Curvale, Carolina//2010-09-23

Does Political Participation Aect Political Stability? A Study of Latin America During the 19th and 20th Centuries

September 2010

Carolina Curvale

Abstract Political stability is necessary for the proper conduct of both private and public life in society. The old-standing question is whether and how political stability depends on political participation: who participates, how many participate, and how do they participate in politics? Based on an original data le that includes information about political institutions and political events since independence, I employ three measures of political stability - coups detat, constitutional change, and executive turnover - to confront the leading hypotheses in the literature that relate political participation to political stability. While the data did not provide denitive support on whether political participation aected political stability, a fascinating nding emerged from the analyses. The empirical evidence indicated that allowing for at least a modicum level of political opposition promoted political stability throughout the independent histories of Latin American countries, regardless of political participation.

Ph.D., Department of Politics, New York University, carolina.curvale@nyu.edu.

Introduction

The acquisition of political rights that enable political participation provides groups with an institutional voice and a chance to participate in policy making. In the political struggle over the control of resources, the inclusion of new stakeholders may plausibly lead to political instability when disputes cannot be resolved within existing institutions. Alternatively, the causes of disorder may be unrelated to the participants in the political game or be mediated by other factors. Although much has been written on the subject, the literature is missing a broad understanding of the evolution of the relationship between these variables that takes into account multiple types of participation of dierent groups over long periods of time. Numerous attempts to understand the eect of political participation on political stability have focused on specic groups. However, since specic groups tend to become politically visible at dierent times, these explanations frequently account for particular periods of time, usually dependent on structural processes of development. On the empirical arena, there exists some fragmentary evidence on specic theories with selected cases but no systematic test has yet been conducted. The purpose of my research is to ll this gap by empirically investigating the explanatory power of theories that relate political participation and political instability employing the entire independent histories of Latin American countries.1 One type of political participation, electoral participation, has been a source of disputes among political actors since the inception of the Latin American nations. The post revolutionary project of state building generally contemplated broad electoral participation. Many of the initial republics of citizens gave way later on to the more restrictive republics of decent people, typically through the introduction of literacy requirements (Irurozqui and Peralta 2003). Developments in electoral legislation were nevertheless not uniform across the region: at the time when some countries contracted the franchise, others enacted reforms that facilitated broad participation.2 Some
Following the denition of the Oxford English Dictionary, I include in Latin America those countries in Central and South America where Spanish or Portuguese are the dominant languages. Hence, the countries included are: Argentina, Chile, Brazil, Uruguay, Bolivia, Paraguay, Peru, Ecuador, Colombia, Venezuela, Costa Rica, Panama, Cuba, Nicaragua, El Salvador, Honduras, Costa Rica, the Dominican Republic and Guatemala (see codebook for time coverage of each country). 2 For example, the Senz Pea law, passed by the Argentinean congress in 1912, provided for secret and compulsory a n surage. According to Valenzuela (1998), the 1890 Chilean electoral law was designed to end ocial intervention in elections by making the vote eective through the installment of a secret chamber.
1

2 nineteenth century advocates of the restriction of the franchise blamed widespread participation for the troubles encountered in the process to attain political order. At the turn of the century, the meaning and plausibility of social strife changed: the state was more capable to repress and the revolutions were lled with novel political content. New actors emerged on the political scene workers and peasants - and the issue of their political incorporation posed a novel challenge to political life. The twentieth century witnessed recurrent military interventions in politics that in the name of order sometimes fostered the political inclusion of the excluded (in the form of national populism) and other times completely suppressed all sorts of popular political involvement. The imminent quest to nd theoretical apparatuses that could explain the pendular regime dynamics in postwar Latin American countries drew on the origins of the social transformations that began in the latter part of the nineteenth century, but did not suciently incorporate the incredibly rich electoral experiences of the nineteenth century nascent republics. This article is organized as follows. In Section 2, I present and discuss the main theoretical approaches to the questions posed above and provide denitions of the concepts of political stability and political participation. Section 3 oers an analysis of the data on the dependent variable in this study political stability, contributing with three measures of the phenomenon, and identifying trends over the last two centuries. Section 4 is centered on describing the changing data on political participation, with emphasis on electoral provisions. In Section 5, I test empirically the theoretical questions and perspectives. Section 6 concludes.

Theoretical Perspectives

Below I provide the denitions of the dependent variable -political stability- and the independent variable -political participation.3 Next, I present the leading two hypothesis in the literature that account for the relationship between these variables, which I have called Exclusion and Inclusion Hypotheses, followed by a few of my conjectures.
Other alternative explanations of political instability include presidentialism (Linz and Valenzuela 1984), the role of the armed forces (Finer 1962, Fitch 1998), and culture (Morse 1954, Dealey 1974, Wiarda 1982, Pike 1982).
3

2.1

Denitions

The concept of political instability emerged in the 1960s as a result of the preoccupation about regime stability in the context of the ideological battles of the cold war. From a Parsonian view of society, predominant at the time, changes were challenging given that they risked the preservation of the social equilibrium (i.e., Ake 1975, Needler 1968). The lack of clarity about concept denition likely led to the emergence of empirical denitions. A number of indexes that lumped together dierent features of political systems - in particular, measures of government change - took over. Unsurprisingly, the focus on quantication did nothing to clarify the concept. On the contrary, the dimensions of political instability often included events that are part of democratic activity (such as cabinet changes and peaceful anti-government demonstrations) and events that imply the disruption of the political system (such as revolutions, coups detat). The confusion lingers until today: an avalanche of research conducted by economists trying to explain the eect of political instability on economic growth continues to show the old imprecision.4 Political stability does not imply lack of change or immobility. It is precisely the merit of an institutional order to be able to absorb and peacefully process potential conicts of interests and values without breaking down. Political institutions must be self-sustaining, that is, they survive and function only if they continually generate outcomes that are preferred to the use of force by each and every group that could impose itself by violating the institutional order. In order to be self-sustaining, a political system must be organized in such a way that the outcomes it generates, whether the distribution of incomes or division of rents or realization of some non-material values, must reect the distribution of the brute, pre-institutional power, including the military force of dierent groups. Political stability entails a degree of predictability; that is, there is room for change and conict, but always within the boundaries imposed by institutions. Hence, political stability refers to the compliance of political behaviors with a pre-existing set of rules. Political participation, according to a classic denition, includes all the actions taken by groups or individuals aimed at inuencing governmental decisions (Milbrath 1965). However, there is a crucial dierence between political participation and the broader notion of political activity. While
4

For example, Cukierman, Edwards, and Tabellini (1992), Alesina and Perotti (1997), etc.

4 the former assumes that certain rules and institutions -from which individuals could be included or excluded- exist, the latter also comprises behaviors that attack the rules of the game themselves. One way to look at this dierence is to think of political participation as entailing legal activities while political activity comprises all political behaviors, even those that are illegal in light of the existing institutions. The incorporation of illegal activities in the denition of the term mixes participation with instability and therefore poses a methodological dilemma to the study of the relationship between these variables. Hence, political participation refers to political activity that is exercised through institutional channels.

2.2

The Exclusion Hypothesis

Perhaps the most vocal theorist of the detrimental eects of increased participation on institutions in countries trying to develop is Samuel Huntington (1965, 1968). He proposed to re-conceptualize political development as institutionalization, and not as modernization as the early political development scholars had done. Although Huntingtons theory of institutional decay was created to explain the post world war instability in the -by then- new countries, his theoretical apparatus does not specify time periods. In his view, the stability of a polity depends upon the ratio of institutionalization to participation.5 The particular case that concerns Huntington is the one in which the denominator is higher than the numerator (praetorian societies). Note that logically rapid mobilization should not necessarily cause instability in this theory since institutions may adapt fast or already be strong, but Huntington argues that during the most intense phases of modernization even societies with high levels of institutionalization suer (1968: 86). Therefore, institutional decay is rather inescapable in the presence of rapid mobilization. Huntington spells out at least three reasons why broadened participation leads to institutional decay, conveying some of the arguments already mentioned above. First, new groups have not internalized the institutional norms in place and do not know how to behave properly: In a praetorian society groups become mobilized into politics without becoming socialized by politics (1968: 83).
Huntington denes institutions as stable, valued, recurring patterns of behavior (1965: 394) which seems inappropriate considering that institutions constitute one of the two elements that dene stability in the theory.
5

5 Individuals may not only have to learn civic ways, but also unlearn behaviors that are dissonant with the constituted order. Second, individuals acquire consciousness of their political opportunities and develop preferences based on the new aspirations and expectations to which they are exposed to in the process of modernization.6 In the absence of chances for social and economic mobility, social frustration ensues. Demands for change proliferate but the existing political institutions cannot cope with them: Rapid or substantial expansions in the membership of an organization or in the participants in a system tend to weaken [institutional] coherence (1965: 403). Third, the political mobilization of the masses may bring to the political arena groups that break the existing consensus among the leading modernized political participants. Political mobilization, reads the argument, may even awake the self-consciousness of previously dormant ethnic fractions that may divide the political system along ethnic lines (1965: 406). A crucial implication derives from Huntingtons theory of institutional decay. Since patterns of behavior are subject to generational transmission, no decay should be expected when the boundaries of institutions remain constant. In Latin America, out of 322 years with at least one coup detat or revolution, 213 took place after 10 years with constant male franchise. 138 years with disruptions of the institutional order occurred after a generation (25 years) with constant enfranchisement. This casts doubts on the assumption that constant membership guarantees stability, at least with regard to electoral inclusion. Additionally, the electoral incorporation of women, in light of Huntingtons theory, is suspiciously harmless. If sudden and large increases in the electorate erode institutions, the inclusion of women should be disruptive since the extension of universal male franchise to women usually doubles the size of the electorate.7 Yet the impact of their incorporation does not appear to be particularly destabilizing of political order. Note that in this line of argumentation, neither the identity of those who will be incorporated nor their channels of participation are specied. There is no indication as to whether the pre-existing level of participation is relevant.
Especially, if the newly mobilized are educated: The faster the enlightenment of the population, the more frequent the overthrow of government. (1968: 47). Thus, the author implies that disorder follows the political mobilization along ideological lines, and not necessarily along class lines. The problem appears to be the activation of real opposition. 7 I assume that women account for half of the population, which is generally true. In cases when male franchise is based on economic or literacy criteria, the impact of women enfranchisement on the size of the electorate depends on other than demographic factors often putting women in a disadvantaged position.
6

2.3

The Inclusion Hypothesis

While in Huntington (1965, 1968) the institutional order is preserved with political exclusion, the theories in this section suggest that inclusion achieves the same end. A classic view sustains that political reform is a strategy to achieve political survival and legitimacy (e.g., Bendix and Rokkan (1962), Przeworski and Cortes (1971), Freeman and Snidal (1982)). More recently, the issue of enfranchisement has been revisited from a strategic perspective.8 In Acemoglu and Robinsons story (2000, 2006), granting voting rights prevents the worst outcome: a revolution. Briey stated, through a model that employs the median voter theorem as the political mechanism, these authors propose that political elites may resort to two strategies to cope with revolutionary threats: temporary policy concessions or enfranchisement. Policy concessions may take place when the disenfranchised poor pose a permanent threat of revolution because the poor can enforce promises of future concessions. If the threat of revolution is transitory, enfranchisement is the elites best reply since the poor are unorganized and future promises of transfers are not credible to them. In the latter case, political inclusion constitutes a credible commitment to ensure future transfers of goods because the median voter will be poor. Thus, redistribution policies should be undertaken when the labor movement is organized and the franchise is restricted or after a franchise extension.9 There are a number of empirical implications derived from the Acemoglu and Robinson (2000) story. First, if franchise extensions occur in the presence of a revolutionary threat, we should observe the occurrence of events showing social unrest (such as strikes and public demonstrations against the government) before franchise extensions. Second, if enfranchisement is indeed a credible commitment to future transfers, one may reasonably expect that the frequency of those events will diminish after the right to vote is granted to the poor. At the extreme, revolutions should not take place under broad franchise; conversely, a revolution should adopt universal franchise where previously absent. More crucially, turnout should increase after the concession of voting rights since the poor are expected to make use of their newly acquired political power to inuence transfers.
For example, Conley and Temini (2001), Jack and Laguno (2003), Lizzeri and Persico (2004), and Llavador and Oxoby (2005). 9 Our approach also predicts that after franchise extension, there should be increased redistribution. (Acemoglu and Robinson 2000: 1191)
8

7 Third, redistribution should occur shortly after the broadening of the scope of the franchise and inequality should decrease thereafter. However, there exists some evidence that the vote of the poor is not necessarily foretold.10 Przeworski (2007) analyzed empirically the determinants of enfranchisement. In agreement with the Acemoglu-Robinson story, his conclusion indicates that surage extensions were the result of instances of unrest: surage was granted to prevent revolution. My concern, however, is quite dierent: once surage rights were granted, did they help ensure political stability?11

2.4

Conjectures and Predicted Eects

Contrary to the Inclusion and Exclusion Hypotheses, I contend that political participation had no eect on political stability during the period of turmoil that followed the declarations of independence. Until some sort of institutions could be adopted after the wars of independence, neither the level of electoral participation nor its changes (in both directions) had any eect on political stability for the rest of the nineteenth century. Politics were characterized by intra-elite conict. In nineteenth century Latin America, it was not uncommon to observe that the predominance of one group or party was terminated by a coup detat perpetrated by another group that in turn inaugurated a period of its own predominance. Frequently, political leaders appealed to popular support by promising polices that fostered the interest of the rural or urban working class groups; followers were frequently enlisted through patronage or some form of compulsion (Bushnell and Macaulay 1994: 35). Even when surage was relatively broad, indirect voting mandated highly restrictive rules for candidacy typically based on property, wealth, literacy, social position and other class-based considerations. The stability that reigned by the end of the nineteenth century until around 1925 was independent of political participation.
10 Vines and Glick (1967) studied the results of a system of bond referenda originating in the 1890 Louisiana constitution that required each voter to cast two ballots simultaneously, one of which contained the amount of the voters property assessment. They found that universal surage inicts no damage to the propertied. 11 In their 2006 piece, Acemoglu and Robinson allow for the possibility of instability after enfranchisement if the poor set a high tax rate. Unfortunately, their predictions about instability after inclusion are mixed: once the masses are incorporated into political life, the elites may produce a military coup depending on how costly the coup is relative to its benets. The theory allows for a strange timing of events. An example illustrates this point. In Argentina, male surage was universal since 1856 and became secret in 1912. Even if we agreed that the universal franchise was eective only after 1912, a military coup did not take place until 1930 - 18 years after eective franchise.

8 This is not to say that political stability depends on the existence of democracy, where all the political parties involved in the political game have some chance to access to power. Given that to be self-sustaining institutions must reect the relations of forces, they may be inegalitarian, restricting political access to those groups that can mobilize military prowess, and still favoring those more powerful among those groups that are admitted into the institutional interplay of interests. The institutions that rst generated political stability in Latin America are often described as oligarchical but they were in fact more often polyarchical.12 They were highly inegalitarian, restricting political rights to a small portion of the male adult population, but also typically pluralistic, allowing a modicum of political opposition to function within the institutional framework. While elections were carefully manipulated by the incumbent governments, thus assuring either their own permanence in oce or the victory of appointed successors, opposition was legally tolerated, allowed to win some seats in the legislature, and sometimes even a share of power. And these incentives were most often sucient for the opposition to participate. Hence, intra-elite conicts were processed according to rules and, even if not without sporadic repression and sporadic rebellions, were peacefully resolved. Following Chile after 1831 (about which see Valenzuela 1995), several Latin American countries established stable systems of succession in power in which incumbent presidents completed their terms and, faithfully obeying term limits, chose their successors, and through various devices assured their victory at the polls.13 The stability of such systems of oligarchical pluralism Chile between 1831 and 1891 and again until 1924, Nicaragua between 1856 and 1890, Brazil between 1894 and 1930, Argentina between 1897 and 1916, Uruguay between 1898 and 1932 was remarkable. Things took a turn in the beginning of the twentieth century, when socioeconomic developments brought about new groups that wanted a share of power. Political conicts ceased to be intra-elite. The development of Latin America during the second half of the nineteenth century led to the emergence of two major distributional conicts: over land (or wages in agriculture) and over wages and working conditions in industry. The nature of the conict became distributional. The long neglected peasantry started to rebel. Worker organizations, fueled by the emergence of communism
12 Dahl (1971) used the term polyarchy to denote any kind of an imperfect democracy. Here it is meant literally, as plural oligarchy. 13 For a summary of devices by which governments controlled results of elections, see Posada Carb (2000). o

9 in Russia, started to realize that they could also participate in decision-making and improve their well being. On the other hand, the elite had reasons to be concerned, as the source of their power -their wealth- came to be challenged by demands of better salaries, wages, and labor conditions. Peasants joined the impetus of the labor movement in order to achieve gains regarding wages in agriculture and land ownership. After the poor became politically active in the twentieth century, so did the military. Prevalent reactions to the political incorporation were the military coups that began in the 1920s, which in turn triggered further instability (Londregan and Pool 1990). In that sense, the exclusion of the poor was destabilizing, not their inclusion. The poor have no incentives to obey rules (nish a term, respect the constitution, etc.) when their basic demands are unfullled and they are politically organized. This in itself may not trigger a revolution (there were only three successful social revolutions) but prompt an elite reaction. When their assets are at stake, the rich do not obey the rules. Each of the three hypothesis presented specify relationships between the dependent and the independent variables in this study. Table 1 summarizes the predicted eects of the Exclusion and Inclusion Hypotheses, as well as the predicted eects of my expectations as described above.

Variable Scope of the franchise Franchise extensions Turnout % population voting % population eligible Franchise contractions Policy concessions to workers and peasants Strength of the military Past instability Organizational strength of workers Political Opposition

Exclusion n/a (-) n/a n/a n/a n/a n/a

Inclusion + + + + + n/a (-) + n/a n/a n/a n/a

My Conjectures Electoral provisions had no eect in the 19th century; negative eect in the 20th century n/a +

Table 1: Predicted Eects on Political Stability

10

The Dependent Variable: Political Stability

There are many ways to think about political stability. Below, I analyze three dimensions of political instability: coups detat, chief executive turnover, and constitutional change.

3.1

Coups detat

Coups detat do not only measure institutional instability, but political instability in general: I have included cases of non-pre-agreed transfers of power within lawless regimes, assassinations, military coups, revolutions, forced resignations (excluding resignations due to public pressure), and civil wars. The dummy variable coup shows whether a coup occurred at any time during each particular year. Hence, I considered as a coup cases where force was used and cases where there were threats to use force that resulted in a resignation.

Figure 1: Proportion of Countries with at Least One Coup. Latin America, 1811-2005

Figure 1 shows the proportion of countries that experienced at least one coup in a given year. If we were to consider illegal closures of the legislature, of 1,008 country years for which I have data, there were 66 instances of this phenomenon (including twice in one year in Paraguay 1940). We nd an outlier in 1829, when 5 out of 8 countries with non missing data experienced at least

11 one coup. During the twentieth century, the peaks occur in 1911 and 1930 when 6 out 19 countries registered at least one coup detat. The country that experienced coups most frequently is Bolivia (36), and the most stable in this sense is Cuba (5). In terms of the timing in the occurrence of coups, the two centuries are almost even, with 14 more coups occurring during the nineteenth century. The average number of years without coups is 18.8 in 1930, being the highest value for the period 1900-1950.14 According to the data on coups, the peak of stability during the nineteenth century occurred in 1888, when 9 out of 17 countries had experienced at least 10 years without coups. Until 1900, the maximum number of accumulated years without coups is held by Chile in 1890 (60 years) followed by Brazil (57 years) in 1888. During the entire twentieth century, the maximum occurs in Uruguay in 1972 (96). Uruguay only experienced one coup detat in the entire twentieth century. Overall, we observe that as we get closer to the present day, the regional mean increases, because countries had time to accumulate after the low of 1963 when 5 countries experienced military coups (Guatemala, Honduras, the Dominican Republic, Ecuador, and Peru). The most recent coup in the dataset occurred in Ecuador in 2005. Coups detat inform us that political behaviors did not comply with pre-existing rules given that the use of force or the threat of the use of force was employed to produce a succession in executive power. However, taking into account the occurrence of coups alone does not describe the entire picture of political instability, given that the political leadership may change institutions without resorting to force.

3.2

Constitutional Endurance

Since constitutions establish political institutions, constitutional endurance is a sign of institutional stability. At the onset of national life, the leaders of the independent movements embarked on the task of organizing the government. There was much hesitation about the institutional choice and also lack of experience with popular rule due to the fact that the colonial administration practically excluded representative institutions.
For example, Argentina had accumulated 68 years without coups by 1929, Panama 27, Colombia 29, Brazil 40, and Uruguay 53. Between 1930 and 1931, 10 Latin American countries experienced a coup.
14

12 The feverish production of written rules in the form of constitutions resulted in a total of 217 national constitutions15 with 216 reforms.16 Most of the time, Latin Americans had some sort of framework that organized their businesses: in only 147 country years there was no constitution in force, representing only 4.6% of the dataset. Given this overproduction of constitutions, one may be led to believe that these documents were mere formalities promoted by tyrants seeking some degree of legitimacy. And sometimes they were. However, the attempts to create an institutional order indicate that political leadership believed that an institutional framework was necessary, particularly during the nineteenth century. According to Gargarella (2004), nineteenth century constitutions were rarely whimsical and expressed dierent views on how to organize society with some of those legal debates continuing until the present date. Even when taking into account these caveats, constitutional endurance is a measure of political stability. A way to think about institutional stability is to assess whether a constitution survived a generation (set at 25 years). Only Uruguays 1830 constitution achieved it within the rst generation after the nal declaration of independence and Brazil and Uruguay almost did.17 Five countries waited a second generation until a constitution lasted at least 25 years, while the remaining ones waited even longer. Table 2 shows the constitutions that rst lasted 25 and 50 years by country. In eight countries, never did a constitution last 50 years, and in Ecuador no constitution ever lasted at least 25 years -although the 1906 charter came close, having lasted for 23 consecutive years. There were periods of constitutional stability that lasted over two generations. Colombias 1886 constitution survived the maximum time without having been suspended: 105 years. Other constitutions that have endured over 50 years (as of 2005) are: Mexico 1917 (89), Argentina 1853 (96), Chile 1833 (92), Uruguay 1830 (88), Costa Rica 1949 (57), Guatemala 1879 (66), Bolivia 1880 (51), and Brazil 1824 (65). Among the unstable countries are Venezuela with 25 constitutions in force, followed by Ecuador (21), the Dominican Republic (21), Bolivia (19), and Peru (17). Institutional instability measured by constitutional age also manifests the heterogeneity that predominates in the region.

15 16

I include only the constitutions that were in eect as of December 31. Any written document that organized government was counted as a constitution. 17 Years in which the constitution was suspended as of December 31 set the counting back to zero.

13 Country Argentina Bolivia Brazil Chile Colombia Costa Rica Cuba Dominican Rep Ecuador El Salvador Guatemala Honduras(a) Mexico Nicaragua Panama Paraguay Peru Uruguay Venezuela Independence Year 1816 1825 1822 1818 1830 1838 1902 1844 1830 1841 1840 1838 1821 1838 1903 1811 1821 1828 1829 Constitutions that rst lasted One generation Two generations 1853 1853 1880 1880 1824 1824 1833 1833 1886 1886 1871 1949 1901 None 1929 None None None 1886 1886 1879 1879 1982 None 1857 1857 1858 None 1904 None 1814 1870 1860 None 1830 1830 1830 None

Note: (a) the Honduran 1982 charter had its 25th anniversary in 2006.

Table 2: First Constitutions that Survived One and Two Generations

In order to identify periods of political stability, we may also refer to the average cumulative number of years with one constitution in force (gure not shown). The maximum of 32 years occurred in year 1916, the peak of a 30-year period of stability, once again conrming that what was deemed the oligarchical period was remarkably stable. In line with what was shown with coups detat, the rst years after independence proved to be unstable for most countries (excluding Cuba and Panama which became independent in the twentieth century), with improved stability around year 1900 and then declining after the rst quarter of the twentieth century.

3.3

Chief Executive Turnover

Chief executive turnover, when changes of heads occur in accordance with the rules of the game, is a sign of the normal development of politics. In Latin America, the presidential terms were typically

14 of 4, 6, and 5 years (in that order), with an average of 4.34.18 If the constitutional term of the chief executive is for non-renewable 4 years and if this norm is always observed, we should expect the typical 4-year moving average value to be 1.25 heads per year. Protracted dictatorships or repeated reelections would generate lower numbers, while coups and other contests for power would be seen as more frequent turnovers. The four-year moving average in Latin America is 1.42, indicating that on average frequent executive turnovers outweigh the cases of long tenured executives. Latin Americans traditionally established terms and term limits. In only 7.69% of 3,171 cases, rulers did not have terms or had life terms, and in 19.46% of 3,108 observations there were no term limits. However, terms were frequently extended and the cases of long lasting rulers are not so few. The longest tenured leader in the entire history of Latin America is Alfredo Stroessner (Paraguay 1954-1988). Another ve leaders held power for entire years during periods longer than 20 years: Pedro II (Brazil 1841-1870), Fidel Castro (Cuba 1976-2005), Porrio D (Mexico az 1884-1910), Jos Gaspar Rodr e guez de Francia (Paraguay 1813-1839), and Manuel Estrada Cabrera (Guatemala 1898-1919). 12 others kept power for over 10 years. Curiously, ve of the six longest tenured leaders mentioned governed with rules and held elections.19 Figure 2 shows the average number of cumulative consecutive years in which the nominal chief executive was the same person for the entire year. The rules are spelled out in the codebook (see the Appendix), but I should briey mention two things. First, the value of this variable decreases when legal changes take place during a year due to illness or the natural death of the incumbent. Second, it decreases when periodic alternations in power take place. The curve describing the average cumulative number of consecutive years with one chief executive varies greatly; its mean is 2.7 and its standard deviation is almost 1. Here too there is a peak around year 1900 and the highest value of the series appears at the end of the 1980s. Even when institutional instability is conceptually distinct from political instability, we observe that the peaks and lows of the dierent measures are similar.

During the early period, there were some two-year terms in Central America. Stroessner changed term limits; Pedro II was a constitutional monarch; Fidel Castro was indirectly elected without term limit; Porrio D eventually modied both term and term limits; and Manuel Estrada Cabrera modied term az limits. Francia, who died in oce, is the sole exception, since he ruled with only a pseudo-constitution and was declared Dictator for Life by a Congress that never again met.
19

18

15

Figure 2: Average Cumulative Years with One Chief Executive. Latin America, 1811-2005

What sense can we make of these data? The institutional experience of colonial Latin America had a profound eect on the development of independent institutions. The Spanish colonial administration was direct and centralized, leaving little space for self-government. No one born in the colonies, a creole, could hold higher posts in the Spanish colonial administration.20 Hence, when the Spanish colonial administration disintegratedand it collapsed not because of any pressure for independence in the Americas, but because of events in Europethe ensuing conicts, whether between territorial units or between landowners and peasants over land or between creoles and peninsulares over political power or just among dierent militias over nothing, could not be resolved within a pre-existing institutional framework. Until one of the forces established its military domination or the opposing forces agreed to process conicts according to some rules, conicts could only take violent forms. Finding a suitable institutional framework was a learning-by-doing process. The choice of the rules of the game was not obvious and new republics experimented with dierent forms of governThe only institution that entailed some modicum of self-government in Spanish Americathe cabildowas an estate body, with oces that could be purchased and kept in perpetuity after 1556, and only few elective posts subject to the conrmation by the Crown and elected under highly restricted surage.
20

16 ment.21 Moreover, independence was not a one-time event for several Latin American countries.22 Once countries became independent, they could and eventually did develop political institutions. The political dust of independence settled around 1870 and the oligarchical republics, which were remarkably stable, ourished. The predominantly liberal governing elites, duly obeying terms, oversaw fast economic growth and the expansion of primary exports (Skidmore and Smith 2001). These highly exclusionary political institutions allowed the peaceful processing of conicts between elites -typically over centralization, taris, or the role of religion (conservadores versus liberales, see Gargarella (2005)). Among those regimes are Central Americas coee republics, Bolivia (18801899 and 1899-1916), and Peru (1886-1914). With manipulated elections, the incumbent frequently guaranteed the election of his chosen successor, and in the unlikely case that he failed to do so, election results were often annulled with a legally questionable argument (for example, during the 1906 Costa Rican presidential elections (Yashar 1997: 54)). While stability reigned by the eve of World War I, the dominance of liberalism was challenged after the war (Peloso and Tenenbaum 1996). A new wave of destabilization occurred around 1930, when several countries experienced military coups. According to Rouquie (1994: 223), Between February and December of 1930, the military were involved in the overthrow of governments in no fewer than six, widely diering Latin American nations - Argentina, Brazil, the Dominican Republic, Bolivia, Peru, and Guatemala. The same year also saw four unsuccessful attempts to
For example, the Colombian monarchy (1811-1812), the Mexican empire (1821-1823), the Argentinean (1811-1812 and 1812-1814) and Venezuelan (1811-1812 and 1812) triumvirates, the Argentinean (1810, 1810-11) and Chilean (1823) early juntas, the Paraguayan Perpetual Dictatorship (1816-1841) and Gran Colombias Dictatorship (18281830), rotating executives in Venezuela (1811-12) and Paraguay (1813-4), and a three-member congreso plenipotenciario in Chile (1823). In later periods, further experimentation took place. 22 Costa Rica, El Salvador, Guatemala, Honduras, and Nicaragua declared independence from Spain in 1821 but were then briey absorbed into Iturbides Mexican Empire until its dissolution in 1823. Afterwards, they constituted the United Provinces of Central America (later called Federal Republic of Central America), with Guatemala as its political center. Under federal provisions, each component state issued its own constitution and elected local ocials. Nicaragua, Costa Rica, and Honduras seceded in 1838, while El Salvador formally did so in 1841. The Dominican Republic declared independence in 1821, but was subsequently occupied by Haiti until 1844. Mexico declared independence from Spain in 1813, but was reconquered and could not establish independent institutions until 1821. Panama declared independence from Spain in 1821 but remained a part of Colombia throughout the nineteenth century. Even before its liberation from Spanish domination in 1822 by Bolivarian troops, the fate of Ecuador had been decided at the 1819 Angostura Congress, and the country remained a part of the Republic of Colombia until 1830. After an eleven-year struggle against Spain, Venezuela abandoned colonial rule as a component state of the Republic of Colombia, leaving it only in 1829. In Uruguay, revolutionary struggle began in 1811, and after conicts with Buenos Aires, Portuguese Brazil annexed it between 1816 and 1825. Having attained independence from Spain in 1821 and 1825, Peru and Bolivia constituted the Peru-Bolivian Confederation in 1836, which died after only three years, largely due to Chilean intervention.
21

17 seize power by force in other Latin American countries. Over the following years, Ecuador and El Salvador in 1931, and Chile in 1932, joined the list of countries in which military-provoked political shifts and unscheduled changes of the executive had taken place. Only after 1980 was political stability restored again.

The Key Independent Variable: Political Participation

The Latin American founding fathers not only had to build new republics but also an entirely new social structure. How could a traditional society coexist with and adapt to modern representation? The search for rules consistent with the revolutionaries liberal ideology and the existing hierarchical social ties resulted in the non-monotonic evolution of surage provisions, a development unique to Latin America. In time, surage ceased to be the most prominent instrument of political participation. Workers, who had emerged as a social class at the turn of the nineteenth century, started to organize. Peasants, who were largely a dormant sector of society, became politically visible by articulating their demands in a coordinated fashion. In this section, I study the evolution of electoral provisions, as well as the developments of organized labor and the peasantry.

4.1

Electoral Participation

Elections were a premature occurrence in Latin America. There were no less than ve elections in Spanish America between 1809 and 1814,23 and this was even before the elections mandated by the 1812 Cadiz constitution took place. While elections were more frequent and regular after independence, the provisions included in the 1812 Spanish charter inuenced surage regulations during the XIX century enormously. The Cadiz constitution embodied a concept of citizenship rooted in the notion of vecino (literally neighbor), which was a sociological rather than a legal concept. It referred to a head of a family with an established standing in the local community and a
23 In 1809, the patriciates of the American provinces elected representatives to the Central Spanish Junta; between 1810 and 1812, Spain and loyalist America designated deputies and substitutes to the extraordinary Cortes. (Demlas-Bohy and Guerra 1996: 35) e

18 stable and independent source of income.24 Upon independence, the Latin American constitutions preserved the requirement of economic independence, which excluded servants, day workers, slaves, and women. Citizenship also tended to leave out those individuals of a renowned decient moral quality.25 Contrary to the experiences of European countries, were surage rights tended to be progressively extended, Latin American countries experienced oscillating processes of incorporation and exclusion. The right to vote was broad in the rst independent constitutions. A number of constitutions mandated a literacy requirement that was not to be applied for a decade in order to give time to the new citizens of the republic to learn how to read and write. Figure 3 depicts the evolution of surage restrictions in Latin America. It shows, by year, the proportion of countries that had economic and/or literacy requirements on male surage (labeled as narrow surage).

Figure 3: Proportion of Countries with Narrow Male Surage. Latin America, 1811-2005
The local sense of the vecino was reinforced by the fact that the parish juntas ruled without appeal on the prerequisites. As Annino summarizes it, [I]n short, in order to vote, a citizen must be settled in a community and be earning an honest living (1996: 65). 25 For example, the Bolivian 1843 constitution excluded those who were alcoholics or gamblers; the electoral regulations for the Ccuta Constituent Congress of 1820 that created Great Colombia excluded those married men u who - without legal reasons - did not live with their wives (Bushnell 1972: 196); and in the 1843 Ecuadorian charter those who were declared lazy lost citizenship. This moralistic spirit lingered in the legal language in Mexico, where having an honest living is a prerequisite of citizenship that was introduced in 1847 and has been kept until today.
24

19 Except for Argentina26 and the two countries that became independent in the twentieth century -Panama and Cuba-, Latin American countries typically incorporated economic requirements, subsequently followed by more restrictive literacy requirements, which aimed to exclude agricultural workers and tenant farmers. 10 out of 19 countries had broad franchise the year of their nal independence and by year 1850, 13 out of 16 countries had narrow surage.

Figure 4: Years When Particular Franchise Categories Were in Force. Latin America, 1811-2005

Figure 4 shows the franchise categories adopted by independent Latin American countries over the period 1811-2005.27 The spread is largest for the categories independent males and uni-

versal male, while it is narrower for propertied men and literate men and some women on a narrower basis. The restrictions on surage were prevalent around year 1850, followed by a progressive removal of requirements in the following decades. By 1950, only Brazil, Peru, and Chile kept literacy restrictions on surage; Brazil was the last country to remove them in 1985. Women were generally incorporated to the electorate when other restrictions had been lifted. The rst country that granted the right to vote to women was Ecuador in 1929 and the last one to
The Argentine constitution of 1853 left the regulation of the franchise to ordinary legislation. In 1856, the National Congress gave the right to vote in national elections to all male adults (Alonso 1996). Buenos Aires, in its provincial constitution of 1821, adopted broad but not universal surage, allowing all free men the right to vote. 27 The vertical lines inside each box show the median. Each box comprises the observations located between the 25th and the 75th percentiles of the distribution of each franchise category (50% of observations symmetrically around the median). The whiskers represent the upper and lower adjacent values, and the outliers are represented by observations outside this range.
26

20 do so was Paraguay in 1961. Once women acquired the right to vote, they almost never lost it.28 Universal surage for men and women existed in all countries after 1961. Universal, secret, equal, and direct surage is a contemporary phenomenon. Democracies29 have found legal ways to control the outcome of elections other than through the extent of the franchise. The regimes that supported elections in early Latin America frequently incorporated democratic lters (Asadurian et al. 2006). Regardless of the extent of the franchise, the lawmaker has other legal options to ensure favorable results, such as gerrymandering, registration procedures, the choice of the electoral formula, and the general administration of elections. Decisions over these matters manipulate the electoral process, and since these types of decisions have to be made, what is left to study is what types of manipulations produce what result. Examples of democratic lters are indirect and public voting.30 In nineteenth century Latin America, indirect and public voting were prevalent. The identity of those elected in rst, second, and third grade elections typically diered from those who were entitled to vote in rst grade elections. This was undoubtedly a way to ensure that decision making resided in the hands of specic groups. As Peloso and Tenenbaum (1996: 11) put it, [...] the voting public of elite and popular-sector men participated in primaries to select candidates, after which the rules called for the elites to sit with their peers and make the real decisions about public representation. Secret voting is supposed to guarantee the free vote: a vote free of physical coercion, patronage, and cohecho (vote buying).31 The data shows that the curves describing the proportion of countries with public voting and indirect presidential elections evolved similarly over time. Since having the right to vote does not mean exercising it, we may want to look at actual
In only two cases did an electoral reform withdraw womens right to vote based on gender. The 1948 military coup that ousted Venezuelan President Rmulo Gallego eliminated the universal surage by reinstating the 1936 o charter -which provided for literate and male surage- but allowed women to vote in municipal elections (Bushnell 1972, ft. 36). Those provisions were only eective during year 1949. The other case is Guatemala, where the 1956 excluded illiterate women but gave the facultative right to vote to illiterate men (OAS 1999). 29 A number of dictatorships have also employed the technology of voting to legitimate their tenure in power or as a way to display their power (Gandhi 2008). 30 Other aspects of surage regulations are also key, such as the electoral system and whether voting was compulsory. 31 A practical problem of secret voting was illiteracy. In Guatemala, two electoral laws proclaimed the secret vote during the nineteenth: in 1838 and 1848. Secret voting was the object of much criticism, which in turn served to delegitimize the 1848 elections. Mej (2000) transcribes an excerpt of an article published on November 1, 1850 in as the Guatemalan newspaper Gaceta: Those elections are a lie, because they are notably determined by indigenous tribes, whose members ignore what is written in the votes they cast [...] they are not and can never be the free and true expression of popular will. (p. 306, my translation). For the detractors of secret voting, the issue was whether those enfranchised would vote correctly.
28

21 voting patterns. The available information on the percentage of the population that actually voted is fragmentary. The percentage of the population that actually voted is a rough proxy of popular involvement in elections. Figure 5 shows the trends of the proportion of the total population that voted (participation), the proportion that was allowed to vote (eligible), and the proportion among those eligible who actually cast votes (turnout).32 The curve in the middle -eligibility- shows the proportion of the total population that was entitled to vote. Note that both curves start to increase steeply after 1900. The two-century average is 32.15%. Similarly, the lower curve depicts a steady increase of the participation rate after 1900.

Figure 5: Participation, Turnout, Eligibility. Latin America, 1846-2000

While the proportion of the population that voted in nineteenth century Latin America likely did not exceed 10%, there is no consensus about the scope of the electoral experience. The traditional position is that politics were an issue of the minorities.33 Some authors contend that elections were events than convoked widespread involvement, either as voters or as external participants.34 With little or widespread popular involvement, many elections were manipulated, vote
The upper curve -turnout- reveals a mean of 68.7%; notably, the Dominican Republic during the Trujillo era reported turnout levels of 100%. 33 For example, Bushnell and Macaulay (1994) point out that during nineteenth century Latin America the lower class was at the margin of political disputes. 34 For example, Muckes (2001) detailed analysis of the Peruvian 1871-72 presidential campaign reveals that the urban lower sectors, under a broad franchise, participated massively in the electoral campaign. Murilo de Carvalho (1995) states that elections under the rst republican constitution in Brazil were about keeping local power more
32

22 buying was widespread, and the results were frequently fraudulent. As Halper Donghi (1973: n 116) observed for the early period, [A]mong the many ways of overthrowing the government practiced in postrevolutionary Spanish America, defeat at the polls was conspicuously absent. I only know of three incumbent Latin American presidents who presented themselves for reelection and lost, two of them relatively recently.35 Denunciations of electoral corruption were not uncommon in later periods either.36 Ultimately, we may think that the existence of a market for votes reveals that voters are free to choose who to sell their vote. As Argentinean president Pellegrini put it, There is no freer vote than the one that is sold (quoted in Posada Carb (2000: 640 note 106)). o

4.2

The Institutional Incorporation of Workers

Many poor and disenfranchised Latin Americans were workers. Although working conditions at the turn of the century were appalling, the urban labor movements that emerged in the major cities in the 1890s were not strong enough to pose a serious challenge to the constituted oligarchical order (Halper Donghi 2002). n While not deadly threatening, worker protests were an unprecedented

form of expression that caused concern. Historical accounts report that the rst decade of the twentieth century witnessed widespread labor protesting, but systematic data on labor disputes for the period prior to 1919 are unfortunately not available. Workers rst dealt with the grievances through mutual aid organizations which later gave rise to dramatic, continent wide wave of worker protest in the late 1910s (Skidmore 1979: 97, as cited in Collier and Collier 2002: 60). Repression was the early government response to protesting, and it was driven by economic reasons. Latin American liberalism enjoyed the benets of the export led growth that allowed for sucient taxthan about exercising a political right, and hence local bosses mobilized their subordinates to the polls. Paula Alonso (1993) contends that in Buenos Aires, Argentina in the 1890s, elections were highly competitive, and involved all sectors of the population. Irurozqui (1996) states that a restricted franchise did not impede the electoral participation of the popular sectors in Bolivia. Valenzuela (1996) showed that the involvement of artisans and other working class groups in electoral politics greatly exceeded the number of men who actually voted in nineteenth century Chile. 35 Hiplito Mej in the Dominican Republic in the 2004 elections and Daniel Ortega Saavedra in the Nicaraguan o a 1990 elections. There is also the case of the Costa Rican Braulio Carrillo, rst elected in 1835 to complete the term of an incumbent who was forced to resign. Braulio Carrillo lost reelection in 1837 but one year later he overthrew the electoral winner and enacted a constitution that declared him president for life. He was deposed in turn in 1843. 36 Irregularities in the electoral process are nowadays common to democracies and dictatorships in dierent parts of the world (Little and Posada Carb 1996). Even old democracies have suered and still suer from controversial o electoral experiences; consider eighteenth century England, the nineteenth century Tammany Hall machine in New York, and the 2000 American presidential election.

23 ation to nance the increasingly more centralized government. The ongoing public works and the new exports meant that governments could not aord to let productive inputs -land and labor- rest (Peloso and Tenenbaum 1996). The often violently expressed labor demands eventually received a response in the form of legislation. At least two international pressure factors contributed, over time, to the legal recognition of workers rights. First, Catholic Church Pope Leo XIII issued the Encyclical Rerum Novarum in 1891, which addressed the issue of working conditions. Second, the International Labor Convention of 1921 adopted the Versailles Treaty of 1919 advancing workers rights (Bronstein 1998). These international trends, combined with domestic changes, led to a period more favorable to the workers cause in Latin America. In the 1920s, the model of the state predominant in Latin America emphasized state intervention in the economy, resulting in labor legislation that was highly protective of the worker.

Country Argentina Bolivia Brazil Chile Colombia Costa Rica Cuba Dominican Rep. Ecuador El Salvador Guatemala Honduras Mexico Nicaragua Panama Paraguay Peru Uruguay Venezuela

First Piece of Legislation 1905 1915 1919 1906 1905 1920 1909 1925 1916 1921 1906 1921 1917 1919 1914 1917 1911 1914 1917

First Labor Code 1949 1939 1943 1931 1950 1943 1984 1951 1938 1963 1947 1959 1931 1945 1947 1961 1945 1934 1936

Dierence 44 24 24 25 45 23 75 26 22 42 41 38 14 26 33 44 34 20 19

Table 3: Labor Legislation

24 Table 3 shows the years when workers rst gained policy concessions through legislation. The pioneer countries are Argentina and Colombia, where Sunday rest laws were passed in 1905.37 Costa Rica, Chile, and Ecuador rst addressed the eight-hour working day; El Salvador, Guatemala, Brazil, Peru, and Uruguay focused on work-related accidents; Venezuela started its labor legislation with a law on workshops and public facilities; Nicaragua issued a law on working hours, maternity, and unemployment in 1919; Panama simultaneously ruled on Sunday rest, child labor, and the eight-hour working day; and Mexicos 1917 constitution addressed labor issues in a more complete manner.38 The adoption of labor codes -a more comprehensive piece of legislation than a law- came much later (see Table 3). Latin American countries took on average 32.6 years to produce labor codes after the rst piece of labor legislation was enacted. Argentina, which was the leader in imple-

menting Sunday rest, took 44 years -the maximum in the region- to adopt a labor code under the Pern administration. o We may think of the dierence shown in Table 3 as the time it took for

labor interests to be fully institutionalized, at least on paper. In Central America, the 1940-50s saw the adoption of labor codes in the context of the consolidation of the import substitution model (Blanco Vado 1996). Surprisingly, the Cuban revolution did not issue a labor code until 1984, thus making Cuba the last country to adopt one in the region.39 The governments response to labor demands largely depended on the political context; however, the predominant strategy was that the incumbent government controlled the organized labor force. For example, pro-labor President Pern of Argentina issued a document stating the rights o of the workers in 1947 that did not mention the right to strike. An example of an even more striking show of a politically driven adjustment is found in Mexico in August 1916, when President Venustiano Carranza came to the extreme of signing a decree that established the death penalty for strikers. Nevertheless, once the political mood changed, he signed the 1917 constitution that granted the right to strike to workers. However, repression against strikers -a small proportion of
37 Argentina and Colombia were only preceded by Germany and Spain in implementing Sunday rest. The Dominican Republic, Honduras, Bolivia, Chile, and Paraguay included Sunday and holiday rests in their rst pieces of labor legislation. 38 The Mexican constitution was the rst one in the world that included a chapter on social issues and it became the Latin American model promoting the rights of workers (Bronstein 1998). 39 The Cuban labor code departs from the assumption that there is harmony between the interest of the only employer in the economy -the state-, workers, and labor unions (Bronstein 1998).

25 the population when compared to peasants- continued (De Buen Lozano 2006). Figure 6 shows the proportion of countries with strikes for the period 1919-1995. The total number of strikes peaked around year 1950 and later started to crawl up consistently until reaching a maximum in 1987, when 60% of countries experienced at least one strike. After that, striking tended to diminish.

Figure 6: Proportion of Countries with Strikes. Latin America, 1919-1995

Once a number of countries abandoned protectionist policies in the 1970s, the argument stating that high labor costs hindered competitiveness gained strength. The focus of the debate became the extent to which the goal of achieving labor exibility was desirable and acceptable for the parties involved. Several countries revised their labor legislation; for example, Argentina, Colombia, Ecuador, Panama, Peru, Venezuela, and Brazil. International credit institutions included as a prerequisite to access credit for structural adjustment plans, the task of labor exibility. However, that policy was later revised.

4.3

The Political Role of Peasants

If industrial workers fought a long ght until they received an institutional answer to their demands, peasants have taken even longer and their tale is as old as agriculture itself. The history

26 of the land tenure systems in Latin American countries is characterized by land concentration. The prototype land tenure system in the pre-Hispanic world was the ethnic seor where the Inca and n o the Aztec ruled (Sempat Assadourian, Challu and Coatsworth 2006). The Incas and Aztecs had imperial right to the lands of subjects and to the labor needed to work them. The Spanish king replaced the Indigenous king, and indigenous communities were allowed to produce and consume the products of the land owned by the Crown. Employment of forced labor was at the center of the colonial practice. The use of mita remained in mining centers. Independence did not bring land reform to Latin America.40 The period of turmoil after the wars of independence provided the opportunity for Creole families to amass giant states -which the colonial rule had forbidden- through a series of mechanisms.41 Hence, while cosmetically different, the land tenure structure suered little change in comparison with colonial times. The new republics implemented other changes in the countryside that were related to the revolutionary ideology. Most former colonies abolished slavery42 and mayorazgos, which were neither economically nor politically advantageous to the new ruling elite (Mirow 2004).43 By the end of the nineteenth century, big commercial farms emerged, usually with imported or foreign capital, devoted to agriculture for export of single crops, such as sugar cane, bananas, coee, wheat and livestock. It was an ecient system that sought to maximize mechanization. Where it succeeded, the unemployed masses resented it. Latin Americans began to advance demands for land reform, usually in association with liberal democratic movements opposed by communism under the banner Bread, Land, and Liberty (Pico 1964: 147-148). In countries with large indigenous populations, abuses persisted.
The notable exception is Uruguay before its nal independence where Artigas two-year rule (1814-1816) issued a plan for radical agrarian reform that Eduardo Galeano has called the rst Latin American land reform in history. 41 Mirow (2004: 150) mentions a few: Some individuals merely needed to hold on to their large estates through independence, having aligned themselves with the winning side. Others saw civil unrest as presenting a prime opportunity to push less powerful people o land during the instability period that followed independence. Still others amassed landed estates through the time-honored practice of dividing up the spoils of war. In some areas, farm after farm and estate after estate were brought together to give individuals ownership and authority over vast regions (latifundios) (...) In the pulling together of massive estates, the wealth of the caudillo system developed to create an almost quasi-governmental authority in some areas. 42 It is not clear if abolitionism was the product of the revolutionaries own convictions or rather a trend of the times. The antislavery movement had gained strength at the time of Latin American independence leading to the British parliament abolition of slavery in all of Britains colonies in 1833. 43 From an economic perspective, mayorazgos tied a set of assets together to the owner and hence removed land from a uid market. Politically, the French revolutionaries had set an example by abolishing them, and since they required royal consent, following the French example was also a show of disapproval of royal institutions.
40

27 The crisis of the liberal agro-export model in the 1930s gave rise to a new socioeconomic order that aected the countryside. This change in the model involved a social pact. Petras and

Veltmeyer describe it as follows: The 1930s crisis of the liberal agro-export system led to the emergence of a new import-substitution model, which harnessed agro-exports to local industrial production without changing the domination of the agricultural elite over the peasantry and the rural labor force. In eect, the ascendancy of the urban bourgeoisie and petty bourgeoisie involved a trade-o in which the agrarian oligarchy accepted its political subordination in exchange for continued control in the rural sector. Agrarian reform supposedly a democratic demand of the progressive bourgeoisie had no part in the social pact between the urban bourgeoisie and the agrarian oligarchy (2002: 50).44 Thus, agrarian reforms would have to wait longer. Since the rst land reforms in the region (Mexico in 1917, Bolivia in 1952, and Cuba in 1959) were also social revolutions, and amidst the cold war, the Kennedy administration sponsored the Alliance for Progress, a program for the socioeconomic development of Latin America. In exchange of introducing programs of comprehensive agrarian reform, Latin American countries would receive American nancial assistance at very low or zero interest rates, totaling one billion dollars for the twelve-month period after the Alliance was announced on March 13, 1961. All Latin American countries, except for Cuba, subscribed to it. In 1962, right after the benets of the Alliance for Progress could be exploited, ve countries (Costa Rica, the Dominican Republic, Honduras, Panama, and Chile) introduced land reforms. In total, the 1960s witnessed twelve such reforms. The hype later tended to wane down. In the 1970s, only eight land reforms are recorded and three of them occurred in Ecuador; ten in the 1980s; and eight in the 1990s. More recently, Venezuelan President Hugo Chvez initiated a land a reform process in 2001. The earliest agrarian reform occurred in Costa Rica in 1909 when the Law of Family Heads granted every male Costa Rican of at least 20 years of age or emancipated, the one-time right to 50 hectares of national vacant land (Royo Aspa 2003). However, the Costa Rican reform was unsuccessful.

Bejarano (1985) agrees with this interpretation: The history of the agrarian ghts of the XX century is a call reaching out to the peasantry as a political object and its rapid conversion into a political subject, which permanently provokes the reunication of the dominant classes to conspire against this challenge (p. 63-4, my translation)

44

28

Figure 7: Percentage of Family Farms. Latin America, 1850-1990

Did land reforms result in a more egalitarian distribution of land? Figure 7 shows the evolution of the percentage of family farms in Latin America for the period 1850-1990. On average, the curve shows an upward trend until 1980 (there is only one data point for 1990 and hence we cannot infer from it that the regional trend declined). Coatsworth (1998, 2005) and Prados de la Escosura (2003) maintain that inequality increased in Latin America only during the second part of the nineteenth century, when improved transportation and the expansion of trade made land more valuable. Williamson (2004: Table 7) shows that, with 1913 as 1:0, the ratio of real wages to land value declined from 6:9 in 1880-1884 to 0:7 in 1930 in Argentina and from 11:1 in 1870-1874 to 1:1 in 1930 in Uruguay. Hence, it is not clear whether land inequality was an original sin of Latin America or it was generated only when the continent embraced the export-oriented development strategy during the second half of the nineteenth century. Between 1850 and 1970, the proportion of family farms increased by 16:5 percent in an average Latin American country, with the range of 2 to 40 percent. Those Latin American countries that had more family farms in 1850 tended to redistribute more land. However, while land inequality has persisted in Latin America, the relative importance of land as an asset has consistently declined over time.

29

Results

The data presented in the previous sections are part of a time series cross sectional dataset, with the time series dimension of a higher magnitude than the cross sectional dimension. The unit of analysis is country-year. Original data on political stability and political participation were collected using a combination of secondary sources and archival information. The coding protocol and data sources are described in detail in the codebook. Table 4 shows the correlation between the three measures of political instability.

Constitutional Change Executive Turnover Coup detat

Const. Change 1.00 0.19 0.19

Exe. Turnover 1.00 0.58

Coups

1.00

Table 4: Correlation Coecients Between the Three Measures of the Dependent Variables

The correlation coecients are positive but not very high. Indeed, while the three variables measure instability, each of them captures a dierent aspect of the phenomenon. Substantially, the variable completed terms tells us whether the system attained some degree of predictability. Successful coups detat and executive turnover do not only measure institutional instability, but political instability in general. Constitutional change measures institutional stability. Hence, in order to better understand the dynamics between political stability and political participation, and also to assess the robustness of the ndings, all three measures are used as dependent variables in dierent model specications. The choice of the statistical model depends on the measure of political stability employed and on how it is expressed. The variable coups detat (COUPS DUMMY) is dichotomous; it is equal to one if at least one coup detat occurred in a given year and it is zero otherwise. I transformed the variable constitutional endurance presented in Section 3 into a dummy (CONST CHANGE); it is equal to one if the constitution was changed or amended in a given year and it is equal to zero otherwise. Hence, either probit or logit models to study the probability of a coup detat and the probability of a

30 constitutional change, respectively, are suitable.45 To account for executive turnover, I constructed the variable HEADCHANGES, which is a linear transformation of the variable HEADS described in Section 3. HEADCHANGES counts the number of changes of chief executives in a given year. HEADCHANGES is count data, with a distribution that approximates a Poisson distribution.46 Therefore, I used a Poisson model to analyze the eects of the independent variables on the number of changes of chief executive in a given year. Wherever needed, I tested for the appropriateness of introducing country xed eects to account for unobserved country characteristics that are time-invariant and may aect the predictor, such as such as levels of socio-economic development, geography, etc. When country xed eects were not used, I corrected heteroskedastic standard errors by estimating the models with robust standard errors clustered by country. I included variables measuring electoral participation (e.g., the percentage of the population voting (PART PR), the scope of the surage (FRANCHISE), franchise extensions (FRANCHISE EXT) and contractions (FRANCHISE CONTR), direct voting in presidential (PRESDIRECT) and legislative elections (LEGDIRECT), secret ballot (SECRET BALLOT), and the occurrence of presidential elections (PRESELEC)) in the right hand side of the equations. Concerns about endogeneity arose frequently; for example, between variables measuring electoral participation and coups or constitutional changes. To illustrate this, consider that the occurrence of a coup47 may aect the level of electoral participation and, more directly, the occurrence of elections. In the absence of suitable instrumental variables, I dealt with these issues by lagging the relevant independent variables whenever possible.48 I incorporated in all regressions a variable measuring the presence or absence of political opposition;49 other independent variables of interest capture the strength of
45 The only cases in which logit and probit estimates may dier substantially is where there are heavy accumulations of observations in the tails of the distribution, when logit models are more appropriate (Liao 1994). 46 It is obvious from an eye inspection of the data - e.g., a histogram - that the variable HEADCHANGES has a Poisson-like distribution (graph not shown). In any case, skewness is equal to 3.355897 and kurtosis equals 22.99602. 47 A coup detat (an irregular transfer of power) includes both revolutions that extended the franchise, as in Bolivia in 1952, and military coups that eliminated voting, as in Guatemala in 1963. 48 The lagged values of endogenous variables are treated as exogenous variables, because for determination of the current periods values of the endogenous variables they are given constants (Kennedy 2003). 49 Political pluralism -understood as the presence of at least a modicum level of opposition- existed in two thirds of the observations in the dataset. The proportion of countries with completed terms and the proportion of countries with opposition correlate fairly well (0.7), as well as the average number of consecutive years with opposition and

31 the military, past instability, the length of the presidential term, and the strength of the labor and peasant movements. A control for economic conditions was also included whenever the data was available. Since the data was described in detail in the previous sections, I dig in the statistical analysis without further preamble.

5.1

The Nineteenth Century

As discussed before, workers and peasants were practically unorganized in the nineteenth century. Hence, electoral participation was the sole expression of political participation during that period. I contended that electoral participation had no eect on political stability during the nineteenth century. Ideally, one would want to incorporate data on electoral participation from actual elections in the model specications. Unfortunately, there are only 11 observations on turnout for the nineteenth century, 22 on the proportion of the population eligible to vote, and 54 on the percentage of the population that voted in national elections.50 Hence, I employed data on formal electoral rights to proxy electoral participation.51 Another disadvantage of the empirical analysis of the nineteenth century is that we lack appropriate controls for economic performance. Table 5 shows logit estimates of the probability of the occurrence of a coup detat during the period 1811-1899. Let us recall that neither the Exclusion nor the Inclusion Hypotheses made distinctions about the eect of political participation on political instability according to time period; they predict instead a signicant negative eect and a signicant positive eect on political stability respectively. The variables measuring the level and the changes in the size of the electorate have no eect on political stability expressed as coups detat during the nineteenth century.

the average number of consecutive years with completed terms (0.9). The analysis of the evolution of opposition and completed terms across countries is indicative, however, of intra regional heterogeneity. 50 The mean of the variable percentage of the population that voted during the nineteenth century is very low, at 6.59%. 51 I run a very rudimentary probit model for the period 1811-1899 with the available data on voting. The dependent variable was coup detat; the independent variables were the lagged scope of the franchise and the lagged percentage of the population voting (N=52). Both coecients were insignicant. Note that the model was poor since it was capable of correctly predicting only half of the positive outcomes.

32 Variable Lagged Scope of the Franchise Lagged Franchise Extension Lagged Franchise Contraction Lagged No. Pres. Elections Lagged Secret Ballot Lagged Direct Voting (Pres.) Lagged Direct Voting (Leg.) Year of Final Independence Post Independence Turmoil Past Coups Lagged Opposition Constant N Prob>chi2 (1) Coef. (Robust SE) 0.03 (0.09) 0.11 (0.54) -0.78 (1.09) -0.13 (0.28) 1.08*** (0.28) 0.13 (0.43) -0.06 (0.32) -0.00 (0.02) 0.01 (0.01) -0.11** (0.04) -0.52* (0.23) -6.77 (30.32) 654 0.0000 (2) Coef. (Robust SE) 0.07 (0.06) 0.23 (0.54) -0.71 (1.02) (3) Coef. (Robust SE) -0.05 (0.08) 0.43 (0.53) -0.73 (1.06) -0.16 (0.27) 0.71** (0.26) 0.24 (0.40) -0.31 (0.27) (4) Coef. (Robust SE) 0.06 (0.07) 0.46 (0.58) -0.56 (0.98) -0.09 (0.28)

0.03 (0.03) 0.00 (0.01) -0.00 (0.03) -0.66* (0.27) -49.26 (29.31) 777 0.0087

-0.56* (0.24) -1.68*** (0.34) 659 0.0000

-0.64* (0.26) -2.07*** (0.35) 771 0.0705

Notes: * p<0.05; ** p<0.01;*** p<0.001.

Table 5: Logit Regression. Dependent Variable: Coups detat, 1811-1899

Results do not change when we include lags up to four periods of the variable Scope of the Franchise (not shown) and the coecients of those lagged variables are insignicant.52 According to the model chi2 of the likelihood ratio statistic, the models in Table 5 are overall signicant. The full model shown in column (1) indicates that the secret ballot contributed to the occurrence of a coup while the number of past coups and the presence of a modicum of opposition had a significant and negative eect on the probability of a coup. A control for the number of presidential
Four lags is an appropriate choice, given that presidential terms in the nineteenth century have a mean of 3.7 and a median equal to 4. In the case of legislative terms, the mean is 3.1 and the median is also 4. The length of presidential terns and the length of legislative terms are proxies to the presumed regularity of elections.
52

33 elections (PRESELEC) occurred in the previous year was insignicant.53 The variable Past Coups (PAST COUPS) counts the number of coups occurred in a country since nal independence until t-1 and it measures past instability. The odd result in column (1) is that the more coups a country experienced in the past, the less likely it is to experience a coup in the current year; however, the eect of past coups disappears in the model shown in column (2). The alternative models shown in columns (2) through (4) reveal that the signicance of the variables Lagged Secret Ballot and Lagged Opposition is not sensitive to model specication.54 The variables that operationalize the democratic lters discussed in Section 4 were removed in column (2). The model in column (3) leaves out the indicators of past instability. The minimal model shown in column (4) shows regressors for the franchise variables and political opposition and the latter is signicant and with a negative sign. The parameter estimate for the variable Lagged Opposition is negative and signicant in all model specications. According to the results of model (1), for a year when there is opposition, the odds of a coup detat are 0.52 smaller than the odds of coup in a year without opposition (odds ratios not shown). The coecient of the variable Lagged Secret Ballot is signicant and positive in models (1) and (3).55 One interpretation of this result is that as voters became harder to control given secret voting, uncertainty over the electoral results may have triggered a coup. Table 6 presents the results of very similar model specications as before, but with Constitutional Change as the dependent variable. The only dierence is that the measure of past instability that I employed accumulated consecutive years with one constitution - ACCCONST) is dierent than the one before.56 As the actual occurrence of presidential elections is not aected in the same time period by a constitutional change, the variable Number of Presidential Elections was not lagged one period. Note that the variables measuring the size of the electorate and its changes
53 The historical narrative in the previous sections suggested that during the nineteenth century it was not uncommon to resolve electoral disputes with violence. However, elections and coups are endogenous. In the absence of a proxy of elections, I took a second-best approach: I lagged the variable that counts the number of presidential elections during a given year (PRESELEC) 54 The Hausman Test tests the null hypothesis that the coecients estimated by the ecient random eects estimator are the same as the ones estimated by the consistent xed eects estimator. For model (1), it reported a Prob>Chi2=0.2206 and hence xed eects are inappropriate. 55 There were very few cases of secret voting in nineteenth century Latin America (28% of country-year observations). 56 Employing this measure of past instability makes more sense because it measures past constitutional stability specically.

34 remain insignicant.

Variable Lagged Scope of the Franchise Lagged Franchise Extensions Lagged Franchise Contractions Number of Presidential Elections Lagged Secret Ballot Lagged Direct Voting (Pres.) Lagged Direct Voting (Leg.) Year of Final Independence Lag Acc. Years with 1 Const. Turmoil Lagged Opposition Constant Country Fixed Eects N Prob>chi2

(1) Coef. (Robust SE) -0.02 (0.06) 0.60 (0.50) -0.93 (1.15) 0.22 (0.47) -0.19 (0.21) 1.87** (0.64) -1.70* (0.70) 0.02 (0.02) -0.03 (0.02) 0.00 (0.00) -0.50 (0.32) -37.29 (31.09) No 654 0.0000

(2) Coef. (SE) 0.02 (0.10) 0.72 (0.64) -0.93 (1.08) 0.11 (0.29) -0.71 (0.49) 1.44* (0.73) -1.11 (0.77) (a) -0.02 (0.02) (a) -0.61 (0.33)

(3) Coef. (Robust SE)

(4) Coef. (SE)

0.43 (0.35)

0.35 (0.24)

Yes 500 0.0776

0.02 (0.01) -0.06** (0.02) 0.01* (0.00) -0.54* (0.25) -28.70 (21.29) No 838 0.0000

(a) -0.04* (0.02) (a) -0.59* (0.27)

Yes 753 0.0006

Notes: * p<0.05; ** p<0.01;*** p<0.001. (a) Omitted because it is constant within group.

Table 6: Logit Regression. Dependent Variable: Constitutional Change, 1811-1899

Incorporating lags of the variable Scope of the Franchise for up to four periods does not change the results (not shown). One may argue that it would be wise to look for determinants of constitutional change outside of constitutional provisions. The models presented in columns (3) and (4) exclude the lagged variables qualifying surage -constitutions determine the values of those variables- and focuses on past instability and lagged opposition. The results of model (3) do not

35 change when we incorporate country xed eects. The coecient of Lagged Opposition is negative and signicant in models (3) and (4); we may interpret this as evidence that when the opposition is allowed to be a part of the political game, even if to a minimal degree, opposition leaders have incentives to play by and maintain the existing rules of the game. The age of a constitution matters in models (3) and (4): the longer a constitution has been around, the less likely it is that it will be changed. A puzzling result in Table 6 is that the sign of direct voting in presidential elections is positive while its counterpart in legislative elections is negative, with both coecients being signicant. Note that this is the case regardless of who is entitled to vote (given the control for the extent of the franchise). A possible interpretation is that direct voting in presidential elections is more eective to bring about changes because a decided chief executive may introduce a reformist agenda, while the legislature is a collegiate body where decisions are not automatically agreed.57 The signicance of the variable Lagged Direct Voting (Legislative) disappears with country xed eects. The variable Number of Chief Executive Changes is the dependent variable in the Poisson regression results presented in Table 7. The model in column (1) includes, as in the previous regressions, variables indicating electoral participation, past instability, and opposition. I added the variable Military Head (NMIL), a dummy equal to one when the nominal head of government is or was a member of the professional military. The variable Number of Presidential Elections has a positive and signicant eect on executive turnover, which reects the institutionally mandated turnover resulted from elections.58 None of the other variables matter for executive turnover, except for opposition. The coecient of the variable Opposition is negative and signicant.

Direct voting in legislative elections was more common than direct voting in presidential elections in nineteenth century Latin America. 58 This, however, does not mean alternation in power. During the nineteenth century, 58% of all alternations in power that resulted from an election did not imply a change of the chief executive (the chief executive remained the same but changed parties) or his party (the candidate of the incumbent party won the election).

57

36

Variable Scope of the Franchise Lag. Franchise Extension Lag. Franchise Contraction No. of Pres. Elections Secret Ballot Direct Voting (Pres.) Direct Voting (Leg.) Length of Pres. Term Military Head Year of Final Ind. Length of Turmoil Lag. Acc. Years with 1 Head Opposition Constant Country Fixed Eects N Prob>chi2

(1) Coef. (Robust SE) 0.01 (0.03) -0.50 (0.50) -0.63 (0.49) 1.30*** (0.16) 0.11 (0.17) 0.07 (0.22) 0.26 (0.28) -0.03 (0.04) -0.12 (0.10) 0.02 (0.01) 0.00 (0.00) 0.00 (0.04) -0.39** (0.13) -38.18 (21.02) No 640 0.0000

(2) 95% CI [-0.04 0.06] [-1.04 0.93] [-1.60 0.34] [0.99 1.61] [-0.23 0.45] [-0.37 0.50] [-0.30 0.82] [-0.11 0.05] [-0.33 0.08] [-0.00 0.04] [-0.01 0.01] [-0.07 0.07] [-0.65 -0.13] [-79.37 3.01] Yes 640 0.0000 0.02 (0.03) -0.58*** (0.13) Coef. (SE) -0.06 (0.05) 0.15 (0.37) -0.58 (0.51) 1.34*** (0.10) 0.05 (0.21) 0.37 (0.25) 0.05 (0.28) -0.01 (0.04) -0.32* (0.13) 95% CI [-0.15 0.04] [-0.58 0.87] [-1.58 0.43] [1.14 1.55] [-0.36 0.45] [-0.11 0.86] [-0.51 0.60] [-0.09 0.07] [-0.58 -0.06] (a) (a) [-0.03 0.07] [-0.83 -0.33]

Notes: * p<0.05; ** p<0.01;*** p<0.001. (a) Omitted because it is constant within group.

Table 7: Poisson Regression. Dependent Variable: Number of Chief Executive Changes, 1811-1899

37 I controlled for country xed eects in model (2) and the results remained the same, except that the coecient of the variable Military Head is signicant. This result suggests that if the chief executive as of the end of the year is aliated with the professional military, it is expected to have a negative impact on executive turnover when compared with civilian counterparts.59 In both models, allowing some opposition makes a high chief executive turnover less likely. To be exact, model (2) predicts that while holding the other variables constant in the model, we should expect a decrease of the executive turnover by a factor of 0.56 when there is political opposition (incidence rate ratios not shown). The empirical evidence suggests that the indicators of electoral participation had no eect on political stability during the nineteenth century. Additionally, the regression results indicate that the presence of at least a modicum of political opposition is a determinant of political stability. The Post Independence Turmoil variable had no consistent negative and signicant eect on political stability. However, the suspected factors that made surage ineective during that century, that is, the democratic lters did not have signicant eects in a consistent manner: secret voting mattered for coups, and direct voting in presidential elections was positive and signicant in the models of constitutional change. Therefore, the sole factor that has a positive and signicant eect on political stability in nineteenth century Latin America is the presence of at least a modicum of political opposition. The decision to study this issue up to 1899 was arbitrary and based on the fact that organized labor and peasant movements were practically inexistent during that time. We now turn to the analysis of twentieth century, were more factors come into play.

5.2

The Twentieth Century

In this section, I included variables measuring electoral participation, labor organizational strength, past instability, economic conditions, and policy concessions to labor and the peasantry on the right hand side of the equations. The dummy variable Broad Franchise (FRANCHISE BROAD) is equal to one when male surage was not conditioned on economic and literacy requirements. Hence,
This result is not intuitively obvious. While the longest tenured chief executives were military in nineteenth century Latin America, on average they lasted in power less than civilian heads (1.98 years vs. 3.33 years).
59

38 Broad Franchise is adequate to measure whether the larger, poorer, and illiterate masses could vote. Including the variable Lagged Percentage of the Population Voting, which is coded per election, implies the inclusion in the sample of the year after elections only. This variable is important for our purposes, since it goes beyond capturing voting rights only; it expresses how people actually exercised the right to vote. To account for actual electoral participation, I used the variable Percentage of the Population Voting (PART PR) instead of the other two candidates -turnout (TURNOUT) and the percentage of the population eligible to vote (ELIGIBLE PR)- because it has less missing information. The correlation coecient between PART PR and ELIGIBLE PR is very high and positive (0.90).60 The variable STRIKES counts the number of strikes occurred during a given year. This variable is a proxy to measure the strength of the labor movement, indicating that workers were suciently organized for collective activity. Labor Legislation (ALL LABOR) is a dummy variable indicating the adoption of important pieces of labor legislation to proxy policy concessions to labor. Agrarian Reform (AGRARIAN REFORM) does the same for the peasantry.61 A control for economic conditions is introduced through the variable Lagged Log GDP per Capita. In addition, a control for the Length of the Presidential Term (PRESTERM) was included whenever it was relevant to do so. Table 8 presents logit estimates of the probability of the occurrence of a coup detat during the twentieth century, specically between 1920 and 1996.62 Column (1) shows the results in a reduced sample, where the variable Lagged Percentage of the Population Voting is included. The coecient of Lagged Opposition is negative and signicant, thus indicating that when there is opposition, coups are less likely. Ample voting rights (variable Broad Franchise) made coups less likely as argued in the Inclusion Hypothesis, but the percentage of the population that actually voted did not matter. Past Instability increases the probability of a coup.63 Column (2) shows the results
The variables PART PR and ELIGIBLE PR evolved similarly after 1850. It is extremely hard to nd data on the organized peasantry. An alternative to measure the size of the peasantry is to take the variable measuring the number of people working in agriculture and like-sectors (see variable AGR ALL in the codebook) as a proxy. For the purposes of my specic analytical endeavors there was a downside to this strategy. First, there are only 98 observations on AGR ALL and due to lack of coincidence with other data; no meaningful regressions could be run with it. Second, the size of the peasantry -and this line of thinking also applies to workersdoes not reveal their organizational capabilities. 62 Data availability on strikes limits the time coverage to the period 1920 - 1996. 63 When we control for country xed eects with the reduced sample (year after elections) (not shown), the coecient of the variable Opposition is the only one signicant and it has a negative sign.
61 60

39 of the model using the full sample and controlling for country xed eects. Voting rights have no impact on the probability of a coup. As the Lagged Number of Strikes increases, coups are more likely. Policy concessions to workers and peasants do not appear to aect the probability of a coup. A puzzling result is that when we control for xed eects with the full sample, past instability continues to be signicant but its sign changes.64

Variable Lagged Broad Franchise Lagged Franchise Extensions Lagged Percentage of the Pop. Voting Lagged Number of Strikes Lagged Labor Legislation Lagged Agrarian Reform Lagged Military Head Past Coups Lagged Opposition Lagged Log GDP per Capita Constant Country Fixed Eects N Prob>chi2

(1) Reduced Sample Coef. 95% CI (Robust SE) -0.94* [-1.84 -0.03] (0.46) 0.49 [-1.92 2.90] (1.23) -0.00 [-0.03 0.03] (0.02) 0.34 [-0.12 0.81] (0.24) 1.10 [-1.53 3.73] (1.34) 0.08 [-2.14 2.31] (1.14) 0.02 [-0.87 0.91] (0.45) 0.06* [0.01 0.11] (0.02) -1.97*** [-2.83 -1.11] (0.44) -0.29 [-1.10 0.52] (0.41) 0.59 [-5.62 6.80] (3.17) No 313 0.0000

(2) Full Sample Coef. 95% CI (SE) -0.17 [-1.01 0.66] (0.43) 0.07 [-1.19 1.33] (0.64)

0.33** (0.12) -1.74 (1.03) -0.20 (0.57) 0.13 (0.28) -0.03** (0.07) -0.70** (0.26) 0.49 (0.45) 1.88 (1.25)

[0.10 0.57] [-3.75 0.27 ] [-1.43 1.03] [-0.42 0.69] [-0.36 -0.09] [-1.22 -0.18] [-0.40 1.37] [-0.57 4.32] Yes 1198 0.0000

Notes: * p<0.05; ** p<0.01;*** p<0.001.

Table 8: Logit Regression. Dependent Variable: Coups detat, 1920-1996

64 Why would results dier across the two samples? The model specication is dierent since the inclusion of the variable Lagged Percentage of the Population Voting reduces the sample to years after an election only. I believe that it is substantially more meaningful to observe the results of the full sample, even if we lose the assessment of the impact of the Lagged Percentage of the Population Voting, because coups occur in non-election years as well.

40 When the dependent variable is Constitutional Change (see Table 9) the only coecient that is signicant in model specications (1) and (2) is that of the variable Opposition, with a negative impact on the probability of a Constitutional Change. Model (1) includes election years only (the variable Percentage of the Population Voting is coded in election years and is missing otherwise) and model (2) uses all the data for the period 1920-1996. Hence, political opposition contributes to constitutional stability in both the reduced and the full sample, with and without country xed eects. Labor legislation appeared to diminish the probability of a constitutional change.

Variable Lagged Broad Franchise Lagged Franchise Extension Percentage of the Population Voting Number of Strikes Labor Legislation Agrarian Reform Military Head Lag Acc. Years with 1 Const. Opposition Log Lagged GDP per Capita Constant Country Fixed Eects N Prob>chi2

Reduced Sample (1) Coef. (2) Coef. (Robust SE) (SE) -0.05 0.40 (0.57) 1.11 1.42 2.62 (1.19) 1.34 -0.02 -0.09* (0.03) (0.04) -0.09 -0.06 (0.37) 0.48 0.67 0.72 (0.88) (1.23) 0.55 -0.25 (0.74) (1.46) -0.11 0.62 (0.69) (0.84) -0.01 0.02 (0.01) (0.02) -2.10*** -2.14** (0.59) (0.68) -0.15 0.88 (0.77) (0.95) 0.60 (5.82) No Yes 310 223 0.0000 0.0026

Full Sample (3) Coef. (4) Coef. (Robust SE) (SE) -0.13 0.68 (0.22) (0.55) 1.07 1.03 (0.58) (0.59)

0.14 (0.15) 0.96* (0.47) 0.65 (0.51) 0.21 (0.28) -0.01 (0.01) -1.14** (0.36) -0.31 (0.28) 0.26 (1.95) No 1189 0.0000

0.23 (0.13) 0.95* (0.44) 0.81 (0.52) 0.67 (0.35) 0.02 (0.01) -0.90** (0.32) -0.75* (0.32)

Yes 1118 0.0000

Notes: * p<0.05; ** p<0.01;*** p<0.001.

Table 9: Logit Regression. Dependent Variable: Constitutional Change, 1919-1996

41 The Poisson regression shown in Table 10 does not provide new results. In the reduced sample models -with and without country xed eects- only the variables Opposition and Strikes matter for Executive Turnover. However, these eects disappear in the full model with xed eects.

Variable Franchise Broad Franchise Extensions Percentage of the Population Voting Length of Presidential Term Number of Strikes Labor Legislation Military Head Agrarian Reform Lag. Acc. Years with 1 Head Opposition Lagged Log GDP per capita Constant Country Fixed Eects N Prob>chi2

Reduced Sample (1) Coef. (2) Coef. (Robust SE) (SE) -0.38 -0.32 (0.20) (0.28) 0.28 0.25 (0.27) (0.42) -0.01 -0.01 (0.01) (0.01) -0.05 -0.08 (0.06) (0.07) 0.30** 0.25** (0.10) (0.09) 0.25 0.43 (0.41) (0.38) -0.15 -0.14 (0.21) (0.20) 0.12 0.07 (0.00) (0.35) 0.02 0.02 (0.25) (0.03) -0.42* -0.40* (0.17) (0.20) -0.05 -0.04 (0.13) (0.27) 0.94 (0.91) No Yes 311 311 0.0000 0.0832

Full Sample (3) Coef. (4) Coef (Robust SE) (SE) -0.31 -0.31 (0.14) (0.19) 0.10 0.00 (0.34) (0.31)

-0.10*** (0.02) 0.16* (0.06) -0.11 (0.30) -0.11 (0.21) 0.40 (0.30) 0.00 (0.03) -0.26 (0.20) -0.04 (0.09) 0.06 (0.73) No 1188 0.0000

-0.14 (0.03) 0.15 (0.04) -0.08 (0.26) -0.16 (0.14) 0.48 (0.21) 0.04 (0.02) -0.30 (0.13) -0.23 (0.14)

Yes 1188 0.0000

Notes: * p<0.05; ** p<0.01;*** p<0.001.

Table 10: Poisson Regression. Dependent Variable: Chief Executive Turnover, 1919-1996

Of the three measures of stability, Executive Turnover is the only one that is not signicantly aected by Opposition in the regressions for the twentieth century. In substantial terms, this

42 should not be so surprising. Consider a long tenured head of government that stays in power for many years governing with an iron st: opposition is zero. A high executive turnover per se does not mean institutional instability; the political instability that it implies could be a sign of eective processes of alternation, as in cabinet changes under parliamentary regimes. This variable captures both the healthy turnover and the unhealthy turnover.

In the previous pages, we investigated the relationship between political participation in the form of electoral participation and political stability during the nineteenth century. Then we studied that relationship over the twentieth century and added variables that proxy the organizational strength of workers and peasants. Partitioning the sample in that fashion was necessary due to substantial and practical reasons. First, workers and peasants were not organized in the nineteenth century and the democratic lters were irrelevant once voting converged to secret and direct over the twentieth century. Second, because workers and peasants were not suciently organized in the nineteenth century, data on their organizational strength does not exist for that period. However, simple regressions including the data that is available for the two enturies65 and employing the three dependent variables maintains the nding that opposition is a determinant of political stability.

Conclusion

Latin Americas political performance varies greatly across and within countries, particularly with regard to their experiences with political participation and political stability. I employed three indicators of political instability: executive turnover, constitutional change, and coups detat. These three variables measure dierent dimensions of political instability. In substantial terms, executive turnover captures the regular executive turnover due to legal ways of transferring power as well as illegal changes. Constitutional change expresses the frequency of the change in the rules of the game, thus indicating institutional instability. Coups detat reveal the breakdown of institutions.
65 The independent variables employed are:FRANCHISE BROAD, NMIL, PAST COUPS, OPPOSITION. Depending on the model, lags were introduced as necessary.

43 Is there an underlying determinant of the dimensions of political stability? The predictions of the Exclusion and the Inclusion Hypotheses, as well as my own conjectures presented in Section 2 were confronted with the data. I divided the statistical analysis into two time periods: the nineteenth and the twentieth centuries. Regression analyses provided support for my conjectures in that electoral participation did not matter for political participation during the nineteenth century. Hence, neither the Inclusion nor the Exclusion Hypothesis was veried in nineteenth century Latin America. In accordance with my expectations, there is some evidence that secret voting and direct voting in presidential elections operated as democratic lters, thus making surage ineective. I incorporated new elements in the analysis of the twentieth century: organized workers and peasants. The statistical analyses employing the three measures of political stability as dependent variables showed mixed evidence in light of the hypotheses at stake. There was no consistent support for the Inclusion and Exclusion hypotheses, given that the results varied across models and most of the time the variables of interest had no eect on political stability. Part of the reason why the indicators of organizational strength of workers and peasant employed had no eect on political stability might be that there is no adequate data to measure their organizational strength. Ideally, one would want to have data on labor union membership and peasant organization membership. Gathering these data and testing once again the hypotheses would be a topic for further research. Overall, I did not nd evidence that political participation aected political stability. The sole consistent and robust result is that the existence of a modicum level of political opposition is key to attain political stability. Now, why does political pluralism aect political stability? Institutions that tolerate some opposition are able to absorb and resolve conicts. Allowing political opposition provides all political actors with incentives to engage in the political contest with the existing rules of the game. To be self-sustaining, institutions must reect the relations of force; they may restrict political access to those groups that can mobilize military prowess and still favoring those more powerful among those groups that are admitted into the institutional interplay of interests. Institutions that do not tolerate some opposition do not absorb and resolve conicts; instead, they outright eliminate manifestations of dissent. The apparent tranquility of periods of dictatorship -when no opposition is allowed- hides the fact that institutions able to process conicts could not be achieved. This

44 was the case in nineteenth century Latin America, when politics were characterized by intra-elite conict. Resorting to electoral fraud to guarantee a favorable election outcome for the incumbent or his party was fairly common. While the incumbent governments carefully manipulated elections, opposition was legally tolerated, allowed to win some seats in the legislature, and sometimes even a share of power. And these incentives were most often sucient for the opposition to participate. Hence, intra-elite conicts were processed according to rules and, even if not without sporadic repression and sporadic rebellions, were peacefully resolved. Allowing some political opposition was also eective to achieve political stability once workers and peasants were incorporated to the political system in the twentieth century. Opening the political game to the masses did not alter the eectiveness that the existence of political opposition had in the previous century. When at least a modicum level of opposition -a legislature and a chief executive elected in contested elections- is allowed, the probability of a coup detat or a constitutional change is lower. The probability of a high executive turnover is also less likely with political opposition than without it. Being allowed to play in the political game gives political players -new and old- incentives to continue participating and to have the expectation of a chance to be in power in future periods. In conclusion, the empirical evidence suggests that the recipe for political stability is allowing a modicum level of political opposition.

Appendix: Codebook

Note: All case-by-case coding decisions for the data collected by the author (see sources in each variable below), as well as coding decisions with the Mitchell (2003) data, are available upon request. Country Identiers

45 Country 66 67 68 69 70 53 154 54 71 55 57 59 61 62 63 73 74 76 77 Country Name Argentina Bolivia Brazil Chile Colombia Costa Rica Cuba Dominican Republic Ecuador El Salvador Guatemala Honduras Mexico Nicaragua Panama Paraguay Peru Uruguay Venezuela Entry Year 1816 1825 1822 1818 1819 1838 1902 1844 1830 1841 1840 1838 1821 1838 1903 1811 1821 1828 1829 Exit Year 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005 2005

COUNTRY: Country code from the ACLP dataset (Przeworski et al. 2000). COUNTRYN YEAR Independence TURMOIL: Length in years of the period between the year of nal independence (YEAR IND FINAL) and the completion of the second presidential term of at least four years. Country invariant. Source: own compilation. YEAR IND DECLARED: Year of declaration of independence from the colonial power. Country invariant. Source: own compilation. YEAR IND EFFECTIVE: Year of eective independence, when the last colonial soldier left the territory or the ghting for independence in the national territory ended. Country invariant. Source: own compilation. YEAR IND FINAL: Year of nal independence. It includes the latest date between the YEAR IND DECLARED and the YEAR IND OTHER. Country invariant. YEAR IND GRITO: Year of the grito de la independencia or rst uprisings against the colonial power. Country invariant. Source: own compilation. YEAR IND OTHER: Year of independence from associations with other states, a date that occurs after YEAR IND DECLARED. It is missing when the country did not form any associations with other states. Country invariant. Source: own compilation.

46 Leaders ACCHEADS: Accumulated number of consecutive years with one and the same person as chief executive. Made from HEADS. ACCNOTERM: Accumulated number of consecutive years with incomplete terms. Made from HEAD 1 TERM. ACCTERM: Accumulated number of consecutive years with completed terms. Made from HEAD 1 TERM. HEADCHANGES: Number of changes in the chief executive in a given year. Made from HEADS. HEADS: Number of distinct spells of chief executives during a given year. Spell refers to a continuous tenure of a person in an oce. If the same person holds oce more than once during a particular year, with someone else serving in between, each tenure of the same person is counted separately. Bearers of provisional titles are included as well as acting occupants of existing oces. Juntas and collective presidencies that govern collectively are counted as one, but rotating presidencies are counted separately. Counted as heads are: (1) prime ministers (includes chief ministers, presidents of government, and equivalent titles) (2) presidents who are either directly elected or are named by constitutions as chief executives, (3) monarchs or regents or governor generals or equivalent, (4) rst (general) secretaries of ruling single parties (only if they do not occupy other positions on this list), (5) collective bodies (juntas, triumvirates, collective presidencies), (6) persons bearing other titles, ranging from dictator to the administrator of the state of emergency, if none of the positions listed above exists or if none is occupied. Note: (1) If there are several governments in dierent regions and there is no national one, the whole period is coded as -1. (2) Juntas are counted as one head. (3) Acting heads are counted in cases of time overlap between the nominal head and the acting head; in such cases, the nominal head is not counted. Includes regents. (4) Governments in rebellion are not counted. (5) In some cases an oce is not occupied during some part of a year. Those vacancies are not counted. (5) Periods of foreign occupation are recorded as -1. Source: own compilation. HEAD 1 TERM: This variable indicates whether the chief executive in question completed a previously specied term, if there is one. Code as missing if a particular oce does not exist and code as 0 if the oce does exist but is not occupied during an entire year. This variable is treated dierently for dierent types of heads and it is not coded for secretaries of single ruling parties. Right-hand censored spells are coded as -1. For presidents or monarchs: 1 For each year of a term during which:

47 (i) the chief executive is elected or appointed or inherits a previously specied term AND completes the term OR (ii) the chief executive who was elected or appointed to a previously specied term dies of natural causes or resigns for reasons of health AND is succeeded by the constitutional successor who then completes the term OR (iii) the chief executive calls for an early election or legally reduces the term in oce and completes it. 0 For each year of a term during which (i) the incumbent extends her current term in oce (changes in term limits do not count as extensions of the current term) (ii) the incumbent is a provisional, acting, or interim chief executive (except for regents under monarchy) (iii) the incumbent serves a term that had not been previously specied (iv) the chief executive is deposed or impeached or assassinated (v) the incumbent resigns for reasons other than poor health (vi) the incumbent resigns or is briey ousted and then comes back to nish the term (vii) the incumbent is elected or appointed for a new term before nishing the current term and the new term overlaps with the current one (viii) the term is specically designed for a particular person. With regard to prime ministers the relevant term is the LEGTERM: 1 For each year of a spell that is not interrupted by a duly chosen prime minister being overthrown by force or the parliament not being dissolved unconstitutionally. If a parliament is dissolved constitutionally before the LEGTERM expires and new elections are help promptly, code as 1. 0 Otherwise. This includes cases in which the prime minister was imposed by force as well as rst ministers who serve a term ad hominem designed for them. It also includes cases in which the current legislature extends its term. Example: In country X, the chief executives term is 2 years. Chief executive A completes his term, which runs from September 1880 to September 1882 (2 year term). Thus, TERM=1 for 1880 and 1881. The value of term for 1882 and 1883 depends on the status of the term that runs from September 1881 to September 1884. Chief executive B enters oce in September 1882 and resigns later that year after generalized riots. The legal successor takes over and nishes the term in September 1884. TERM equals zero for 1882 and 1883. Source: own compilation. HEAD 1 FIRST TERM: Year in which the chief executive completed a term for the rst time since nal independence. Made from HEAD 1 TERM. HEAD 1 SECOND TERM: Year in which the second consecutive term of at least four years was completed. Calculated from HEAD 1 TERM. The second term referenced was counted from the year of nal independence (YEAR IND FINAL). NMIL: Dummy equal to one whenever the nominal chief executive is or ever was a member of the military by profession, 0 if civilian. Retired members of the military as NMIL=1 since the shedding of a uniform is not necessarily enough to indicate the civilian character of a leader. Rulers who come to power as head of guerilla movements are not considered as military. Sources: Gandhi (2008) for the period 1946-2002 and own compilation for earlier dates and for the years 2004 to

48 2005. PRESIDENT: Dummy equal to one whenever the system is presidential. Made from TITLE CE. TITLE CE: Title of the nominal chief executive according to the constitution. Nominating and dismissing powers are not executive powers. Presiding over meetings of the council of ministers or promulgating laws are also not considered executive powers. These rules apply even if these powers are specied in a section on executive powers. -1 If the constitution is not in force or its status is ambiguous 0 If monarch or governor general (if the constitution says that the king is the chief executive, regardless of the status quo) 1 If prime minister (president of the government or equivalent, as long as the president can be removed by the legislature), 2 If president who cannot be removed in good behavior during a xed term, whether directly or indirectly elected, 3 If the constitution gives executive power both to president, who cannot be removed during a xed term, and prime minister (this includes separate executive powers enumerated for the president, with the remaining reserved for the government) 4 If collective. Source: own compilation. Political Events ACCNOCOUPS: Accumulated number of consecutive years with no coups. COUPS: Number of coups detat during a given year. Note: Coups are hard to dene, although one normally knows when one occurred. The standard denition irregular transfer of power does not work when there are no rules. Coded as coups are: (1) Any change of government, constitutional or not, that entails the use or the threat of force by a militarily organized group. Resignations under the pressure of popular mobilization are not coups. Neither are irregular transfers of power among civilians that do not entail force (Hence, changes of communist leaders are not coups). (2) Any replacement of a military chief executive by another military, unless the succession occurred according to pre-established rules (e.g. Figueiredo replacing Geisel in Brazil). Only successful coups are counted. Coups from above are coded separately. Any military intervention of foreign power(s) resulting in overthrowing or restoring a leader (even elected ones) is coded as a coup. Sources: own compilation drawing from Banks (2002), Nohlen (2005), Lambert (1968), Smith (2005), Dean (1970), and political histories of several countries. COUPS BANKS: Number of Coups dEtat. The number of extraconstitutional or forced changes in the top government elite and/or its eective control of the nations power structure in a given year. The term coup includes, but is not exhausted by, the term successful revolution. Unsuccessful coups are not counted. Source: Banks (2002), variable labeled S21F1. COUPS MILITARY: Dummy equal to one if a military coup took place that year. Coverage 1990-2005. There is no data on Cuba. Source: Smith (2005). OPPOSED: Dummy equal to one if the chief executive is elected with some opposition, that is, there was at least one vote for other candidates. Whenever there is political consensus about a

49 candidate and thus the vote is unanimous, OPPOSED equals 1. If the incumbent is not elected, opposed equals zero. In cases of legal succession, the successor is coded the same as the incumbent that he is replacing. Source: own compilation. OPPOSITION: This variable is intended to indicate whether the institutional system allows at least some political pluralism. 1 If as of December 31 there is a legislature that was at least in part elected by voters facing more than one choice. Note: Even if the legislature consisted of multiple parties but in the election all the candidates were presented to voters on the same, single list, we consider that voters had no choice. If there was only one party while some candidates ran as independents, we also regard it as a situation in which there was no partisan pluralism. If, however, everyone ran without party labels, then we regard this situation as pluralistic. The idea is that situations in which elections are non- partisan, frequent during the early period, and situations in which more than one party competes in elections represent political pluralism, while monopoly of one party is a political monopoly. 0 If the above is not true OR, only in presidential systems, if it is true BUT either the chief executive is not elected OR was elected unopposed (unless there was consensus among all parties; see Chile 1891). -1 If the legislature was closed at some point during the year and the state as of December 31 is unclear. Source: own compilation. PAST COUPS: Accumulated number of coups detat at t-1. Made from COUPS. PAST MILITARYCOUPS: Accumulated number of military coups detat at t-1. Since the data on military coups begins in 1900, the value of PAST MILITARYCOUPS was equal to one if there was at least one military coup before 1900. Made from COUPS MILITARY. SALTEREL: Dummy for strong alternation resulting from an election: 1 If the party (and the person) of the nominal chief executive changes as a result on an election or, where the CE is a monarch, the eective head of government changes along partisan lines as a result on an election or if the entire cabinet changes when there is no prime minister, 0 If both the old and the new chief executive head elected governments but the party of the chief executive remains the same or the person remains the in oce having changed parties or, where there is no prime minister, the partisan composition of the government remains the same, -1 If there are elections but no discernible parties or two successive independents or the change is between a non-partisan (caretaker, technical) government and a partisan one, unless the only role of a non-partisan government was to ensure impartial administration of an election, -2 If the current chief executive assumes power in a non-constitutional way, but always following an election. Note: This code applies only to the year of election or the year of inauguration if not the same. Left as missing otherwise. It is not always possible to tell if the alternation followed or preceded the election when both occurred during the same year. In case of doubt, alternations that occur during years of elections are coded as having resulted from the election. Alternation can occur only if there was more than one party in the previous election and if the current election did not follow a coup. Partial elections are counted as elections. Alternations that result from a previous agreement between parties count as alternations.

50 Source: own compilation. STRIKES: Number of general strikes, including any strike of 1,000 or more industrial or service workers that involves more than one employer and that is aimed at national government policies or authority. Source: Banks (2002), variable labeled S17F2. STRIKES DUMMY: Dummy variable equal to one if there was any strike during the year. Made from STRIKES. Elections and Surage COMPULSORY VOTE: Refers to any national level elections. 0 Optional 1 Compulsory 2 Compulsory for some, facultative for others. ELECPARTY PRES: Number of candidates presented to voters in presidential elections. 0 If one 1 If more Note: Applies whether the president is directly or indirectly elected. If there is more than one election during a year, the last one is coded. ELIGIBLE: Registered or eligible voters in national elections. Sources: Banks (2002) -who does not distinguish between registered and eligible-, Payne et al. (2002) and additions from dierent sources. ELIGIBLE PR: Ratio of the number of people eligible to vote to the total population. Coded only for years of legislative or presidential elections. Note: In some cases, registration is automatic, so that the number of registered voters is equal to the number of eligible voters. Whenever they dier and a series for registered is available, this is the series used. If it is not available, numbers of eligible voters are used. Finally, in several instances we do not know whether the numbers refer to registered or eligible voters. EXSELEC: Dummy for the chief executive having been elected (whether directly or indirectly; indirectly means elected by people who have been elected; if indirectly elected, election by a partly elected body counts as being elected). Constitutional successors of elected chief executives are coded as elected. Code each year of the current chief executive. Prime ministers are always coded as elected as long as the legislature is open (OPENL=1). If there is a president and a prime minister, EXSELEC=1 only if both have been elected. If a chief executive is elected, overthrown by force, and then reinstated, code as of the year of reinstatement. Source: own compilation. FRANCHISE: Qualications for the right to vote in national elections. Preliminary notes: (i) to account for exclusions not captured under FRANCHISE, the variable OTH EXCLUSIONS is used; (ii) if the qualications are set at a sub-national level, we code the variable as missing; (iii) regulations concerning immigrants and foreign residents are not considered. Coding of FRANCHISE: The codes 0 to 7 are for males only: 0 No legal provisions for surage. This code is applied:

51 (a) before rst elected representative institutions were established, (b) whenever constitutions or other legal acts providing for surage were either abrogated without being replaced or formally suspended, (c) if (b) is unclear, whenever the legislature was closed and neither legislative nor presidential elections were held. 2 Property only 3 (Property OR income OR taxes OR exercise of profession OR educational titles) AND literacy. 4 Property OR income OR taxes OR exercise of profession OR educational titles. This code applies whenever the law species a minimum threshold, even if this threshold is very low (if this requirement is not observed, as in post-1917 Mexico, the code is 6.) 5 Literacy only OR (Literacy OR Property OR income OR taxes OR exercise of profession OR educational titles). 6 All the economically independent (not personal servants, not debtors, only residence requirement). Includes: (a) phrases that explicitly refer to economic independence. Example: Tener una propiedad, o ejercer cualquiera profesion, o arte con titulo publico, u ocuparse en alguna industria util, sin sujecion a otro en clase de sirviente o jornalero (Peru in 1823). (b) phrases such as known means of subsistence,honest livelihood Example: Son ciudadanos todos los habitantes de la Republica naturales de pais o naturalizados en el que fueren casados, o mayores de diez y ocho anos, siempre que exerzan alguna profesion util o tengan medios conocidos de subsistencia (Costa Rica 1824) (c) phrases that suggest clearly broad qualications. Example: Son ciudadanos todos los salvadoreos mayores de veintin aos que sean padres de familia, o cabezas de casa, o que sepan leer y escribir, o que tengan la propiedad que designa la ley (El Salvador 1841) (d) cases where surage is universal but only for free men (i.e., not slaves) (e) cases that fall under 4 without explicit thresholds. Example: The 1860 Peruvian constitution qualied anyone who could read and write OR paid any taxes OR owned a workshop OR owned some land. The last provision was intended to allow the indigenous population to vote. (f) cases in which franchise is declared to be universal but excludes those who have failed to pay taxes or those permanently receiving public assistance or those under legal bankruptcy. 7 All. The only exclusions admitted in this category are: (1) having been convicted of a crime (2) being legally incompetent (3) short, less than two-year local residence requirement. There are some cases, however, where the law contains restrictive phrases, such as the requirement of appearing on the tax rolls or having fullled military obligation, these restrictions were not enforced. In such cases franchise was coded as 7. For females, I use the second digit, distinguishing only situations in which: 0 no women can vote 1 women are qualied on narrower basis than males (for example, at older age, in local elections only, etc.) Auxiliary code: -1 More than one constitution, any other kind of political chaos that makes the electoral rules inoperative. For years of foreign occupation we either set as missing or extend the restriction in eect before the occupation. Source: own compilation.

52 FRANCHISE AGE: Age at which those otherwise qualied to vote can exercise this right. Whenever age thresholds are dierent for dierent groups, lowest age is coded. Source: own compilation. FRANCHISE CONTR: Dummy equal to one if the right to vote was contracted and equal to zero if it was not contracted during a particular year. Constructed from FRANCHISE. FRANCHISE EXT: Dummy equal to one if the right to vote was extended and equal to zero if it was not extended during a particular year. Constructed from FRANCHISE. FRANCHISE BROAD: Dummy equal to one if the male franchise is broad. Constructed from FRANCHISE. FRANCHISE NARROW: Dummy equal to one if the male franchise qualications include economic and/or literacy restrictions. Constructed from FRANCHISE. LEGDIRECT: Mode of election of the legislature. Refers only to the lower house. 0 Indirect 1 Direct Source: own compilation. LEGELEC: Number of legislative elections. It includes general elections, elections in which only a part of the legislature is renewed, and annulled elections, but not by-elections to replace representatives who for some random reason did not continue in oce. Elections to the upper house only and elections to the constitutional assembly are excluded. Second rounds are not counted. LEGPART PR: Ratio of voters in legislative elections to the total population. Coded only for years of legislative elections. If elections are indirect, only the number of primary voters is used. Source: own compilation. LEGPARTY: Number of parties in the legislature. By party I take here any recognizable, named grouping, regardless whether and how they appeared in elections. This variable is coded regardless whether the legislature was elected or appointed. -1 If no legislature 0 If none (including cases where parties are ocially banned) 1 If one (including cases where other parties are ocially banned 2 If more Note: During the early period candidates typically competed in elections on an individual basis, without any kind of centralized party organizations or shared programs. Within legislatures, however, they often coalesced into groups, currents, factions, etc. Hence, subjective judgments are inevitable in coding this variable. Our rule of thumb was to code as parties any kind of groups that bore a label that survived over some period of time. This would be true in several countries of Liberals and Conservatives. In other countries, however, the divisive issues, the parliamentary groups, and their labels changed frequently, and such cases are coded as none. In turn, a legislature consisting of one party and independents is coded as one party. LEGSELEC: Dummy for the lower house of the legislature having been at least partly elected. Source: own compilation. LEGVOTNR: Number of voters in legislative elections. Sources: Banks (2002), Nohlen (2005),

53 Payne et al (2002) and additions from dierent sources. OPENL: A dummy variable for the existence of a legislature, by December 31. A legislature is a body that satises two conditions: (1) It issues at least some laws and (2) It does not perform executive functions. Legislative Councils which advise the ruler but have no authority to proclaim laws are not considered legislatures. Military Juntas that issue legal acts but also perform executive functions are disqualied by the second criterion. PART: Number of voters in presidential and legislative national elections. Made from LEGVOTNR and PRVOTNR. PART PR: Proportion of the total population voting in national elections. Constructed from PRESPART PR, LEGPART PR, and additions by country from miscellaneous sources. PART PR CHANGE: Change in the proportion of the total population voting in national elections. The pair of elections considered were not separated by more than ve years. PRESDIRECT: Mode of election of the president. 0 Indirect 1 Direct Source: own compilation. PRESELEC: Number of presidential elections that took place during the year. Includes any kind of elections, direct and indirect, in the latter case regardless of the constitutional status of the electing assembly. Second rounds are not counted, while annulled elections are counted. Presidential plebiscites are included. Source: own compilation. PRESPART PR: Ratio of voters in presidential elections to the total population. Coded only for years of presidential elections. If elections are indirect, only the number of primary voters is used. PRESTERM: Constitutionally prescribed duration of the chief executives tenure (in years). -1 If no term is specied 0 If life term Note: If the term of the president is extended while an incumbent is in oce, it changes in the year it was extended. PRESTERM LIMIT: Term limits for presidents. -1 If no term is specied 0 Unlimited number of consecutive terms 1 Out after 1 term, can never come back 2 Out after 1 term, can come back 3 Out after more than one term, can come back 4 Out after more than one term, can never come back Source: own compilation. PRVOTNR: Number of voters in presidential elections. Sources: Banks (2002), Nohlen (2005), Payne et al (2002) Vanhanens Polyarchy Dataset, and additions from dierent sources.

54 REGNR: Registered or eligible voters in national elections. Sources: Banks (2002) -who does not distinguish between registered and eligible-, Nohlen (2005) -registered only-, Payne et al. (2002) and additions from dierent sources. SECRET BALLOT: Refers only to legislative elections. 0 Open, public 1 Secret Source: own compilation. TURNOUT:Turnout in national elections. Equal to PART/ELIGIBLE. Note: This seems to be a more consistent way of getting turnout that the numbers oered in the sources. WOMEN VOTE: Dummy equal to one when women had the right to vote on an equal or a narrower basis as men. Made from FRANCHISE. Population FPOP20: Female population over 20 years old, in thousands. Calculated by adding up the population of 20 and over. In some cases, the break down did not match our needs, in which case the maximum proximity was privileged. See Mitchells notes for a detail of those cases. In the few cases where only one or both brackets [75-79] and [80 and over] were missing, the observation was recorded anyways. If other brackets were missing, the observation was recorded as missing. The unknown were not included in the sum. Source: Mitchell (2003). MPOP20: Male population over 20 years old, in thousands. Calculated by adding up the population of 20 and over. In some cases, the break down did not match our needs, in which case the maximum proximity was privileged. See Mitchells notes for a detail of those cases. In the few cases where only one or both brackets [75-79] and [80 and over] were missing, the observation was recorded anyways. If other brackets were missing, the observation was recorded as missing. The unknown were not included in the sum. Source: Mitchell (2003). POP C: Population, in thousands. These data incorporate variations in population due to changes in geographical boundaries. Constructed by taking the source that contained the larger amount of non-missing data from either Maddison 2003 (present borders as of 1990) or Global Financial Data by country. For countries with no signicant border changes, the possibility of using the variable tpop in the COW2 dataset whenever it had more data was assessed on a case by cases basis and implemented if appropriate. Sources: Correlates of War 2, Maddison (2003), and Global Financial Data. POP C FILL: Population, in thousands. It was constructed by linearly interpolating POP C on YEAR for each country. POP20: Population over 20 years old, in thousands. Calculated from MPOP20, FPOP20, and POP C. POP20 PR: Proportion of the total population that is over 20 years old. Calculated from POP20 and POP C.

55 Other Institutional Features ACCCONST: Constitutional endurance. This variable counts the number of years that a constitution was in eect. Source: own compilation. AGRARIAN REFORM: Dummy equal to 1 during the year in which a land reform was adopted. Source: Anjali Thomas. ALL LABOR: Dummy equal to one whenever FIRST LABOR LAW, FIRST LABOR CODE, ILO C29, ILO C87, or ILO C98 are equal to one. It is zero otherwise. This variable captures labor legislation. CONST CHANGE: Dummy equal to one if the constitution in force was changed or amended during a particular year. It was made from CONST INFORCE and CONST REFORM. CONST INFORCE: Year of adoption of the constitution that was nominally in force as of the end of each year. Constitutional provisions are coded even if they have fallen into disuse or were never used. Constitutions are dened nominally, that is, a document is a constitution if it is called that or if there is any other document that establishes how the government will be organized. Coded as 0 if the constitution was formally suspended without being replaced, there is no constitution, or there are multiple constitutions. Source: own compilation. Source: own compilation. CONST REFORM: Dummy equal to 1 if a constitutional reform is carried out during the year and 0 if no reform took place. Source: own compilation. FIRST LABOR CODE: Year of adoption of the rst labor code. Source: own compilation. FIRST LABOR LAW: Year of adoption of the rst law regulating labor. Source: Bronstein (1998) and multiple historical accounts. ILO C29: Year in which the country ratied the International Labor Organization Convention C29, on Forced Labor Convention (1930). Source: International Labor Organizations web site. ILO C87: Year in which the country ratied the International Labor Organization Convention C87, on the Freedom of Association and Protection of the Right to Organize Convention (1948). Source: International Labor Organizations web site. ILO C98: Year in which the country ratied the International Labor Organization Convention C98 on the Right to Organize and Collective Bargaining Convention (1949). Source: International Labor Organizations web site. SLAVERY: Year of abolition of Negro slavery. Country invariant. Source: own compilation. Economics and Development FAMILY FARMS: Area of family farms as a percentage of the total area of holdings. Source: Vanhanen (1979, 1997).

56 GDP: Gross Domestic Product in 1990 Geary-Khamis dollars, in millions. Source: Maddison (2003). GDPCAP: Income per capita in 1990 Geary-Khamis dollars. Source: Maddison (2003).

References

Acemoglu, Daron and James Robinson. 2006. Economic Origins of Dictatorship and Democracy. Cambridge, MA: Cambridge University Press. Acemoglu, Daron and James Robinson. 2000. Why Did the West Extend the Franchise? Democracy, Inequality, and Growth in Historical Perspective. Quarterly Journal of Economics, CXV: 1167-1199. Ake, Claude. 1974. Modernization and Political Instability: A Theoretical Exploration. World Politics, 26 (4), pp. 576-591. Alesina, Alberto and Roberto Perotti. 1997. The Welfare State and Competitiveness, American Economic Review, American Economic Association, vol. 87(5), pages 921-39, December. Alonso, Paula. 1996. Voting in Buenos Aires (Argentina) before 1912. In Eduardo Posada Carb o (ed.). America. New York: St. Martins Press. Alonso, Paula. 1993. Politics and Elections in Buenos Aires. 1890-1898: The Performance of the Radical Party, Journal of Latin American Studies, vol. 25, no. 3. Annino, Antonio. 1996. The ballot, land and sovereignty: Cadiz and the origins of Mexican local government, 1812-1820, in Eduardo Posada Carb (ed.), Elections before Democracy: The History o of Elections in Europe and Latin America, London: Macmillan Press Ltd. Asadurian, Tamar, Carolina Curvale, Sunny Kaniyathu, and Anjali Thomas. The Elusive Ballot Box: Democratic Filters, 1776-2004, Paper presented at the 64th Midwest Political Science Association, Chicago, April 20-23, 2006. Banks, Arthur. 2002. Cross-National Time Series Data Archive. Bejarano, Jesus Antonio. 1985. Campesinado, luchas agrarias e historia social en Colombia: notas para un balance historiogrco. En Pablo Gonzlez Casanova (coordinador), Historia pol a a tica de los campesinos latinoamericanos, volumen 3: Colombia, Venezuela, Ecuador, Per, Bolivia, u Paraguay, Instituto de Investigaciones Sociales, Universidad Nacional Autnoma de Mexico, Mexo ico: Siglo Veintiuno Editores. Bendix, Reinhard, and Stein Rokkan. 1962. The extension of national citizenship to the lower classes: a comparative perspective. Paper presented at the Fifth World Congress of Sociology, Washington, D.C.

57 Blanco Vado, Mario A. 1996. Las relaciones laborales en el marco de la globalizacin econmica: o o los principios del derecho laboral, su regulacin y sus tendencias. Primera edicin, San Jos: Corte o o e Suprema de Justicia, Escuela Judicial de Costa Rica. Bronstein, Arturo S. 1998. Pasado y presente de la legislacin laboral en Amrica Latina, Costa o e Rica, San Jos, Junio de 1998, manuscrito, Organizacin Internacional del Trabajo. e o Bushnell, David and Neill Macaulay. 1994. The Emergence of Latin America in the Nineteenth Century. Second Edition. New York: Oxford University Press. Bushnell, David. 1972. La evolucin del derecho del sufragio en Venezuela, Bolet Histrico, o n o Vol 29, Mayo, Caracas, Venezuela. Coatsworth, John H. 2005. Structures, Endowments, and Institutions in the Economic History of Latin America. Latin American Research Review 40: 33, pp. 126-144. Coatsworth, John H., 1998. Economic and Institutional Trajectories in Nineteenth-Century Latin America,and Alan M. Taylor. In John Coatsworth and Alan Taylor (eds.), Latin America and the World Economy Since 1800. Cambridge, MA: Harvard University Press, pp. 2354. Collier, Ruth Berins, and David Collier. 2002. Shaping the Political Arena. Princeton: Princeton University Press. Colomer, Josep M. 2004. Taming the Tiger: Voting Rights and Political Instability in Latin America, Latin American Politics and Society, 46, 2: 29-58. Conley, John and Akram Temimi. 2001. Endogenous Enfranchisement when Groups Preferences Conict, Journal of Political Economy, V. 109, pp. 79-102. Correlates of War 2, National Material Capabilities Dataset, http://cow2.la.psu.edu/ [Accessed in April 2006]. Data rst presented in: Singer, J. David, Stuart Bremer, and John Stuckey. 1972. Capability Distribution, Uncertainty, and Major Power War, 1820-1965, in Bruce Russett (editor) Peace, War, and Numbers, Beverly Hills: Sage, 19-48. Updated and expanded in: Singer, J. David. (1987). Reconstructing the Correlates of War Dataset on Material Capabilities of States, 18161985 International Interactions, 14: 115-32. Cukierman, Alex, Sebastian Edwards and Guido Tabellini. 1992. Seigniorage and Political Instability, American Economic Review, vol. 82(3), pages 537-55, June. Dahl, Robert A. 1971. Polyarchy: Participation and Opposition. New Haven: Yale University Press Dealy, Glen. 1974. The Tradition of Monistic Democracy in Latin America. Journal of the History of Ideas, Vol. 35, No. 4 (Oct. - Dec.), pp. 625-646. Dean, Warren. 1970. Latin American golpes and Economic Fluctuations, 1823-1966. Social Science Quarterly, Vol 51, No. 1, June.

58 De Buen Lozano, Nstor. 2006. El derecho a huelga, en Kurczyn Villalobos, Patricia (coord.). e Evolucin y tendencias recientes del derecho del trabajo y de la seguridad social en Amrica. Mexico: o e Instituto de Investigaciones Jur dicas, Universidad Autnoma de Mexico. o Demlas-Bohy, M. D. and F. X. Guerra. 1996. The Hispanic Revolutions: the adoption of modern e forms of representation in Spain and America (1808-1810), in Eduardo Posada Carb (ed.), Elections o before democracy: the history of elections in Europe and Latin America, London: Institute of Latin American Studies. Finer, S.E. 1962. The Man on Horseback. New York: Frederick A. Praeger, Inc. Fitch, J. Samuel. 1998. The Armed Forces and Democracy in Latin America. Baltimore: The Johns Hopkins University Press. Freeman, John R., and Duncan Snidal. 1982. Diusion, Development, and Democratization: Enfranchisement in Western Europe. Canadian Journal of Political Science 55: 299-329. Gandhi, Jennifer. 2008. Political Institutions under Dictatorship. Cambridge: Cambridge University Press. Gargarella, Roberto. 2005. Los fundamentos legales de la desigualidad: El constitucionalismo en Amrica (1776-1860). Madrid: Siglo XXI. e Gargarella, Roberto. 2004. Towards a Typology of Latin American Constitutionalism, 1810-60, Latin American Research Review Vol. 39, No. 2, pp. 141-153. Global Financial Data, http://www.globalnancialdata.com/index.php3. Halper Donghi, Tulio. 2002. n Durham: Duke University Press. Halper Donghi, Tulio. n Harper&Row. 1973. The Contemporary History of Latin America. Sixth edition. The Aftermath of Revolution in Latin America. New York:

Huntington, Samuel P. 1968. Political Order in Changing Societies. New Haven: Yale University Press. Huntington, Samuel P. 1965. Political Development and Political Decay. World Politics, 27: 386-430. Irurozqui, Marta y V ctor Peralta. 2003. Elites y sociedad en la Amrica andina: de la repblica e u de ciudadanos a la repblica de gente decente 1825-1880, en Juan Maiguashca (editor), Historia u de Amrica Andina, Quito: Universidad Simn Bolivar. e o Irurozqui, Marta. 1999. La ciudadan clandestina - Democracia y educacin ind a o gena en Bolivia, 1826 - 1952, Estudios Interdisciplinarios de Amrica Latina y el Caribe, vol 10 (1), enero - julio. e Irurozqui, Marta. 1996. Ebrios, vagos y analfabetos. El sufragio restringido en Bolivia, 18261952, Revista de Indias, vol. 56, no. 208, pp. 697-742.

59 Jack, William, and Roger Laguno. 2003. Dynamic Enfranchisement. Department of Economics, Georgetown University. Kennedy, Peter. 2003. A Guide to Econometrics. The MIT Press; 5th edition. Lambert, J. 1968. Amerique Latine, Presses Universitaires de France. Liao, Tim Futing. 1994. Interpreting Probability Models: Logit, Probit, and Other Generalized Linear Models. Sage Publications. Linz, J. J. and Arturo Valenzuela (eds.). 1984. The Failure of Presidential Democracy. Baltimore: Johns Hopkins University Press. Little, Walter and Eduardo Posada Carb (eds.). 1996. Political Corruption in Europe and Latin o America. New York: St. Martins Press. Lizzeri, Alessandro y Nicola Persico. 2004. Why Did the Elites Extend the Surage? Democracy and the Scope of Government, With an Application to Britains Age of Reform, Quarterly Journal of Economics 119, 707 - 765. Llavador, Humberto, and Robert Oxoby. 2005. Partisan Competition, Growth and the Franchise. Quarterly Journal of Economics 119: 1155-1189. Londregan, John, and Keith Poole. 1990. Poverty, the Coup Trap, and the Seizure of Executive Power. World Politics, 42 (2), p. 151-183. Maddison, A. 2003. The World Economy: Historical Statistics. Paris: OECD. Malamud, Carlos. 2000a. Introduccin, en Carlos Malamud (coordinador), Legitimidad, repreo sentacin y alternancia en Espaa y Amrica Latina: las reformas electorales (1880-1930), Mxico: o n e e Fondo de Cultura Econmica. o Mej Sonia Alda. 2000. La consolidacin de la republica restrictiva ante las demas de as, o as la representacin popular en la Guatemala del siglo XIX, en Carlos Malamud (coordinador), o Legitimidad, representacin y alternancia en Espaa y Amrica Latina: las reformas electorales o n e (1880-1930), Mxico: Fondo de Cultura Econmica. e o Milbrath, Lester W. 1965. Political Participation. Chicago: Rand Mc Nally. Mirow, Matthew Campbell. 2004. Latin American Law: A History of Private Law and Institutions in Spanish America. Austin: University of Texas Press. Mitchell, B. R. 2003. International Historical Statistics: The Americas, 1750-2000. Fifth edition. London: Palgrave MacMillan. Morse, Richard. 1954. Toward a Theory of Spanish Government. Journal of History of Ideas, 15: 71-93.

60 Mucke, Ulrich. 2001. Elections and Political Participation in Nineteenth-Century Peru: The 1871-72 Presidential Campaign, Journal of Latin American Studies 33, 311-346. Murilo de Carvalho, Jose. 1995. Desenvolvimiento de la ciudadan en Brasil. Mexico: Fondo de a Cultura Econmica. o Needler, Martin. 1968. Political development in Latin America; instability, violence, and evolutionary change. New York, Random House. Nohlen, Dieter (editor). 2005. Elections in the Americas. A data handbook Volume I: North America, Central America, and the Caribbean and Volume II: South America. New York: Oxford University Press. Organization of American States (OAS). 1999. Obsevacin Electoral en Guatemala 1999. Obsero vaciones Electorales, Serie Amricas, No. 25 Department for Democratic and Political Aairs. e Payne, J. Mark, Daniel Zovatto G., Fernando Carrillo Florez and Andrez Allamand Zavala. 2002. Democracies in Development. Politics and Reform in Latin America. Published by the InterAmerican Development Bank and International Institute for Democracy and Electoral Assistance. Distributed by The Johns Hopkins University Press, Washington DC. Peloso, Vincent y Tenenbaum, Barbara (eds.). 1996. Liberals, Politics, and Power: State Formation in Nineteenth-Century Latin America. Athens: University of Georgia Press. Petras, James and Veltmeyer, Henry. 2002. The peasantry and the state in Latin America: A troubled past, an uncertain future, Journal of Peasant Studies, 29:3, 41 - 82. Pico, Rafael. 1964. Problems of Land Tenure Reform in Latin America. Journal of InterAmerican Studies, Vol. 6, No. 2 (Apr., 1964), pp. 143-156. Pike, Frederick B. 1982. Spanish America 1900-1970: Tradition and Social Innovation. In Howard Wiarda (ed.), Politics and Social Change in Latin America: The Distinct Tradition, Second Edition, Amherst: University of Massachusetts Press. Posada Carb, Eduardo. 2000. Electoral Juggling: A Comparative History of the Corruption of o Surage in Latin America, 1830-1930. Journal of Latin American Studies, Vol. 32, No. 3 (Oct.), pp. 611-644. Prados de la Escosura, Leandro. 2003. Assessing the Economic Eects of Latin American Independence. Working Paper 03-12. Madrid: Universidad Carlos III. Przeworski, Adam and Carolina Curvale. 2007. Instituciones polticas y desarrollo econmico en las Amricas: el largo plazo. En Visiones sobre el Desarrollo en Amrica Latina, CIDOB/CEPAL. Przeworski, Adam. 2007. Conquered or Granted? A History of Surage Extensions, manuscript, New York University. Przeworski, Adam, Michael Alvarez, Jos Antonio Cheibub and Fernando Limongi. 2000. ACLP e Dataset in Democracy and Development: Political Institutions and Material Well-Being in the

61 World, 1950-1990. Cambridge: Cambridge University Press. Przeworski, Adam and Fernando Corts. 1971. Sistemas partidistas, movilizacin electoral y la e o estabilidad de sociedades capitalistas. Revista Latinoamericana de Ciencia Pol tica, II (2). Rouquie, Alain, 1994. The military in Latin American Politics since 1930, The Cambridge History of Latin America, vol. VI: 1930 to the Present, Leslie Bethell (ed.), Cambridge, Cambridge University Press, pp. 233-306. Royo Aspa, Antoni. La Reforma Agraria en Costa Rica (1962-2002): balance de las intervenciones estatales en el cantn de Osa, Revista de Historia, Julio de 2003. o Sempat Assadourian, Carlos, Am lcar Challu and John H. Coatsworth. Agriculture and Land Tenure. The Colonial Era and the Short Nineteenth Century. Eds. V ctor Bulmer-Thomas, John H. Coatsworth and Roberto Corts Conde. Cambridge University Press, 2006. Cambridge Histories e Online. Cambridge University Press. Skidmore, Thomas and Peter H. Smith. 2001. Modern Latin America. New York: Oxford University Press. Skidmore, Thomas. 1979. Workers and Soldiers: Urban Labor Movements and Elite Responses in Twentieth Century Latin America. In Virginia Bernhard (ed.)Elite, Masses, and Modernization in Latin America, 1850-1930. Austin: University of Texas. Valenzuela, J. Samuel. 1998. La ley electoral de 1890 y la democratizacion del regimen politico chileno. Estudios Pblicos, 71 (invierno). u Valenzuela, J. Samuel. 1996. Building aspects of democracy before democracy. Electoral practices in nineteenth century Chile, Working Paper #223, Kellogg Institute, University of Notre Dame. Valenzuela, J. Samuel. 1995. The Origins and Transformation of the Chilean Party System. Working Paper #215. The Helen Kellogg Institute for International Studies, University of Notre Dame. Vanhanen, Tatu. The Polyarchy Dataset, Norwegian University of Science and Technology. http://www.prio.no/page/Project detail//9244/42472.html [Accessed in April 2006]. Vanhanen, Tatu. 1997. Prospects of Democracy. A study of 172 countries. London: Routledge. Vanhanen, Tatu. 1979. Power and the Means of Power. A Study of 119 Asian, European, American, and African States, 1850-1975. Ann Harbor: University Microlms International. Vines, Kenneth N. and Henry Robert Glick. 1967. The Impact of Universal Surage: A Comparison of Popular and Property Voting. American Political Science Review, 61 (4), pp. 1078-1087. Wiarda, Howard (ed.). 1982. Introduction and Conclusion. Politics and Social Change in Latin America: The Distinct Tradition, Second Edition, Amherst: University of Massachusetts Press.

62 Williamson, Jerey G. 2004. Real Wages, Inequality and Globalization in Latin America before 1940. Revista de Historia Economica 17: 101-142. Yashar, Deborah J. 1997. Demanding Democracy. Reform and Reaction in Costa Rica and Guatemala 1870s 1950s. CA: Stanford University Press.