Economic Origins of Dictatorship and Democracy

by Daron Acemoglu

  Possibly because of the reliance on reduced-form models, this game-theoretical literature has adopted the same dichotomy between structural and political approaches to explaining regime transitions first advocated by Linz and Stepan in the 1970s. For example, Colomer (2000), in a chapter entitled “Structural versus Strategic Approaches to Political Change,” argues that

  Two basic approaches can be distinguished in the literature on regime change and transitions to democracy. One emphasizes the structural, socioeconomic or cultural requisites of democracy.... The other approach looks at political regimes as outcomes of strategic processes of change. The main role is given here to choices and interactions by the actors. (p. 133)

  That such a dichotomy exists seems to be widely accepted by political scientists. Shin (1994) argues

  ... the establishment of a viable democracy in a nation is no longer seen as the product of higher levels of modernization, illustrated by wealth, bourgeois class structure, tolerant cultural values, and economic independence from external actors. Instead, it is seen more as a product of strategic interactions and arrangements among political elites, conscious choices among various types of democratic constitutions, and electoral and party systems. (pp. 138-9)

  The framework we develop is game theoretic and individuals and groups behave strategically based on individual motivations and incentives. Yet, individuals function within social and economic systems that both constrain their actions and condition incentives. In fact, there is no dichotomy at all between structural and strategic approaches - they are one and the same.

  Our approach, then, is to build much richer political-economy models from which we can derive empirical predictions about the incidence of democracy. We treat individuals’ preferences as given but allow people to differ with respect to their income, wealth, the form in which they hold their wealth, or their options and alternatives. From these fundamentals, we derive individual preferences over regime types. Thus, if a member of the elite is a hardliner, it is because we can show that “hardline” behavior is optimal for him given his preferences, endowments, and opportunities. We do not define people by their behavior.

  Although we know of no work of the same scope as ours, our results have been complemented by a number of other recent formal models of democratization. Most related is the research of Rosendorff (2001), who developed a model to argue that democratization occurred in South Africa because falling inequality made democracy less threatening for whites, an idea clearly related to one of the building blocks of our approach. Boix’s (2003) recent book develops a simple static version of a democratization model derived from our papers (2001) and close to the model sketched by Dahl (1971), and applies it to historical instances of democratization, particularly in Switzerland and the United States. Because his book uses the framework we developed in our published articles, it suggests several of the comparative statics we analyze in this book. For instance, Boix verbally discusses ideas about how trade, exit, and the structure of the economy influence redistributive politics and thus democratization. The research of Ellman and Wantchekon (2000) is also related to our analysis of coups; they show how the threat of a coup may influence the policies that political parties offer in an election. This is one element of our analysis in Chapter 7. Other papers by Feng and Zak (1999), Justman and Gradstein (1999), and Conley and Temimi (2001) provide different formal models of democratization. Another stream in the political economy literature - including both nonformal work by Kiser and Barzel (1991) and Barzel (2001) and theoretical models by Green (1993), Weingast (1997), Gradstein (2002), Bueno de Mesquita, Morrow, Siverson, and Smith (2003) and Lizzeri and Persico (2004) - builds on the idea that democracy is voluntarily granted by political elites because it solves some sort of market failure or contractual incompleteness. For instance, Green (1993) argues that the creation of legislative institutions was a way for rulers to credibly signal information. The other research, though differing in details, is based on the idea that rulers face a severe commitment problem because they cannot use third parties to enforce their contracts. Creating democracy, therefore, can be Pareto-improving because, by giving away power, a ruler can gain credibility.

  An alternative formal approach to democratization was proposed by Ades (1995), Ades and Verdier (1996), and Bourguignon and Verdier (2000). These papers assume that only wealthy citizens can vote and they study how, for a fixed-wealth threshold, changes in income distribution and economic development influence the extent of the franchise and, hence, the equilibrium policy. Another approach was developed by Ticchi and Vindigni (2003a), who analyze a model in which countries are engaged in interstate warfare and political elites democratize in order to give their citizens greater incentives to fight.

  PART TWO.

  MODELING POLITICS

  4

  Democratic Politics

  1. Introduction

  In this chapter, we begin to analyze the factors that lead to the creation of democracy. As discussed in Chapter 2, our approach is based on conflict over political institutions, in particular democracy versus nondemocracy. This conflict results from the different consequences that follow from these regimes. In other words, different political institutions lead to different outcomes, creating different winners and losers. Realizing these consequences, various groups have preferences over these political institutions.

  Therefore, the first step toward our analysis of why and when democracy emerges is the construction of models of collective decision making in democracy and nondemocracy. The literature on collective decision making in democracy is vast (with a smaller companion literature on decision making in nondemocracy). Our purpose is not to survey this literature but to emphasize the essential points on how individual preferences and various types of distributional conflicts are mapped into economic and social policies. We start with an analysis of collective decision making in democracies, turning to nondemocratic politics in Chapter 5.

  The most basic characteristic of a democracy is that all individuals (above a certain age) can vote, and voting influences which social choices and policies are adopted. In a direct democracy, the populace would vote directly on the policies. In a representative democracy, the voters choose the government, which then decides which policies to implement. In the most basic model of democracy, political parties that wish to come to office attempt to get elected by offering voters a policy platform. It may be a tax policy, but it may also be any other type of economic or social policy. Voters then elect political parties, thereby indirectly choosing policies. This interaction between voters’ preferences and parties’ policy platforms determines what the policy will be in a democracy. One party wins the election and implements the policy that it promised. This approach, which we adopt for most of the book, builds on a body of important research in economics and political science, most notably by Hotelling (1929), Black (1948), and Downs (1957).

  Undoubtedly, in the real world there are important institutional features of democracies missing from such a model, and their absence makes our approach only a crude approximation to reality. Parties rarely make a credible commitment to a policy, and do not run on a single issue but rather on a broad platform. In addition, parties may be motivated by partisan (i.e., ideological) preferences as well as simply a desire to be in office. Voters might also have preferences over parties’ ideologies as well as their policies. There are various electoral rules: some countries elect politicians according to proportional representation with multimember districts, others use majoritarian electoral systems with single-member districts. These electoral institutions determine in different ways how votes translate into seats and, therefore, governments. Some democracies have presidents, others are parliamentary. There is often divided government, with policies determined by legislative bargaining between various parties or by some type of arrangement between presidents and parliaments, not by the specific platform offered by any party in an election. Last but not least, interest groups influence po
licies through nonvoting channels, including lobbying and, in the extreme, corruption.

  Many of these features can be added to our models, and these refined models often make different predictions over a range of issues.10 Nevertheless, our initial and main intention is not to compare various types of democracies but to understand the major differences between democracies and nondemocracies. For instance, although the Unites States has a president and Britain does not, nobody argues that this influences the relative degree to which they are democratic. Democracy is consistent with significant institutional variation. Our focus, therefore, is on simpler models of collective decision making in democracies, highlighting their common elements. For this purpose, we emphasize that democracies are situations of relative political equality. In a perfect democracy each citizen has one vote. More generally, in a democracy, the preference of the majority of citizens matter in the determination of political outcomes. In nondemocracy, this is not the case because only a subset of people have political rights. By and large, we treat nondemocracy as the opposite of democracy: whereas democracy approximates political equality, nondemocracy is typically a situation of political inequality, with more power in the hands of an elite.

  Bearing this contrast in mind, our treatment in this chapter tries to highlight some common themes in democratic politics. Later, we return to the question of institutional variation within democracies. Although this does not alter the basic thrust of our argument, it is important because it may influence the type of policies that emerge in democracy and thus the payoffs for both the elites and the citizens.

  2. Aggregating Individual Preferences

  In this subsection, we begin with some of the concepts and problems faced by the theory of social or collective choice, which deals with the issue of how to aggregate individual preferences into “society’s preferences” when all people’s preferences count. These issues are important because we want to understand what happens in a democracy. When all people can vote, which policies are chosen?

  To fix ideas, it is useful to think of government policy as a proportional tax rate on incomes and some way of redistributing the proceeds from taxation. Generally, individuals differ in their tastes and their incomes, and thus have different preferences over policies - for example, level of taxation, redistribution, and public good provision. However, even if people are identical in their preferences and incomes, there is still conflict over government policy. In a world where individuals want to maximize their income, each person would have a clear preference: impose a relatively high tax rate on all incomes other than their own and then redistribute all the proceeds to themselves! How do we then aggregate these very distinct preferences? Do we choose one individual who receives all the revenues? Or will there be no redistribution of this form? Or some other outcome altogether?

  These questions are indirectly addressed by Arrow’s (1951) seminal study of collective decision making. The striking but, upon reflection, reasonable result that Arrow derived is that under weak assumptions, the only way a society may be able to make coherent choices in these situations is to make one member a dictator in the sense that only the preferences of this individual matter in the determination of the collective choice. More precisely, Arrow established an (im)possibility theorem, showing that even if individuals have well-behaved rational preferences, it is not generally possible to aggregate those preferences to determine what would happen in a democracy. This is because aggregating individual rational preferences does not necessarily lead to a social preference relation that is rational in the sense that it allows “society” to make a decision about what to do.

  Arrow’s theorem is a fundamental and deep result in political science (and economics). It builds on an important and simpler feature of politics: conflict of interest. Different allocations of resources and different social decisions and policies create winners and losers. The difficulty in forming social preferences is how to aggregate the wishes of different groups, some of whom prefer one policy or allocation whereas others prefer different ones. For example, how do we aggregate the preferences of the rich segments of society who dislike high taxes that redistribute away from themselves and the preferences of the poor segments who like high taxes that redistribute to themselves? Conflicts of interest between various social groups, often between the poor and the rich, underlie all of the results and discussion in this book. In fact, the contrast we draw between democracy and nondemocracy precisely concerns how they tilt the balance of power in favor of the elites or the citizens or in favor of the rich or the poor.

  Nevertheless, Arrow’s theorem does not show that it is always impossible to aggregate conflicting preferences. We need to be more specific about the nature of individuals’ preferences and about how society reconciles conflicts of interest. We need to be more specific about what constitutes power and how this is articulated and exercised. When we do so, we see that we may get determinate social choices because, although people differ in what they want, there is a determinate balance of power between different individuals. Such balances of power emerge in many situations, the most famous being in the context of the Median Voter Theorem (MVT), which we examine in the next subsection.

  To proceed, it is useful to be more specific about the institutions under which collective choices are made. In particular, we wish to formulate the collective-choice problem as a game, which can be of various types. For instance, in the basic Downsian model that we consider shortly, the game is between two political parties. In a model of dictatorship that we investigate in Chapter 5, the game is between a dictator and the disenfranchised citizens. Once we have taken this step, looking for determinate social choices is equivalent to looking for the Nash equilibrium of the relevant games.

  3. Single-Peaked Preferences and the Median Voter Theorem

  3.1 Single-Peaked Preferences

  Let’s first be more specific about individual preferences over social choices and policies. In economic analysis, we represent people’s preferences by a utility function that allows them to rank various alternatives. We place plausible restrictions on these utility functions; for example, they are usually increasing (more is better) and they are assumed to be concave — an assumption that embodies the notion of diminishing marginal utility. Because we want to understand which choices individuals will make when their goal is to maximize their utility, we are usually concerned with the shape of the utility function. One important property that a utility function might have is that of being “single-peaked.”

  Loosely, individual preferences are single-peaked with respect to a policy or a social choice if an individual has a preferred policy; the farther away the policy is from this preferred point, in any direction, the less the person likes it. We can more formally define single-peaked preferences. First, with subsequent applications in mind, let us define q as the policy choice; Q as the set of all possible policy choices, with an ordering “>” over this set (again, if these choices are simply unidimensional [e.g., tax rates] this ordering is natural because it is simple to talk about higher and lower tax rates); and Vi(q) as the indirect utility function of individual i where Vi : Q → R. This is simply the maximized value of utility given particular values of the policy variables. It is this indirect utility function that captures the induced preferences of i. The ideal point (sometimes called the “political bliss point”) of this individual, qi, is such that Vi(qi) ≥ Vi(q) for all other q ∈ Q. Single-peaked preferences can be more formally defined as follows:

  Definition 4.1 (Single-Peaked Preferences): Policy preferences of voter i are single-peaked if and only if:

  Strict concavity of Vi(q) is sufficient for it to be single-peaked.11

  It is also useful to define the median individual indexed by M. Consider a society with n individuals, the median individual is such that there are exactly as many individuals with qi < qM as with qi > qM, where qM is the ideal point of the median person.

  To assume that people have single-peaked prefere
nces is a restriction on the set of admissible preferences. However, this restriction is not really about the form or nature of people’s intrinsic tastes or utility function over goods or income. It is a statement about people’s induced preferences over social choices or policy outcomes (the choices over which people are voting, such as tax rates); hence, our reference to the “indirect utility function.” To derive people’s induced preferences, we need to consider not just their innate preferences but also the structure of the environment and institutions in which they form their induced preferences. It usually turns out to be the features of this environment that are crucial in determining whether people’s induced preferences are single-peaked.

  We often make assumptions in this book to guarantee that individual preferences are single-peaked. Is the restriction reasonable? Guaranteeing that induced preferences over policies are single-peaked entails making major restrictions on the set of alternatives on which voters can vote. These restrictions often need to take the form of restricting the types of policies that the government can use - in particular, ruling out policies in which all individuals are taxed to redistribute the income to one individual or ruling out person-specific transfers. Assuming preferences are single-peaked is again an application of Occam’s razor. We attempt to build parsimonious models of complex social phenomena and, by focusing on situations where the MVT or analogues hold, we are making the assumption that, in reality, democratic decision processes do lead to coherent majorities in favor of or against various policies or choices. This seems a fairly reasonable premise.