Economic Origins of Dictatorship and Democracy

by Daron Acemoglu

  Figure 3.3. Evolution of Democracy 1960-1995.

  Figure 3.4. Evolution of Democracy 1840-2000.

  Figure 3.5. Democracy and Income 1990s.

  Figure 3.6. Democracy and Income 1990s.

  Figure 3.7. Democracy and Education 1990s.

  Figures 3.7 and 3.8 show another well-known correlation: more educated countries (i.e., those with higher levels of average years of schooling as reported in the data set by Barro and Lee 2000) also tend to be more democratic.

  Both of these patterns have been influential in the thinking of scholars working on democracy. In particular, the positive association between income and democracy (and, to a lesser extent, between education and democracy) has been the cornerstone of the famous modernization theory advocated by Lipset (1959) and many others since. Building on the insights of the modernization theory, many scholars today believe that democracy is only possible in sufficiently educated and rich societies. Furthermore, a common view both in the literature and the popular press is that an increase in economic prosperity and the level of education will naturally bring a process of democratization. Although influential, these views suffer from a lack of a well-articulated theory explaining when and how democracies emerge and consolidate. The purpose of this book is to develop such a theory and use it to understand, among other things, the potential links between economic prosperity and democracy.

  Figure 3.8. Democracy and Education 1990s.

  Another pattern in the data is emphasized by Przeworski et al. (2000). These authors argue and document that the positive association between income and democracy is largely driven by the tendency of rich countries to remain democratic, whereas poor countries have a greater tendency to experience decline in their democracy score (i.e., suffer coups and other actions against democracy). Figures 3.9 and 3.10 illustrate this tendency using the Przeworski et al. (2000) data. Figure 3.9 is a histogram of the fraction of countries of different income levels that start as nondemocracy and transition to democracy. The sample includes countries that were nondemocratic in 1965, 1970, 1975, 1980, 1985, and 1990, and measures democratizations in each case during the next five years. Countries are placed in income quintiles constructed according to the average world income distribution between 1965 and 1990. This figure shows that countries in the top two quintiles have a greater tendency to transition to democracy; however, there is no monotonic relationship between income and the fraction of nondemocracies that transition to democracy. Figure 3.10 is constructed analogously but for transitions from democracy to nondemocracy, rather than the reverse. There is a more striking relationship between transitions and income quintiles. Although countries at the bottom two quintiles face a high likelihood of transitioning into nondemocracy in any five-year period, this probability is much lower for those in the third quintile, and zero for those in the top two quintiles. These histograms make it clear that whereas the likelihood of transitioning to democracy is weakly correlated with income, there is a big difference between the fraction of rich and relatively poor democracies falling back to nondemocracy.

  Figure 3.9. Transitions to Democracy and Income 1970-1995.

  Figure 3.10. Transitions to Nondemocracy and Income 1970-1995.

  Figure 3.11. Democracy Growth and Income Growth 1970-1995.

  Although this is not the correct forum for reevaluating the existing empirical evidence, we emphasize that the patterns shown in Figures 3.5 through 3.10 do not correspond to causal effects of income and education on democracy and democratic transitions. More explicitly, these correlations do not establish that as a country becomes richer, it will necessarily tend to become more democratic. The major problem with a causal interpretation of these patterns is that countries that differ in income levels (or levels of educational attainment) also differ in histories and other institutional characteristics. Our recent work (Acemoglu, Johnson, Robinson, and Yared 2004) investigates this issue in detail and establishes that there is little causal effect of income (or education) on democracy or democratic transitions. Instead, other historical factors seem to determine both the economic and political development paths of various societies, leading to the types of correlations shown in Figures 3.5 through 3.10.

  It is sufficient to give a glimpse of this pattern by showing how changes in income are related to changes in democracy during the period covered by Figures 3.5 through 3.10. This is shown in Figures 3.11 and 3.12 for the Freedom House and Polity indexes. In both figures, the horizontal axis shows the change in log GDP per capita between 1970 and 1995, and the vertical axis shows the change in the democracy score between the same dates (for the Freedom House and Polity indexes, respectively). This way of looking at the data is useful because it differences out potentially fixed characteristics that are simultaneously affecting income and democracy (thus bringing us closer to the causal relationship between income and democracy). Both figures show a clear pattern: there is no relationship between changes in income per capita and changes in democracy. In other words, although richer countries are more democratic, there is no evidence that countries that grow faster than others tend to become more democratic, at least over this period. A natural interpretation of the patterns shown in Figures 3.5 and 3.6 in light of these results is that they are largely driven by some fixed country characteristics. Consequently, conditional on these characteristics, countries that have grown faster during the past twenty-five to thirty years have not become more democratic.

  Overall, a salient pattern in the data is the positive correlation between income and democracy, but this does not necessarily correspond to the causal effect of income on democracy. Therefore, part of the challenge to models of democracy and democratization is to understand how the world might have this positive correlation without a large causal effect. We return to this issue in Chapter 9.

  Figure 3.12. Democracy Growth and Income Growth 1970-1995.

  3. Democracy, Inequality, and Redistribution

  As discussed in Chapter 2, our approach to democracy emphasizes the role of social conflict, especially between different groups. One implication of this approach is that inter-group inequality should have an effect on the equilibrium of political institutions and thus on the likelihood that a society ends up as a democracy. The problem, however, is that the relevant notion of inter-group inequality is often difficult to measure (e.g., when it is between two different ethnic groups). Nevertheless, when the major conflict is between the rich and the poor, one variable that captures inter-group inequality is the share of labor income in GDP. The reasoning here is that, whereas the poorer segments of society obtain most of their income from labor, capital income (and sometimes land income) accrues largely to a smaller rich elite. Therefore, a high labor share corresponds to a low level of inter-group inequality when conflict is between rich and poor.

  Figure 3.13. Democracy and Inequality 1990s.

  Figures 3.13 and 3.14 show the relationship between the labor share in the 1990s and the relevant democracy indexes. The labor share data is from the United Nations, is also used by Rodrik (1999), and covers only the manufacturing sector, so it may be less than fully representative for the entire economy. Both figures show a positive association between the labor share and democracy.

  Figures 3.15 and 3.16 show the relationship between democracy and another measure of inequality: the Gini coefficient, which is the most common index of inequality in the literature and has a greater coverage of the various sectors of the economy than the labor share from the manufacturing sector (see Dollar and Kraay 2002 for more on these data). A higher value of the Gini coefficient corresponds to greater inequality. The relationship is similar to the one with the labor share, although less pronounced with the Polity data: countries that are more unequal and, consequently, have higher Gini coefficients tend to be less democratic.

  Figure 3.14. Democracy and Ineqeality 1990s.

  Figure 3.15. Democracy and Inequality 1990s.

  Figure 3.16. Democracy and Inequality 1990s.
>
  As emphasized in the context of the relationship between income and democracy, these correlations do not correspond to the causal effect of labor share or inter-group inequality on democracy. Moreover, these correlations are not always robust to the inclusion of other variables in a regression model, and a relatively large literature has not reached a consensus on the relationship between inequality and democracy. Whereas the claim that democracy is not possible in highly unequal societies is common in the nonquantitative literature (e.g., Dahl 1971; Huntington 1991; and the review in Bollen and Jackman 1985), the empirical evidence is more mixed. Using cross-sectional econometrics, Bollen and Jackman (1985) found no relationship between measures of inequality and democracy. Muller (1988, 1995) presented empirical evidence suggesting that higher inequality made dictatorships more stable and reduced the propensity of a society to democratize, although his results were criticized for being nonrobust by Bollen and Jackman (1995). More recently, Przeworski et al. (2000) investigated the effects of three measures of inequality on transitions to and from democracy using probit analysis. The measures they used were the Gini coefficient, the ratio between the share of total income going to the richest 10 percent of the population and the share going to the poorest 10 percent (the higher this ratio is, the greater is the inequality), and the share of income produced by manufacturing that accrues to workers. They found no relationship between democratization and either of the first two measures of inequality, noting that (p. 120) “the durability of dictatorships is unaffected by income distribution.” However, for the third measure, they found (p. 122) “dictatorships ... are much more vulnerable when the functional distribution of income is more unequal.” They also found that (p. 122) “democracies are less stable in societies that are more unequal to begin with, in societies in which household income inequality increases [when inequality is measured by the Gini coefficient or the ratio of top to bottom income shares], and in societies in which labor receives a lower share of value added in manufacturing.” Using a similar methodological approach, Boix (2003) reports results in which higher inequality reduces the propensity of a society to democratize.

  Figure 3.17. Democracy and Tax Revenue 1990s.

  Other scholars have examined the relationship among inequality, revolution, and political instability, which is also relevant to our approach. Here again, the findings are mixed, although Muller and Seligson (1987) and Alesina and Perotti (1996) found that greater inequality leads to greater political instability (see Lichbach 1989 for a review of this literature).

  The existing empirical literature is, therefore, rather contradictory and, more important, as already emphasized, focuses on correlations, not causal relationships. The correlations shown in Figures 3.13 through 3.16 are nonetheless informative. They suggest, for example, that models in which democracies are more redistributive and hence have a higher labor share, as well as models in which democracies can survive better in less unequal societies, can do a reasonable job of matching this pattern in the data.

  Figures 3.17 and 3.18 further suggest that at least part of the positive correlation among democracy and labor income and the Gini coefficient might be due to the greater tendency for redistributive policies in democracies. These figures show a positive association between the share of tax revenues in GDP and the democracy scores during the 1990s. Again, this is a correlation and should not be interpreted as a causal relationship.9

  Figure 3.18. Democracy and Tax Revenue 1990s.

  The historical evidence is also consistent with the notion that the patterns of redistribution change after democratization. Here, we briefly discuss some of the evidence; the reader is referred to Lindert (2004), for a more detailed and satisfactory discussion of the European experience. Although Figures 3.17 and 3.18 emphasize the association between democracy and fiscal redistribution, in practice, many other instruments- ranging from labor-market policies to educational policies - appear to be important in governments’ attempts to influence the distribution of income in society (DiNardo, Fortin, and Lemieux 1996; Wallerstein 1999).

  In Britain, the Reform Acts of 1867-84 were a turning point in the history of the British state. In 1871, Gladstone reformed the civil service, opening it to public examination and thus making it meritocratic. Liberal and Conservative governments introduced a considerable amount of labor-market legislation, fundamentally changing the nature of industrial relations in favor of workers. From 1906 to 1914, the Liberal Party, under the leadership of Asquith and Lloyd George, introduced the modern redistributive state to Britain, including health and unemployment insurance, government-financed pensions, minimum wages, and a commitment to redistributive taxation. As a result of the fiscal changes, taxes as a proportion of GNP more than doubled in the thirty years following 1870, and then doubled again in the subsequent thirty years. In the meantime, the progressivity of the tax system also increased (Lindert 2004).

  Meanwhile, the educational system, which was either primarily for the elite or run by religious denominations during most of the nineteenth century, was opened up to the masses; the Education Act of 1870 committed the government to the systematic provision of universal education for the first time which was made free in 1891. The school-leaving age was set at eleven in 1893 and increased to twelve in 1899; special provisions for children of needy families were introduced (Mitch 1993). As a result of these changes, the proportion of ten-year-olds enrolled in school that stood at a disappointing 40 percent in 1870 increased to 100 percent in 1900 (Ringer 1979, p. 207). Finally, a reform act of 1902 led to a large expansion in the resources for schools and introduced the grammar schools that subsequently became the foundation of secondary education in Britain.

  In France, the situation was similar. During the Second Empire, there was a significant expansion of government support for education; illiteracy fell from 39 to 29 percent of adults, and the primary-school enrollment rate increased from 51 to 68 percent (Plessis 1985, Table 14, p. 100). In 1881, the government abolished fees in public primary schools and, in 1882, it introduced seven years of compulsory education for children. The primary-school enrollment rate increased from 66 percent in 1863 to 82 percent in 1886. The “liberal” phase of the Second Empire saw significant labor-market legislation with strikes legalized in 1863, and unions were finally officially tolerated in 1868. Moreover, central-government expenditure as a percentage of GDP increased by one third from 9.4 percent in 1872 (a figure inflated by the Franco-Prussian War of 1870) to 12.4 percent in 1880 (Flora 1983).

  In Germany, a large increase in redistribution in the 1920s was initiated by the Weimar state (Flora 1983). Also, in Sweden, major redistribution appears to have started only after democratization. Lindert’s (1994) data show that before 1920, there was no redistribution in Sweden; after this date, it increased sharply. More generally, Lindert (2000b) shows that there is a strong historic relationship between democratization and educational expansion in Western Europe.

  Overall, we can summarize our discussion, especially the relationship between democratization and educational reforms, by quoting Easterlin (1981, p. 14):

  ... to judge from the historical experience of the world’s 25 largest nations, the establishment and expansion of formal schooling has depended in large part on political conditions and ideological influences and a major commitment to mass education is frequently symptomatic of a major shift in political power and associated ideology in a direction conducive to greater upward mobility for a wider segment of the population.

  4. Crises and Democracy

  An important element of our theory of democratization, as discussed in Chapter 2, is that transitions to democracy (and, similarly, transitions away from democracy) are more likely to occur amid economic and political crises, when there is a transitory shift in political power. The reason goes to the heart of our framework: changes in political institutions take place as a way of turning transitory de facto political power into more durable de jure political power. This reasoning suggest
s that we may expect a correlation between severe crises and transitions to and from democracy.

  Haggard and Kaufman (1995), in particular, emphasized that both democracies and nondemocratic regimes are destabilized by economic and political crises. They argue, for example, that “in Argentina, Bolivia, Brazil, Peru, Uruguay and the Philippines, democratic transitions occurred in the context of severe economic difficulties that contributed to opposition movements” (p. 45). Przeworski et al. (1996, p. 42), on the other hand, point out that: “the fragility of democracy ... flows largely from its vulnerability in the face of economic crises.” Przeworski et al. (2000, pp. 109-10) find that “most deaths of democracy are accompanied by some economic crisis; in twenty-eight out of thirty-nine instances, deaths of democracies were accompanied by a fall in income during at least one of the two preceding years.” (See also Londregan and Poole 1990, 1996; and Gasiorowski 1995 on the relationship between crises and coups.) Our historical discussion in Chapter 1 and the following section also illustrates that many of the key transitions to democracy during both the nineteenth and twentieth centuries have happened in periods of unusual social unrest and turbulence. Here, we show some additional evidence consistent with this pattern.