Democratic Government and Development: A Survey

Recent decades have witnessed an unprecedented expansion of democracy. During the third wave of democratization, as described by Samuel Huntington, democracy spread well beyond its historical boundaries and it is now adopted in all major regions of the world. Yet, not all democracies are equally effectual in delivering good governance and progrowth policies. Why do democratic institutions induce good governance and prosperity only in some economies? This paper presents an overview of the dimensions along which successful and unsuccessful democracies differ. It argues that four socioeconomic variables are of critical importance to create and maintain a well-functioning democracy: (i) social capital, (ii) information, (iii) education, and (iv) equality. History also plays an important role as do the contingencies characterizing the collapse of authoritarian regimes and the emergence of democratic institutions.


I. Introduction
Amartya Sen won the Nobel Prize in Economics, in part, for assessing the importance of democracy as a universal value in the sense that a vast majority of individuals, independent of their geographic location and specific characteristics, and at almost all times, may have reason to see democracy as valuable. Its value does not rest on just one particular merit, but rather on a plurality of virtues. Sen (1999) emphasizes the intrinsic importance of political participation and freedom in human life, and the constructive role of democracy in the formation of values and the understanding of needs, rights, and duties.
Beyond these widely accepted virtues of democratic institutions, Sen also considers a more disputed dimension of democracy: the instrumental importance of political incentives in keeping governments responsible and accountable. Periodic, free, fair, broadly participatory, and genuinely contested elections, together with a system of norms and institutions that support the right to vote and make it meaningful, give people the power to control and discipline the government. This ensures the implementation of policies favoring the population as a whole rather than specific (political or economic) segments of the population. In this sense, democratic institutions are both responsive to the demands of constituents and effective in using limited resources to address these demands.
The available empirical evidence, however, suggests only a weak correlation between the extension of democratic liberties and good policy making. Various societies that are considered democratic, according to standard measures, are seriously marred by institutional malpractices of different kinds (see, for example, Acemoglu, Robinson, and Santos-Villagran 2013). Figure 1 plots the correlation between a commonly used index of democracy and a standard indicator of government quality. 1 It shows that while most 1 As is standard in the literature, this paper measures the quality of government with an aggregate index that jointly considers corruption and competency indicators. The Quality of Government Index is the simple average of the following International Country Risk Guide variables: control of corruption, law and order, and bureaucratic quality. (Corruption and bureaucratic quality are highly correlated in the data.) The Quality of Government Index ranges between 0 and 1, with higher values associated with a higher quality of government. Democracy is measured using an average of the Polity IV Project and Freedom House indexes of democracy. The Polity IV Project classifies different political regimes on the basis of the formal rules and institutions in place; it looks at the formal inclusiveness and contestability of the political system-the quality of political representation-and at the constraints on executive power. Freedom House takes the view of "democracy as freedom" and classifies political regimes on the basis of well-governed states also have well-established democratic institutions, many equally democratic states perform poorly on governance. 2 This paper examines some of the reasons behind the variation in performances of modern democracies and presents recent developments in research on the functioning of democratic institutions. This paper will also probably open more questions than it will offer answers, which is partly its intention. Despite the growing interdisciplinary work on democratic governance, we still have a limited understanding of the subject. This paper is organized as follows. Section II presents the debate on the relative merits of democratic institutions and discusses the critical socioeconomic dimensions along which modern democracies differ. Section III looks at one additional dimension: the role of history and of different paths to democracy to understand current and observable differences in governance. Section IV examines in some detail the case of East Asian economies and how democracy evolved in the region. Section V concludes by advancing policy suggestions to foster the functioning of democratic institutions in developing economies.

II. Democracy and Good Government
In the last 40 years, democracy has spread well beyond its historical boundaries in the Western hemisphere. Between 1973 and 2003, more than 60 economies all over the world adopted democratic institutions either after the collapse of larger states (e.g., the Soviet Union) or the end of dictatorships and authoritarian dominantparty systems (e.g., the Republic of Korea). By the beginning of the 21st century, the only major region of the world lacking democratic economies was North Africa and the Middle East, but recent events in this region are apparently putting an end to this Arab anomaly. This visible globalization of democracy has fomented a revival of the old debate on the relative merits of democratic and autocratic regimes dating back to Plato and Aristotle. Acemoglu, Johnson, and Robinson (2001); Acemoglu et al. (2008); and Acemoglu and Robinson (2012) emphasize the importance of inclusive political institutions as a fundamental factor in explaining cross-economy differences in income per capita. observable outcomes in terms of political and civil rights. The combined measure employed in this paper takes into account both conceptions and ranges between 0 and 10, with higher values associated with greater levels of democracy.
2 Interestingly, the correlation is overall positive for consolidated democracies (+6 to +10) confirming the instrumental importance of democratic institutions for governance, but it turns out to be negative for autocratic regimes (-10 to -6). The quality of government is, in fact, relatively high in fully institutionalized autocracies and tends to decrease as we move toward relatively more democratic regimes. This U-shape suggests that economies ruled under a consolidated regime (either democratic or autocratic) tend to be better governed than economies in transition.
In line with this optimistic view on the interactions between democratic institutions and economic dynamics, North (1990) argues that modern democratic societies with universal suffrage offer the most favorable institutional structure for the functioning of the market. Other authors focus on the benefits of democracy for the functioning of the sociopolitical system. Olson (1993) suggested that the democratic electoral process is essential to solving commitment issues, while according to Wittman (1989), democratic institutions minimize the transaction costs of sociopolitical organizations. 3 More skeptical theories of democracy emphasize instead the inefficiencies of representative government. In his foundational study on political development, Huntington (1968) argued that populist demands for consumption and taxation might be channeled easily through democratic institutions with negative consequences on economic performance. 4 Becker (1983) models how a democratic polity can yield inefficient outcomes by enabling various interest groups to compete for political influence, while Besley and Coate (1998) synthesize a vast literature that models the distortions caused by incumbent politicians running excessive deficits to guarantee re-election.
These different views on the complex linkages between democracy and the economy triggered a lively debate in the empirical literature as well. Rodrik and Wacziarg (2005), and Papaioannou and Siourounis (2008a) find democracy (or democratization) has a positive effect on growth rates, and Persson and Tabellini (2008) find, on average, a negative effect on growth of leaving democracy. 5 On the contrary, Barro (1996 and2000) and Glaeser et al. (2004) find no evidence for a direct causal effect of democracy on economic growth and, consistent with public choice theories, Tavares and Wacziarg (2001) show that democracies are associated with low levels of private investment and high government spending.
The debate in the literature shows no sign of an emerging consensus on the merits of democracy and its effects on economic growth. Rather than studying the average effect of democracy, this paper looks more carefully to its variation: when are democratic institutions more likely to breed success and when are they associated with economic failure? So far, a number of different dimensions have been identified by economists, political scientists, and sociologists to distinguish democratic winners and losers.

A. Social Capital
The literature in political science and sociology has long argued that social capital is of critical importance to create and maintain a democracy. 6 When citizens interact often, join groups, and trust each other, their relationships aid democratization by crystallizing and organizing opposition to a nondemocratic regime. 7 Once a democracy is established, these relationships help consolidate democratic institutions by affecting both the quantity and quality of citizens' political participation (Paxton 2002).
The historical tradition of civil society and democracy holds that an individual's participation in civil associations creates feelings of duty, increases a sense of interdependence with others, and produces a habit of participation. Membership in voluntary associations therefore increases the amount (or quantity) of political participation. This view is supported by vast empirical evidence that documents the linkages between membership in voluntary associations and political participation (see, for example, Verba, Schlozman, and Brady 1995).
The bonds of social capital also affect the nature (or quality) of participation. Because relationships increase communication and the flow of information, they increase the exposure of individuals to political ideas and foster the creation of an informed and reasoned public opinion. New ideas and opinions are therefore more quickly disseminated throughout the population, while extremist ideas are more easily challenged, as they have less chance of remaining isolated (Habermas 1989).
Another strand of literature focuses on the role of social capital in the formation of common expectations. According to Fukuyama (1995), a community that regularly expresses collaborative behavior based on a set of common norms develops mutual expectations. In turn, these expectations, or social trust, facilitate profitable exchanges by reducing risks and making behavior more predictable.
Along the same lines, Putnam (1993) argues that generalized social trust is based on norms developed within networks in which members of the community are engaged, and that democracy requires generalized social trust to operate fruitfully. He shows that the newly empowered regional governments introduced in Italy in the early 1970s with fairly similar organizational structures succeeded only in those regions blessed with vibrant networks and norms of civic engagement. Where social life was fragmented and a culture of distrust diffuse, the performance of the 6 More recently, the economic literature has recognized the importance of social norms and civic culture, and their effects on economic performance (see, for example, Tabellini 2010). 7 Fortunato and Fugazza (2014) show that certain forms of social capital may instead contribute to the stabilization of nondemocratic regimes. Their results emphasize the need to distinguish carefully between the mechanics of trust (the existence of mutually accepted norms) and the political content embodied in such norms. new executive bodies was less responsive to public demands as politics took on patron-client characteristics.
Blending together empirical and anecdotal data, Putnam (1993) describes two radically different Italian political traditions. While the Normans laid the foundations for patron-client politics in the South, the cities and towns of the North governed themselves as communal republics with an emphasis on community trust and mutual security. Putnam (1993) argues that these different civic traditions have radically affected Italians' relations with modern democratic institutions and, in turn, the functioning of these institutions. In the North, residents have inherited what he calls horizontal traditions, emphasizing civic associations and egalitarian political participation. In contrast, Southern Italians have received only vertical traditions, characterized by a distrust of institutions and engagement in patron-client relationships. 8 This scenario resembles Banfield's idea of amoral familyism, defined as a social equilibrium in which people care exclusively about their nuclear family (and expect everybody else to behave likewise) and disregard common goods, thereby preventing the development of well-functioning institutions (Banfield 1958).
More recent contributions confirm the existence of a positive association between the quality of democratic institutions and social capital by replicating Putnam's results with up-to-date and more disaggregated data (Rice and Sumberg 1997).

B. Information
While the political science literature emphasizes the interaction between democracy and social capital, the literature in economics has traditionally examined democratic governance focusing on the distortions brought about by the presence of informational asymmetries between citizens and politicians. These asymmetries reflect the nonperfect observability of the actions undertaken by politicians in office and the characteristics of the candidates running in an electoral competition (e.g., ability and honesty).
The first class of informational problems gives rise to moral hazard and rentseeking, and is the object of the classic political economy models of democracy. Barro (1973) and Ferejohn (1986) consider principal-agent settings where voters are the principals who seek to control the actions of the agents they elect (policy makers). In a world of scarce information in which the actions of policy makers are only partly observable by citizens, elected officials face strong incentives to appropriate rents. As information dissemination increases, however, voters are better able to use elections to screen and discipline politicians. Hence, political accountability rises and the space left for rent-seeking declines.
A vast series of papers has offered empirical support to these predictions by showing the positive effects of information diffusion on democratic governance. The literature on the role of mass-media invariably finds that an increased media presence is associated with the dissemination of information and improvement of electoral accountability Burgess 2002, Djankov et al. 2003). Alternative sources of information have been shown to be important for democracy ;Djankov et al. (2010), for example, focus on the disclosure of politicians' finances and business activities. They examine the relationship between disclosure rules and several measures of the quality of government using a database of the financial and business disclosures of members of the lower house of parliament in 175 economies. They find that high-and upper-middle-income economies require disclosure and make the information publicly available more often than the rest of the world. They also find that public disclosure is associated with better government. According to their estimates, the introduction of public disclosure norms increases an economy's government effectiveness score in the World Bank's Worldwide Governance Indicators by 0.242 points.
The idea that a lack of information constrains electoral accountability is also consistent with the higher incidence of clientelism in low-income settings. Using evidence from the introduction of conditional cash transfers for the poor in Mexico, De La O (2013) argues that voters use the incidence of clientelistic policies to learn about politicians' behavior. Villages that were randomly chosen for the program 21 months before the 2000 presidential election saw an increase in turnout of 7% and an increase in the incumbents' vote share of 16% as compared to villages that entered the program only 6 months prior to the election. She interprets these findings as evidence that the amount of time since program entry predicts the extent of preferences for incumbents, suggesting that citizens who learned the most about the program changed their voting behavior the most.
The second class of informational asymmetries-related to politicians' characteristics in terms of talent, virtue or preferences, generates problems of adverse selection. The literature in this area concentrates on the way public officials are recruited rather than on the decisions they take once in office. It suggests that bad politicians are relatively more likely to enter when formal returns to politics are low and more likely to be selected if information about candidates quality is limited (Caselli and Morelli 2004;and Besley, Pande, and Rao 2005).
When information about politicians' quality is limited, ethnic politics tend to prevail. Ethnic networks provide informal insurance and enable information flows (Miguel and Gugerty 2005). Furthermore, a politician's ethnic identity is often a good predictor of redistributive preferences (Besley, Pande, and Rao 2007). All else being equal, both reasons will cause voters to favor politicians belonging to their own ethnicity when the information about the candidates' quality is limited. This will provide the party that represents the majority ethnic group with an electoral advantage and can reduce elections to a mere head-counting exercise, lessening their role as a source of accountability.
Consistent with this hypothesis, Banerjee and Pande (2009) demonstrate the existence of a positive correlation between ethnic politics and economy-level measures of political corruption. They argue that the forces that make a politician's ethnic identity more salient reduce voter emphasis on honesty, and provide empirical support for this thesis using panel data from a North Indian state.

C. Education
One of the classical theories of political sociology, formulated by Martin Lipset and often referred to as the modernization hypothesis, maintains that as all societies grow they are headed toward more modern and democratic ways of cohabitation. Higher levels of education, in particular, naturally lead to a betterfunctioning democracy and stronger institutions (Lipset 1960).
Throughout history, educated people have shown a remarkable willingness and ability to fight for (and protect) democracy because of their interest in ideas and their ability to organize. Nowhere is this clearer than in the remarkable propensity of students to battle for political and civil rights, as recently shown by the nature of the protest movements that paved the way for the Arab Spring. 9 Democracies in the Netherlands, the United Kingdom, and the United States (US) have also regularly been defended by people steeped in legal and political learning. It is not surprising then that recent empirical research identifies a positive correlation between levels of education and the extent of democratization (Glaeser et al. 2004, Papaioannou andSiourounis 2008b). Inspired by this evidence, Glaeser, Ponzetto, and Shleifer (2007) developed a model in which democracies have more, albeit less-motivated, supporters than dictatorships, since under a dictatorial regime insiders earn large political rents while in democracy the political rents are shared among many people. They show that relatively more people fight for a more inclusive regime as human capital increases.
The positive interaction between democracy and education can also be explained by looking at the impact of education on political engagement and participation (social capital), and on citizens' ability to make good ex ante electoral choices and to evaluate ex post the actions of elected officials (informational asymmetries). 10 The idea that education might positively affect social capital dates back to the socialization hypothesis first raised by Samuel Bowles and Herbert Gintis. Bowles and Gintis (1976) suggested that education is primarily about teaching people how to interact successfully with others (see also Knack and Keefer 1997). Successful interaction includes understanding and appreciating others' point of view, as well as being able to effectively communicate one's own. When people communicate successfully, they may control any innate antisocial tendencies and become more productive participants in social and political activities. 11 In fact, in most empirical studies, education emerges as the strongest predictor of political participation.
Using data on 324 national lower house elections between 1972 and 1995, Blais and Dobrynska (1998) find that turnout increased by 16 percentage points when illiteracy declined from 85% to 12%. The cross-country correlation between education and turnout is paralleled by a positive correlation between years of schooling and propensity to vote in the microdata for rich economies. For the US, the classic survey data study by Wolfinger and Rosenstone (1980) shows that citizens with a college degree are significantly more likely to vote than people with fewer than 5 years of schooling. (They estimate a difference of around 38 percentage points.) Education can also attenuate the informational problems that threaten the quality of governance by equipping citizens with the cognitive skills they need to be effective participants in a representative democracy. Education is indeed likely to increase citizens' ability to select able leaders, understand the issues upon which they vote, and recognize corrupt public officials (Milligan, Moretti, and Oreopoulos 2004;Ostrom 2006). In short, the severity of both adverse selection and moral hazard problems seem to be linked, at least in part, to the level of education of the citizenry. Fortunato and Panizza (2015) empirically examine the role of education in explaining the success or failure of democratic governance without looking at the specific channel of interaction; that is, without decomposing the impact of education into problems of civic culture, selection, or incentives. Using cross-country and panel data regressions they show that the correlation between democracy and the quality of government is not statistically significant in countries with low levels of education and is positive and statistically significant in countries with high levels of education. In particular, they find a negative but not statistically significant relationship for countries with low levels of education that becomes positive and statistically significant when average education reaches 8 years. They conclude that the success of democratic institutions is closely related to the educational attainment of the population. 12 In discussing their findings, Fortunato and Panizza (2015) acknowledge that causality is hard to establish because their key explanatory variables are likely 11 Almond andVerba ([1963] 1989, 315) famously wrote that "the uneducated man or the man with limited education is a different political actor from the man who has achieved a higher level of education." 12 Fortunato and Panizza (2015) also find a positive and statistically significant marginal effect of education on the quality of government in democratic countries, and no significant effect of education in nondemocratic countries. This ancillary result suggests that democracy is of critical importance to channel the political benefits of education into government performance. to be endogenous and there are not good external instruments for education and democracy. 13 They deal with this issue by running a set of Monte Carlo simulations aimed at testing the robustness of the estimates and show that even the presence of extreme endogeneity does not reverse the findings; in the worst case scenario, severe endogeneity leads to statistically insignificant estimates of the parameters of interest.
The problem of endogeneity is, of course, more general than that. It goes beyond the relationship between education and quality of government in democratic settings and equally affects the other factors discussed in this survey. At the end of the day, the argument of this paper is that democratic institutions are endogenous; their form and their functioning depend on the conditions under which they emerge and endure. Conditions shape institutions and institutions partly transmit the causal effects of these conditions. The question, thus, is how to distinguish the effects of institutions from those of the conditions that give rise to them.
In econometrics jargon, this represents an identification problem; solving it ultimately relies on finding a credible and transparent exogenous source of variation in the data. This is what most of the empirical papers reviewed here have done. Given the aim of this survey, however, the focus is on the economics behind the results, rather than on the specific empirical tests employed. Interested readers are requested to seek out the original sources for the econometric details.

D. Inequality
In his seminal investigation on the virtues of democratic institutions, Alexis de Tocqueville ([1835] 1998) warned of the possible dangers associated with democracy in societies characterized by large economic disparities. He described an ideal state with democratic institutions in place coupled with perfect economic equality, but also suggested that equality of political rights might lead to a "tyranny of the majority" in which the distribution of resources is uneven. In unequal societies, so the argument goes, democracies granting more equal political influence may experience excessive redistributive pressure, and therefore curtail property rights protection and distort the incentives for individual entrepreneurship. Low inequality and correspondingly low redistributive pressure, on the contrary, would facilitate the implementation of good economic institutions and policies under a democracy.
Along the same lines, the modern endogenous fiscal policy literature models the interaction between polity type, income inequality, and economic performance. The models proposed by Alesina and Rodrik (1994) and Persson and Tabellini (1994) show that high levels of income inequality raise the demand for redistribution. If under democracy fiscal policy and taxation are selected endogenously by majority voting, then high inequality will lead to lofty taxation, lower investment, and (consequently) slower growth. Autocratic regimes would be more insulated from such a mechanism since citizens in the lower tail of the income distribution are denied a political voice. Cervellati, Fortunato, and Sunde (2008) expand this reasoning and, in line with de Tocqueville's view, explicitly model the impact of inequality on the quality of institutions, in particular on property rights protection, under different political regimes. They show that efficient democracies, which are able to enforce property rights, cannot emerge and consolidate if economic inequality is sufficiently large. Under these conditions, and in line with Hobbes' Leviathan metaphor, all social groups prefer to leave control over policies in the hands of a rich and powerful elite. Conversely, a social contract emerges only under democracy if inequality is sufficiently small. Sunde, Cervellati, and Fortunato (2008) bring these predictions to the data. In their cross-sectional analysis, they find evidence for a significant interaction effect between democracy and inequality in determining the quality of growth-promoting institutions like the rule of law. They find that democracy is associated with better rule of law only when inequality is low.
In an influential contribution, Piketty (2014) highlights an additional channel through which inequality may dampen the functioning of democratic institutions. Extremely high levels of inequality raise the risks of institutional capture by part of those (few) in the upper tail of the income distribution, thereby seriously threatening one of the pillars of modern liberal democracies: the principle of representation.

III. Democratization and Good Government
The previous section examined four critical elements that partly explain the huge variation in the performances of democratic regimes across space and time. A fifth element, however, has been so far left aside: history.
Historically, authoritarian regimes have collapsed and new democracies emerged under very different scenarios involving varying levels of conflict. While in some economies democratization has been essentially peaceful and well accepted (or even actively promoted) by the elite, in others the establishment of democracy was the result of violent social conflicts triggered by an uprising of the politically and economically deprived classes and fueled by failed attempts at repression. The threat of revolution and social unrest, for example, played an essential role in the establishment of voting rights for the disenfranchised people in many Latin American economies (e.g., Uruguay in 1919, Colombia in 1936, Venezuela in 1945, and Nicaragua in 1979, while historical examples of consensual regime transitions include the Nordic economies. Different transition scenarios, in turn, may affect the functioning of the new democratic institutions. Transitions, defined as periods of regime change, are formative and founding moments. As such, they set a society on a path that shapes its subsequent political (and economic) development. After discussing at length the occurrences of violence during the democratic transitions of the third wave of democratization, Huntington (1993) concludes that violent uprisings are less likely to lead to a regime change and such uprisings when successful may lead to worse democracies. 14 A.
Type of Democratization Cervellati, Fortunato, and Sunde (2012) propose a model of endogenous democratization that analyzes the determinants of different transition scenarios and their consequences for the features of emerging democracies. The model predicts that democratic transitions taking place amid broad consensus in the population foster the establishment of democracies with good institutional quality, since individual behavior can be coordinated during the transition process. A transition initiated by part of the ruling elite, in fact, can function as a signal and facilitate the settlement of a social contract. Cervellati, Fortunato, and Sunde (2014) broaden this view by investigating the incentives of different social groups to engage in violence, and by studying the role of openly violent conflict during the democratization process in shaping the features of emerging democracies. The theory predicts that violent transitions are less likely to lead to well-functioning democracies with high levels of protection of civil and economic liberties. The results also suggest that a minimum consensus among the population is required for the emergence of sound democratic institutions, and that consensual and peaceful transitions are more likely if economic interests are fairly aligned in the population. This is the case if, for example, economic power is not concentrated among a very small group of elites.
The prediction that the level of violence characterizing the regime transition may persistently affect the future prospects of democracy is consistent with the empirical evidence from the third wave of democratization over the period 1970-2003. The comparison of civil liberties between economies that experience either a peaceful democratization or a violent transition to democracy reveals that the peaceful route leads to a substantially, and persistently, larger improvement in civil liberties. Figure 2 illustrates this finding by plotting the evolution of average civil liberties before and after a permanent democratization. Lower index scores depicted in the graph imply more civil liberties. The plot distinguishes between economies with a violent transition to democracy and economies in which the transition to democracy was peaceful. 15 While the overall pattern in both groups is similar, there 14 As Huntington (1993, 207) stated: "Governments created by moderation and compromise are ruled by moderation and compromise. Governments created by violence are ruled by violence." 15 Civil liberties are represented by an aggregate index provided by Freedom House. It includes the following categories: freedom of assembly, demonstration, and public discussion; the possibility to form political parties and organizations; religious institutions; and the existence of an independent media. The aggregate index takes a value is a clear difference in the level of civil liberties after the transition. After a nonviolent democratization, civil liberties are about one index point higher than the world average, corresponding to an improvement of about 1.5 index points. Economies with a violent transition experience a much less pronounced improvement.
The political science literature also recognizes the importance of the contingencies characterizing the transition to democracy. Remmer (1990), among others, suggests that the opportunities and possibilities that follow a change of regime will make a real difference only when the great majority of the population participates actively in the process of democratization. This analysis is corroborated by a study by Freedom House (2005) examining the political dynamics of 67 economies that had experienced transitions from authoritarian rule over the last 3 decades. The report studies the relationships between the mode of civic involvement in democratic regime change and the post-transition state of freedom; that is, the degrees of political rights and civil liberties that the citizens of these economies were experiencing in 2005. It revealed the likelihood that a transition from authoritarian rule would lead to liberal democracy was over four times higher for transitions supported by strong and nonviolent civic coalitions. from 1 to 7, with 1 representing the most free and 7 representing the least free. Data on the extent of violence during democratic transitions are provided by the Uppsala Conflict Data Program at the Peace Research Institute Oslo (Armed Conflict Dataset 9), as researched and reported on by Harbom and Wallensteen (2010). Using this data, violence is coded as a binary variable if an economy experiences any incidence of civil conflict with more than 25 battle-related deaths in a given year.

B. Natural Resources
Many factors can affect the contingencies of democratic transitions, but one is particularly important: natural resources. When natural resources are abundant and their distribution is highly concentrated, the opportunity cost is often too high to induce a small group of elites to willingly give up power. Under this scenario, elites prefer an authoritarian regime, which allows them to keep full control of the resources, and firmly oppose any kind of democratic reform. This is especially the case for economies richly endowed with easy-tograsp natural resources-such as alluvial diamonds and drugs like coca and opium poppy-or with resources requiring high operational costs-such as oil, bauxite, and mineral gas-that are subject to public expropriation. In these economies, democratization is likely to be enforced by the disenfranchised population under (the shadow of) conflict and against the will of the ruling elite.
Cross-sectional regression analysis confirms that the propensity for observing violence during democratization is significantly higher in economies with abundant natural resources and high levels of inequality. 16 In turn, relative to nonviolent transitions, significant or high levels of violence during a regime change reduce institutional quality in the long run (Cervellati, Fortunato, and Sunde 2015). The empirical results also show that natural resource abundance has a long-lasting negative effect on institutional quality even after controlling for the type of transition. There also appears to be a direct effect of natural resources on institutions beyond the one mediated through the transition scenario. These findings are fully consistent with the recent literature on the political economy of the "natural resource curse" that has extensively documented the linkages between resource abundance and institutional quality (Torvik 2009).

C. Democratizing for Development, or the Other Way Around?
The discussion so far has focused exclusively on the conditions that can favor the emergence and success of democratic institutions. However, growth in less-than-democratic economies, such as the People's Republic of China (PRC), has surged in recent years. This brings us to whether authoritarian or democratic states are more conducive to economic development. It also brings us to whether it is the introduction of democratic institutions that breeds economic success or, on the contrary, it is an economy's brightening economic picture that boosts the chances of democratization. 16 In particular, it is the dependence on natural resources (as measured, for example, by the share of natural resources in total exports) that is positively associated with violence. In general, recent empirical evidence suggests that it is not an abundance of natural resources, but rather an excessive dependence on natural capital, including oil and other mineral resources, that blunt incentives to build up other types of capital that are essential to sustain development over long periods (Gylfason 2008).
There is a lively debate on this topic. An important strand of literature asserts that modern capitalist growth requires not only secure property rights, but also creative destruction; that is, the entry of new firms with new ideas and technologies that replace the successful firms of the past (see, for example, Acemoglu and Robinson 2012). While autocratic regimes can effectively secure the first, and therefore generate growth for certain periods of time, democracies are inherently better at favoring the latter; creative destruction requires a level playing field, which democracies are better at providing because they have more equal distribution of political power than autocracies or monarchies. According to this view, democratic institutions in the US were essential to create the environment for new businesses to enter and flourish in the marketplace, spurring the industrial growth of the 19th century, while rich autocracies-such as Cuba, Haiti, and Jamaica-stagnated because they were unable or unwilling to change collective decision-making processes and democratize. 17 In line with the Lipset's seminal hypothesis, other scholars emphasize the primacy of economic development, particularly human capital accumulation, over institutions (see, for example, Dahl 1971;Glaeser et al. 2004;Glaeser, Ponzetto, and Shleifer 2007). Human capital boosts labor productivity and, by providing skills to the population at large, favors wage increases and a reduction of economic inequality. Human capital development, therefore, affects both national income level and income distribution. But education, as discussed in detail in the previous section, also facilitates the diffusion of interpersonal trust, empowers the disadvantaged, and promotes prodemocratic attitudes. As a matter of fact, many of the recent development success stories have been driven by less-than-democratic regimes, such as Lee Kwan Yew's Singapore and Park Chung Hee's the Republic of Korea, which invested heavily in human capital and then commenced their first steps toward democracy only after an economic takeoff. These successes underscore the capacity of human capital (education) to affect both economic and institutional development. Glaeser at al. (2004) find that autocratic regimes with a more highly educated populace are more likely to shift toward democracy over time. Education also makes existing democracies more resilient: 95% of the democracies ranked as "well-educated" in 1960 stayed democratic for the next 40 years; by contrast, 50% of those ranked as "less well-educated" became dictatorships within a decade (Glaeser et al. 2004).

IV. Democratization in East Asia
The experience of East Asian economies in recent decades can be used to illustrate the pattern of human-capital-driven economic and institutional development discussed in the previous section. Some of the most successful economies in the region, like the Republic of Korea and Taipei,China, experienced fast growth and industrialization under authoritarian regimes and later transited smoothly to democracy; others, like Malaysia, Singapore, and the PRC, are currently experiencing controlled and progressive political and institutional changes after impressive growth over the last couple of decades. In a sense, while fast economic growth provided legitimacy to many nondemocratic regimes in East Asia, it also undermined the very foundation of such regimes by educating the population, pluralizing society, and broadening and politically empowering the urban middle class.
The Republic of Korea represents probably the most notable example of this pattern of transition. It was exceptionally poor at the end of the civil war in 1953 and was subsequently ruled by an authoritarian regime for over 30 years. Over the period 1954-1980, the average value of its Polity IV democracy index, which is distributed over a range between perfect autocracies (-10) and perfect democracies (+10), was -3, only marginally higher than the Democratic People's Republic of Korea (Marshall, Gurr, and Jaggers 2014). Yet, the autocratic regime of the Republic of Korea implemented an effective set of economic measures that led to modernization of the economy and fast economic growth accompanied by sustained wage increases. In 1961, 80% of the population of the Republic of Korea comprised poor farmers. By 1980, when the economy reached a per capita level of income of $1,589, industrial workers (over 50%) and the middle class (about 40%) dominated the social structure (Ohno 2013).
The regime also heavily invested in education; from 1945 to 1960, enrollment in primary schools tripled, secondary school enrollment increased more than eightfold, and enrollment in higher education increased 10 times. In 1945, only 40% of school-aged students were enrolled in grades 1-6, while 15 years later, 96% were. Enrollment continued to grow steadily so that almost all school-aged students were attending middle school in the 1970s and the high school graduation rate reached 90% by the end of the 1980s (Kim and Lee 2002). This massive expansion of education, along with the expansion of the middle class, laid the basis for the emergence of a new and vibrant civil society that constituted the bulk of the emerging democratization movement (Gwon 2006). In the 1980s, the Republic of Korea began transforming itself into a full-fledged democracy and the progressive process of democratization proceeded hand-in-hand with sustained economic development. 18 On the contrary, and despite a relatively more developed industrial base at the end of the civil war, the Democratic People's Republic of Korea stagnated. 19 In 2000, it remained an autocratic state with a level of income equal to only around onesixteenth of the Republic of Korea's. While it is obvious that since the end of the civil war the Republic of Korea has had the better political institutions of the two, these institutions seem to be the outcome of effective economic policy making and growth after 1953 rather than the cause.
The other poster child for successful democratization in the region, Taipei,China, followed a similar pattern. The authoritarian regime of Taipei,China obtained domestic legitimacy from sustained economic growth and prodevelopment policies over many decades.
Rising real wages, land reform, and educational progress eventually produced a broad middle class that grew increasingly uncomfortable with the autocratic institutional setup in Taipei,China. In the 1980s, when this emerging middle class joined a coalition with other regime opponents, the ruling party, the Kuomintang, gradually loosened its grip on power and initiated democratic reforms (Freedom House 2005). 20 While the Republic of Korea and Taipei,China successfully completed their transitions and developed democratic institutions, the PRC, the largest and most successful Asian economy in the recent decades, remains under the rule of a lessthan-democratic regime. 21 Yet, much like the Republic of Korea and Taipei,China, its growth was fostered by a major change in economic policy away from a rigid central planning and collective ownership structure to a system that allowed price incentives to encourage agricultural production for the market, at least at the margins. This pattern continued when economic reforms spread to the urban sector and industry.
Recognizing the absence of market-supporting institutions, Chinese policy makers, starting with the Deng Xiaoping administration, adopted a cautious and strategic approach to economic reforms (Vogel 2011). 22 They gradually established the necessary institutions for longer-term economic reform while experimenting with institutional arrangements to address constraints. Various dual-track industrialization strategies were adopted, such as a combination of support for import substitution in selected sectors with export-processing activities considered new combination) of Noguchi, which accounted for one-third of Japanese investment in the Republic of Korea and the Democratic People's Republic of Korea, was centered in the north (Acemoglu et al. 2008). 20 The first step toward systematic liberalization came in 1986, when martial law was lifted and the people of Taipei,China gained freedom of assembly and association. After years of steady liberalization, popular elections were introduced: first for the national assembly in 1992, then for provincial governorships in 1994, and finally for president in 1996. 21 The PRC has enjoyed a spectacular economic rise over the past quarter-century: its average annual per capita gross domestic product growth of 9.5% has resulted in an increase in per capita income by a factor of seven and raised the PRC's status from that of a low-income to an upper-middle-income economy (World Bank 2015). 22 However, as pointed out by Nee and Opper (2012), some private sector-led experiments with market production and the ending of collective incentives had started before Deng Xiaoping's reforms. For example, in Anhui province, peasant households had dissolved communes and collectives and started a land-lease system before any nationwide reforms were introduced. Therefore, the Communist party was pushed to catch up with developments on the ground that had resulted from the earlier political vacuum.
to the domestic economy (McMillan and Rodrik 2011), thereby blending an East Asian model of national enterprise-led growth with a Southeast Asian model of global value chain-led growth primarily orchestrated by multinational corporations (UNCTAD 2014).
The overall package of economic reforms undertaken greatly facilitated the entry of new businesses and increased access to education and economic opportunities for a vast part of the populace. Using the language introduced by Acemoglu and Robinson (2012), the PRC's economic institutional setup became substantially more inclusive, thereby creating the basis for economic prosperity.
Whether this new institutional setup can survive and progress in the long run without the support of more inclusive and democratic political institutions remains an open question. According to Acemoglu and Robinson (2012), inclusive economic institutions can emerge and encourage growth in the short run but cannot survive in the long run in a nondemocratic regime; the future success of the PRC would therefore rely on its capacity to accelerate political reforms. In a Lipsetian perspective, however, these political reforms are likely to follow as a by-product of the economic and social progress that has taken place in recent decades. The impressive income growth and educational advancements achieved by the PRC, in fact, should naturally smooth the path toward democratization in the years to come.
The key to East Asian growth, which in many cases has preceded a political transition, is that it is broad-based and is shared with the populace. Economic institutions became progressively more inclusive in most economies in the region, thereby providing a more level playing field and more equal access to education and productive opportunities. The leaders of the region's "developmentalist" states provided education and social goods-such as housing, pensions, and health carefacilitating the achievement of an inclusive and inequality-reducing growth path that ultimately paved the way for successful transitions to democracy (Ginsburg 2008). 23 Many East Asian economies, especially the Republic of Korea and Japan, were also "blessed" with few natural resources. Export orientation represented the only plausible route out of poverty for these economies and they were therefore insulated against the pathologies of the natural resource curse discussed in the previous section. Others-such as the PRC, Malaysia, and Thailand-achieved economic diversification despite being endowed with substantial mineral resources. One common theme among these (different) success stories is the need to reduce costs for the non-natural resource sectors, whether through macroeconomic management and exchange rate policy, trade policy, or well-focused public investments. Another is the importance of openness to foreign investors, skills, and new markets, and the efforts to supplement market incentives in various ways to encourage diversification (Gelb 2010).
It was the combination of the effective development of human capital and efficient deployment of physical capital for economic diversification that made the difference in East Asia. The lesson from this paper's perspective is that it is not growth alone, but the type of growth that matters for democracy to emerge and thrive. Only if growth is widely shared among a broad social base, and then only if it brings sustained educational improvements, are healthy democratic institutions likely to arise.

V. Concluding Remarks
The literature discussed in this paper suggests that democratic institutions do not always guarantee good governance. Democracy does indeed require important ancillary conditions to operate fruitfully and foster economic development. Where social life is fragmented and a culture of distrust diffuse, evidence shows that democratically elected bodies are less responsive to public demands and politics takes on patron-client characteristics. A well-functioning democracy also requires voters to be informed about the political process and politicians' actions. In places where the media does not facilitate access to unbiased information and where the rules and practices of information disclosure are not part of the political culture, elected officials are likely to be only partially accountable and serious moral hazard problems may arise. Education is likely to affect both the strength of social networks and access to information; it therefore represents a key pillar of democratic governance. Education may also have a direct effect on the functioning of democratic institutions, since it empowers citizens to engage with government institutions and facilitates the negotiated resolution of social and economic disputes. Finally, inclusive and efficient democratic institutions also require a certain commonality of (economic) interests across the population. Economies heavily relying on natural resources whose rents are unevenly distributed are more likely to experience violent conflicts and, in turn, to develop a dysfunctional institutional setting after the transition to democracy. Again, education may play an important role here; modern economies in which a significant share of value-added accrues to human capital development are less likely to be trapped in this institutional facet of the natural resource curse.
This complex network of direct and indirect effects is illustrated in Figure 3. The diagram exemplifies the interplay among the different factors analyzed in the paper and the functioning of democratic institutions. Solid lines represent direct channels of influence on the quality of democracy, while dotted lines symbolize the existence of linkages among our different explanatory variables (and thus of indirect effects on governance). 24 Low-income economies score poorly on many of the dimensions that determine the quality of a democracy. Therefore, identifying the priorities on the policy agenda to create and/or consolidate democratic institutions represents a critical development challenge.
Policy makers should try to increase stock of social capital. Yet, social capital is frequently a by-product of religion, tradition, shared historical experience, and other factors that lie outside the control of any government. It is extremely difficult to identify effective levers for creating trust and other forms of social capital. The area where governments probably have the greatest direct ability to generate social capital is education. As discussed above, educational institutions do not simply transmit human capital, they also pass on social capital in the form of social rules and norms.
Governments can also foster the creation of social capital indirectly by efficiently providing necessary public goods, particularly property rights and public safety. There are important externalities associated with the enforcement of property rights; people cannot associate, volunteer, vote, or take care of one another if they fear for their lives when walking down the street. Given a stable and safe environment for public interaction and property rights, it is more likely that trust will arise spontaneously as a result of the iterated interactions of rational individuals.
Public policy should also seek to increase the information readily available to voters. Sustaining the diffusion of independent mass media and fostering political competition is of critical importance in this regard. However, in the short to medium run, relying on the media may not suffice. Many low-income economies are far from having an independent media whose role is to provide citizens with up-to-date and objective information. Equally, increases in political competition may not suffice since information transmission by political parties may simply not be credible.
The potential inefficacy of "good" prodemocratic reforms, such as facilitating the diffusion of an independent media and increasing the contestability of the political arena, brings us to the final point of this paper: not only the choice of specific political and economic reforms matter for the functioning of democratic institutions, but also their sequencing. Discussing recent African economic and institutional advancements, Miguel (2011) asked why some African economies, such as Ghana, developed successful democracies in the 1990s, while others, such as Côte d'Ivoire, tried and failed. And why did democracy take root only recently, while earlier democratic experiments in the 1960s collapsed within an election cycle or two. The answer to that question may lie (once again) in education. In the 1990s, unlike in the 1960s, education was widespread on the continent, and it was this rapid expansion of schooling, which took place during the otherwise lost decades of the 1970s and 1980s, that paved the way for effective and sustainable political reform. As a matter of fact, the same sequencing characterized the successful democratic transitions in East Asia discussed in the previous section.
A complex combination of factors is needed to sustain the functioning of democratic institutions and guarantee their success. Despite the voluminous research to date, we are still far from a complete understanding of the interplay among these factors. Education seems to be at the core of this network and future advances in the field will inescapably rely on a deeper understanding of the channels through which it influences the mechanics of modern democracies.