Economic Origins of Dictatorship and Democracy

by Daron Acemoglu

  Perhaps paradoxically, a high q makes franchise extension less likely. A high q corresponds to an economy in which the citizens are well organized so they frequently pose a revolutionary threat. Alternatively, if µL is sufficiently less than one, then even in this state, the elites have to redistribute to the citizens. In this case, a low value of µL would also lead to the same result. A naive intuition may have been that in this case franchise extension would be more likely. This is not the case, however, because with a frequent revolutionary threat, future redistribution becomes credible. When the citizens have the power to oversee the promises made to them, there is less need for the elites to undertake a change in institutions to increase the future political power of the citizens.

  This result may explain why in the nineteenth century, Germany instituted the welfare state while allowing only a highly circumscribed democracy, whereas Britain and France democratized much more unconditionally. Social unrest against the existing system was as strong in Germany as it was in Britain and France. However, there were significant differences between the three countries in terms of the strength of the working class under the existing regime. Whereas there were no strong Socialist parties in Britain and France and trade unions were of little importance, the Social Democratic Party in Germany was by far the largest left-wing party in Europe at that time and the labor movement was strong (although not allowed to participate effectively in elections because of voting restrictions). For example, Nolan (1986, p. 354) explains the strength of the German workers movement as follows: “Although Britain experienced the first industrial revolution and France developed the first significant socialist associations, Germany produced the largest and best-organized workers’ movement in the late nineteenth century.” An alternative theory of democratization based purely on the strength of the working class would predict franchise extension in Germany before Britain and France. Proposition 6.3, which constructs a theory of democratization as a transfer of political power, in contrast, predicts that German elites should have had more flexibility in dealing with social unrest by promising future redistribution. This is also in part consistent with the actual evidence. Whereas Britain and France democratized and then increased redistribution toward the poor, Germany undertook redistribution without changing its nondemocratic regime. There is also little doubt that these redistributive measures were taken as a response to the potential revolutionary threat from the working class. Williamson (1998, p. 64), for example, writes that “the main aim of [the German] welfare programme was to avoid revolution through timely social reform and to reconcile the working classes to the authority of the state.”

  In addition, the distinction between the high and low state emphasizes that regime changes happen during unusual periods, perhaps economic crises or recessions. This is also in line with the evidence discussed in Chapter 3 (see also Acemoglu, Johnson, Robinson, and Yared 2004). Although in this book we capture these ideas using the reduced-form parameter µ so that the costs of revolution fluctuate directly, in Acemoglu and Robinson (2001) we showed how the same results follow from a model in which the cost of revolution is constant but total factor productivity fluctuates, as in standard models of the business cycle. In that model, changes in productivity change the opportunity costs of revolutions (and coups) and this has the same effects.

  8. Subgame Perfect Equilibria

  In the previous section, we characterized a subset of the subgame perfect equilibria of G∞(β). In this section, we analyze our basic dynamic model of democratization without the restriction to Markovian strategies. More specifically, we look for subgame perfect equilibria. In general, there are many subgame perfect equilibria of this game that are supported by various history-dependent strategies and our analysis mirrors that of Chapter 5. We are interested in understanding the extent to which punishment strategies can make redistribution in state µL credible. Thus, we look for the best possible equilibrium for the elites, which will be the one that prevents democratization for the largest set of parameter values. Therefore, implicitly we are interested in the maximum possible amount of credible redistribution to the citizens in the nondemocratic regime. To simplify, we abstract from the use of repression, although this can be easily added. As in Chapter 5, Section 7, the analysis in this section focuses on showing that a cutoff level of µ,** < µ* exists such that when µ ≥**, there will be redistribution without democratization, preventing revolution. In contrast, when µ <**, the equilibrium features democratization when µt, = µH.

  Exactly as in the analysis of Chapter 5, we study the circumstances under which the elites can redistribute at some tax rate τL > 0 in state µL, thus avoiding the transition away from the nondemocratic regime even when µ < µ*. There, we saw that the limitation on such redistribution was that it had to be incentive-compatible for the elite - that is, it had to be such that the payoff to the elites from redistributing according to the vector [τL, τH], given by the value Vr(N, µL, [τL, τH]), had to be greater than the payoff from deviating,

  There is only one substantive difference between the game we studied in Chapter 5 and this one: as long as (6.7) holds, when the nondemocratic regime collapses, there will be a transition to democracy. Therefore, the value(N, µL) here takes into account that when the elites deviate in state µL, their “punishment” in state µH is democratization instead of revolution as before. This is because it is not a subgame perfect strategy for the citizens to threaten a revolution after the elites democratize because they obtain greater payoff from democracy than revolution. Consequently, if the elites democratize, it forestalls a revolution. This implies that the valuefor the elites is given by the following recursion:

  where Vr(D) is as in (6.14).

  As before, only redistribution at the tax vector [τL, τH] such that:

  is credible. In addition, it is obvious that the derivations leading up to Vp (N, µH, [τL, τH]) in (5.38) in Chapter 5 still apply. So, the incentive-compatibility constraint for the elites will only differ from before because of the change inN, µL).

  As in Chapter 5, in general, the best equilibrium for the elites needs to consider the incentives to smooth taxes over time. However, to simplify the discussion and because the concept of tax-smoothing is not central to our analysis, we focus on characterizing the minimum value of µH such that the elites can avoid democratizing. We denote this** such that when µ ≥**, nondemocracy can be maintained with promises of redistribution. It is still the case that the maximum tax rate in the state µH is τp. So, we only need to find the maximum incentive compatible redistribution in state µL, which we now denote by’ By an identical argument, it is given by:

  Because Vr (D) > 0, the citizens can punish deviation less when the elites can democratize, which implies that deviation is more attractive for the elites. In consequence, it is immediate that’ <’, which satisfies (5.39).

  In addition, because the value of revolution to the citizens is also the same, the formula for the critical value of the cost of revolution,**, must be identical to the one derived for µ** in Chapter 5, with the value of’ derived there replaced by the new value of’. Thus, the critical value** can be easily found so that

  (6.22)

  The value of** implied by (6.22) is greater than the value of µ** in Chapter 5 because here, the potential punishments on the elites are less severe.

  More important, it is clear that** < µ* (where µ* is given by (6.17)) and we have as before that if µ ≥**, the elites can stay in power by redistributing. Equally important, when µ <**, contrary to Chapter 5, there is no revolution because the elites have an extra instrument - they can democratize.

  In summary, allowing the elites and the citizens to play non-Markovian strategies has implications in this model similar to those in Chapter 5. The threat of punishments by the citizens - in particular, the threat that they will undertake revolution - implies that some amount of redistribution can be sustained in state µL. It is interesting that this amount is actually lower here b
ecause the possibility for the elites to democratize limits the punishment that the citizens can inflict on them. Most important, however, is that the main thrust of the analysis of Chapter 5 applies. Although the ability to use punishment strategies increases the circumstances under which the elites can stay in power by making concessions, this does not eliminate the problem of credibility. When µ <**, concessions do not work because of the absence of sufficient future credibility, and the elites will be forced to democratize.

  9. Alternative Political Identities

  We now return to the model in Chapter 4, Section 4.4, in which we considered political conflict along the lines not of socioeconomic class but in terms of group X versus group Z. Recall that when group X is the majority, and taxes and the form of transfers are determined sequentially by majoritarian voting, there are two types of subgame perfect equilibria. In both types, redistribution is from group Z to the more numerous group X and, if> ½, the equilibrium tax rate will be the ideal point of poor members of X; if< ½, the equilibrium tax rate will be the ideal point of rich members of group X. We now discuss how that model can be embedded in our static model of democratization presented in Section 6 of this chapter.

  We think of nondemocracy as rule by group Z, who we will think of as the elites. Clearly, rule by the elites is no longer rule by the rich because some of the members of group Z are relatively poor. The first issue is the determination of the tax and transfer rates in nondemocracy and how key decisions, such as repression and democratization, are made. We assume that they are determined by majority voting in group Z, which implies that there are two cases to consider: one in which> δZ/2 and one in which the opposite holds. In this section, we do not attempt a comprehensive analysis of all possible cases; we proceed by assuming> δZ/2, which implies that it is the preferences of the rich members of Z that determine the social choices in nondemocracy. We also assume> ½, thus dealing with only one of the democratic equilibria outlined in Chapter 4. With respect to the tax rate, we maintain the notation τN for nondemocracy.

  All members of group Z prefer to set Tx = 0 and if there is no threat of revolution, then the unconstrained tax rate will be the one set by the median member of Z, a rich agent. Hence, the tax rate in nondemocracy is the ideal point of a rich member of Z, which satisfies the first-order condition:

  (6.23)

  which we assume to have an interior solution and in which we have used the fact that

  Therefore, in this case, redistribution goes from group X to group Z, with the equilibrium tax rate on incomeMoreover, no redistribution is given to group X, Tx = 0, and Tz =/δZ. Clearly, members of group Z prefer nondemocracy to democracy, whereas the opposite is true for members of group X.

  If the elites choose to repress, then we assume - following our analysis earlier in this chapter - that members of both Z and X incur costs of repression. The payoffs to members of group Z after repression are:

  (6.24)

  These equations follow because if the elites use repression, they will stay in power and they will also be able to transfer income from group X to themselves. The optimal tax rateis independent Of κ. The payoffs to members of group X after repression are:

  Imagine now that members of group X can engage in collective action and mount revolution against nondemocracy. Assume that this leads to the expropriation of all members of group Z but that, as in our main analysis, revolution is costly. Assume that after revolution, all income (not just the income of Z) is divided equally between members of group X. Because there is now heterogeneity within group X, we have to decide how to solve the social-choice problem that the group faces. To see where this problem originates, first note that the payoff to all members of group X from revolution is:

  for i = p, r, whereas without a revolution, the payoffs to the poor and rich members of X areandThus, there are now two revolution constraints:

  Recall that incomes are defined as(1 — α)andα), so thatis the fraction of the income of group X accruing to the rich in this group. Substituting these into the revolution constraints, we find

  (6.25)

  It is now immediate from the assumption that>which impliesthat the revolution constraint binds first for the poor. Thus, there can be situations in which the poor in group X favor a revolution whereas the rich do not. We solve this social-choice problem by assuming that group X makes decisions according to majority voting, which implies that the preferences of the poor, because they are more numerous, determine whether a revolution takes place. An equivalent alternative would be to simply assume that the poor in group X can undertake a revolution on their own.

  Faced with the threat of revolt by group X, the median voter of group Z wishes to make concessions by reducing the amount of redistribution toward himself and, in the limit, even giving redistribution to group X (i.e., set Tx > 0). As before, one can calculate the maximum amount of utility that group Z can credibly promise to group X. This involves setting Tz = 0; setting the tax rate preferred by a poor member of X, and setting Tx =Considering that any promise of redistribution is only upheld with probability p, this gives members of X the expected payoffs:

  These expressions incorporate the fact that with probability 1 - p, the elites will be able to reset the tax rate and, therefore, because the revolution threat has passed, they will be able to set their preferred tax rate,and members of group X will get no redistribution.

  We can use this to define a new µ* such that if µ < µ*, then concessions do not stop revolution. µ* is defined by the equationwhich implies:

  (6.26)

  The first main point to emphasize is that similar to our analysis in the case of conflict between rich and poor, if µ < µ* defined by (6.26), then the elites cannot stay in power by offering redistribution or concessions; they either have to repress or democratize. Thus, the basic mechanism around which our book is built-namely, that promises may not be credible without fundamental changes in the structure of political power - functions no matter what the nature of political identities are.

  All other trade-offs are qualitatively similar to before as well. For example, when µ < µ*, whether the elites democratize depends on how costly democracy is compared to repression, whereas if µ ≥ µ*, the elites have to decide whether to make concessions or repress.

  The main point of divergence is the comparative statics of this model, especially with respect to inequality. As discussed in Chapter 4, an increase in inter-group inequality can be captured by an increase in α. Consider the effects of α. If µ < µ*, the trade-off for the elites is between democratization or repression. A higher α leads the median voter in group X to favor higher tax rates, which makes democracy worse for members of group Z, favoring repression. If µ ≥ µ*, higher α increases the amount of redistribution that the elites have to offer group X to make it indifferent between a revolution and nondemocracy, again favoring repression. These results, with respect to inter-group inequality, are basically the same as those derived in Section 5. Changes in inter-group inequality in this section, however, do not necessarily map into changes in observed measures of inequality.

  Moreover, now consider the effects of an increase inthe share of group Z income that accrues to rich members of the group, holding α andconstant. An increase inunambiguously increases measured inequality. First, observe that whengoes up, the equilibrium tax rate levied in nondemocracy falls. Second, because the left side of (6.25) does not change, the benefit from having a revolution does not change. Therefore, because the tax rate levied in nondemocracy falls, a revolution becomes less attractive even though measured inequality has certainly increased.

  This brief analysis of conflict between two noneconomic groups illustrates that the basic mechanisms of democratization apply whichever political identities are relevant and also highlights that the comparative statics with respect to inequality may be quite different. This emphasizes that the robust predictions of our approach are those concerning the role of political institutions in affecting the future dis
tribution of power when promises are not credible.

  10. Targeted Transfers

  We now briefly discuss how the introduction of targeted transfers (see Chapter 4, Section 4.3) changes our results in the static model of Section 5. What we showed there and in Chapter 5 was that allowing for targeted transfers increased the burden of democracy on the elites, making it worse for the elites but better for the citizens. At the same time, this effect is reinforced by the fact that the elites could redistribute from the citizens to themselves in nondemocracy. Thus, the burden of nondemocracy on the citizens increases. Citizens dislike nondemocracy more, whereas elites like it better and fear democracy more. More generally, when transfers can be targeted, there will be greater distributional conflict in society (not only between rich and poor but also between any groups) because those in power can use the fiscal system more effectively to redistribute resources to themselves.

  The impact of increased conflict in our framework is obvious. First, targeted transfers make a revolution more attractive for the citizens because in nondemocracy, the citizens now pay taxes that are redistributed to the elites. The same argument also implies that nondemocracy is more attractive for the elites, and they are more willing to use repression.