Economic Origins of Dictatorship and Democracy

by Daron Acemoglu

  The key in this section is that because land is supplied more inelastically, when allowed, the citizens impose higher taxes on land than on capital. Thus, everything else being equal, the elites are more opposed to democracy when land is more important for their incomes. This gives us another reason for land-intensive economies to be less likely to consolidate democracy (and also to transition to democracy).

  Let us now discuss this issue by assuming that there can be separate taxes on income from different sources: in particular, a tax rate on capital income, τk, and one on income from land, τL. Throughout, we simplify the discussion by assuming that there is no tax on labor income (i.e., the tax on labor, τN, is equal to 0). Clearly, the citizens would not like to tax their own incomes but, more generally, in a nondemocratic regime, the elites might like to tax the citizens and redistribute to themselves (as in discussions of targeted transfers in previous chapters). To simplify the exposition, we ignore this possibility by restricting attention to the case in which τN = 0.

  How do we model the costs of taxation when there are separate taxes on capital income and land income? The costs of taxation originate, in large part, from the fact that factors of production are supplied elastically. For example, labor taxation is “costly” because individuals take more leisure instead of supplying work to the market. There are two aspects to these costs, both of them relevant for this discussion. First, as less labor is supplied to the market, measured income and therefore tax revenues decline. This constitutes a cost for those who use tax revenues because there are fewer revenues now. Second, there is also a cost of allocative efficiency; without the taxation, labor was being allocated to its best use: market work. Taxation discourages this and creates a distortion by creating an incentive for time to be reallocated away from its most efficient uses, forcing it to be used where it is less valuable, in leisure or home production. Capital taxation is similarly costly, especially because capital can flee to other activities, or even abroad, and avoid taxes. Again, this response of capital is costly both because there are substantially less revenues from taxation and the allocation of capital between various activities is distorted. More generally, in all cases, distortions from taxation result because in its effort to avoid taxes, each factor is not being allocated to its most productive use, and measured market income on which taxes are collected is declining. It is also important that both of these costs relate to the “elasticity of the supply” of various factors. When a factor is supplied inelastically, it cannot be easily withdrawn from market activity; hence, measured income does not change and there are few distortions. Thinking of the supply elasticities as the major factor determining the costs of taxation immediately reveals that taxing capital should be more costly than taxing land. After all, capital can easily go to other sectors, but land is set in its place; at best, it can be withdrawn to inactivity.

  Motivated by these considerations, we think that when the tax on capital is τK, there is a cost of taxation equal to CK (τK) r K; when the tax on land is τL, the cost of taxation is CL (τL) vL. As before, we assume that both of these functions are continuous, differentiable, and convex. Moreover, we impose the usual boundary condition that(0) =(0) = 0 and a slightly different boundary condition(1) > 1 and(1) > 1 (the reason for this difference becomes clear later). The crucial assumption we make is that:

  This assumption implies that the marginal cost of taxing capital is always higher than the marginal cost of taxing land, which is equivalent to capital being supplied more elastically than land. The important implication of this assumption is that the citizens would like to impose greater taxes on land than on capital.

  To further simplify the discussion, we now depart in one more respect from our baseline model. As in our targeted transfers model, we assume that in addition to lump-sum transfers, there are transfers targeted to specific groups - in particular, to the citizens, Tp - as well as a lump-sum transfer to the elites, Tr.

  Given all these pieces, we can write the total post-tax incomes of the elites and the citizens as follows:

  which incorporates our assumption that all capital and land are equally owned by each member of the elite, and there are δ of them.

  The government budget constraint can now be written as:

  (9.18)

  The left-hand side of (9.18) is total expenditure on transfers. Tr is the lump-sum transfer that members of the elite receive and is thus multiplied by δ; Tp is the transfer to a citizen and is thus multiplied by 1 — δ. The right-hand side is total tax revenue from the taxation of capital and land. At the tax rates τK, τL, capital owners pay a total of τKτ K in tax and landowners pay τLv L. From these amounts, we subtract the costs of taxation, CK (τK)r K and CL (τL)v L.

  Given the availability of a targeted transfer to themselves, the citizens would simply redistribute all the income they raise from capital and land using this targeted transfer; hence, we have Tr = 0 in democracy.

  Next, because the citizens are no longer taxing themselves, their most preferred taxes are those that maximize the net tax receipts, the right-hand side of (9.18) - in other words, the citizens would now like to be at the top of the Laffer Curve, which relates total tax revenue to tax rate. Therefore, the citizens’ most preferred taxes can be computed simply by solving the following maximization problem:

  The first-order conditions are straightforward and give the most preferred taxes for the poor,implicitly as:

  (9.19)

  which maximize their net tax revenues. The assumption that(τ) <(τ) immediately implies that

  We next compute the net burden of democratic taxation on the elites. As in Chapter 4, we define the burden as the net amount of redistribution away from the elites. Because they receive no transfers now, this is simply equal to taxes they pay; hence:

  Using (9.3), we can write this relative to total income and in terms of capital intensity as:

  (9.20)

  First, note that from (9.19), the tax ratesandare independent of k. Then, (9.20) implies that as the economy becomes more capital-intensive, the burden of democracy on the elites will decrease. This reflects the fact that capital is less attractive to tax than land. Analytically, the burden of taxes, B, is decreasing in capital intensity:

  which follows immediately from the fact thatThis result implies that elites are less opposed to democracy for another reason when they are invested more in capital than in land; this is because democracy taxes capital less than it taxes land.

  There is another interesting interpretation ofSo far, we have emphasized the different tax rates imposed on incomes generated by land and capital. Another possibility is redistribution of assets. Because asset redistribution has not been explicitly considered in this chapter, we might think that the potential for asset redistribution is also incorporated into these taxesandAre there any reasons to think that the potential for asset redistribution is different for capital and for land? The answer is yes. Although democracy can easily redistribute land via land reform, redistribution of capital is more difficult because capital, in the form of factories, is not easily divisible. More important, when these factories are taken away from their owners and given to new parties, they typically are not very productive. This is because the complex relationships necessary for capitalist production - the specific investments, and the know-how - are all in the hands of the original owners and difficult or even impossible to transfer. One could argue that rather than redistribute the capital itself, shares in firms could be redistributed ; yet, the modern theory of the firm (Hart 1995) suggests precisely that the incentives of agents within a firm depend on the ownership structure so that capital cannot be arbitrarily redistributed without damaging productivity. Indeed, if capital markets are perfect, one would expect the initial ownership structure to be efficient (although if they are not, then the effects of redistribution are more complex; e.g., Legros and Newman 1996).

  Land is much easier to redistribute without creating distortions. When land is taken from b
ig landowners and redistributed to agrarian workers, the loss of efficiency may not be significant and, in fact, according to some estimates, there might even be a gain in efficiency because many of the big farms are owned by major landowners who farm more land than is efficient (Binswanger, Deininger, and Feder 1995 discuss evidence that land reforms may have efficiency gains; Besley and Burgess 2000 show that land reforms in India have had little adverse effect on aggregate economic performance). This suggests that land reform is often an attractive policy tool for democracies to achieve their fiscal objectives without creating major distortions. Naturally, this implies a greater burden of democracy on landowners than on capital owners. This consideration implies that when land is a more important asset of the rich, they have more to fear from democracy and typically they expect greater redistribution away from them and a greater burden. This could be captured by our result that

  We now put these two pieces together and analyze the likelihood of coups in a world with different taxes on capital and land. Consider the economic model described herein and the political model depicted by the game in Figure 7.1. We further simplify the discussion by assuming that the same fractions of capital and land are destroyed in the process of a coup (i.e., ϕK= ϕL or that ξ = 1). This assumption isolates the channel we want to emphasize in this section.

  If the citizens get to set their most preferred taxes and transfers, taxes on capital and land are given by (9.19), and we also have Tr = 0. This implies that the transfer to each citizen is given by:

  (9.21)

  The superscript p onindicates that it is the preferred value of the citizens. Therefore, the corresponding values are those in an unconstrained democracy :

  (9.22)

  with factor prices w, r, and v given by (9.2); withandgiven by (9.19); andgiven by (9.21 ).

  Whether the elites mount a coup depends on the continuation values in democracy and nondemocracy. The citizens again set taxes on capital and labor income, which are potentially different from their most preferred tax rates,anddenoted byK andL. The corresponding redistribution to a citizen is:

  (9.23)

  That the citizens would decide to cut taxes on capital and land rather than redistribute lump sum to the elites is obvious because these taxes are distortionary. If we had allowed labor income to be taxed, the citizens could find it optimal to tax themselves and transfer resources to the elites to avoid a coup.

  After this, the elites decide whether to undertake the coup. If they do, society switches to nondemocracy, and the elites set the tax rate. Naturally, they choose their most preferred tax rates,0. As a result, the game ends with respective payoffs for the citizens and the elites, Vp(C, ϕ) and Vr(C, ϕ), where:

  (9.24)

  Alternatively, if the elites decide not to undertake a coup, the political system remains democratic. In this case, nature moves one more time and determines whether democracy gets to reset the tax from that promised by the citizens in the previous stage. As before, this continuation game captures the fact that democracy may be unable to commit to less redistribution (i.e., to not adopting pro-citizen policies) once the threat of a coup disappears. Nature determines with probability p that the tax rates promised by the citizens remain, and the citizens and the elites receive values V (yp| andwhere:

  whereis given by (9.23).

  If, on the other hand, nature allows democracy to reset the tax, they both receive the (unconstrained) democracy values, VP(D) and Vr(D), as given by (9.22). Therefore, the values resulting from a promise of less redistribution only at the tax rates (K,L) by the citizens in democracy are VP(D,and Vr(D,such that:

  (9.25)

  with w, r, and v given by (9.2);andgiven by (9.19);given by (9.21); andgiven by (9.23). These expressions take into account that with probability 1 — p, the citizens get to reset the tax, in which case they are unconstrained and choose their most preferred taxesandas given by (9.19).

  We can now characterize the subgame perfect equilibrium of this game by backward induction. The crucial issues are whether undertaking a coup is in the interests of the elites and whether the citizens can prevent a coup by promising concessions.

  Whether a coup is attractive depends on whether the coup constraint, Vr(C, ϕ) > Vr(D), binds. The answer is yes when the burden of taxation on the elites is sufficiently high. Using (9.22) and (9.24), the coup constraint can be expressed as:

  (9.26)

  When this constraint does not bind, democracy is fully consolidated.

  In contrast, when this constraint binds, democracy is not fully consolidated: if the citizens do not take an action, there will be a coup along the equilibrium path. The action that the citizens can take is to reduce the burden that democracy places on the elites by reducing taxes on both capital and land. In particular, the best that the citizens can do is promise zero taxes on both Vr(D, to the elites. As in the previous analysis, we can then define a threshold value for ϕ, ϕ*, such that when ϕ < ϕ*, the promise of limited distribution by the citizens is not sufficient to dissuade the elites from a coup. Therefore, we must have that at ϕ*, Vr(D,= Vr(C, ϕ*). Solving this equality gives the threshold value ϕ* as:

  (9.27)

  Given this discussion, we can summarize the subgame perfect equilibrium of this game as follows:

  Proposition 9.5: In the game described above, there is a unique subgame perfect equilibrium such that:

  • If the coup constraint (9.26) does not bind, democracy is fully consolidated. The citizens set their most preferred tax rates on capital and land, > 0 and > 0, as given by (9.19).

  • If the coup constraint (9.26) binds and ϕ ≥ ϕ*, democracy is semiconsolidated. The citizens set taxes below and

  • If the coup constraint (9.26) binds and ϕ < ϕ*, democracy is unconsolidated. There is a coup, the elites come to power, and set their preferred tax rates, =

  Let us again define two threshold levels of capital intensityand k*, such that as the economy passes these threshold levels, it first becomes a semiconsolidated and then a fully consolidated democracy. These threshold values are:

  (9.28)

  and

  (9.29)

  Then, Corollary 9.2 applies exactly as before with k* andas given by (9.28) and (9.29). The result is, therefore, similar to before: as capital and industry become more important relative to land and agriculture, the elites become less averse to democracy and the threat against democracy diminishes. The reason this happens is different from before, however. In the model of the previous section, the burden of democracy was independent of the composition of assets of the elites; their different attitudes toward coups originated from the different costs that the disruption due to a coup would cause. Perhaps more important in practice is that not all segments of the elite suffer equally in democracy. This section emphasizes this by constructing a model in which land is taxed more heavily (or perhaps redistributed more radically by land reform); therefore, the elites have more to fear from democracy when land is an important source of income for them. As the degree of capital intensity increases, their opposition to democracy declines and consolidation is more likely.

  The implications of the model in this section carry over immediately to democratization. Because the burden of democracy falls more heavily on landowners than on capitalists, as the capital intensity of the economy increases, repression becomes less attractive relative to democracy and democratization becomes more likely to arise. Indeed, by analogy to the previous analysis, there exists a level of capital intensity that is sufficiently high to ensure that repression is never attractive to the elites.

  7. Conflict between Landowners and Industrialists

  The previous analysis showed how the increased capital intensity of an economy made coups against democracy less likely. To simplify the discussion, we allowed the composition of assets to change but we assumed that the elites were homogeneous, with each member holding the same share of capital and land. In practice, there are distinct groups - landowners and industrialists -
and certain groups are more opposed than others to democracy. Such distinctions are an enduring theme of the literature stemming from Moore (1966) and have emerged in the more recent literature on democratization under the guise of “hardliners” and “softliners.” In the previous chapter, we discussed how the distinction between a hardliner and a softliner could be given some content and microfoundations in the context of a model with both rich and middle-class agents. Nevertheless, in Chapter 8, incomes were still exogenous and the only difference between such agents was their income level.

  The models of this chapter provide another approach to this issue. In particular, because both the costs of repression and coups fall more heavily on capital holders than land holders and the burden of democracy is greater on the latter than the former, we expect capitalists and industrialists to be less opposed to democracy than landowners. Thus, we can imagine situations in which the elites split, capitalists are in favor of conceding democracy, and landowners are opposed to it.