Economic Origins of Dictatorship and Democracy

by Daron Acemoglu

  North and Weingast’s explanation is compelling and provides a good description of the various issues involved in one of the major examples of institutional change in European history. Why is it that these new institutions make repayment credible? Why, if Parliament was strong enough to remove from office the legitimate king, James II, did it need to alter institutions to ensure that future kings would not renege on their debt? A full exploration of the answer takes us to political power and the relationship between political power and institutions. When it deposed James II, Parliament used its de facto political power and that of the Dutch, who had sent an army to help. However, this situation was transitory; the Dutch were not going to send an army every time Parliament asked for it (for one thing, they were busy fighting the French). So, Parliament changed the political institutions in Britain to try to lock in their transitory de facto power. The new institutions allocated de jure political power to Parliament - if not completely, then much more so than previously. Moreover, this new allocation of power guaranteed that the king would not be able to default on his debt because much of it was held by Parliament, which therefore had an interest in making sure it was paid off (Stasavage 2003).

  Similar issues will be important in our theory of democratization: the elites will be forced to democratize to prevent revolution by the disenfranchised. Once established, democracy will create durable changes in the political arena and these changes will constitute a sufficiently credible commitment to give the citizens power and the policies they want in the future.

  4.2 Institutions and Commitment

  Our discussion of North and Weingast (1989) raises a fundamental question: Why do institutions provide commitment at all? In our model, this is because de jure political institutions determine who can take which actions and when. For instance, in a democracy, policies are determined by majority voting, which means that the citizens can get what they want if the elites do not have de facto power to challenge the citizens. When democracy is created, the citizens understand that the institutions will give them de jure political power, which serves as a commitment to more pro-majority policies, even if they do not have de facto power in the future.

  Moreover, there are natural reasons for why it will be costly to replace democracy once it has been created - most obviously because groups invest in particular sets of institutions (Brainard and Verdier 1997; Coate and Morris 1999; Acemoglu and Robinson 2001). To take one example, it was only after the Second Reform Act in 1867 in Britain that the Conservative and Liberal Parties began to organize themselves as mass parties and create the institutions needed to compete as national organizations. They created conservative and liberal clubs and countrywide networks of organizers who were needed to mobilize the new mass electorate. These are specific investments whose value would be destroyed if democracy ceases to function. This makes democracy persist because it gives people a greater incentive to fight for it ex post. Moreover, the creation of these organizations specific to democracy makes it easier to solve the collective-action problem once they have been created. These are fundamental reasons why democracy, once created, is difficult (though not impossible) to reverse and why it, as a set of political institutions, has commitment power.

  4.3 Political Power

  The discussion thus far emphasizes that political power has different facets. Obviously, political institutions bestow political power on those who control the presidency or the legislature. For example, the constitution of the United States allocates power to propose and make laws, which gives groups who are successful in elections the power to determine policies in their favor. Yet, there is clearly more to political power than this. Consider the case of Venezuela. Hugo Chávez was elected president by an overwhelming majority in 1998 and was able to closely control a process of rewriting the constitution in 1999, which increased his powers substantially. Chávez, therefore, has a lot of de jure political power. Yet, other groups, who neither control the presidency nor had any impact on the process of redrafting the constitution, also have significant de facto political power. Forces that oppose the policies that Chávez prefers - for example, the managers of the state oil company - can organize strikes that bring the economy to its knees, as they did for two months after December 2002. Political opponents can also organize street demonstrations to demand that the regime changes its policies, even if they have no de jure political power with which to influence such policies. Such economic decisions and collective actions are costly for the regime.

  Nevertheless, such power to challenge regimes is, by nature, transitory. Although the striking oil workers imposed heavy costs on the economy and hurt the regime, they simultaneously hurt themselves and their families. Strikes must, by necessity, be transitory. Moreover, strikes are difficult to organize and sustain, and their power depends on other factors that change over time, such as the world price of oil. The power of the oil workers in Venezuela also depends on geopolitical factors and the fact that the United States imports 15 percent of its oil from Venezuela. This induces the U.S. administration to intervene in Venezuelan politics to keep the oil flowing. However, the nature of such interventions depends on the character of the U.S. administration, which changes over time, again making de facto power transitory.

  One could argue that the threat of strikes or demonstrations is continually present, which would be sufficient to induce Chávez to change his policies. Yet, it is clear that Chávez did not make any concessions until these threats actually manifested in strikes and demonstrations. Generally, it will be unclear whether threats to organize strikes are credible because the actions of many people have to be coordinated and a strike may fail because the regime can organize strike-breaking activities. Even after a strike or demonstration has occurred, there is no guarantee that another one can be easily orchestrated in the future. These factors indicate why the opponents of Chávez were not content with policy concessions because they anticipate that they can be reversed. They would only be satisfied with the removal of the president and, thus, a change in the allocation of de jure power.

  In the context of democratization, one of the best examples of the relationship between transitory shocks and switches in political power was pointed out by Therborn (1977), who observed that many democratizations took place following wars. This fits well with our theory because war is a time when the citizens, who comprise the armed forces, have significant temporary power until they are demobilized. This threat is clearly seen in the democratizations in countries such as Germany after the First World War.

  An important point about de facto political power, therefore, is that it is not necessarily “stationary” - which group has political power changes over time because of economic and political shocks and social changes. We discussed earlier an example of transitory political power in our simple model of dictatorship. It is interesting that the transitory nature of de facto power has been explicitly noted in the transitions literature by O’Donnell and Schmitter (1986) who describe the dynamics of collective action in opposition to an authoritarian regime as follows:

  ... this wave crests sooner or later ... A certain normality is subsequently reasserted as some individuals and groups depoliticize themselves again, having run out of resources or become disillusioned, and as others de-radicalize themselves ... Still others simply become tired of constant mobilization and its intrusion into their private lives. (p. 26)

  5. A Static Model of Democratization

  We now build a model that features all the essential elements of our approach to democratization. As well as political conflict and the commitment role of institutions, this approach features transitory political power for the disenfranchised coming from a revolution threat. Under certain circumstances, the elites are induced to democratize as a credible commitment to future pro-citizen policies in order to prevent a revolution. In this chapter, we proceed by assuming that, once created, democracy is consolidated. We defer a study of coups against democracy to the next chapter.


  There are two groups, the rich and the poor, with fractions δ and 1 — δ. The elites are the rich and the citizens are the poor, although in Section 9 we show that results of the analysis are robust to alternative structures of political identities. Individual preferences are defined over post-tax incomes, given by:

  and society starts in a nondemocracy in which government policy is decided by the elites.

  Recall that when the elites have uncontested political power, they choose zero taxes and no redistribution of income (i.e., τr = 0). In contrast, the most preferred tax rate for the citizens is τp > 0, given by (4.11). The comparative statics of τp also play an important role. Recall from our previous discussion that a greater level of inter-group inequality (i.e., a higher level of θ) increases the desired tax rate of the citizens; hence, dτp/dθ > 0.

  Let us now summarize the timing of the extensive-form game between the elites and the citizens in which the sequence of moves is depicted in the game tree in Figure 6.1. Following the discussion of the game depicted in Figure 5.3, we can conceive of the initial choice being made by “nature,” which determines the value of a shock that affects how attractive it is to challenge the regime. However, as discussed in Chapter 5, in the static model there is no loss in suppressing the state L, dropping this branch from the tree, and simply focusing on the one state in which the nondemocratic regime is challenged. This being the case, we also suppress the notation H exactly as we did before. Hence, Figure 6.1 differs from Figure 5.3 in that the left side of the tree, that following the branch L, is dropped.

  Figure 6.1. The Democratization Game.

  The elites have political power initially and move before the citizens. They first decide whether to create a democracy, the branch labeled D, or not, the branch N. As in the last chapter, we denote the tax rate set by the elites in nondemocracy by τN and use the notation τD to refer to the tax set in democracy by the median voter. If the elites choose D, democracy is established and the median voter, a poor agent, sets the tax rate. If they do not democratize, then the tax rate is determined by the elites. Following this policy decision, the citizens decide whether to initiate revolution. Following the discussion in Chapter 5, revolutions generate private benefits for individuals who take part in them and there is, therefore, no collective-action problem. If revolution is attempted and a number ξp ≤ 1 — δ of the citizens take part, it always succeeds. After revolution, poor citizens expropriate the income of the elites. However, during revolution, a fraction µ > 0 of the income of the economy is destroyed. A high value of µ implies that revolution is relatively costly.

  These assumptions, as in the analysis of Chapter 5, imply that after revolution, each citizen receives a payoff of:

  (6.2)

  The elites are expropriated in revolution and we assume that they receive nothing (i.e., Vr(R, µ) = 0).

  We again say that the revolution constraint is binding if the citizens obtain more in revolution than when the elites implement their ideal policy, τr. Therefore, the revolution constraint is binding if VP(R, µ) = (1 — µ)/(1 — δ) > yP, or if:

  (6.3)

  As in Chapter 5, greater inequality (i.e., higher θ) makes the revolution constraint more likely to bind. Also, naturally, a low level of µ (i.e., greater income for the citizens after a revolution) makes revolution more attractive, and the revolution constraint (6.3) is more likely to bind. If the citizens undertake a revolution, branch R, then the game ends with payoffs to the citizens and to the elites of (Vp(R, µ), Vr(R, µ)).

  If democracy has been created and there is no revolution, we are along the branch (D, NR). In this case, the game ends with the tax rate preferred by the median voter being implemented. In this case, the citizens and the elites obtain payoffs of (Vp(D), Vr(D)) where, as before:

  (6.4)

  The alternative is for the elites not to choose democratization and set the tax rate themselves. In this case, the issue is whether the elites can credibly commit to certain concessions. We again model this in a simple way by introducing a “continuation game” in which with probability 1 — p the elites can reset the tax rate, whereas with probability p, they cannot and the tax rate chosen before the revolution decision is implemented. This allows us to model the idea that in a nondemocratic society, the elites may make a promise of high redistribution in the future but cannot necessarily commit to it - the crucial transitory nature of de facto political power.

  As discussed in Chapter 5, a more satisfactory approach is to have a repeated game, in which the elites can deliver the policy they promised today but can make no promises for the policies in the future, once the threat of revolution disappears. This is precisely the model we develop in Section 7, and we shall see that the current setup is similar to but, in many ways, much simpler than that dynamic game. Therefore, we prefer to start with this simpler setup to highlight the basic issues, returning to the more satisfactory framework later.

  To prevent a revolution, the elites may try to set a tax rate τN= , different from their ideal tax rate. This is the tax rate that will be effective when the elites do not democratize and are not able to reset the tax. Therefore, if the elites promise redistribution at the tax rate, the citizens choose not to revolt and nature does not allow the elites to reset the tax; the game ends with payoffs V (yp | τN =) and V (yr | τN= ). In contrast, if nature allows the tax rate to be reset, the elites will set their most preferred tax rate, τr. In this case, the payoffs are Vp(N) and Vr(N), where:

  Consequently, the expected payoffs from the promise of income redistribution can be written as (VP(N, τN), Vr(N, τN)), such that:

  which takes into account the fact that redistribution at the tax rate τN happens only with probability p. (Notice the difference between the notation Vi (N), which refers to values when the society is nondemocratic and unconstrained, and V’(N, τN), which refers to the case when the society is nondemocratic but the elites are forced to set a tax rate to avoid revolution. We use this type of notation later as well.)

  We now analyze the subgame perfect equilibria of this extensive-form game. To do so, we start at the end of the game tree and apply backward induction, as in Chapter 5. We refer to the actions of the elites and the citizens as σr = {φ, τN, N} and (σp= {ρ(·), τD}. The elites determine a tax rate τN ∈ [0, 1] and decide whether to create democracy φ ∈ {0, 1}, where φ = 1 indicates that democracy has been created. If there is no revolution and nature chooses ν = 1, then the elites get to reset the tax rate. Because the elites do not make a decision when ν = 0, we represent this as a choiceN ∈ [0, 1]. The citizens decide whether to initiate revolution, p ∈ {0, 1} (with p = 1 representing revolution); this decision is conditioned on the actions of the elites; hence, p : {0, 1} x [0, 1] → {0, 1}. Here, ρ(φ, τN) is the revolution decision when the elites make the democratization decision φ and set the tax rate τN. Finally, if φ = 1, then democracy is created and the poor get to set the tax rate τD ∈ [0, 1]. Then, a subgame perfect equilibrium is a strategy combination, {r,p}, such thatp andr are best responses to each other in all proper subgames.

  First, consider the situation in which the elites do not create democracy, promise a specific tax rate of τN =, and there is no revolution. This generates expected payoffs of:

  (6.5)

  If Vp(N, τN =) ≥ Vp(R, µ), then such a concession would stop revolution. Following the analysis in Chapter 5, we can define µ* such that at µ = µ*, we have Vp(R, µ*) = Vp(N, τN = τ p); that is, the citizens get the same payoff from revolution as from the elites promising the best tax rate for them, τp. (Of course, VP(N, τN = τp) < Vp(D) because in the former case, the elites are only promising this tax, and their promise is realized only with probability p.) This critical value of the revolution cost, µ*, is given from the equation Vp(R, µ*) = Vp(N, tN = τp) by:

  (6.6)

  When µ < µ*, then revolution is not costly and we have from the definition of µ* that Vp(R, µ) > Vp(N, τN= τp). Thus, even at the best tax rate, the promises of t
he elites are not sufficient to prevent revolution. The elites must, therefore, democratize to stop revolution. The strategy of democratization is feasible if democracy generates enough redistribution that the citizens do not revolt after democracy. This is the case when Vp(D) > Vp(R, µ), which is equivalent to:

  (6.7)

  When µ ≥ µ*, then revolution is sufficiently costly that the elites can prevent democratization by redistributing. In this case, they can stay in power by setting the tax rate at a level where the poor are just indifferent between revolting or not - that is,satisfies Vp(R, µ) = Vp(N, τN= ), which implies:

  and they do not democratize.

  Now we can see that there is a unique subgame perfect equilibrium; however, the character of this equilibrium depends on parameter configurations. First, when θ > µ and µ ≥ µ*, the elites can stay in power by setting a tax rate. More interesting, the unique pair of strategies that constitute an equilibrium when θ > µ and µ < µ* (and (6.7) holds) involve democratization by the elites to avoid revolution. It is useful to write the strategy profile for just this one case in full. Here, the following strategy profile is the unique equilibrium: for the elites, τN = 0, φ = 1, and τN = 0. For the citizens, ρ(φ = 0, ·) = 1, ρ(φ = 1, ·) = 0, and τD = τp. In this equilibrium, the elites create democracy and the citizens set the tax rate τD = τp. If democracy is created, then the citizens do not revolt (ρ(φ = 1, ·) = 0); but, off the equilibrium path, the citizens play ρ(φ = 0, ·) = 1 - that is, if democracy is not created, the citizens choose to mount a revolution. It is this credible threat of a revolution that induces the elites to democratize.