We now have the following result:
Proposition 6.1: There is a unique subgame perfect equilibrium {r, p} in the game described in Figure 6.1, and it is such that:
• If θ ≤ µ, then the revolution constraint does not bind and the elites can stay in power without democratizing or redistributing income.
• If θ > µ, then the revolution constraint binds. In addition, let µ* be defined by (6.6). Then:
(1) If µ > µ*, the elites do not democratize and set the tax rate to redistribute enough income to avoid a revolution.
(2) If µ < µ* and (6.7) holds, concessions are insufficient to avoid a revolution and the elites democratize.
(3) If µ < µ* and (6.7) does not hold, there is a revolution.
The most important conclusion to be drawn from Proposition 6.1 is that democracy arises to avoid a revolution when the promises of the elites to make policy pro-citizen are not sufficiently credible. Note that the lower is p, the less credible are such promises, the higher is µ* and the less likely it is that concessions will avoid a revolution. Thus, it is lack of credibility that forces the elites to democratize. Moreover, inequality must be sufficiently high (θ > µ) that a revolution becomes attractive in the first place. Before investigating the comparative statics of this model in detail and discussing more of its implications, we introduce repression.
6. Democratization or Repression?
So far, we have studied the trade-off between concessions and democratization when the citizens can challenge the power of a nondemocratic regime. However, as mentioned in Chapter 2, rather than make any type of concession, nondemocracies often respond with force to block political change. There are many examples of this. In December 1989, the Ceausescu regime in Romania attempted to block democratization by using the military. This tactic backfired when the army decided to side with the demonstrators, leaving only the secret police loyal to the regime. Similarly, in Tiananmen Square in June 1989 in China, the Communist Party used tanks to crush the pro-democracy movement rather than make any type of concession. Another relevant example is the military junta in Burma (Myanmar) maintaining its power by using force to repress all opposition. We now introduce repression into the model of the previous section and study the circumstances under which democracy emerges when repression is an option. The analysis initially begins by assuming that if the elites decide to repress the citizens, this always succeeds. In line with this assumption, O’Donnell and Schmitter (1986) noted:
... no transition can ever be forced purely by opponents against a regime which maintains the cohesion, capacity, and disposition to apply repression. (p. 21)
Nevertheless, later in this chapter we consider situations where repression may fail, in which case revolution can happen in equilibrium.
Pre-tax incomes are given by (4.7), except that now there can also be costs due to repression that affect net income. In particular, the post-tax net return of agent i is:
(6.8)
where Δ is the cost due to repression, with ω = 0 denoting no repression and ω = 1 denoting repression. We model the cost of repression as we did the cost of revolution. If the elites decide to repress, then all agents lose some fraction of their income in the period of repression. We assume that Δ = 1 — K, which makes the effective cost of repression equal to κ yi. We adopt the assumption that the citizens lose the same fraction of income as the elites only for symmetry; this plays no major role in the analysis because the repression decision is made by the elites.
Figure 6.2. Democratization or Repression.
The game is identical to that depicted in Figure 6.1 except that now the elites first choose among promising redistribution, using repression, or creating a democracy - see Figure 6.2. If they use repression, it always succeeds and the game tree ends with payoffs (Vp(O | κ), Vr(O | κ)), where the letter O refers to “oppression” (because R is already taken for revolution). With repression, the elites maintain power and can set their most preferred tax rate:
If the elites opt against repression, they can choose democracy, and the rest of the tree is the same as in Figure 6.1.
The analysis closely mirrors that of the previous section. First, the calculations leading to µ* are unchanged so that, exactly as before, if µ ≥ µ*, the elites can maintain power by making concessions, whereas if µ < µ*, they cannot. However, whatever the value of µ, the elites have the choice to repress. To understand what will happen in equilibrium, we have to compare the payoff to the elites from repressing to the payoff from democracy or concessions. Bearing this in mind, we can define two threshold levels for the cost of repression,and, such that the elites are indifferent between their various options at these threshold levels. More specifically, letbe such that:
or, in other words,
(6.9)
Therefore, at, the elites are indifferent between redistribution and repression. As a result, for all κ <, they prefer repression to promising redistribution. Recall that K is the fraction of income destroyed by repression so, the lower it is, the more attractive repression will be. This implies that one set of parameter configurations in which repression emerges is when µ ≥ µ* and κ <.
Next, define the other threshold such that:
or, more explicitly,
(6.10)
At, the elites are indifferent between democratization and repression. As a result, for all K <, they prefer repression to democratization. Therefore, another set of parameter values in which repression will be an equilibrium outcome is when µ < µ* and k <.
Both threshold levelsandare increasing in inequality - that is, increasing in θ. For example, totally differentiating (6.10), we have:
To see why this is so, notice that (τp- C(τp))is the per capita transfer from the government budget constraint; we must have C(τp) — τp < 0, which gives -δ (C(τp) - τP) /θ2 > 0. Next, — δ + θ > 0 follows from yr> yP and we also know that dτP/dθ > 0. Hence, d/dθ > 0.
That greater inequality increasesand is intuitive. Greater inequality makes redistribution more costly for the elites and, all else being equal, makes repression more attractive relative to democracy and relative to the promise of redistribution. This makes the elites more willing to undertake repression even if it is more costly.
We can now state a proposition outlining the nature of the equilibria in this game. To do this, we again adopt the intuitive approach. The nature of the strategies is similar to that discussed in Proposition 6.1, the only differences being that the elites initially have to decide whether to repress, ω ∈ {0, 1}, and the revolution decision of the citizens is conditioned on ω in addition to φ and τN. Again, a subgame perfect equilibrium is a strategy combination {r, p}. Democracy results when θ > µ, µ < µ*, and κ ≥ .
We now have the following result:
Proposition 6.2: There is a unique subgame perfect equilibrium in the game described in Figure 6.2, and it is such that:
• If θ ≥ µ, then the revolution constraint does not bind and the elites can stay in power without repressing, redistributing, or democratizing.
• If θ > µ, then the revolution constraint binds. In addition, let µ* be defined by (6.6) and and be defined by (6.9) and (6.10). Then:
(1) If µ ≥ µ* and K ≥ , repression is relatively costly and the elites redistribute income to avoid a revolution.
(2) If µ < µ* and k < or k ≥and (6.7) does not hold, or if µ ≥ µ* and K < , then the elites use repression.
(3) If µ < µ*, (6.7) holds, and K ≥ , concessions are insufficient to avoid a revolution and repression is relatively costly so the elites democratize.
As in Proposition 6.1, democracy arises as a credible way to make policy more pro-citizen. Whether democratization will happen depends on the values of µ and κ. When θ > µ and µ is lower than µ*, revolution is relatively attractive and, given that the promises made by the elites are only imperfectly credible, it is unlikely that any tax rate that the elites promise before a revolution will ever be implemented. In thi
s case, even when the elites offer the most desirable possible tax rate, τp, the citizens prefer revolution. Anticipating this, the elites must either repress or democratize to avoid being expropriated in a revolution. Repression is attractive when κ is relatively low, so democracy arises when a revolution is sufficiently remunerative to the citizens and repression costly enough to the elites. Repression is also used when the creation of democracy is insufficient to stave off a revolution.
When concessions do not work because they are not credible, the elites must democratize or repress. In Acemoglu and Robinson (2000b), we showed that there may be another important reason why concessions do not work. We developed a model in which the elites’ strength and ability to repress is private information. Strong types can easily repress a revolution whereas weak types cannot. When faced with a revolution, we showed that there are circumstances where an elite that does not repress but instead makes concessions such as income redistribution may be inferred to be weak. In this case, concessions can actually encourage a revolution. We showed, therefore, that concessions are not used because of the information they may transmit to the citizens and the elites must repress or democratize.
6.1 Comparative Statics
We now investigate the comparative statics of the equilibrium in more detail. It is interesting to analyze the relationship between inequality and democratization.
For low levels of inequality, in particular for θ ≤ µ, democratization never occurs because the threat of revolution is not binding. Democratization, therefore, requires that the society be sufficiently unequal (i.e., θ > µ) so that revolution is a threat. Intuitively, in highly equal societies, the citizens do sufficiently well under the status quo distribution of assets that they never wish to contest power and democratization never occurs (unless, perhaps, as we discuss later in the chapter, the elites have a strong intrinsic preference for democracy that outweighs the loss from redistribution). Moreover, inequality has to be high enough that the promise of redistribution is not sufficient to stave off the revolutionary threat; in particular, θ > θ*, where:
Here, we use the notation τp(θ*) to emphasize that the tax rate preferred by the median voter depends on the extent of inequality. This needs to be considered when calculating the comparative statics. Clearly, θ* > µ because p (τp (θ*) (θ* — δ) — ( 1 — δ)C(τP (θ*))) > 0. Therefore, an increase in inequality starting from low levels makes democratization more likely. From (6.7), we can define another critical value of θ,, such that:
where> θ*. This inequality follows from the fact that p < 1 and τp(θ)(θ — δ) — (1 - δ)C(τp(θ)) is increasing in θ. To see this latter result, note that the derivative of this expression is:
This is so because by the envelope theorem (i.e., the first-order condition that defines τp), (θ - δ - (I - δ)C’(τp(θ))) = 0 and also τP > 0. Thus, there is a range of inequality levels θ ∈ (θ*, ] where democracy will be conceded, avoiding revolution.
However, when inequality is very high,andare relatively high, and the elites prefer repression rather than suffer high levels of redistribution. Therefore, democratization only occurs for intermediate levels of inequality. The important theoretical point here is that the citizens prefer democracy to nondemocracy because it is more redistributive, and this preference becomes stronger as inequality increases. By the same token, the elites prefer nondemocracy, and they do so more intensely when inequality is higher and they expect more redistribution away from them in democracy. The higher the inequality, the more attractive nondemocracy is relative to democracy for the elites. Therefore, in a highly unequal society, the elites will use their resources to garner force and prevent revolution without democratizing.
For a given cost of repression, K, we can implicitly define a critical threshold of inequality,(κ), such that
Then, democratization requires that inequality is less than this threshold, or θ ≤(κ). Define θmin = min{,(κ)}. We now state:
Corollary 6.1: There is a nonmonotonic relationship between inequality and democratization. In particular, when θ ≤ θ*, the society remains nondemocratic and the elites maintain power; when θ > θmin, the society remains nondemocratic with repression. Democratization occurs when θ ∈ (θ*, θmin].
If≤ (κ), then before repression becomes attractive, (6.7) does not hold and - given that θ > θ* so that concessions do not work - the elites are forced to repress to avoid revolution. If> (κ), then when the critical level of inequality(κ) is reached, although it would be feasible to avoid revolution by democratizing, the elites find it more attractive to repress.
The results in Proposition 6.2, especially those in Corollary 6.1, may help us understand some comparative patterns of democratization discussed in Chapters 1 and 3. Although all Western European countries democratized by the early twentieth century, in parts of Latin America, such as Paraguay, Nicaragua, and El Salvador, dictatorial regimes survived practically the entire century by using repression to avoid democratization. This was also the case in African countries such as Zimbabwe (Rhodesia) until 1980 and South Africa until 1994. Such outcomes are explicable in our model because the extent of inequality in those societies made democratization very costly to the elites, leading them to prefer repression.
It may also be the case that repression was relatively cheap in those countries - for example, in Central America - because the disenfranchised were Amerindians who were ethnically distinct from the elites who were primarily descendents of Spaniards. Similarly, in Rhodesia and South Africa, the enfranchised were white whereas the disenfranchised and repressed were black Africans. In Chapter 2, Section 6.1, we discussed how the organization of civil society is important for democratization. If civil society is disorganized and ineffective, then it may be difficult to solve the collective-action problem to form threats to the existing regime, and any such attempt may be easier to repress. The long history of racial domination in both Central America and Southern Africa may be important in explaining the evolution of civil society. In Guatemala, for example, forced labor was still used until 1945, and government policies restricted labor mobility and the ability to organize collectively (McCreery 1994). In South Africa, the apartheid regime issued banning orders and pass laws and placed restrictions on the educational and career opportunities of black Africans. In both cases, these factors helped to fragment civil society and allowed the nondemocratic regimes to persist.
Figure 6.3. Gini Coefficient (Korea-Taiwan and Singapore). Sources: Singapore: Economic Growth Research, Deininger and Squire Data Set Korea-Taiwan: Bourguignon and Morrison (2002).
When the model is made even richer, the costs of repressing may also be influenced by such things as the form of wealth held by political elites. Later, we show that it may be significant that in all these countries, the political elites were primarily landowners. Indeed, the creation of democracy in these countries may have coincided with important changes in the elites’ assets.
Proposition 6.2 also suggests the reason why there seem to be so few pressures toward political change in Singapore. For instance, Case (2002) notes
... “despite the emergence of a large middle class and suggestions that society is generally growing more participatory, social forces have failed to cumulate in any strong pressures for democracy.” (p. 81 )
Our analysis suggests that this absence may be due to the low levels of inequality in Singapore. Figure 6.3 shows data on inequality in Singapore from the Deininger and Squire dataset. This dataset, compiled by the World Bank,17 gives measures of inequality only from 1973 because there are no historical data on inequality in Singapore from the colonial period. The data show that inequality has been persistently low in Singapore since independence and has shown no tendency to rise. Figure 6.3 also shows data from Bourguignon and Morrisson (2002) on the historical pattern of inequality in Taiwan and South Korea, two other Asian countries that experienced delayed democratization. The picture is similar to that of Singapore, except for the large fall between 1950 a
nd 1960 when agrarian reforms were implemented.
Finally, two recent empirical papers by Epstein, Bates, Goldstone, Kristensen, and O’Halloran (2004) and Papaioannou and Siourounis (2004) find tentative support for this nonmonotonic relationship between democratization and inequality that we first proposed in Acemoglu and Robinson (2001).
The costs of taxation also affect the form of the equilibrium and whether democratization will arise. When C(·), especially C’(·), is low, τp can be higher and there will be more redistribution in democracy. Although this makes democracy more attractive for the citizens, somewhat paradoxically it may also make it less likely to arise in equilibrium. This is because as the tax that the elites can promise increases, they can prevent revolution without democratization.
Economic Origins of Dictatorship and Democracy Page 28