually make the democratic proletariat myopically worse off, since in the absence of a coup threat, they would prefer to leave the land with the elite, tax them, and transfer income; under some parameters, this is less costly than that caused when assets are destroyed under redistribution. Acemoglu and Robinson also consider the case where redistribution takes another period to implement; if a recession occurs during the implementing period, the elite may wish to mount a coup and reverse the land reform. For some parameters, the democratic government will choose to implement land reform which will fully stabilize democracy forever if the following period is nonrecessionary, but which will lead to a coup otherwise. Finally, if the current state is dictatorship, and inequality is high, the dictators themselves may find it cheaper to redistribute assets rather than to extend the democratic franchise; with the post-redistribution state more equal, the threat of revolution is less salient, and hence following redistribution, taxes can be lower.
I’m not totally convinced either that political transitions are particularly driven by the business cycle (try explaining 1989 or 2011 that way!) or that, in such a model, the punishment-free Markov perfect equilibrium is the right refinement. That said, this is a solid parsimonious model in the usual Acemoglu fashion – that is one productive dude.
http://econ-www.mit.edu/files/4121 (Final AER version; three cheers to Acemoglu for putting the final version on his website!)
Reading your post made me wonder: which criteria should I use when assessing which refinement to use? Why Markov perfect equilibrium would not be the right refinement? Could you explore this idea in more detail?
MPE is great for tractability, but it does allow strategies that condition on more than one period of history. For example, in this paper, the dictator never gives transfers to the poor when the economy is not in a recession. Various grim trigger strategies thus aren’t possible. In a lot of Acemoglu’s papers, he shows that MPE is not necessary, and that the described strategy is still unique in any subgame refinement. They don’t do that in this paper, however.