COMPARATIVE TURNOUT LEVELS IN LATIN AMERICA SINCE Paper Presented at the 21st World Congress of Political Science, Santiago, Chile, July 2009

Transcription

1 COMPARATIVE TURNOUT LEVELS IN LATIN AMERICA SINCE 1990 Paper Presented at the 21st World Congress of Political Science, Santiago, Chile, July 2009 MAY 2009 DRAFT Alan Siaroff Department of Political Science The University of Lethbridge 4401 University Drive Lethbridge, Alberta, Canada T1K 3M4 INTRODUCTION The level of voter turnout is a, if not the, central measure of political participation. Within Latin America there are sharp variations in election turnout. Yet with the exception of Fornos et al. (2004) the cross-national variations in the region have not been analysed. In this paper I shall look at turnout in both parliamentary and presidential elections in eighteen Latin American countries since Choosing that year gives us a two-decade period of study. It also comes after almost all countries in the region had (re)turned to democracy (Mexico being the main exception), so this issue of founding elections with their higher turnout can be excluded. More generally, I shall assess whether the factors that shape turnout variations in the industrial democracies hold for Latin America. The analysis is a cross-sectional time series analysis; that is, the dependent variables are not countries but specific elections in individual countries. This vastly increases the sample, to some 90 parliamentary elections and 74 presidential ones. That said, some independent variables are country-specific. Turnout is relevant to a polity if for no other reason than the legitimacy of the political system in general. Where citizens cannot be bothered, or inspired, to play an active political role, then there is little positive endorsement of the political 1

2 order and the specific policy outputs of the government of the day. Consequently, victorious political parties and presidential candidates cannot rightfully claim the same mandate in a situation of low turnout as in one of high turnout. Of course, the desirability of high turnout assumes that this turnout involving support for competing democratic parties. However, as for example Weimar Germany showed, high turnout can go hand in hand with strong support for extremist parties. Fortunately, though, such parties are rarely strong in Latin America today. THE RANGE OF VOTER TURNOUT IN LATIN AMERICA Table I shows the average voter turnout (vote participation as share of the electorate; that is, registered voters) for parliamentary and presidential elections in each of the eighteen countries. These data are taken from the International Institute for Democracy and Electoral Assistance (IDEA), Nohlen 2005, and sometimes Psephos. 1 In terms of cross-national variations, two groups of countries stand out in a positive sense. In the first group are those countries that have maintained very high turnout, defined as averaging 80 percent or more. This group consists of Brazil, Chile, and Uruguay, as well as Peru for its presidential elections. In the second group are those countries that have maintained high turnout, defined here as being in the 70s and thus above the regional average. This group consists of Argentina, Bolivia, Costa Rica, Nicaragua, Panama, the Dominican Republic for its presidential elections, and Peru for its parliamentary ones. Overall, all of the countries in these two groups stand out from the rest in a positive sense. Conversely, very low turnout (defined as being below 60 percent) occurs in Colombia (which in fact has the lowest turnout), El Salvador, and Guatemala, as well as Venezuela for its parliamentary elections. Low turnout (defined as being in the 60s) exists in Ecuador, Honduras, Mexico, Paraguay, the Dominican Republic for its parliamentary elections, and Venezuela for its presidential elections. That said, both Ecuador and Paraguay could also been deemed to have average turnout for the region. 1 Wherever presidential elections involved two rounds of voting, the higher of the two turnout values was used. 2

3 Overall the parliamentary elections have had lower turnout, by some three percent. That said, in part this is because the two lowest turnout elections overall have been parliamentary ones with extenuating circumstances. The Guatemalan parliamentary election of 1994 was an interim election to elect the parliament for only a one-year term, at which point the 1995 parliamentary election would be held simultaneously with the presidential election the latter from then on set for every four years as opposed to five in the past. The turnout in 1994 was the lowest in this analysis, at 21.0 percent. The second lowest turnout overall, at 25.3 percent, came in the Venezuelan parliamentary election of 2005 which was boycotted by the opposition. Table I Mean Turnout Levels Since 1990 Parliamentary Presidential Elections Elections Argentina Bolivia Brazil Chile Colombia Costa Rica Dominican Republic Ecuador El Salvador Guatemala Honduras Mexico Nicaragua Panama Paraguay Peru Uruguay Venezuela unweighted average

4 EXPLANATIONS OF CROSS-NATIONAL TURNOUT LEVELS There is a vast and growing literature on comparative voter turnout (for example, Blais and Carty 1990, Blais and Dobrzynska 1998, Franklin 2002, Siaroff and Merer 2002, Franklin 2004, and Geys 2006). This academic literature suggests several areas of explanatory factors for why some countries have had higher (lower) turnout than others. These are discussed by area, and bivariate regressions are provided for their effects on parliamentary turnout (Table II) and presidential turnout (Table III). Table II Bivariate Regressions on Parliamentary Turnout Unstandardized significance N B t-score level adjusted r 2 population cultural fragmentation indigenous share of population population density urbanization Human Development Index for Freedom House score year of universal suffrage simultaneous election compulsory voting scale compulsory voting dummy variable presidential powers legislative powers unicameral legislature dummy variable parties with 5 percent of the vote vote difference between first and second parties party system polarization dummy variable party system institutionalization party system programmatic nature

5 Table III Bivariate Regressions on Presidential Turnout Unstandardized significance N B t-score level adjusted r 2 population cultural fragmentation indigenous share of population population density urbanization Human Development Index for Freedom House score year of universal suffrage simultaneous election compulsory voting scale compulsory voting dummy variable presidential powers legislative powers unicameral legislature dummy variable first ballot candidates with 5 percent of the vote vote difference between first and second candidates party system polarization dummy variable party system institutionalization party system programmatic nature The first theoretical area is socio-economic / demographic. In this regard, factors such as a high level of development, small size, and high population density have been found to affect positively turnout (Blais and Carty 1990; Blais and Dobrzynska 1998). It is hypothesized that in more developed countries people are more informed and have more resources (including time) to devote to politics, thus turnout will be higher. Indeed, the countries globally with very high or high turnout are generally quite developed, be this in terms of income per capita or broader measures of development such as the United Nations Human Development Index. Yet there are some highly developed nations with very low turnout, most specifically Switzerland. It is also hypothesized that turnout will be higher in smaller countries, since citizens will feel more 5

6 powerful as individuals versus the whole (see Dahl and Tufte 1973: 43). A related assumption is that turnout will be higher in more densely populated (urbanized) settings, as individuals will not only be but also feel closer to their neighbours as fellow citizens, and thus be more likely to take an interest in collective matters. Certainly in Western Europe many of the countries with very high or high turnout are small or tiny in terms of population: Austria, Belgium, Denmark, Greece, Norway, Sweden, and especially Cyprus, Iceland, Luxembourg, and Malta (the latter four each having less than a million people and each having very high turnout). Most of these very high or high turnout countries also have a high population density and high level of urbanization. On the other hand, large populations have not prevented high turnout in such countries as Germany and Italy. Finally in terms of demographic factors, Geys (2006: 644: 645), suggests that, since cohesion increases group solidarity (and social pressure ), more homogeneous societies should have higher turnout. These demographic factors are measured in terms of various independent variables: population in the election year (from IDEA), the cultural fragmentation of the population (from Fearon 2003), the indigenous share of the population (from the CIA World Factbook), the population density in the election year and the level of urbanization in the election year (United Nations Population Division data), and the United Nations Human Development Index score for 2000 (to take a fixed year at the middle of the time period). Population levels range from 2.58 million in Panama in 1994 to million in Brazil in Cultural fragmentation ranges from zero in Argentina, the Dominican Republic, and Uruguay to (out of 1) in Bolivia. The indigenous share of the population ranges from zero in the Dominican Republic, Paraguay, and Uruguay (and close to zero in several other countries) to 55 percent in Bolivia, 45 percent in Peru, and 41 percent in Guatemala. Population density ranges from 7 people per square kilometre in Bolivia in the 1970s to 287 people per square kilometre in El Salvador in Urbanization ranges from 41.1 percent in Guatemala in 1990 to 92.6 percent in Venezuela in Lastly, the 2000 Human Development Index score ranges from in Guatemala and in Honduras to in Argentina. As Tables II and III show, in terms of bivariate regressions population itself, cultural fragmentation, and the indigenous 6

7 share of the population are all statistically insignificant in terms of explaining either parliamentary or presidential turnout. Conversely, population density, urbanization, and the Human Development Index value are all statistically significant in terms of explaining levels of both parliamentary and presidential turnout. That said, the coefficient for population density is negative in both cases, thus it is lower population density which leads to higher turnout (higher population density thus leading to lower turnout). This goes against the general argument based on the industrial democracies. It may well be that in Latin America population density is not so much an indicator of social cohesion as it is of less development (crowded favelas or equivalents). The next area involves the level of democracy in a country. It is hypothesized that the more democratic a country is and the more rights people have, the more they are likely to exercise those rights. Annual Freedom House scores are used as the independent variable here, and indeed they are also reported in the IDEA turnout data. Since a lower score indicates greater freedom, it is hypothesized that the relationship with turnout will be negative. Actual Freedom House scores range from 1.0 (the lowest possible) in Costa Rica at times and in Uruguay in 2004 to 4.5 in Peru in 1995 (when it was clearly an autocracy under Fujimori) and in Guatemala in Related to the level of democracy is the year of universal suffrage, since this indicates a long (or not so long) tradition of all adults being able to vote in elections. Specifically by universal suffrage we mean the granting of suffrage to women and the removal of any literacy requirements. For example in Europe suffrage is low in Switzerland which was the last major country to grant women the vote (in 1971). In Latin America the year of universal suffrage ranges from 1934 in Uruguay to 1988 in Brazil (when literacy requirements were removed), with a country mean of As Tables II and III show, the Freedom House score is a highly significant predictor of both parliamentary and presidential turnout and the coefficients are negative in each case, indicating that a lower Freedom House score (a higher level of democracy) leads to higher turnout. As for the year of universal suffrage it relates significantly to parliamentary turnout but not presidential turnout; however the coefficient for parliamentary turnout is positive, indicating (against the hypothesis) that parliamentary turnout is higher in those countries that were late to achieve universal suffrage. 7

8 Certainly Brazil and Chile stand out as such countries, leading us to conclude that such a variable does not have a valid utility in Latin America. The next overall theoretical area regarding voter turnout involves institutional aspects of the vote itself. The first issue here is whether the parliamentary and presidential elections (the first round thereof if two rounds) are held simultaneously, that is, on the same day. This is hypothesized to increase turnout since the voter gets double the effect for showing up at the polls, and there is also greater media focus. The most illustrative case here is the Dominican Republic: through 1994 its parliamentary and presidential elections were simultaneous, then they shifted to be two years apart. Subsequent parliamentary turnout clearly dropped, whilst nothing changed for presidential turnout. A dummy variable is used to indicate simultaneous elections. As Tables II and III show, having a simultaneous election is a statistically strongly significant factor for higher parliamentary turnout but not really for higher presidential turnout (as a single explanation). This may be because the presidential election is perceived as the more important one, so it affects parliamentary turnout but not the other way around. The second institutional aspect is compulsory voting, perhaps the central solution to increasing turnout (Lijphart 1997: 8-9). There are two ways to categories this. The first is the continuum used by Fornos et al from the data in Payne et al (their Table 3.2), namely countries with voluntary voting = 0, countries with compulsory voting but no sanctions = 1, countries with compulsory voting and legal sanctions but where these sanctions are not usually enforced = 2, and countries with compulsory voting and legal sanctions that are enforced in practice = 3. The second way to categorize compulsory voting is as a dichotomous dummy variable where the latter two categories are deemed to constitute compulsory voting since there are legal sanctions. It is hypothesized that it is only sanctions for non-voting that will make a appreciable difference, due to the monetary fines and/or other punishments levied. Indeed, all three European countries with strictly enforced compulsory voting (Belgium, Cyprus, and Luxembourg; see Gratschew 2001) have extremely high turnout, as does Australia, so it can be argued that enforced compulsory voting definitely works ; in other words, it is a sufficient (but not necessary) 8

9 causal factor producing / maintaining very high turnout. Whether one uses the continuum scale or a dummy variable, compulsory voting is a highly significant predictor of both parliamentary and presidential turnout in Latin America, as Tables II and III show. For example, using the dummy variable of compulsory voting with legal sanctions, one sees that this leads to 18 percent higher turnout in parliamentary elections and 11 percent higher turnout in presidential elections, all other things being equal. The next area of causal factors involve the powers of the president and parliament and the importance of the national parliament and thus its elections for a country s political system. As scholars such as Blais and Dobrzynska (1998: 245) argue, other elected bodies could weaken the power of the national parliament. The absence of such bodies in Malta has been noted by Hirczy (1995: 260). (Multiple elections also lessen parties ability and will to mobilize for any one specific election see Cox 1999: 408.) Of course, all Latin American countries have an elected president so none approach the pure concentration of parliamentary power of a Malta. In this regard three questions are asked about Latin American countries: how powerful is the president, how powerful is the parliament, and is the parliament unicameral? Presidential power is measured by the index of Stein et al. 2005: 49 (their Table 3.5) which calculates the overall legislative powers of presidents. The power of the legislature is calculated using the new parliamentary powers index of Fish and Kroenig Both of these are on a zero to one scale. The strongest presidents are those of Chile (0.66) and Brazil (0.62); the weakest are those of Nicaragua and Paraguay (both 0.19). The strongest legislatures are those of Nicaragua (0.69), Peru (0.66) and Uruguay (0.66); the weakest legislatures are those of the Dominican Republic (0.41), Bolivia (0.44), and Mexico (0.44). Since these are different scales (and authors) the patterns are not perfect inverses of each other; nevertheless, for extreme cases like Nicaragua the contrasts are clear. It is hypothesized that where the president or the legislature is comparatively strong, turnout will be higher since the institution has more power. Likewise, it is hypothesized that parliamentary turnout (at least) will be higher in unicameral systems since there is no senate to act as a separate veto point (that is, control over the single chamber produces complete legislative control). As Tables II and III show, neither 9

10 greater presidential powers nor greater parliamentary powers affects turnout in a statistical sense. This may be because such measures compare across countries, but voters are not picking which country to vote in. As for unicameralism, in terms of bivariate regressions this is statistically significant for parliamentary turnout but not presidential turnout, leading as expected to lower turnout in parliamentary elections (by over eight percent). The last five theoretical areas regarding voter turnout involve separate aspects of the party system. The first of these is party system fragmentation / concentration. As Blais and Dobrzynska (1998: 248) hypothesize, the greater the number of parties, the more choice electors are offered, and the higher the turnout. The alternative hypothesis here is that concentration provides the greater chance of a clear winner and thus a decisive outcome, which makes voting more important. Either way, a cut-off of five percent support is used to determine relevance in this regard. Thus for each parliamentary election the number of parties winning five percent or more of the vote is calculated, likewise for presidential elections the number of candidates with at least five percent support on the first ballot (even if that is not the ballot with the highest turnout). The number of relevant parties / candidates thus ranges from one in the 2005 Venezuelan parliamentary election (which as noted was boycotted by the opposition) and two in many cases to eight in Brazil in its 1990 and 1994 parliamentary elections, Ecuador in its 2002 parliamentary election, and Panama in its 1994 parliamentary election. As Table II shows, more parties relates significantly to increased parliamentary turnout; however, as Table III shows, having more size-relevant candidates is not significant in terms of presidential turnout. Related to aspects of party system fragmentation are those of party system competitiveness. Here the issue, or debate, is whether or not turnout is higher in close elections (Blais and Dobrzynska 1998: 249). This measure is especially relevant when one is comparing specific elections rather than longstanding national patterns. Blais and Dobrzynska (1998: 249) also ask whether an election produces a single party majority government. It is hypothesized in this regard that turnout will be lower where one party wins an outright majority, since the result is predictable (Siaroff and Merer 2002: 921). Malta clearly speaks against this argument, since it is a very high turnout country with consistent single party majorities; moreover, going 10

11 back to the previous point, Malta is also a pure two party system. In relatively few cases in Latin America do legislative elections produce a single party majority, conversely a presidential election will by definition always produce a single winner so let us focus just on the closeness of the result, measured in terms of the vote difference between the first and second parties in parliamentary elections and the first and second candidates in presidential elections. Such a difference ranges from less than one percent in several parliamentary elections and more crucially since they were decisive 0.2 percent in the 2006 Costa Rican presidential election and 0.6 percent in the 2006 Mexican presidential election up to 80 percent in the 2005 Venezuelan parliamentary election. As Table II shows, closer results do lead to higher parliamentary turnout; however, as Table III shows, the closeness of results is not a statistically significant factor for presidential turnout. The third traditional area of party system comparison is that of party system polarization, that is, the ideological gap between the (main) parties / candidates, and the resulting policy difference between a government of (or led by) party / president A versus one of party /president B. This point specifically has been made for the United Kingdom by Heath and Taylor (1999). As they argue (1999: ): On rational choice grounds, if there is a great deal of difference between the parties, then the outcome is likely to make more difference to the country s political future and hence there is more reason to turn out and vote. Conversely, if there is little difference between the parties, then the election result will not make much difference either. We would, therefore, expect to see the highest turnouts when the parties are furthest apart ideologically from each other. Strictly speaking, the standard rational choice model of turnout is a multiplicative one, and we would expect to see the highest turnout when there was both a large difference between the parties and a close race. Certainly the evidence supports their argument: postwar Britain s highest turnouts (and the only ones above 80 percent) were in the elections of 1950 and 1951, both of which involved major policy differences between the Labour government and the 11

12 Conservative opposition and a close national race. More recently, the 1992 British election also combined both factors (at least in terms of the opinion polls re closeness), and it too had relatively high turnout (Heath and Taylor 1999: ). For our purposes a dummy variable is used to indicate high to moderate polarization, the top category (of three) of Payne et al. (2002: 151; 2003: 162). The countries so classified by them are Brazil, Ecuador, Nicaragua, Venezuela, and El Salvador, each of which receives this country dummy variable. To this list we can add Bolivia since 2005 with the polarization that has occurred for and against President Morales. That said, as Tables II and III show, party system polarization does not relate significantly to either parliamentary or presidential turnout. Indeed, El Salvador is one of the countries with very low turnout (Table I). Next, although not specifically emphasized in the literature on comparative voter turnout, one can and should discuss here the issue of party system institutionalization; that is, the extent to which parties and party competition are stable. Mainwaring and Scully (1995) distinguish ultimately between institutionalized and inchoate party systems in Latin America in terms of whether or not the parties are durable, rooted in society, and ideologically consistent with respect to each other, whether electoral volatility is low, and whether parties and elections have broad legitimacy. A more current classification is that of Jones (2005), who uses the same criteria as Mainwaring and Scully but who scores countries out of 100. Jones country scores on party system institutionalization (2005: his Table 7) thus range from a high of 76 in Uruguay to a low of 48 in Guatemala. As Table II and III show, this is not a statistically significant variable in bivariate regressions for parliamentary turnout but it is for presidential turnout, with greater party system institutionalization leading to higher presidential turnout. Related and last is the issue of programmatic party systems, that is, the extent to which parties exist and compete on the basis of clear and stable ideological and policy differences (as opposed to clientelism et cetera) and whether voters are aware of these distinctions. Jones (2005: his Table 15) has also developed a scale for Latin American countries in this regard, which ranges from 0 to 8 (the highest possible score). Chile and Nicaragua have such a score of 8 and El Salvador and Uruguay 12

13 have a score of 7; all four countries thus clearly have programmatic party systems. At the opposite extreme, Colombia, the Dominican Republic, and Panama all score 0. 2 As Tables II and III show, this is a statistically significant variable in both bivariate regressions: greater party system programmatic nature relates to both higher parliamentary and higher presidential turnout. MULTIVARIATE ANALYSIS AND OVERALL CONCLUSIONS Tables IV and V show the best fit multiple regressions for each of parliamentary and presidential turnout. For parliamentary elections the relevant factors for higher turnout as shown in Model One of Table IV are population density (but, again, turnout is higher with lower population density against the standard hypothesis), a higher level of human development (measured by the 2000 HDI value), a lower Freedom House score in the election year (thus indicating a higher level of democracy), having the election held simultaneously with the (first round of the) presidential election, ever greater enforcement of compulsory voting, a close election in the sense of an ever smaller vote difference between the first and second parties, and the party system having a more programmatic nature. Model Two simplifies the analysis a bit by removing the Freedom House score, as it was the least significant of the previous variables. In both cases (Models One and Two) about two-thirds of the variance in parliamentary turnout is explained. For presidential elections, the relevant factors for higher turnout as shown in Model One of Table V are population density (again, turnout is higher with lower population density), higher urbanization, a lower Freedom House score (higher level of democracy), having the election held simultaneously with the parliamentary election, ever greater enforcement of compulsory voting, having a unicameral system (which is significant in a multiple regression), and having an institutionalized party system. There are some differences here compared to the multivariate explanation of parliamentary 2 Jones then further modifies his country scores to take into account the level of electoral volatility, however for our purposes we are using only the raw subtotal. 13

14 turnout: urbanization matters, the level of development does not matter, a unicameral system matters, political competition in the sense of a close election does not matter, and the party system needs to be institutionalized rather than programmatic. Indeed, Model Two adds in the party system programmatic nature but this has no effect. The biggest difference between the two models, though, is apparently not the presence of absence of any one variable but the overall variance explained: for presidential elections this is only just under half the variance. One contributing factor here appears to be having a simultaneous election: whilst this boosts parliamentary turnout by almost 13 percent (Table IV), it only increases presidential turnout by four percent (Table V). This shows that the presidential election is perceived to matter more. Table IV Multivariate Regressions on Parliamentary Turnout significance Model One B error t level population density Human Development Index for Freedom House score simultaneous election compulsory voting scale vote difference between first and second parties party system programmatic nature constant N 85 F significance level adjusted r

15 significance Model Two B error t level population density Human Development Index for Freedom House score simultaneous election compulsory voting scale vote difference between first and second parties party system programmatic nature constant N 86 F significance level adjusted r Table V Multivariate Regressions on Presidential Turnout significance Model One B error t level population density urbanization Freedom House score simultaneous election compulsory voting scale unicameral system party system institutionalization party system programmatic nature constant N 72 F significance level

16 adjusted r significance Model Two B error t level population density urbanization Freedom House score simultaneous election compulsory voting scale unicameral system party system institutionalization party system programmatic nature constant N 72 F significance level adjusted r REFERENCES Blais, André, and Ken Carty (1990), Does proportional representation foster voter turnout?, European Journal of Political Research, Volume 18, pp Blais, André, and Agnes Dobrzynska (1998), Turnout in electoral democracies, European Journal of Political Research, Volume 33, pp Cox, Gary W. (1999), Electoral rules and the calculus of mobilization, Legislative Studies Quarterly, Volume 24, pp Crewe, Ivor (1981), Electoral participation, in David Butler, Howard R. Penniman, and Austin Ranney (eds.), Democracy at the Polls: A Comparative Study of Competitive National Elections. Washington, D.C.: American Enterprise Institute. Dahl, Robert A., and Edward R. Tufte (1973), Size and Democracy (Stanford: Stanford University Press). 16

17 Fearon, James D. (2003), Ethnic and Cultural Diversity by Country, Journal of Economic Growth, Volume 8, pp Fish, M. Steven, and Matthew Kroenig (2009), The Handbook of National Legislatures: A Global Survey (New York: Cambridge University Press). Fornos, Carolina A., Timothy Power, and James C. Garand (2004), Explaining Voter Turnout in Latin America, 1980 to 2000, Comparative Political Studies, Volume 37: 8 (October), pp Franklin, Mark N. (2002), The Dynamics of Electoral Participation, pp in Lawrence LeDuc, Richard G. Niemi, and Pippa Norris (eds.), Comparing Democracies 2: New Challenges in the Study of Elections (London and Thousand Oaks: Sage). Franklin, Mark N. (2004), Voter Turnout and the Dynamics of Electoral Competition in Established Democracies since 1945 (New York: Cambridge University Press). Geys, Benny (2006), Explaining voter turnout: A review of aggregate-level research, Electoral Studies, Volume 25, pp Gratschew, Maria (2001), Compulsory voting. Stockholm: International IDEA < Heath, Anthony, and Bridget Taylor (1999), New Sources of Abstention?, pp in Geoffrey Evans and Pippa Norris, eds., Critical Elections: British Parties and Voters in Longterm Perspective (London: SAGE). Hirczy, Wolfgang (1995), Explaining near-universal turnout: The case of Malta, European Journal of Political Research, Volume 27, pp International Institute for Democracy and Electoral Assistance, International IDEA Voter Turnout Website < Jones, Mark P. (2005), The Role of Parties and Party Systems in the Policymaking Process, paper prepared for the Inter-American Development Bank Workshop on State Reform, Public Policies, and Policymaking Processes, Washington, DC, 28 February to 02 March. Lijphart, Arend (1997), Unequal participation: Democracy s unresolved dilemma, American Political Science Review, Volume 91, pp Mainwaring, Scott, and Timothy R. Scully (1995), Introduction: Party Systems in Latin America, pp in Mainwaring and Scully (eds.), Building Democratic Institutions: Party Systems in Latin America (Stanford: Stanford University 17

18 Press). Nohlen, Dieter, ed. (2005), Elections in the Americas: A Data Handbook [two volumes] (New York: Oxford University Press). Payne, J. Mark et al. (2002), Democracies in Development: Politics and Reform in Latin America (Washington, DC: Inter- American Development Bank). Payne, J. Mark et al. (2003), La política importa: Democracia y desarrollo en América Latina (Washington, DC: Inter-American Development Bank). Psephos Adam Carr s Election Archive < Siaroff, Alan, and John Merer (2002), Parliamentary Election Turnout in Europe since 1990, Political Studies, Volume 50: 5 (December), pp Stein, Ernesto, et al. (2005), The Politics of Policies (Washington, DC: Inter-American Development Bank and Cambridge, MA: David Rockefeller Center for Latin American Studies, Harvard University). 18