Scholars and policy makers have advanced conflicting hypotheses about the dynamics of voter participation in nascent democratic regimes. The authors advance the research program by examining 108 parliamentary elections in postauthoritarian Latin America and post-Communist Europe from 1978 through 2003. Institutional, political, and demographic variables shape turnout in new democracies, but there is also a strong temporal effect: voter turnout drops sharply after founding elections and continues to fall through the fourth electoral cycle. Moreover, after appropriate controls, rates of turnout in Eastern Europe are consistently higher than the equivalent rates for Latin America. The authors attribute these differences to historical legacies and the mode of transition to democracy.
Keywords: voter turnout; disengagement; founding elections; Third Wave democracies; electoral dynamics
TO what extent do citizens in nascent democracies exercise their newfound right to vote? Contemporary political regimes feature a bewildering variety of forms of political participation. However, in democratic regimes, the one form of political participation that is definitionally connected to the concept of democracy is the act of voting in competitive elections. In the final quarter of the twentieth century, several dozen countries around the world underwent democratic transitions that made this unique form of participation available to their citizens for the first time in memory. Regardless of the prior regime type, the underlying popular demand in these political transitions was to offer citizens the chance to choose their rulers-a choice that is exercised via participation in direct, competitive elections.
Although many observers have applauded the struggles that overthrew dictatorships, few have examined the dynamics of electoral participation under new democracies. In the mid-1980s, Guillermo O'Donnell and Philippe Schmitter suggested that pressures for popular participation would build irresistibly under authoritarian rule, only to be discharged in a "founding election" that would inaugurate the new democracy. Voter participation should then be expected to decline in subsequent elections as the excitement of the transition wears off and voters learn that elections are not a panacea (O'Donnell and Schmitter 1986; see also Turner 1993). However, a recent study by the International Institute for Democracy and Electoral Assistance (IDEA) claimed that the "founding election" hypothesis is wrong and that there is no significant difference in voter turnout between the first postauthoritarian election and subsequent elections in new democracies (International IDEA 2002). Given that O'Donnell and Schmitter did not test their expectation empirically, and given that International IDEA'S counterclaim is based only on a simple difference-ofmeans test,1 researchers have begun to probe these questions more deeply using multivariate approaches. For example, in examining post-Communist elections in Eastern Europe, Kostadinova (2003) found strong evidence that the second and third elections under democratic rule were associated with significant declines in voter turnout. Similarly, a study of Latin America in the 1980 to 2000 period by Fornos, Power, and Garand (2004) reported that founding elections were associated with a 16 percent higher turnout rate than other elections, net of all other variables in their model. These singleregion studies suggest that citizens begin to disengage from electoral participation almost immediately after the democratic transition has been completed. However, to date we know of no sophisticated, multivariate studies that examine the dynamics of posttransitional electoral participation using a large number of elections and covering more than one world region.
In this article, we advance the research program on turnout dynamics in new democracies by comparing electoral participation in posttransition Eastern Europe and Latin America. By examining more than one hundred elections in twenty-seven new regimes spread across two world regions during the global Third Wave of democratization, we are able to subject both the "founding election" hypothesis and the "disengagement" hypothesis to more rigorous empirical testing. In the first section of the article, we describe our case selection, our data set, and our dependent variable. In the next section, we revisit some previous crossnational studies of electoral participation, and from these studies we derive a roster of well-established control variables that are crucial to our multivariate analyses. We then review the empirical evidence on turnout dynamics in new democracies. Finally, we place our findings in theoretical and comparative perspective.
Comparing Turnout: Case Selection in the Third Wave
To test disputed hypotheses about voter turnout in new democracies, we collected data on 108 parliamentary elections in postauthoritarian Latin America and post-Communist Europe in the period from 1978 through 2003.2 Some 48 of these elections occurred in Latin America, while 60 took place in Eastern and Central Europe. The relatively large W allows for a more fully specified model of turnout patterns in new democracies, incorporating a much wider range of historical, social, institutional, and political-process variables than could be achieved in a single-region study.
For inclusion in our data set, an election had to be associated with a Third Wave democratization process, in accordance with the historical framework devised by Huntington (1991). According to Huntington, the Third Wave of regime change began in Southern Europe in 1974 with the collapse of the Portuguese dictatorship, a seminal event that was soon followed by other democratic transitions in Spain and Greece. Spreading outward from Southern Europe, the tide of democratization washed over Latin America in the late 1970s and 1980s and then toppled the Communist systems of Eastern and Central Europe between 1989 and 1991. This global process of democratization includes a number of other political transitions in Asia and Africa, but the new democracies of Latin America and post-Communist Europe account for the bulk of the Third Wave cases. Not only are the new regimes of these two regions part of a single historical process of political change (Huntington 1991; Diamond 1999), but they also share many similarities at the domestic level: uneven or declining economic performance, painful structural reforms, a decline in state capacity, difficulties in political institution building, electoral volatility, and the continued presence of some antisystem actors in national politics. For these reasons, the new democracies of Eastern Europe and Latin America have been systematically compared against one another for the past fifteen years, enlarging and enriching the comparative research program on democratization (e.g., Przeworski 1991; Linz and Stepan 1996).
Because the Third Wave came to Latin America in 1978 and to Eastern Europe only after 1989, the electoral processes studied here are not perfectly coterminous. The earliest election in our data set took place in the Dominican Republic in 1978, and the most recently included election was that of Paraguay in 2003. Because we are interested in the dynamics of political participation in the early years of democracy, in our case selection we opted for decision rules based on electoral cycles rather than on specific years or dates. This method allows us to focus on equivalent phases of posttransitional political development. For each of our twenty-seven country-units, we examined the first four electoral cycles under democracy, collecting turnout data for the lower house or single chamber of the national legislatures. Using the same number of elections per cross-section has the methodological advantage of providing us with a balanced panel for statistical analysis. The choice of four electoral cycles is admittedly somewhat arbitrary. However, a lower number of cycles would provide at best an incomplete perspective on the dynamics of voter turnout after democratic transition, while requiring a higher number (five or more consecutive free elections) would censor an unacceptably large number of democratic transitions from the dataset, thus reducing the effective N for analysis. Using four electoral cycles per cross-section allows us to include twelve new democracies in Latin America3 and fifteen in post-Communist Europe, while at the same time affording us a broad temporal perspective on posttransitional politics. This early phase of democracy ranges from six years in Argentina, where staggered elections for the lower house were initially held every two years, to fifteen years in Uruguay, where the electoral cycle is five years long (see Appendix A for a complete list of elections used in this study).4
Turnout is measured here as total ballots cast (including blank and spoiled ballots) in elections for the lower house or single chamber of the national legislature, calculated as a percentage of the voting-age population (see Appendix B for descriptions of all variables and sources).5
Comparing Turnout: Theories and Variables
Comparative studies of voter turnout have tended to emphasize institutional, socioeconomic, or political process variables. Each of these three rival approaches to explaining cross-national differences in turnout has highlighted a number of well-established predictors that we employ as control variables in our analysis.
Institutional accounts of electoral participation were launched in the mid-1980s by the seminal articles of Powell (1986) and Jackman (1987), and twenty years later institutionalism remains the dominant approach to turnout (Biais 2006). Powell argued that singlemember district plurality (SMDP) electoral rules do not provide parties with incentives to run candidates everywhere, meaning that some districts are simply "written off' and parties do not mobilize voters uniformly across the national territory; in contrast, proportional representation systems with higher district magnitudes are conducive to broad-based mobilization on the ground. Jackman examined nineteen industrialized democracies, regressing voter turnout on five institutional variables: nationally competitive districts (following Powell), multipartyism, electoral disproportionality, unicameralism, and compulsory voting laws. Jackman's institutionalist approach generates strong results, with his model he was able to explain up to 75 percent of cross-national variation in turnout (Jackman 1987; Jackman and Miller 1995). In our analysis below, we include the first four of Jackman's key variables in all models; the fifth, mandatory voting, is included only in the model for Latin America. All of our Latin American cases use compulsory voting and none of our Eastern European cases do.
Here we improve on Powell's (1986) operationalization of nationally competitive districts through four categories. We argue that a more precise way to measure the national competitiveness of an electoral system is by using its (logged) average district magnitude: the higher the average M, the greater the incentive for parties to mobilize voters to win seats across the national territory.
The three other institutional variables are relatively straightforward. With regard to electoral disproportionality, highly disproportional systems punish minor parties. Supporters of these parties will thus have less of an incentive to appear at the polls, and therefore disproportionality should depress electoral participation. Here we use the simplified index of disproportionality devised by Lijphart (1984), which takes the "average vote-seat share deviation of the two largest parties in each election" (p. 163).
Jackman (1987) argued that multipartyism should be inversely related to voter turnout. The expectation is that due to the unpredictability of coalition formation, the composition of the postelection cabinet is often unknown on election day, and voters in fragmented party systems will be less efficacious because they perceive that their votes are not perfectly translated into the formation of governments. A related expectation about multipartyism is that a saturated political market with many parties may confuse and alienate potential voters (Power and Roberts 1995; Kostadinova 2003); this expectation is especially relevant in new democracies where party systems have not yet institutionalized and labels are still unfamiliar.6 We measure multipartyism using Laakso and Taagepera's (1979) effective number of parties in the lower house or single chamber.
Jackman (1987) also argued that legislative structure should affect turnout. According to Lijphart (1984), bicameralism can be "asymmetrical" in a manner that favors the legislative powers of the lower house or "symmetrical" when both houses are roughly equal. Also, both houses may be "congruent" in their social bases of representation (i.e., representing the same units) or they can be "incongruent" when the constituencies of the two chambers differ. Lijphart claimed that congruent chambers weaken bicameralism (p. 213). In bicameral systems, legislation must be negotiated between the two houses, implying a weaker role for each chamber and some agency loss on the part of the electorate. Consequently, the more a system approximates the unicameral model, the greater the perception of decisiveness and efficiency, thereby fostering higher voter turnout. The sign on this variable should be positive, indicating that legislatures approaching the unicameral model encourage greater electoral participation by citizens. (See Appendix B for details on coding.)
We also add one institutional variable not considered by Jackman (1987) or Powell (1986) but recently advocated by Norris (2004): concurrent elections. In presidential systems, turnout in parliamentary elections should naturally be higher when there are simultaneous elections for the national executive and lower when legislative elections are held alone. Concurrence periodically raises the level of electoral salience for parliamentary elections (Franklin 1996); this has been the pattern for more than two hundred years in the United States (Franklin and Hirczy de Miño 1998). Given that we are pooling the presidential systems of Latin America together with the mostly parliamentary and semipresidential systems of Eastern Europe, it is essential that we control for this effect.
Moving to socioeconomic models of turnout, we draw upon a vast literature that emphasizes how wealth and education stimulate political participation (Almond and Verba 1963; Wolfinger and Rosenstone 1980; Fuer, Kenny, and Morion 1993; Rosenstone and Hansen 1993; Norris 2004). Here we include three control variables that are critical measures of national-level socioeconomic modernization: urbanization, literacy, and gross domestic product per capita. Urbanization exposes citizens to national politics and exposes them to diverse sources of political information. Literacy empowers citizens to digest sources of information and thus enhance their political skill levels, including the subjective competence needed for electoral participation. Per capita GDP is the most widely used cross-national measure of economic development, which we measure at purchasing power parity (PPP) using the 1995 international prices.7
Some cross-national studies of turnout have emphasized aspects of the domestic political process, one of these being the level of democratization. We recognize that within the genus of democracies (and especially among nascent ones) there are still observable differences in the level of pluralism and political freedom. To capture this, we rely on data from Freedom House, which has rated support for political rights and civil liberties in each country and year since the early 1970s. We reversed the original coding and expect the coefficient for the Freedom House rating to be positive, that is, turnout should be higher in more democratic country-year cases.
We note that democracy can be measured not only as a level but also as a legacy: a nation's historical experience with democracy has been claimed to have a strong effect on the prospects for success of present-day transition and consolidation (Huntington 1991; Linz and Stepan 1996; Crawford 1996; Smith 2005). Latin America and Eastern Europe are two regions with very different histories of democratic governance: most post-Communist states held their last multiparty elections more than forty years prior to their Third Wave transitions, while Latin American countries had more recent experiences with democratic procedures punctuated by return(s) to authoritarianism in the 1960s and 1970s. Furthermore, particularly in Latin America, strong intraregional variation in previous experiences with democracy may be also responsible for varying levels of mobilization. Therefore, we include a variable capturing the democratic legacy for each country for roughly the two generations preceding the Third Wave transition.8 We expect that voter turnout should be higher in cases featuring a stronger democratic legacy.
Another aspect of the political process that should affect turnout is the closeness of elections (Biais 2006). Both Downs (1957) and Aldrich (1993) argued that potential voters take into account the likelihood that their votes will affect the election outcome, and some studies have found empirically that closer elections are in fact associated with higher levels of turnout (e.g., Cox and Munger 1989; Biais 2000). We measure electoral competition as the difference in the vote share between the first- and second-placed parties, subtracted from 100. For example, a score of 99 indicates that the first-placed party bested the secondplaced party by only 1 percentage point, whereas a score of 85 would indicate that the margin of victory was 15 points (a significantly lower level of competitiveness). Therefore, we expect the sign on this variable to be positive.
Another critical variable is that of compulsory voting.9 Mandatory voting is almost always measured as a dichotomous variable in cross-national work (e.g., Jackman 1987; Jackman and Miller 1995; Franklin 1996, 1999; Radcliff and Davis 2000; Pérez-Liñán 2001). Given that it would be scored as a 1 for all Latin American cases and O for all Eastern European cases, it is represented in our pooled models by a dummy variable for Latin America-a variable that necessarily subsumes a large number of other crossregional differences. Only within the Latin American subsample do we control for the extent to which compulsory voting regulations are enforced in practice, using a polychotomous measure of sanctions against nonvoters (see Appendix B).
Having controlled for a wide variety of factors thought to affect cross-national differences in electoral participation, we can now introduce the independent variables of interest that best capture our concerns with posttransition dynamics. Rather than rely on a single measure of time effects, we employ two specifications. Election sequence is a count variable scored as 1 for the first election held in a new democracy, 2 for the second election, 3 for the third election, and 4 for the fourth and final election included in our data set. If citizens gradually withdraw from electoral participation after the instauration of democracy, as Kostadinova (2003) found for Eastern Europe, then the coefficient for this variable should be negative. To take the closest look possible at the dynamics of posttransition turnout, we then include separate dummy variables for the second, third, and fourth elections under democracy. The coefficients for these variables will reveal how specific posttransition conjunctures are associated with the propensity of citizens to turn out at the polls.
Model and Data Analysis
To evaluate the effects of the various determinants of turnout in transitional democracies identified in our theoretical discussion, we estimate a statistical model based on four clusters of independent variables: institutional, socioeconomic, political, and dynamic. Thus, the basic functional form of the model is
Y^sub it^ = a + ∑b^sub k^X^sub it^ + ∑c^sub i^E^sub it^ + ∑d^sub m^P^sub it^ + ∑z^sub n^D^sub i^ + e^sub ir^.
where Y is voter turnout rate in country i at time t; a is the intercept; k, I, m, and n are the numbers of independent variables; X is the common designation for institutional factors (including district magnitude, disproportionality, multipartyism, unicameralism, and concurrent elections); E is the cluster of economic and social predictors (including per capita GDP, literacy, and urbanization); P stands for political context determinants (level of democracy, prior democratic experience, and electoral closeness); D represents the transitional dynamics factors; e is the error term, and b, c, d, and z are regression coefficients. Since we expect that the two regions under study are different in ways that remain not accounted for by the variables in the general model, we include a dummy variable for Latin America to control for those effects.10 While this variable is intended to account for cross-regional differences broadly defined as culture and historical legacies, it will also help us model the effect of the compulsory voting procedures introduced in all of the Latin American and none of the East European countries in our sample.11
We begin the analysis with some simple descriptive statistics revealing variation in our dependent variable, voter turnout, across regions and over time. Our initial expectations are that (1) turnout rates will be different in Eastern Europe and Latin America, and (2) electoral participation will be the highest in founding elections and will decline in subsequent electoral contests. Examining Table 1, we can see that overall, voter mobilization in founding elections has been considerably lower (more than 17 percent) in Latin America than in Eastern Europe, even though our Latin American cases use mandatory voting and our European cases do not. In subsequent elections, average participation still remains higher in postCommunist countries but the difference with regard to Latin America diminishes steadily, falling to less than 1 percent by the fourth cycle.
Moreover, the mean values illustrate well the dynamics of voter disengagement over time in the Eastern European context. The difference in turnout between founding and second elections is almost 12 percent in the post-Communist cases, followed by a gradual decline at the rate of 3 to 4 percent. In Latin America, things appear quite different. Electoral participation, as revealed by the descriptive statistics in Table 1, does not look significantly different in the founding, second, and third elections, staying within the range of 64 to 65 percent at all times. These findings are consistent with what some observers have already noted, namely, that the theoretically predicted dynamics of voter turnout decline during democratic transition has received empirical support in the post-Communist world but not elsewhere (International IDEA 2002; Kostadinova 2003). While our univariate analysis appears to support this previous observation, we are also curious whether these results will be confirmed after we control for all other nontemporal factors in a multivariate analysis.
Our choice of an appropriate estimation technique is guided by both methodological and theoretical factors. Our estimation procedure needs to take into account the pooled structure of the data (Sayrs 1989). In such cases, problems related to possible temporal and spatial correlation in the error structure are of major concern. Since we are interested in tracing turnout rates over time, we include temporal variables on the right-hand side of the equation. The temporal variables both model the theoretical expectations for decline of turnout over time and account for the dynamics of the data in a methodological sense. Furthermore, we need to consider panel heteroskedasticity and correlation between pairs of countries separated by artificial borders - legitimate concerns associated with any panel-structured data. Our data set has a pooled time-series cross-sectional design where elections are the units of analysis and countries are the cross-sections; in the data matrix, variables change both over time and across space. Thus, we have a very broad and too shallow pool, where N > T(N= 21 and T= 4). To avoid underestimation of the parameter estimates, we opt for Beck and Katz's (1995) ordinary least squares (OLS) with panel-corrected standard errors (PCSE) technique. This method produces accurate estimates of β in the presence of panel heteroskedasticity and contemporaneous correlation in the error structure and is preferred for pools with T much smaller than N.
The second column in Table 2 presents the estimates for the predictors of turnout in all countries included in our data set. The results support our hypotheses about the effects of electoral disproportionality, multipartyism, democratic experience, and election sequence on voter participation. The estimated parameters of these four independent variables have the correct signs and are statistically significant. They reveal that more disproportional systems depress participation, a larger number of parties running in the race discourages participation, more people turn out to vote in countries with a stronger historical legacy of political democracy, and enthusiasm about voting declines over time after the Third Wave transition. Surprisingly, the estimated coefficients for several factors previously found to be important determinants of participation in other cross-national studies - such as concurrent elections, socioeconomic attributes, and present level of democracy - carry the correct sign but fail to reach statistical significance.
To check the robustness of our initial results, we proceed as follows. First, we run bivariate correlations to identify possible high levels of muticollinearity. The highest correlation coefficients are produced by the socioeconomic indicators (.685 between GDP and literacy, .579 between literacy and urbanization). Dropping one of these variables, however, produces results very close to the ones reported for the full model. Second, we employ the fractional pooling technique advanced by Larry Bartels (19%) to establish whether the robustness of findings is dependent on our choice to pool the Latin American and the East European cross-sections. We weight all observations made on the East European elections by a discount fraction λ that takes the values of .7, .5, and .3, which reflect different rates of belief that the subsample belongs to the larger data set. Then we examine the parameters obtained after running the regression with the discounted observations and test a null hypotheses that the coefficients obtained from the fractionally (O < λ < 1) and the fully (λ = 1) pooled data are equal. The results from this multiple procedure show that the null hypothesis of parameter equality can be rejected for at least four independent variables (see Appendix C for detailed results). For example, after a 50 percent discount, the concurrent elections and freedom rating parameters become statistically significant, while multipartyism fails the significance test only when we approach complete elimination of the East European cases. Running the regression for each region separately, that is, λ = 0, produces two sets of results that are fully consistent with the pooled analysis only with regard to the coefficients for district magnitude, vote-seat share disparity, and transitional dynamics. This exercise raises some concerns about the stability of regression parameters across the two subsamples. Therefore, we continue the analysis by running tests and reporting results for the pooled sample while controlling for regional differences (second column in Tables 2 and 3), as well as reporting and discussing separate results for the Latin American and the East European subsamples (third and fourth columns in Tables 2 and 3).12 This search for stable parameters will help us identify the powerful determinants of voter turnout and generate robust estimates of the magnitude of their effects.
The first message from the separate regression analyses is that electoral rules do matter. Positive and statistically significant coefficients for district magnitude are obtained for each region, suggesting that voters tend to participate more often when a larger number of representatives are elected in their constituencies. This effect is more pronounced in Latin America than in Eastern Europe. Furthermore, as shown by the positive and statistically significant estimates consistently obtained for concurrent elections, holding presidential elections on the same day boosts voter mobilization in parliamentary races by nearly 7 percent in Latin America and nearly 10 percent in Eastern Europe. The results further suggest a statistically significant decrease in turnout rates in countries with election systems producing disproportional results. The model predicts that all else kept the same, if the vote-seat share deviation for the two largest parties increases by one percentage point, turnout will decrease by about 1 percent in Latin America and 0.7 percent in Eastern Europe. Most important, the estimated parameters for election sequence are consistent in polarity and in achieving statistical significance in both subsamples. The coefficients suggest that with each successive Third Wave election in Latin America and Eastern Europe, enthusiasm for voting declines by 4.5 percent.
For five of our independent variables, the hypothesized impact is supported in only one of the two democratizing regions. The coefficient for multipartyism is statistically significant only for the postCommunist countries, where 2.4 percent fewer voters show up to vote when the effective number of parties rises by one. Also in Eastern Europe, a statistically significant increase in turnout rates of nearly 2 percent is associated with a legislative body that is one category closer to unicameralism. Furthermore, the data provide strong support for the mobilizational effects of national wealth and democratic legacy in post-Communist Europe. At the same time, the expected effects for present-day civil and political liberties find conclusive empirical support only in the Latin American subsample. There, a one-unit increase in the Freedom House rating (signifying a strengthening of democratic commitments) is associated with a 2.0 percent increase in electoral participation, net of all other variables in the model. Consistent with previous research, stricter provisions for enforcement of mandatory voting appear to have some mobilizing effect in Latin American democracies, but the evidence is not conclusive after controlling for all other factors.
Next, we model temporal dynamics of transitional elections by controlling for the separate effect of each one of the four postauthoritarian elections. In a study of post-Communist voter turnout, Kostadinova (2003) argued that a gradual decline in participation follows founding elections due to voters' fatigue and disappointment with transitional hardships. Here, we have an opportunity to test this argument for the Latin American countries and also to repeat the test for the East European context while controlling for a more comprehensive set of institutional, socioeconomic, and political variables. Consequently, we introduce three dummy variables for the second, third, and fourth democratic elections, thus keeping the founding elections as a referent category. Our expectation is that the estimated coefficients will be negative in sign. Moreover, we expect that the third election coefficients will be larger than those for the second election and that the fourth election coefficients will, in turn, be larger than the ones obtained for the third elections. If obtained, these results would indicate a gradual decline in participation, signifying voter withdrawal and disengagement after the democratic transition.
Our results, reported in Table 3, offer convincing evidence in support of the argument that electoral participation declines after the instauration of democracy.13 All of the estimated coefficients for second, third, and fourth elections have negative signs, suggesting that turnout in these contests has been lower than in the founding elections (the referent category), and all but one (the second electoral cycle in the pooled test) reach statistical significance. There is strong empirical evidence for a continuous drop in turnout rates in both Latin America and Eastern Europe. The estimated parameters suggest that on average and holding all other factors constant, second elections in Latin America are predicted to experience a much lower turnout rate (by about 9.7 percent), then another drop of about 2 percent will follow in the third, and a new drop of 4.6 percent in the fourth electoral contest. The results for Eastern Europe are consistent with Kostadinova's (2003) findings for declining trends in post-Communist electoral participation.
Moreover, we note the striking similarity in the parameter estimates for the second and fourth elections in Eastern Europe and Latin America. Despite key differences in history, culture, and prior regime type, the dynamics of voter disengagement under Third Wave democracy are remarkably consistent across the two regions. The separate regional models in Table 3 suggest that by the time of the fourth election under democracy, voter turnout will have suffered a drop of nearly one-sixth in relation to the historic election that inaugurated the regime.
Discussion and Conclusions
Does democratization depress participation? To answer in the affirmative would indeed be counterintuitive. Before proceeding, however, we note that our analysis is restricted to electoral participation in politics (voter turnout). We recognize that democracy - especially in its contemporary, Third Wave incarnation - is conducive to a wide variety of forms of political participation, and the act of voting is only one of them. Still, the right to vote in free and fair elections was one of the key demands of the antiauthoritarian struggle in both Eastern Europe and Latin America, and it is somewhat troubling to see electoral abstention rising so consistently and so rapidly in the new democracies of these two regions. Our conclusions concerning this phenomenon focus on three aspects of turnout: (1) enduring differences between Latin America and Eastern Europe, (2) performance of the traditional predictors of turnout in our respective analyses, and (3) the importance of posttransition dynamics.
A striking finding of this article concerns the consistently lower turnout rates in Latin America as compared to Eastern Europe. Table 1, which presents only simple means, suggests that Latin American turnout is significantly lower only in the very early years of democracy, but that after two electoral cycles the two regions begin to converge. But this is misleading. Tables 2 and 3, which present multivariate analyses, show a quite different picture - after controlling for the many other factors thought to affect voter participation, turnout in Latin American elections is more than 20 percentage points lower than in post-Communist elections. Why is this so?
Comparing the democratization experiences of the two regions, we see at least four reasons that could explain the relatively higher rates of voter participation in Eastern Europe. First, in terms of the origins of the nondemocratic regimes, most Communist systems were either established and/or supported by a foreign state, the USSR. This imparted an element of nationalism - far less important in Latin America - to the East European transitions of 1989: citizens mobilized to overthrow foreign influence and develop indigenous electoral institutions and practices. Second, in terms of the nature of the nondemocratic regimes, the Latin American regimes were authoritarian whereas the Eastern European regimes were totalitarian or posttotalitarian (Linz and Stepan 1996).14 While in Latin America there was limited pluralism and a more autonomous civil society, the Communist systems were rigidly ideological and monopolistic. The sharp break with totalitarianism in Eastern Europe may have generated more incentives to immediately exercise newfound democratic freedoms such as voting. Third, we could make a parallel argument about the length of the nondemocratic experiences in the two regions: generally shorter and more punctuated in Latin America, far longer and uniformly uninterrupted in Eastern Europe. This again would generate stronger incentives for immediate and sustained voter mobilization in the post-Communist cases. Fourth, there was also cross-regional variation in the mode of transition: generally longer and more gradual in most Latin American cases (with a few important exceptions, e.g., Argentina) but almost always approximating the "rupture" or "collapse" model in Eastern Europe. Gradual transitions "from above" afford greater space for manipulation by proregime elites, which may in turn have a demobilizing effect on mass participation (Share and Mainwaring 1986). The ruptura model of transition, on the other hand, often features mass mobilization "from below" - just as Eastern Europe witnessed in late 1989. In short, there are powerful reasons to believe that these discrete features of the complex process of political democratization - which unfolded differently in Latin America and Eastern Europe between 1978 and 1991 - may explain some of the variance in posttransitional electoral participation across the two regions.
Turning to more traditional arguments about voter turnout, this article has examined four clusters of independent variables (institutional, socioeconomic, political, and temporal) thought to predict voter turnout in nascent democracies. In terms of institutional factors, we have found that electoral disproportionality and unicameralism are important predictors of voter turnout in Latin America, whereas multipartyism appears to be more important in Eastern Europe. The stronger depressing effect of multipartyism in Eastern Europe may be related to the volatile, inchoate party systems that emerged from the ashes of monopolistic Leninism.15 With unfamiliar party labels and rising fragmentation, Eastern European voters could easily be confused and alienated by their new choices. In contrast, even with their well-known deficiencies, most Latin American party systems contained at least one or two parties with prior roots in a Second Wave democratic experiment, and some democratic transitions (Argentina, Chile, Paraguay) were characterized by a partial resurrection of a previous party system. Moreover, the Latin American experiences with nondemocratic rule were shorter, and many voters were familiar with multipartyism from the past. This may explain the differing effect of party fragmentation on voter participation across the two regions.
Electoral concurrence is an institutional device that has strong positive effects on voter turnout in both Eastern Europe and Latin America. When a president is being elected on the same day, voters are much more likely to cast ballots for Parliament. This effect has already been documented for the presidential systems of Latin America (Fornos, Power, and Garand 2004), but we are among the first to detect it in the more institutionally diverse systems of post-Communist Europe. Another institution that has similar effects in both regions is district magnitude: the higher the average M, the more likely citizens are to participate in elections. This finding is likely related to the electoral incentives proposed by Powell (1986): the lower the district magnitude, the more likely parties are to abandon some districts as hopeless, thus failing to mobilize their supporters to the polls. By dropping Powell's ordinal measure of this variable - which Pérez- Liñán (2001, 286) correctly described as "a very crude measure of district size" - and using the interval measure of average district magnitude, we were able to improve on the unexpected findings of Pérez-Liñán (2001) and Fornos, Power, and Garand (2004) for Latin America (both of these studies found nationally competitive districts to have the "wrong sign" when they employed the Powell measure). Using average district magnitude, our Latin American results are statistically significant and have the positive effect on turnout that Powell expected.
In terms of socioeconomic predictors, our Latin American results support the finding by Fornos, Power, and Garand (2004) that social and demographic variables are overwhelmed by institutional variables in explaining turnout. In Eastern Europe, per capita GDP has a positive and significant effect on voter mobilization, in full accordance with traditional theories of political participation. However, we find that urbanization and literacy rates are negatively associated with electoral participation in post-Communist elections, contrary to our expectations.16 What could explain this puzzle? It could be that citizens in rural areas, where local ex-Communist nomenklatura often retain the greatest strength, are under relatively higher pressure to vote. There is ample correlational evidence that Communist successor parties have found it easier to mobilize voters in rural provinces (Troxel 1992; Crawford 1996, 166-67), but the implications of ex-Communist mobilization await further analysis in appropriate multivariate models.
In terms of political factors, we find that "closeness of elections" is a wash - everywhere, all the time. This contrasts with the recent results of Franklin (2004, 37), who found that larger margins of victory are associated with significant drops in voter participation. Our different results may be due the fact that Franklin examined turnout only in advanced industrial democracies, with much more stable patterns of party competition, including a number of systems that correspond to the Westminster two-party model - wherein the electoral distance between party A and party B is both politically meaningful and easily understandable to ordinary voters. However, our sample is composed primarily of fluid, fragmented party systems in transitional democracies. Electoral closeness is not so easy to measure under conditions of volatile multipartyism, both for political scientists and for the electorate (Biais 2006, 1 19-20). Our use of the margin of victory of the first-placed party over the second-placed party does not capture any effects of closeness on voter participation, if indeed there are any in Eastern Europe and Latin America.
A somewhat more promising political factor is the level of political rights and civil liberties as measured by Freedom House. But the performance of this variable is not consistent: the level of democracy has a strong positive effect on turnout in Latin America and a strong negative effect in post-Communist Europe. The same finding occurs when we look at simple bivariate correlations between turnout and Freedom House rating across the two regions. This is because a number of post-Communist cases - most notably Albania, Croatia, and Romania - had high turnout rates while maintaining some restrictions on civil and political liberties in their early posttransition years. To a certain extent, their high turnout rates and low (i.e., less liberal) Freedom House ratings are actually two sides of the same coin: these countries had Communist successor parties or personalistic machines that crowded out other competitors in the early years of democracy. Thus, they may have earned a lower Freedom House rating precisely because their turnout rates were high (read: "suspiciously" high) in their early elections, thus accounting for the inverse relationship between electoral participation and our democracy variable here. More research needs to be done on this unexpected finding.
Finally, we now turn to dynamic factors, the central focus of our approach to Third Wave turnout. This article has found persuasive evidence that the "founding election" hypothesis (that turnout should soar in the inaugural election of a new democracy) and its corollary (that turnout should gradually decline thereafter) are both essentially correct for Latin America and Eastern Europe between 1978 and 2003. O'Donnell and Schmitter (1986) - offering their "tentative conclusions" when the Third Wave had barely commenced - were right. The counterclaim of International IDEA (2002), that there is no significant difference in turnout between founding elections and subsequent elections, is not supported by our analysis. The simple bivariate analysis of International IDEA misses the multidimensional nature of electoral participation. With a fully specified model controlling for various institutional, socioeconomic, and political factors, we are able to isolate the effect of time, which hitherto has been mostly ignored in studies of voter turnout.
Our research shows that in the new Third Wave democracies of Eastern Europe and Latin America, founding elections are dramatic contests of great historical import - but that voter disengagement from electoral participation begins almost immediately thereafter. We have found this effect to be very strong as far as the fourth election under democracy, where we stop our analysis. Despite the robust results presented here, we do not believe that this trend will continue in linear form far into the future, since we assume that somewhere ahead there is a natural "floor" for turnout - a point at which participation is unlikely to drop further. (For example, after several decades of decline, presidential turnout in the United States stabilized at just over 50 percent.) However, even just a medium-term continuation of turnout decline could be problematic for democratic legitimacy, especially given the many hopes and aspirations that ordinary citizens attached to the democratic transition. Thus, we urge researchers to replicate our approach in other regions of the world, extending the temporal dimension as much as possible, and crafting research designs that are sensitive to the complex, multifaceted nature of voter turnout.
Authors' Note: We would like to thank Geoffrey Evans, James C. Garand, Sarah Poggione, Laurence Whitehead, and the PRQ reviewers for their helpful comments and suggestions. We are also grateful to Dan Nielson for providing us with data.
1. This methodological shortcoming also affects the analysis by Turner (1993), who examined eleven parliamentary elections held in nine European countries and Japan. An additional problem is that Turner's definition of "postauthoritarian elections" is restricted only to what might be better labeled instances of nedemocratization - that is, cases where elections were resumed after an interruption of democracy. In this article we examine Third Wave founding elections regardless of whether the nascent regime is a case of redemocratization or first-time democratization.
2. See Appendix A for elections. Missing from the East European group are Serbia and Montenegro, Bosnia and Herzegovina, Georgia, Armenia, and Azerbaijan, all states that were in a state of ethnic war or civil unrest for most of the 1 99Os.
3. Our restriction of the analysis to Third Wave cases means that some older Latin American polyarchies, such as Costa Rica or Colombia (where democracy began in 1949 and 1958, respectively), are excluded from the data set. Moreover, our requirement of four electoral cycles means that some cases where democracy commenced late in the Third Wave (e.g., Mexico) are also excluded.
4. In a recent study of advanced industrial democracies, Franklin (2004) argued that the entry of new cohorts of voters has caused a decline in electoral participation after World War II. Our study, by way of contrast, focuses only on a very early phase of democracy in Latin America and Eastern Europe. Given that we examine only the first four electoral cycles under democracy, intergenerational population replacement is incomplete and therefore not relevant to our research design.
5. For a few recent elections (Chile 2001, Hungary 2002, Macedonia 2002, Paraguay 2003, and Ukraine 2002) where voting-age-based turnout rates are not available, we calculate total ballots cast as a percentage of the voting-age public in the previous election.
6. Biais (2006, 1 19) did not find persuasive the argument that increasing multipartyism increases information costs to citizens: "Voters do not have to inform themselves about each party." However, he did accept that information costs may rise when the overall party system is passing through a transitional or exceptionally inchoate phase. Such period effects are typical of postauthoritarian and especially posttotalitarian democracies, and they tend to coincide with the early electoral cycles we examine here.
7. We also estimated our models using economic performance (measured as year-on-year change in GDP) to test whether our sample of East European and Latin American cases supports Radcliff' s (1992) and Pacek and Radcliff's (1995) findings that economic downturns in developing countries foster participation while growth depresses it. This alternative operationalization, however, did not produce improvement in either the coefficients of the economy variable or the overall fit of the models.
8. See Appendix B, "Prior Democratic Experience." exceptionally, here we use Polity scores rather than Freedom House data because the latter are unavailable for years prior to 1973. In terms of their operationalization of democratic principles. Polity scores are more sensitive to constraints on the national executive whereas Freedom House scores are more sensitive to civil and political liberties. Because we see civil and political liberties as more directly linked to the electoral context, we use Freedom House data whenever possible.
9. Morris (2004) suggested that mandatory voting provisions in new democracies may not have the same effect as in consolidated democracies, but we feel it is necessary to account for their possible impact.
10. Two Latin American countries, Colombia and Nicaragua, use voluntary voting, but they are excluded from the analysis based on our case selection criteria (Colombia is not a Third Wave democracy and Nicaragua had not yet completed four democratic electoral cycles by 2003). We coded sanctions for nonvoting for the joint sample, but the resulting variable has a very skewed distribution since all East European cases take the value of zero. In the presence of very high level multicollinearity between region and mandatory voting (r = .897, significant at .0001), regression coefficient parameters became unstable. We opted not to enter both variables in the joint sample model, and to control for the effect of compulsory voting (i.e., sanctions for nonvoting; see Appendix B) in Latin America only.
1 1 . To explore the possible behavior of the determinants of turnout "as if compulsory voting existed in Eastern Europe, we ran separate regression tests on data that combine actual observations for Latin America with simulated values of turnout for the post-Communist elections. We generated two sets of simulations. First, we used Franklin's (2004) estimated magnitude of the effect of compulsory voting in established democracies (1 1.252) and applied it to Eastern Europe. However, because Norris (2004) argued that the introduction of mandatory voting is not as effective in transitional settings, for our second simulation we substituted an inflator that is half of the Franklin estimate (5.626). The regression results on the simulated data are consistent with what we report in Tables 2 and 3, with the only exception being that the magnitude of the regional dummy variable increases.
12. One possible source of parameter instability is that we pool together data from two different geographic regions, one of which (Latin America) uses compulsory voting, an institution absent in Eastern Europe. Our first step was to include a binary variable to control for these regional effects. Provisions for mandatory participation, however, may condition the effects of other factors and further shape voter turnout rates (Franklin 2001), thus confounding standard theoretical expectations. For example, voters who are not well educated will still turn out and cast a ballot to avoid sanctions for abstention; similarly, some voters may not be discouraged to show up and vote even though many parties run in the election and composition of the future cabinet is difficult to predict. An alternative strategy to improve parameter stability in the pooled data tests would be to add interaction terms and isolate the conditional effects of compulsory voting in Latin America. One caveat concerning this approach is that our Latin America dummy variable captures not only variation produced by compulsory voting regulation but other regional differences as well. Nevertheless, we tried this strategy and obtained results which are consistent with our main findings for the dynamics of voter disengagement in the tests on the two regional subsamples. These results are available and can be provided by the authors upon request.
13. We also estimated a specification where election sequence is modeled only through a dummy variable for founding election. The results provided support for the hypothesis that founding elections are different from other elections in producing much higher rates of participation in transitional countries. On the average, founding elections have brought 11.1 and 12.2 percent more voters to the polling stations in Latin America and Eastern Europe, respectively. This result is consistent with conclusions reached by previous research using a different sample of Latin American elections (Fornos, Power, and Garand 2004).
14. Among our ex-Communist cases, Poland was the only regime that developed some features approximating the authoritarian model. Interestingly, turnout rates in Poland have been considerably lower than elsewhere in the post-Communist world.
15. Our result is consistent with Norris's (2004, 166) specification that extreme party fragmentation depresses turnout Indeed, we find a negative effect of multipartism in Eastern Europe where the effective number of parties scores are much higher (mean = 4.13, maximum = 12.35) than in Latin America (mean = 3.49, maximum = 8.69).
16. We note that nearly all of our European cases have literacy rates above 97 percent, thus compressing the variance on this variable. The outstanding exception is Albania, which had literacy rates of 77 to 85 percent under democracy and a share of urban population of 36 to 43 percent. These are the lowest values for urbanization and literacy in the Eastern European subsample. However, repeating the analysis with a control variable for Albania generates results consistent with those discussed in the text.
Aldrich, John. 1993. Rational choice and turnout. American Journal of Political Science 37:246-78.
Almond, Gabriel A., and Sidney Verba. 1963. The civic culture: Political attitudes and democracy in five nations. Princeton, NJ: Princeton University Press.
Banks, Arthur S. Various years. Political handbook of the world (various issues). Washington: Congressional Quarterly Press.
Bartels, Larry M. 1996. Pooling disparate observations. American Journal of Political Science 40:905-42.
Beck, Nathaniel, and Jonathan N. Katz. 1995. What to do (and not to do) with time-series cross-section data in comparative politics. American Political Science Review 89:634-47.
Birch, Sarah, Francis Millard, Marina Popescu, and Kieran Williams. 2003. Embodying democracy: Electoral system design in post-communist Europe. London: Palgrave Macmillan.
Biais, André. 2000. To vote or not to vote: The merits and limits of rational choice theory. Pittsburgh, PA: University of Pittsburgh Press.
_____ . 2006. What affects voter turnout? Annual Review of Political Science 9: 1 1 1-25.
Cox, Gary, and Michael Munger. 1989. Closeness, expenditures, and turnout in the 1982 U.S. House elections. American Political Science Review 83:217-31.
Crawford, Keith. 1996. East Central European politics today. Manchester, UK: Manchester University Press.
Diamond, Larry. 1999. Developing democracy: Towards consolidation. Baltimore: Johns Hopkins University Press.
Downs, Anthony. 1957. An economic theory of democracy. New York: HarperCollins.
Filer, John E., Lawrence W. Kenny, and Rebecca B. Morton. 1993. Redistribution, income, and voting. American Journal of Political Science 37:63-87.
Fornos, Carolina A., Timothy J. Power, and James C. Garand. 2004. Explaining voter turnout in Latin America, 1980 to 2000. Comparative Political Studies 37 (8): 909-40.
Franklin, Mark N. 1996. Electoral participation. In Comparing democracies: Elections and voting in global perspective, ed. Lawrence LeDuc, Richard G. Niemi, and Pippa Morris. Thousand Oaks, CA: Sage.
_____ . 1999. Electoral engineering and cross-national turnout differences: What role for compulsory voting? British Journal of Political Science 29:205-24.
____ . 2001. How structural factors cause turnout variations at European Parliament elections. European Union Politics 2 (3): 309-28.
-. 2004. Voter turnout and the dynamics of electoral competition in established democracies since 1945. Cambridge: Cambridge University Press.
Franklin, Mark N., and Wolfgang P. Hirczy de Miño. 1998. Separated powers, divided government, and turnout in U.S. presidential elections. American Journal of Political Science 42 (1): 316-26.
Huntington, Samuel. 1991. The third wave: Democratisation in the late twentieth century. Norman: University of Oklahoma Press.
International Institute for Democracy and Electoral Assistance (IDEA). 2002. Voter turnout since 1945: A global repon. Stockhohn, Sweden: International IDEA, http://www.idea.int/vt/ survey.
Jackman, Robert W. 1987. Political institutions and voter turnout in the industrial democracies. American Political Science Review 8 1:405-423.
Jackman, Robert W, and Ross A. Miller. 1995. Voter turnout in the industrialized democracies during the 1980s. Comparative Political Studies 27:467-492.
Jones, Mark P. 1995. A guide to the electoral systems of the Americas. Electoral Studies 14:5-21.
Keesing 's record of world events (various years). Bethesda, MD: Keesings Worldwide.
Kopecky, Petr. 2003. Structures of representation: The new parliaments of Central and Eastern Europe. In Developments in Central and East European politics, vol. 3, ed. Stephen White, Judy Batt, and Paul G. Lewis. Durham, NC: Duke University Press.
Kostadinova, Tatiana. 2003. Voter turnout dynamics in post-communist Europe. European Journal of Political Research 42 (6): 741-59.
Laakso, Markku, and Rein Taagepera. 1979. Effective number of parties: A measure with application to Western Europe. Comparative Political Studies 12:13-27.
Lijphart, Arend. 1984. Democracies: Patterns of majoritarian and consensus government in twenty-one countries. New Haven, CT: Yale University Press.
Linz, Juan J., and Alfred Stepan. 1996. Problems of democratic transition and consolidation: Southern Europe, South America, and post-communist Europe. Baltimore: Johns Hopkins University Press.
Nielson, Daniel J. 2003. Supplying trade reform: Political institutions and trade policy in middle-income democracies. American Journal of Political Science 47 (3): 470-91.
Nohlen, Dieter, ed. 1993. Enciclopedia electoral latinomericana y del Caribe. San José, Costa Rica: Instituto Interamericano de Derechos Humanos.
Noms, Pippa. 2004. Electoral engineering: Voting rules and political behavior. Cambridge: Cambridge University Press.
O'Donnell, Guillermo, and Philippe C. Schmitter. 1986. Transitions from authoritarian rule: Tentative conclusions about uncertain democracies. Baltimore: Johns Hopkins University Press.
Olson, David M., and Philip Norton, eds. 1996. The new parliaments of Central and Eastern Europe. London: Frank Cass.
Pacek, Alexander, and Benjamin Radcliff. 1995. The political economy of competitive elections in the developing world. American Journal of Political Science 39 (3): 745-59.
Payne, Mark J., Daniel Zovatto, Fernando Carrillo-Flórez, and Andrés Allamand Zavala. 2003. La política importa: Democracia y desarrollo en América Latina. Washington: Inter-American Development Bank.
Pérez-Liñán, Aníbal. 2001. Neoinstitutional accounts of voter turnout: Moving beyond industrial democracies. Electoral Studies 20:281-97.
Powell, G. Bingham. 1986. American voter turnout in comparative perspective. American Political Science Review 80: 17-43.
Power, Timothy J., and J. Timmons Roberts. 1995. Compulsory voting, invalid ballots, and abstention in Brazil. Political Research Quarterly 48:795-826.
Przeworski, Adam. 1991. Democracy and the market: Political and economic reforms in Eastern Europe and Latin America. New York: Cambridge University Press.
Radcliff, Benjamin. 1992. The welfare state, turnout, and the economy. American Political Science Review 86:444-56.
Radcliff, Benjamin, and Patricia Davis. 2000. Labor organization and electoral participation in industrial democracies. American Journal of Political Science 44 (1): 132-41.
Rosenstone, Steven J., and John Mark Hansen. 1993. Mobilization, participation, and democracy in America. New York: Macmillan.
Sayrs, Lois W. 1989. Pooled time series analysis. Newbury Park, CA: Sage.
Share, Donald, and Scott Mainwaring. 1986. Transitions through transaction: Democratization in Brazil and Spain. In Political liberalization in Brazil, ed. Wayne Selcher, 175-215. Boulder, CO: Westview.
Smith, Peter H. 2005. Democracy in Latin America: Political change in comparative perspective. New York: Oxford University Press.
Troxel, Luan. 1992. Socialist persistence in the Bulgarian elections of 1990-1991. East European Quarterly 26 (4): 407-20.
Turner, Arthur. 1993. Postauthoritarian elections: Testing expectations about "first" elections. Comparative Political Studies 26:330-49.
Wolfinger, Raymond E., and Steven J. Rosenstone. 1980. Who votes? New Haven, CT: Yale University Press.
Florida International University, Miami
Timothy J. Power
University of Oxford, United Kingdom
Countries and Elections Used in this Study
Latin America (W = 48):
Argentina: 1983, 1985, 1987, 1989
Bolivia: 1980, 1985, 1989, 1993
Brazil: 1986, 1990, 1994, 1998
Chile: 1989, 1993, 1997, 2001
Dominican Republic: 1978, 1982, 1986, 1990
Ecuador: 1979, 1984, 1986, 1988
El Salvador: 1985, 1988, 1991, 1994
Guatemala: 1985, 1990, 1995, 1999
Honduras: 1985, 1989, 1993, 1997
Paraguay: 1989, 1993, 1998, 2003
Peru: 1980, 1985, 1990, 1992
Uruguay: 1984, 1989, 1994, 1999
Eastern Europe (N = 60):
Albania: 1991, 1992, 1996, 1997
Bulgaria: 1990, 1991, 1994, 1997
Croatia: 1990, 1992, 1995, 2000
Czech Republic: 1990, 1992, 1996, 1998
Estonia: 1990, 1992, 1995, 1999
Hungary: 1990, 1994, 1998, 2002
Latvia: 1990, 1993, 1995, 1998
Lithuania: 1990, 1992, 1996, 2000
Macedonia: 1990, 1994, 1998, 2002
Poland: 1989," 1991, 1993, 1997
Romania: 1990, 1992, 1996, 2000
Description of Variables and Sources
Turnout in Legislative Elections: Total votes cast (including blank and spoiled ballots) for the lower house, or for the unicameral legislature, as a percentage of the voting-age population. Sources: International Institute for Democracy and Electoral Assistance (IDEA; 1997 and www.idea.int), supplemented by data from the Political Database of the Americas (www.georgetown.edu/pdba), from the PARLINE database of the Interparliamentary Union (www.ipu.org), and from the Elections Around the World Web site (www.electionworld.org).
District Magnitude: Average district magnitude of the electoral districts used in lower house elections, logged. For mixed electoral systems, we use the average district magnitude on the proportional side. Sources: data set supplied by Nielson (2003) and supplemented by data in Payne et al. (2003).
Electoral Disproportionality: Distortion of representation caused by translation of votes into seats. Following Lijphart (1984), measured as the average vote-seat share deviation of the two largest parties in each election. Sources: same as turnout sources plus Birch et al. (2003), Banks (various years), and Keesing's Record of World Events (various years).
Multipartyism: Laakso and Taagepera (1979) effective number of parties in lower house or unicameral legislature. Sources: calculated from sources used for previous variable.
Unicorne ralism: Degree to which the national legislature approaches the unicameral model. Scoring follows Lijphart (1984). Countries with unicameral legislatures receive a score of 4, countries whose chambers are congruent and assymetrical in a manner that favors the lower house receive a score of 3, countries with incongruent bicameralism receive a score of 2, countries with bicameral legislatures receive a score of 1, and countries with strong bicameralism receive a score of O. Sources: coded by authors based on Nohlen (1993), Jones (1995), Olson and Norton (1996), and Kopecky (2003).
Concurrent Elections: Dummy variable, scored as 1 when presidential and legislative elections are held on the same date. Sources: same as for turnout.
Urbanization: Percentage of the total population living in the urban areas. Source: World Bank, World Development Indicators Online, http://www.worldbank.org/data/wdi.
Uteracy: Percentage of the population aged fifteen years and older that can read and write. Source: World Bank, World Development Indicators Online, supplemented by data from United Nations, Human Development Report (various years), http://hdr.undp.org.
Per Capita GDP: Per capita gross domestic product, measured at purchasing power parity (PPP) using the 1995 international prices. Source: World Bank, World Development Indicators Online.
Freedom House Rating: Freedom House employs separate political rights and civil liberties indexes, both of which are measured from 1 (high freedom) to 7 (low freedom). We summed these into a composite index representing the level of democracy and receded them so that higher scores reflect more democratic environments. The composite scale ranges from 2 (least democratic) to 14 (most democratic). Source: www.freedomhouse.org/ratings.
Election Sequence: For each country, the first parliamentary election associated with the country's Third Wave transition to democracy is scored as a 1, the second election as 2, the third election as 3, and the fourth election as 4. Source: coded by authors.
Prior Democratic Experience: Experience with democratic governance in the fifty years preceding the first election in our data set. This is measured as the sum of the annualized observations for the Revised Combined Polity Score (variable name POLITY2) as coded in the Polity IV data set. Higher scores are associated with stronger democratic experience. In post-Communist Europe, the Czech Republic and Slovakia share a single score for the democratic legacy of the former Czechoslovakia; Croatia, Macedonia, and Slovenia share the score of the former Yugoslavia; and Estonia, Latvia, Lithuania, Russia, and Ukraine share a score for the former USSR. Source: Polity IV Project, available http://www.cidcm.umd.edu/
Sanctions for Nonvoting: Degree to which citizen appearance at the polls is compelled by national legislation (in Latin America). Countries that have a compulsory voting statute, but which have no sanctions against nonvoters written into law, receive a score of 1 (Dominican Republic, El Salvador, Guatemala); compulsory systems possessing such legal sanctions but leaving them generally unenforced receive a score of 2 (Argentina, Bolivia, Brazil, Honduras, Paraguay); and compulsory systems with legal sanctions that are enforced in practice are given the highest value of 3 (Chile, Ecuador, Peru, and Uruguay). Source: Payne et al. (2003, Table 3.2).
Electoral Competition: Electoral competition is measured as the difference in vote shares between the first- and second-placed parties, subtracted from 100. Higher values represent closer elections. Sources: same as for turnout and disproportionality.…