Economic Origins of Dictatorship and Democracy

by Daron Acemoglu

  The utility of an individual i is now (1 — τ)yi + T or i = p, r, where the government budget constraint again implies that:

  Incorporating the costs of taxation, we have the indirect utility of a poor agent as V(yp|τ) = (1 — τ) yp + (τ - C(τ)). The first-order condition of maximizing this indirect utility is identical to that which we derived before and, because we know from Chapter 4 that preferences are single-peaked, we can apply the medianvoter theorem to determine the (unconstrained) democratic equilibrium tax rate, again denoted τp. Using the fact that the incomes of the poor are given by (9.5) and average income is given by (9.4), this equilibrium tax, τp, is identical to the baseline tax rate in Chapter 4: namely, (4.11).

  4. Capital, Land, and the Transition to Democracy

  In this section, we embed the economic and political models of the previous two sections into our basic democratization model of Chapter 6, Section 6, which incorporated repression. The first issue we examine is how the structure of the economy influences the costs of repression. Following our previous discussion, we assume that repression creates costs for the elites depending on the sources of their income - in particular, whether they rely more on income from capital or income from land. As already discussed, it is plausible to presume that the disruption associated with putting down the threat of a revolution and an uprising by the citizens is more costly for industrialists, factories, and commerce than for land and landowners. As a result, when land is important for the elites, they are more willing to bear the cost of repression to avoid democratization. In a society in which income from capital becomes more important than income from land, it is more likely that the potential costs of repression exceed those of democracy and the elites prefer to give democracy to the dissatisfied citizens rather than use force against them.

  Given the parallels to the analysis we conducted before, we simply outline the model here. The rich elites have to decide whether to repress, democratize, or promise redistribution; if there is no repression, no democratization, and no revolution, nature decides once more whether the elites get to reset the tax they have promised. The game tree for this model is identical to Figure 6.2.

  The underlying economic model is the same as the one described in Section 2. The elites own capital and land. Moreover, all members of the elite have identical endowments so there is no heterogeneity among the elites (we return later to the distinction between industrialists and landowners). As before, the payoff to the citizens from revolution is Vp(R, µ) = (1 — µ)/ (1 — δ), whereas the elites always have Vr(R, µ) = 0.

  Repression is costly for the elites. So far, because income was exogenously determined, we simply assumed that repression (and coups) destroyed some fraction of income. The previous arguments, however, suggest that it may be more useful to imagine conflict actually destroying capital - this is what we assume in this chapter. However, all the results of this chapter apply when it is income that is destroyed as long as the fraction of income coming from capital that is destroyed is larger than the fraction that comes from land. Moreover, it can be verified that all results of previous chapters could be restated if we allow assets rather than income to be destroyed by repression, revolution, or coups.

  Consequently, if the elites choose to repress to avoid revolution and democratization, they will lose a fraction KK of the capital stock and a fraction KL of land. Moreover, we assumed:

  To reduce notation, we set κL = κ and κK =K where Q ≥ 1.

  The values to the citizens and to the elites if there is democracy are given by:

  (9.7)

  where the factor prices w, r, and v are given by (9.2) and the most preferred tax rate of the citizens is τp. These expressions take into account that once there is democratization, the citizens set their most preferred tax rate unconstrained.

  If, on the other hand, the elites choose repression, the payoffs are:

  (9.8)

  Finally, the elites could offer redistribution under the existing regime without democratizing and without resorting to repression. The best they can do in this case is offer redistribution at the favorite tax rate of the citizens, τp, given by (4.11); in this case, the values are:

  (9.9)

  which incorporates the fact that this promise is realized only with probability p.

  As before, if θ ≤ µ, the revolution threat is absent. The more interesting case for this discussion is the one in which θ > µ which, for simplicity, we assume to be the case. The promise to redistribute prevents a revolution if we have that Vp(N, τN= τp) ≥ Vp (R, µ). Using the same arguments as those in Chapter 6, this is equivalent to µ ≥ µ*, in which µ* is given by (6.6).

  If µ < µ*, the elites cannot prevent a revolution by promising redistribution, so they have to resort either to democratization or repression. We assume as usual that Vp(D) ≥ Vp(R, µ) so that democratization prevents a revolution; the formula for this is identical to (6.7).

  When do the elites prefer repression? This depends on whether µ ≥µ *. When µ ≥ µ*, the relevant comparison is between redistribution and repression because, for the elites, redistribution is always preferable to democratization when it is feasible. The case that is more interesting is when µ < µ* so that there is a trade-off between repression and democratization. In this case, the elites simply compare Vr(D) and Vr(O|K) as given by (9.7) and (9.8). It is clear that they prefer repression if Vr(D) < Vr(O|κ) or if:

  (9.10)

  It is useful to rewrite (9.10) in terms of the capital-to-land ratio, k = K/L. This gives:

  (9.11)

  as the condition under which repression takes place. We say that when k is higher, the economy is more “capital-intensive,” whereas low values of k correspond to relatively “land-intensive” societies. Equation (9.11) makes it clear that capital intensity of a society is a crucial determinant of whether repression is attractive for the elites. The key comparative statics arising from this condition are discussed in the next section. For now, we summarize the analysis in the following proposition:

  Proposition 9.1: Assume that (6.7) holds, θ > µ, and µ < µ*, where µ* is given by (6.6). Then, we have that:

  • If (9.11) does not hold, democratization happens as a credible commitment to future redistribution by the elites.

  • If (9.11) holds, the elites use repression to prevent a revolution.

  This proposition is therefore similar to the main results in Chapter 6. The interest here is that whether the condition determining if repression is desirable, (9.11 ), holds, depends on how capital-intensive the economy is (i.e., the level of k). The easiest way to see this is to consider the case where= 1 so that the costs of repression fall equally on capital and land. In this case, we have the following proposition:

  Proposition 9.2: Consider the game with = 1. Then, (9.11) is independent of k so the political equilibrium is unaffected by the capital intensity of the economy.

  In contrast, if> 1, it is straightforward to verify that (9.11) is less likely to hold as k increases. Therefore, let us define k* such that:

  (9.12)

  Then we can state:

  Corollary 9.1: Consider a society described by the game with > 1 and define k*by (9.12). Then, in the unique subgame perfect equilibrium, we have that if k < k*, then the elites will meet the threat of a revolution with repression, and if k ≥ k*, they will democratize in response to the threat of a revolution.

  This corollary is the main result of this section. It shows that a more capital-intensive society is more likely to become democratic. This is because the use of force by the elites is more costly in such a society compared to a land-intensive society or, expressed differently, capital investments make the elites more prodemocractic than land holdings (i.e, as discussed in the next section industrialists are more prodemocratic than landowners).

  5. Costs of Coup on Capital and Land

  We now move to extend these ideas to coups. Because of the parallels between using repression and mounting coups, there appear to b
e natural reasons for these costs to also depend on how capital-intensive the economy is. In particular, suppose that during a coup a certain fraction of the productive assets of the economy gets destroyed. Let the fraction of physical capital destroyed be ϕK and land be ϕL if a coup is undertaken. It is natural to think that:

  In other words, the disruptions associated with a coup are more destructive to capital than to land. The reasons that this is plausible are similar to those discussed previously. Coups and the associated turbulence and disruption lead to the breakdown of complex economic relations. These are much more important for capitalist production than agrarian production. This is natural because there is less concern about the quality of products in agriculture than in manufacturing. Moreover, the importance of complex relationships between buyer and supplier networks, and of investments in skills and in relationship-specific capital, is far greater in more industrialized activities. Therefore, land will be hurt less than capital as a result of a coup.

  Let ϕL = ϕ and ϕK = ξϕ where ξ ≥ 1. Given this assumption, we can write the incomes after coups as:

  (9.13)

  (9.14)

  Clearly, both expressions are less than the corresponding ones before the coup, (9.5) and (9.6), because the disruptions associated with a coup typically lead to the destruction of a certain fraction of the productive assets of an economy.

  Armed with this specification of the costs of coups, we can now analyze the impact of economic structure on coups and democratic consolidation. The game tree for the model in this section is identical to the one depicted in Figure 7.1.

  Whether the elites wish to mount a coup depends on the continuation value in democracy and nondemocracy. Faced with the threat of a coup, the median voter wishes to make a concession to avoid a coup (i.e., set τD < τp). After this, the elites decide whether to undertake the coup. If they do so, society switches to nondemocracy and the elites set the tax rate. Naturally, after a successful coup, they choose their most preferred tax rate, τN = 0. As a result, the game ends with respective payoffs for the citizens and the elites, VP(C, ϕ) =p and Vr(C, ϕ) =r, wherep andr are given by (9.13) and (9.14). Alternatively, if the elites decide not to undertake a coup, the political system remains democratic and with probability 1 — p, the median voter may get to reset the tax from that promised by the citizens in the previous stage. Therefore, with probability p, the tax promised by the citizens, τD, remains, and the citizens and the elites receive values V (yp| τD) and V (yr |τD), where:

  If, on the other hand, nature allows democracy to reset the tax, the citizens and the elites both receive the values pertinent to (unconstrained) democracy, Vp(D) and Vr(D), as given by (9.7). Therefore, the values resulting from a democratic promise of lower taxation at the rate τD are VP(D, τD) and Vr(D, τD), such that:

  (9.15)

  These expressions take into account that with probability 1 — p, the citizens get to reset the tax, the coup decision is already a bygone and, consequently, they choose their most preferred tax rate, τp.

  We can now characterize the subgame perfect equilibrium of this game by backward induction. Whether a coup is attractive for the elites depends on whether the coup constraint, Vr(C, ϕ) > Vr(D), binds. This states that a coup is more attractive than living under an unconstrained democracy. This coup constraint can be expressed as:

  (9.16)

  where we again write the expression in terms of the capital intensity of the economy k = K/L. When this constraint does not bind, coups are sufficiently costly that the elites never find a coup profitable - democracy is fully consolidated. Equation (9.16) is fairly intuitive and responds to changes in parameters in the way expected. For example, a greater democratic tax rate, τp, makes it more likely to hold because only the left-hand side depends on τp and is decreasing in it. A greater level of ϕ makes it less likely to hold because a greater fraction of the assets of the elites is destroyed in the process of a coup.

  In contrast, when this constraint binds, the democratic regime is not fully consolidated : if the citizens do not deviate from their most preferred to tax rate, there will be a coup along the equilibrium path. Therefore, we can define a critical value of the fraction of assets destroyed in a coup, denoted ϕ*, such that when ϕ < ϕ* (i.e., a coup is not too costly), the promise of limited redistribution by the citizens is not sufficient to dissuade the elites from a coup. Of course, the most attractive promise that the citizens can make to the elites is to stop redistribution away from them totally (i.e., τD = 0). Therefore, we must have that at ϕ*, Vr(D, τD = 0) = Vr(C, ϕ*), or:

  (9.17)

  This expression implies as usual that a higher level of τP makes democracy worse for the elites and, therefore, increases ϕ*; that is, the elites are willing to undertake more costly coups when τp is higher. We now have the following result:

  Proposition 9.3: In the game described above, there is a unique subgame perfect equilibrium such that:

  • If the coup constraint (9.16) does not bind, democracy is fully consolidated and the citizens set their most preferred tax rate, τp> 0.

  • If the coup constraint (9.16) binds and ϕ ≥ ϕ*, democracy is semiconsolidated. The citizens set a tax rate, τD =< τp, such that Vr (D, τD =) = Vr(C, ϕ).

  • If the coup constraint (9.16) binds and ϕ < ϕ*, democracy is unconsolidated. There is a coup, the elites come to power, and set their most preferred tax rate, τN = 0.

  The novel part of this result is that the likelihood of a coup is now affected by the economic structure - in particular, whether society is capital- or land-intensive. However, the only reason the degree of capital intensity affects the propensity of the elites to mount coups is that different fractions of capital and land are destroyed in the process of the coup (i.e., ϕK > ϕL) To emphasize this, we state an analogous result to Proposition 9.2:

  Proposition 9.4: Consider this game with ξ = 1. Then, (9.16) is independent of k, so the political equilibrium is unaffected by the capital intensity of the economy.

  The proof of this result follows from (9.17) because when ξ = 1, the term (k + σ)/(ξk + σ) = 1 and cancels from the right side. This proposition states that in the model here there is no link between economic structure and capital intensity when costs of coups are the same for capital and land holders.

  This picture changes substantially when ξ > 1, however. With a greater cost of coups on capital than land, (9.16) implies that as k increases, the coup constraint becomes less tight and from (9.17), ϕ* decreases. This implies that we can define two threshold levelsand k* such that at k =, (9.16) holds with equality. On the other hand, k = k* is such that when democracy promises τD = 0, the elites are indifferent between a coup and living in democracy. Naturally, k* <. This discussion establishes the next result:

  Corollary 9.2: Consider a society described by this game and assume that ξ > 1. Let and k*be as described. In the unique subgame perfect equilibrium, we have that if k < k*, then society is an unconsolidated democracy. If k* ≤ k < , then society is a semiconsolidated democracy. If k ≥, then society is a fully consolidated democracy.

  Therefore, in a land-intensive society where k is low, there will be coups during periods of crises. However, when the structure of production is different - that is, when capital is relatively more important in the production process and in the asset portfolios of the elites, as captured by the threshold level of capital intensity k* - then coups no longer happen along the equilibrium path and democracy persists. But, because k <, democracy is not a fully consolidated political institution and survives only by making concessions to the elites who pose an effective coup threat. As society becomes even more capital-intensive and k increases, it eventually becomes a fully consolidated democracy without the shadow of a coup affecting equilibrium tax rates and redistributive policies.

  This model, therefore, illustrates how the structure of the economy, in particular the extent of capital-intensity, influences the propensity of democracy to consolidate. The under
lying idea is that in a more industrialized society with a greater fraction of the assets of the elites in the form of physical capital, the turbulence and disruption associated with coups - like those created by repression - are more damaging. In consequence, coups as well as repression are less attractive in a capital-intensive society.

  6. Capital, Land, and the Burden of Democracy

  An even more important channel via which the economic structure may affect democracy is that the elites’ attitudes toward democracy also vary with the structure of the economy because there are typically different burdens of taxation on capital and land. In this section, we analyze a model with this feature. For brevity, we focus only on coups and democratic consolidations. Given the results in the preceding two sections, it is clear that the analysis of transition to democracy is similar; factors discouraging coups also discourage repression, facilitating transition to democracy.