Disenchanted or Discerning: Voter Turnout in Post-Communist Countries
Abstract
Voter turnout in post-communist countries has exhibited wildly fluctuating patterns against a backdrop of economic and political volatility. In this article, we consider three explanations for this variation: a “depressing disenchantment” hypothesis that predicts voters are less likely to vote in elections when political and economic conditions are worse; a “motivating disenchantment” hypothesis that predicts voters are more likely to vote in elections when conditions are worse; and a “stakes” based hypothesis that predicts voters are more likely to vote in more important elections. Using an original aggregate-level cross-national time-series data set of 137 presidential and parliamentary elections in 19 post-communist countries, we find much stronger empirical support for the stakes-based approach to explaining variation in voter turnout than we do for either of the disenchantment-based approaches. Our findings offer a theoretically integrated picture of voter participation in the post-communist world, and, more broadly, contribute new insights to the general literature on turnout.
In the first decade and a half of after the collapse of communism, observers noted with concern the apparent dramatic decline of voter turnout in post-communist countries. From initial rates of 80% and higher in the first wave of open and competitive elections, average turnout rates in some countries have fallen to below 50% in recent years (Bernhagen and Marsh 2007). Popular explanations have focused on the role of disenchantment as a result of continued social and economic hardship, corruption, and a sense of exclusion as communist-era elites reconsolidated political and economic power (Krastev 2002; Kobach 2001; Mason 2003/04).
Nevertheless, tremendous variation in voter participation across both space and time can be found in the post-communist world. Turnout as low as 43% can be found in 1991 (Poland) and 2004 (Slovakia); turnout over 85% can be found in 1990 (Romania) and 2004 (Georgia). We propose an integrated explanation for this variation. Our principal finding is that rather than being driven by disenchantment, patterns of voter participation in post-communist countries are largely a function of what is at stake in a given election. Specifically, turnout is higher in elections for more important institutions, when countries face fewer external constraints on policymaking, and when the long-term political future of the country is more questionable. To provide empirical support for our argument, we rely on a data set that covers the broadest possible set of competitive elections in post-communist countries, encompassing 137 national elections from 19 countries over a 15-year period.1
We begin by elaborating on our theoretical perspectives and presenting testable hypotheses. We then briefly discuss our data and statistical methods before presenting our empirical evidence. We conclude by assessing the implications of these findings for our understanding of political behavior in post-communist countries specifically, as well as the more general turnout literature.
Disenchantment versus Electoral Stakes
The received wisdom regarding electoral participation in post-communist countries is that declining turnout necessarily follows citizen disenchantment. Mason writes that “citizens of the post-communist states in Eastern Europe and the former Soviet Union remain discontented, dissatisfied with the economy, and cynical about politics, and are increasingly staying away from the polls on election day” (2003/04, 48). The legacy of sudden and persistent double-digit unemployment, rising prices, and increasingly unaffordable goods and services along with the state's retreat from the provision of social services are all offered as reasons for voter disenchantment (Bell 2001, 45; Mason 2003/04, 48–49; Pacek 1994; Tworzecki 2003, 168–69). Scholars also cite rising disillusionment and falling levels of political efficacy as another set of reasons why post-communist citizens are staying away from the polls; post-communist politics is, for many, marked by a sense of exclusion, corruption, and indeed outright criminality in some cases (Hutcheson 2004; Kostadinova 2003; White and McAllister 2004).
Of course, the more general literature on social mobilization suggests two alternatives for citizens disenchanted with their governments, stated must famously by Albert Hirshman (1970) as Exit and voice. The broad literature on turnout also reflects this divide, with scholars arguing for voter “withdrawal” due to adversity (Rosenstone 1982), “mobilization” due to adversity (Nannestad and Paldam 1997; Southwell 1988), or both, depending on levels of development (Radcliff 1992). Therefore, even though most analyses of turnout in post-communist countries have focused on disenchantment leading to exit (e.g., lower turnout), it is possible that disenchantment could have the opposite effect of leading to greater voice (e.g., higher turnout) in the context of post-communist transitions. For ease of interpretation, we will refer to these two approaches for the rest of the article as “depressing disenchantment” (where disenchantment leads to lower to turnout) and “mobilizing disenchantment” (where disenchantment leads to higher turnout).
In studies of turnout in more established democracies, however, the question has often been approached from a different perspective, with scholars considering voter turnout as a cost-benefit analysis to be resolved by individuals in a decision or game-theoretic manner.2 Here, the decision to vote is a function of whether a voter believes the costs of voting outweigh the perceived benefit of her candidate winning the election, weighted by the likelihood that her vote will actually have an effect (be “pivotal”) on whether or not her preferred candidate is victorious.3 While debate continues as to whether there can ever be a satisfactory decision theoretic explanation for why people vote given the generally low likelihood of ever being pivotal in any given election (e.g. Franklin 2004; Friedman 1996; Green and Shapiro 1994; Grofman 1993; Morton 2006), a central insight from this approach is a useful counterpoint to the standard disenchantment story in transition countries: post-communist voters may be more predisposed to participate in elections when they care more about the outcomes.4 If, as Aldrich (1993) has argued, the voting decision is a low-cost, low-expected benefit type of decision, then changes in the stakes of an election might be enough to tip significant numbers of voters into either staying home or casting a ballot.
In the remainder of this section we develop a number of testable hypotheses that stem from the stakes-based and disenchantment-based approaches to predicting variation in aggregate turnout levels in post-communist countries. Given that plausible arguments can be made for disenchantment increasing or decreasing turnout levels, we will posit hypotheses in both directions for variables that we believe are good proxies for disenchantment. Support for either version of the disenchantment approach would of course be seen as falsifying the other (e.g., if we find disenchantment systematically increases turnout, that would falsify the argument that disenchantment decreases turnout), but we will also look for an underlying null-hypothesis that would falsify both disenchantment hypothesis. This would either take the form of variables relating to disenchantment having no systematic effect on turnout, or else an uneven pattern of some variables having a positive effect and some having a negative effect.5
Disenchantment
The standard way to think about disenchantment in post-communist countries is in terms of economic developments. Here we consider three different ways to identify the conditions under which citizens are likely to be more disenchanted with the state of the economy: (1) macroeconomic conditions in the time leading up to a given election; (2) overall levels of economic development and/or quality of life in a given country at the time of an election; and (3) changes in economic conditions since the start of the transition.6
However, it is possible that in transition countries political disenchantment could be just as serious a problem. One potential indicator is the level of democracy in a country, with the assumption that flawed or partial democracy would lead to greater disenchantment. Another form of political disappointment could stem from a country's relationship to European expansion. If the European Union functions as the ultimate guarantor of democratic consolidation, then we might posit that citizens of countries that are farther from EU membership might be more disenchanted with political developments.7
Thus, the depressing disenchantment approach predicts lower turnout when current economic conditions are worse, people are worse off, the economy has not improved as much since the start of the transitions, democracy is less advanced, and EU membership is less likely. Conversely, the mobilizing disenchantment approach would predict higher turnout in view of the same factors (see Table 1).
Electoral Stakes
Our alternative framework suggests that post-communist citizens ought to participate in greater numbers in elections where there is more at stake.8 While there are numerous ways to observe variability in the stakes of different elections, we focus on a number of theoretically motivated hypotheses that are suitable for systematic comparisons across space and time: institutional arrangements; international constraints on policymaking; the ethnic composition of society; the effect of democratization; and overall levels of economic wealth.
Prior research on turnout has focused a great deal of attention on the question of variation in turnout across elections for different types of institutions (e.g., executive vs. legislative elections, elections in presidential vs. parliamentary systems).9 From our perspective, the question of whether a given election is for a parliament or president is much less important than the combinations of governing system and election type. So rather than hypothesize that presidential or parliamentary elections ought to have higher turnout, we instead expect to see higher turnout in elections for “dominant” institutions (presidents in presidential systems, parliaments in parliamentary systems) than in elections for “dominated” institutions (parliaments in presidential systems and presidents in parliamentary systems).10
Another way in which the stakes of a given election could be lowered would be if international agreements significantly reduced the policy making options of any incoming government. Perhaps most importantly, joining the EU requires the adoption of a significant amount of European law, which restricts options available to national level governments (Vachudova 2005). Another international constraint on domestic policymaking in many post-communist countries was participation in International Monetary Fund (IMF) programs in return for access to IMF funding and policy advice (Pop-Eleches 2009; Stone 2002).
We might also expect ethnic heterogeneity to increase the stakes of any election. To the extent that losing elections in ethnically divided societies can lead to more permanent shifts in the balance of power, we would expect elections in ethnically heterogeneous countries, ceteris paribus, to be seen as more important in the eyes of voters than elections in more ethnically homogenous countries.
Additionally, a stakes-based approach suggests that turnout should be higher in elections following periods of democratization. To the extent that democratization makes politics more accessible to more citizens, we could expect more of them to exercise their right to vote. Simultaneously, democratization increases the chances that more entrenched actors could lose positions of prominence, thus raising the stakes for supporters of the status quo as well.
Finally, one could also argue that the stakes of any election are higher the poorer a country is. While citizens in wealthy countries certainly care about elections, stakes are magnified in less developed countries where the resources to mitigate against the hardships of economic downturns are minimal. As Colton succinctly notes, “Office seekers in the quiescent West fuss over…whether to add pennies to the gasoline tax. In Russia the battle is about graver and more incendiary concerns—dysfunctional and insolvent institutions, individual freedom, nationhood, property rights, provision of the basic necessities of life in an economic downswing…” (2000, viii). Furthermore, the idea that economic and political success may breed a certain amount of complacency is reinforced by Roller et al. (2005), who report that in a survey administered from 1998 to 2001 in 13 of the ex-communist countries in our sample almost two thirds of respondents agreed with the statement that “As long as things are getting on well, I'm not really interested in who is in power.”
Table 1 concisely summarizes these hypotheses.
Data and Methods
To test these hypotheses empirically, we constructed a dataset consisting of 137 parliamentary and presidential elections in 19 former communist countries over a 15-year period.11 We only excluded elections in countries which, after achieving independence, did not have at least three elections in years rated by Freedom House as either “Free” or “Partially Free.” In doing so we avoid drawing inferences from elections that were so clearly flawed in the eyes of outside observers so as to cast doubt not only on the fairness of the election but on the very credibility of basic electoral statistics, such as turnout figures. This criterion excluded elections in the Central Asian former Soviet republics, Azerbaijan, and Belarus, which had no or too few reasonably free elections in the post-communist era. We also excluded elections in Bosnia, Serbia, and Montenegro, because the former is still largely governed as an international protectorate, whereas the latter two only experienced democratic elections after the fall of Milošević in 2000, and did not gain formal independence until 2006.12
The unit of analysis is the individual election. We include in our data set all elections for national office in which the whole electorate had the opportunity to vote: any first-round parliamentary and any first- or second-round presidential election.13 Concurrent parliamentary and presidential elections are treated as a single election but different rounds of presidential elections are counted as two elections, since turnout can (and often does) differ significantly. The dependent variable in all of the analyses is turnout as a percentage of registered voters.14
There are only four variables in the analysis for which we are missing data.15 Rather than list-wise delete observations in a dataset with an N of 137, we employ the following strategy to deal with missing data.16 We replace all instances of missing data in a given variable with a constant value (we use 0), and we simultaneously create a dummy variable identifying all of the cases where we have made this replacement. By including this dummy variable in any analysis using the variable in question, we can interpret the coefficient for the original variable as the substantive effect of the variable for the cases for which we actually have data.17 For clarity of presentation, we omit the coefficients and standard errors of the dummy variables identifying cases of missing data from the presentation of our results. A series of dummy variables identifying the electoral sequence number are also included in all regressions and are similarly omitted from the tables presented in the text.
Considering that our dataset is a cross-sectionally dominated panel and given the presence of serial auto-correlation and panel heteroskedasticity, we ran a series of Prais-Winsten regression models with heteroskedastic panels corrected standard errors.18 Since we were interested in both the cross-country and the within-country effect of different variables, the statistical tables presented in the text of the article are all random-effects models (but the key results also hold using fixed effects models).19 We also subject the data to a wide variety of robustness checks, the results of which are presented in the appendix. These include varying the specification of the model (Table A1) and the composition of the sample (Table A2), adding a series of additional institutional control variables (Table A3), and rerunning the analysis excluding each individual country in turn (Table A4). While there are of course differences across these tests, our principal findings are quite consistent.
Empirical Analysis and Discussion
The Stakes-Based Approach
We begin our analysis by considering the empirical evidence at the heart of the stakes-based argument: do more citizens participate in elections for more important institutions? Table 2 presents a simple cross-tab that breaks our data down by the type of election and governing system.20
The results confirm our predictions. In both presidential systems and parliamentary systems, average turnout is approximately 12% higher for dominant elections than for dominated elections. Moreover, in semi-presidential systems turnout was similar for both parliamentary elections, in line with the balance of power between parliaments and presidents.
In Table 3, we reexamine this question in a multivariate framework. Models 1 and 2 introduce the type of election and type of governing system separately (with presidential elections as the reference variable in the former and presidential systems as the reference variable in the latter). None of these variables achieve statistical significance. However, when we interact these variables to identify high- and low-importance elections in the manner predicted by the stakes-based approach, we find the same results suggested by Table 2. High-importance elections (parliamentary elections in parliamentary systems, presidential elections in presidential systems, and simultaneous elections in mixed systems) have significantly higher turnout than low-importance elections (parliamentary elections in presidential systems or presidential elections in parliamentary systems), the omitted category in the analysis.21 While the magnitude of the effect is lower after controlling for other factors, the difference remains both substantively and statistically significant. Indeed, even after controlling for other factors, a high-importance election is still predicted to lead to more than 8% higher turnout than a low-importance election. Model 4 demonstrates that similar effects are present when we fully specify all of the possible interaction effects between type of election and type of governing system instead of relying on the more general high-importance election variable.22 Judging by the conditional effects in Model 4, the difference between high- and low-importance elections was larger in presidential systems, where presidential elections had a 12.2% higher predicted turnout than parliamentary elections (significant at .001), whereas in parliamentary systems predicted turnout was only 4.8% higher in the institutionally more important parliamentary elections (marginally significant at .1 one-tailed). As the R-squared values of models 3 and 4 are practically identical, we use the simpler and more intuitive high-importance election variable for the remainder of the article.
Microlevel support for these macrolevel findings about the salience of institutional importance can be found in Pammett's (1998) account of the 1995–96 Russian election cycle. By our coding, as well as by all conventional accounts, Russia is a strong presidential system (Fish 2000). In Pammett's study, respondents were asked to evaluate how important the parliamentary elections of December 1995 and the presidential elections of 1996 would be for Russian democracy. For the parliamentary elections 34% of respondents thought they would definitely or probably be important for Russian democracy, while for presidential elections the share was 55%. This observation is certainly consistent with the claim that at least one reason turnout was higher in the 1996 presidential elections than in the 1995 parliamentary elections was that Russian voters were aware of the greater institutional importance of the presidency in Russia.
Pushing on this point a bit further, we might suspect that if the difference in turnout between high and low importance elections truly was driven by voters’ understanding of the difference in importance between different institutions, the effect should increase over time as voters acquire more evidence as to whether the presidency or the parliament is in fact the dominant institution. In Figure 1, therefore, we return to the raw data and plot the average turnout by the importance of the election over time. The pattern is very clear: while turnout declines across both high-importance and low/medium-importance elections, the magnitude at which it drops off is significantly larger for the latter.23 Put another way, as the transition progressed, post-communist citizens were likely learning that it was less important to participate in low/medium-importance elections than in high-importance elections. Over-time turnout change by election importance
Taken together, Tables 1 and 2 and Figure 1 present strong empirical support for the institutional component of the stakes-based approach. More post-communist citizens turn out to vote in elections for high-importance electoral institutions than for other electoral institutions, and this trend increases over time. Furthermore, Table A1–A4 in the appendix demonstrates that these institutional effects are robust across a wide variety of specifications of model and sample.
Table 4 presents further tests of stakes-based hypotheses. Model 3 of Table 4 directly tests the external constraints argument that turnout ought to be lower in countries with more time remaining in IMF programs and in countries more likely to join the EU. We find this to be the case. To test the effect of likelihood of EU membership, we divided countries into three “tiers.” First-tier countries are those that had strong reason to assume they would be admitted to the EU in the first-expansion wave (or, in 2004, have already been admitted). Third-tier countries are those that have minimal chances of joining the EU in the foreseeable future. Second-tier countries are those that are somewhere in between: they may have a chance of joining, but it is unclear how realistic that chance is and when it might occur.24 Model 3 of Table 4 reveals that, as predicted, turnout is lower in the first-tier likely admits to the European Union than in either the second- or third-tier EU countries.
One might argue, however, that first- and second-tier EU status has a different meaning at different points in time. In 1992, for example, first-tier EU status meant that a country was likely to join the EU at some distant point in the future, whereas in 2003 first-tier status carried with it very real demands in terms of adopting EU requirements as national law. In Figure 2, therefore, turnout is plotted by EU tier over time.25 Over-time turnout change by EU integration tier
For the period of time in which EU membership was a conceivable but distant goal (1990–96), there is little relationship between EU tier and turnout. However, as accession to the EU becomes more imminent, turnout drops much more substantially in first-tier countries. Indeed, although the linear relationship identified in Model 3 of Table 4 suggested that turnout in first-tier countries was likely to be about 5% lower than in third-tier countries, by the 2000–04 period, the gap is actually almost three times as large: elections in first-tier EU countries in this period on average have more than 14% lower turnout than elections in third-tier countries. And while on average turnout in second-tier countries is statistically indistinguishable from turnout in third-tier countries, a closer look at the temporal evolution reveals a pattern consistent with the stakes-based theory. During 1997–2000, the time period which was decisive for the European integration prospects of many second-tier countries, these same countries experienced a rebound in turnout compared to the 1993–96 period. However, once this objective was achieved (or at least was within reach) turnout in second-tier countries dropped between 2001 and 2004, starting to resemble the turnout patterns in the first-tier integration candidates. These findings overlap nicely with the anecdotal evidence from the Bulgarian elections of 1997 and the Slovak elections of 1998, which were widely regarded as the last opportunity to “catch the train to Europe.”
Returning to Table 4, we can also assess the empirical support for three additional stakes-based implications. First, Model 4 reveals that, as predicted, turnout is higher following periods of democratization.26 Second, as predicted by the stakes-based approach, we see clear evidence that overall development—as measured by the United Nations Development Program's Human Development Index (HDI)27—coincides with lower turnout. Finally, we note that the coefficient for ethno-linguistic fractionalization is positive, as expected. The standard errors, however, are sufficiently large in many of the models that we do not have strong confidence in this effect; this pattern is present throughout the robustness tests as well (see Tables A1–A4). It is worth noting, though, that HDI is correlated with ethno-linguistic fractionalization: when we use other proxies for wealth (discussed in greater detail below), we often have much more confidence in the larger positive coefficients we find for ethno-linguistic fractionalization (see Table 5).
The Disenchantment Approach
The evidence related to the two disenchantment approaches is markedly less consistent and less supportive. Models 1 and 2 of Table 4 present empirical tests of the crux of the two disenchantment arguments: is turnout systematically lower or higher when economic conditions are worse? Recall that we posited three different ways to consider disappointing economic conditions: traditional measures of macro-economic performance; change in the state of the economy since the start of the transition; and overall wealth and development. The first of these categories—as illustrated in Model 1 of Table 4—results in a mixed set of findings. Turnout is lower in the presence of higher inflation and higher unemployment rates, in line with the depressing-disenchantment hypothesis.28 However, GDP growth has a statistically significant negative effect on turnout, in line with the mobilizing-disenchantment hypothesis.
Turning to GDP in the current year as a percentage of a country's GDP in 1989, we find a statistically insignificant effect (see below, Table 5, Model 5). In Model 2 of Table 4, however, we interact GDP as a percent of a country's GDP in 1989 with the number of years that have passed since the beginning of the transition and the results are striking: whereas during the early transition years turnout was higher (but statistically insignificant) in countries with less dramatic output declines, the trend was reversed as the transition progressed; starting in 2000 countries with more vigorous economic recoveries actually experienced significantly lower electoral turnout. So again, we end up with some evidence in favor of the depressing-disenchantment hypothesis (early on in the transition) and some evidence in favor of the mobilizing-disenchantment hypothesis (late in the decade).29
The final manner in which we consider the state of the economy concerns overall levels of wealth and development. As noted previously, we use HDI to tap into this question of overall standard of living. Of course, HDI is not the only variable one could use to tap into overall levels of development. In Table 5, therefore, we consider four other potential development proxies: the percentage of the country that lives in urban areas, GDP per capita, the number of phones per 1000 residents, and the number of TVs per 1000 residents; we also include GDP as a percentage of 1989 GDP in this table. All of these measures have various advantages and disadvantages—e.g., percentage urban varies little within countries, the number of TVs could be tapping into other politically relevant phenomena as well—but what is striking is that there is not a single case where we find the relationship predicted by the depressing-disenchantment theory. No matter what proxy for wealth we employ, we always find that higher degrees of development result in lower turnout, as predicted by the mobilizing-disenchantment (and stakes-based) hypotheses. In some of these cases the standard errors are sufficiently high so as to cast doubt on how strongly we ought to believe that this negative relationship holds, but clearly there is no evidence to support the positive relationship between wealth and turnout that is in accordance with the depressing-disenchantment framework.
Returning to Table 4, we can examine support for the final observable implications of the disenchantment hypotheses, the political disenchantment indicators of EU status and overall level of democratization. In terms of degree of democratization, the coefficient for lagged Freedom House democracy scores in Model 4 of Table 4 does not approach statistical significance, and thus offers no support for either disenchantment hypothesis.30 As noted earlier, all else being equal, third-tier EU countries have higher, not lower, turnout rates than first-tier countries, and this effect only increases in size as the actual accession of the first-tier countries to the EU approaches. This finding is in accordance with the mobilizing disenchantment hypothesis, but in opposition to the depressing-disenchantment hypothesis.
Across a variety of tests, therefore, we find contradictory evidence for the two disenchantment hypotheses. Certainly, there is a good deal of evidence that we should interpret as falsifying the depressing-disenchantment hypothesis. Lower GDP-growth led to higher, not lower, turnout. There is no evidence from any of our proxies for wealth that lower turnout is more prevalent in poorer countries. We also find no evidence of political disenchantment depressing electoral participation; neither lower Freedom House scores nor exclusion from the EU result in lower rates of turnout.31 While there is somewhat more support for the mobilizing disenchantment hypothesis, it is undermined by the contradictory evidence regarding current economic conditions and the fact that there is no relationship between levels of democratization and turnout. Moreover, two of the factors that provide support for the mobilizing-disenchantment hypothesis—the effects of the likelihood of EU membership and overall levels of wealth on turnout—are also predictions of the stakes-based approach.
In contrast to the two disenchantment hypotheses, the empirical support for the stakes-based hypothesis is consistently in the predicted direction across a wide variety of indicators. Most importantly, we are very confident that voters turn out in greater numbers for elections for more important institutions. But we can also conclude that turnout is higher when there are fewer international constraints on policymaking following the election, in periods of time following democratization, and when a country's long term democratic future is less assured. Indeed, the only variable used to test the hypothesis that does not consistently achieve conventional levels of statistical significance is ethnic heterogeneity, and even here the coefficient is always in the predicted direction (and is often significant when we remove HDI from the equation). In short, in the aggregate, post-communist voters are more likely to participate in elections when the stakes are higher.
Conclusion and Implications
The primary contribution of this article is to provide an explanation for aggregate-level variation in turnout in national elections in post-communist countries, which we have just summarized in the preceding paragraphs. However, the article also makes a number of other contributions to the academic literature on post-communist politics, as well as to the study of electoral participation more generally. Our argument that participation in an election is a function of electoral stakes is of course dependent on an information environment that allows voters to discern the policy implications of election results. The extant voting literature makes clear the importance of political information in shaping voter decisions in mature democracies (Fearon 1999; Zaller 1992). As Duch (2001) succinctly argues, however, the new post-communist polities are characterized by “novel information demands and opportunities associated with democracy… (citizens are) exposed to increasingly numerous, heterogeneous, and conflicting messages regarding both the economy and politics…they also face a much broader set of political choices for which this information is relevant” (897). In short, new and rapidly expanding information creates profound challenges for electorates unaccustomed to sifting through it. In time, however, such barriers are overcome as voters become better informed about critical issues. As successive elections provide more information about policy choices, voters become more acutely aware of what “matters” and what does not. Thus, in finding increasing stakes-based effects over time in terms of both electoral institutions and European integration (see Figures 1 and 2), we show the important role of learning in the development of post-communist political behavior.
Our analysis also helps to link previous unconnected findings on the importance of different sets of factors—institutional, economic, and developmental—on voter turnout. We have presented three such theoretical approaches here in the form of the stakes-based and the two disenchantment arguments. We have argued that these are particularly appropriate for the post-communist context, but they should be applicable to other contexts as well. Either way, we hope that this analysis illustrates the value of providing general theoretical perspectives on turnout that can encompass institutional, political, and economic variables, as opposed to merely including a common laundry-list of variables in turnout studies.
Finally, discussions of mass political behavior in transitional societies often assume unique and/or idiosyncratic patterns conditioned by those societies’ singular experiences in moving beyond their authoritarian pasts. Post-communist societies represent a particularly attractive opportunity to test for such uniqueness: the depth and the nature of the dual economic and political transition is an unambiguously singular event, as are the accompanying multiple traumas. Initial accounts of demoralized populations increasingly alienated from democratic politics fed this position and painted an especially bleak picture for the region. Our analysis suggests that this appraisal of post-communist turnout as a function of initially high optimism (and consequently high turnout) followed by disenchantment (and consequently lower turnout) may warrant reconsideration. What we have proposed in this study is a somewhat more complex view that brings together a variety of factors into a coherent explanation of voter participation patterns. Rather than “tuning out” with disappointment from the election process, portions of post-communist mass publics are learning to pick and choose which elections are worth their time and effort. Moreover, the greater decline in turnout in the region's reform frontrunners suggests that lower turnout may primarily be a signal that many of the tougher economic and political issues of the post-communist transition have been resolved. From this perspective, a more optimistic view of long-term democratic prospects in the region may be in order.
Future research will be needed to pursue some of the issues that have been raised here but are beyond the scope of our current analysis. One interesting question is why people are more likely to turn out. Two stories would be consistent with our cross-national findings. On the one hand, this could be part of a completely individual calculus about when it is or is not worth taking the time to go to the polls and vote. As discussed earlier, the admittedly anecdotal survey evidence from Russia in 1995 (Pammett 1999) does suggest that post-communist voters were quick to realize institutionally based differences in election stakes. On the other hand, judging by prior research on political participation in the United States (Rosenstone and Hansen 1993), it could also be the case that political parties and candidates make more of an effort to mobilize voters in more important elections. In the post-communist context, this interpretation would imply that it is political elites who learn when and where it is worth time and resources to mobilize the public. However, such an interpretation is somewhat undermined by the notorious weakness of East European political parties, and by the fact that the gradual increase in resources and campaigning experience of post-communist parties has actually coincided with an over time decline in turnout. Ultimately, case studies and multilevel analysis of survey data will be necessary to adjudicate between these two alternatives. Both of these approaches are beyond the purview of our current analysis but should be pursued by future research. Nonetheless, the current analysis has provided rather consistent evidence that post-communist citizens are very much attuned to the stakes in the electoral process even after the novelty of competitive elections has faded away. Indeed, after a decade and a half of competitive elections, post-communist voters seem to have become increasingly competent at navigating the turbulent waters of post-communist politics—even if at times that means staying away from the polls.
Acknowledgments
We thank Chris Achen, Larry Bartels, Robert Erickson, Achim Goerres, Sandy Gordon, Marty Gilens, Bernard Groffman, Kim Hill, John Jackson, Amaney Jamal, Rebecca Morton, Jason Lyall, Nolan McCarty, Benjamin Radcliff, Jessica Trounstine, and William Zimmerman for helpful comments and suggestions and Andrew Owen, David Rossbach, and Benjamin Freeman for excellent research assistance. All statistical analysis conducted using Stata 9.2.
Notes
Alexander C. Pacek is associate professor of political science, Texas A&M University, College Station, TX 77843.
Grigore Pop-Eleches is assistant professor of politics and international affairs, Princeton University, Princeton, NJ 08544.
Joshua A. Tucker is associate professor of politics, New York University, New York, NY 10012.
1 We describe the data and the case universe in greater detail below, but our principal findings are robust to including either more cases that did not quite make our cut-off or to restricting the sample to fewer countries by using a more demanding cutoff; see Appendix Table A2.
3 Given the generally low likelihood of ever being pivotal, subsequent work along this line has focused on whether voters gain ancillary benefits simply from the act of voting, such as a feeling of fulfilling one's civic duty (Aldrich 1993; Morton 2006).
4 There are of course other insights one can take from the cost-benefit approach to turnout. We could explore whether voters are more likely to participate in elections where they feel they have a greater chance of being pivotal (e.g., close elections), but this raises enormous methodological challenges in terms of aggregate-level measurement in multiparty systems and would almost certainly require the use of survey data. Alternatively, we could explore whether citizens are more likely to participate in elections where they have a larger variance in utility across different candidates or parties. Such an approach would also certainly require the use of survey data to test. Both approaches would represent interesting avenues for future research, but are beyond the scope of our current analysis.
5 One could argue that perhaps both effects are at work and null results indicate these patterns cancelling each other out, but that would leave us with a nonfalsifiable hypothesis by definition (e.g., positive effects, negative effects, and no effects would all be seen as confirming that disenchantment matters in one way or another) so we set aside this possibility for future research designed explicitly to examine this possibility.
6 There is little consensus in the turnout literature writ large on the role of economic factors. Some have found that economic hardships encourage people to enter the political world to redress grievances (Aguilar and Pacek 2000; Radcliff 1992; Schlozman and Verba 1979); while others have suggested that a sour economy encourages citizens to quit the political process to focus on less peripheral concerns (Caldeira, Patterson, and Markko 1985; Jesuit 2003; Radcliff 1992; Rosenstone 1982; Sniderman and Brody 1977; Wolfinger and Rosenstone 1980); still others have found no effect at all for economic downturns (Arcelus and Meltzer 1975; Blais and Dobrzynska 1998; Fiorina 1978; Fornos, Power, and Garand 2004; Lehoucq and Wall 2004). For better or worse, the scant literature devoted to voter participation in the post-communist world is a microcosm of the larger field of turnout research, in that it is characterized by inconsistency more than anything else. Pacek's (1994) aggregate-level analysis of four post-communist elections found that economic adversity depressed turnout, a finding also reported by Fowler (2004) in Hungary. Kostadinova's (2003) more detailed aggregate study showed no economic effect on turnout at all. Individual-level analyses report similarly mixed results, with Wyman and White (1995) finding little impact of economics on the 1993 Russian Duma election, and Bahry and Lipsmeyer (2001) finding positive evaluations of the economy actually decreased the likelihood of voting in the 1995 Russian Duma contest.
7 Another potential source of political disenchantment could be corruption (Cokgezen 2004; Hellman, Jones, and Kaufmann 2000; Tucker 2007). We do not include corruption in our list of observable implications of the disenchantment approach simply because we do not have adequate data to test this hypothesis across our entire sample. We did, however, test the effect of corruption on turnout across a limited subsample of the cases for which corruption data were available and found no relationship between levels of corruption and turnout.
8 Throughout the article, we refer to these interchangeably as stakes-based hypotheses, electoral stakes hypotheses, or electoral importance hypotheses.
9 See, for example, Powell 1986; Jackman 1987; Jackman and Miller 1995; Blais and Dobrzynska 1998; Perez-Linan 2001; Perea 2002; and Fornos, Power, and Garand 2004.
10 For more on “dominant” and “dominated” elections, see Tucker (2006).
11 Our database consists of the following elections. For legislative contests: Albania (1991, 1992, 1996, 1997, 2001), Armenia (1990, 1995, 1999, 2003), Bulgaria (1990, 1991, 1994, 1997, 2001), Croatia (1990, 1992, 1995, 2000, 2003), the Czech Republic (1990, 1992, 1996, 1998, 2002), Estonia (1990, 1992, 1995, 1999, 2003), Georgia (1990, 1995, 1999, 2004), Hungary (1990, 1994, 1998, 2002), Latvia (1990, 1993, 1995, 1998, 2002), Lithuania (1990, 1992, 1996, 2000, 2004), Macedonia (1990, 1994, 1998, 2002), Moldova (1990, 1994, 1998, 2001), Mongolia (1990, 1992, 1996, 2000), Poland (1989, 1991, 1993, 1997, 2001), Romania (1990, 1992, 1996, 2000, 2004), Russia (1990, 1993, 1995, 1999, 2003), Slovakia (1990, 1992, 1994, 1998, 2002), Slovenia (1990, 1992, 1996, 2000, 2004), Ukraine (1990, 1994, 1998, 2002). For presidential contests: Armenia (1996, 2003), Bulgaria (1992, 1996, 2001), Croatia (1992, 1997, 2000), Georgia (1995, 2000, 2004), Lithuania (1993, 1997, 2002–2003), Macedonia (1994, 1999), Moldova (1991, 1996), Mongolia (1993, 1997, 2001), Poland (1991, 1995, 2000), Romania (1990, 1992, 1996, 2000, 2004), Russia (1991, 1996, 2000, 2004), Slovakia (1999, 2004), Slovenia (1990, 1990, 1997, 2002), Ukraine (1991, 1994, 1999, 2004).
12 The most difficult decision was to exclude the Serbian and Montenegrin elections, as both countries have had at least three reasonably free elections since 2000. We ultimately decided against including them to avoid the complication of having one time series that started 10 years later than the rest of the sample. Also, since Montenegro only declared independence in 2006, both Serbian and Montenegrin elections were actually subnational elections for the time period covered by our article and—prior to independence—and thus there are very limited economic and developmental statistics available separately for the two federal states. Nevertheless, we considered these elections to be an important robustness test: according to Model 4 of Appendix Table A2, the inclusion of these additional elections actually strengthens our primary finding regarding the importance of institutional arrangements and further underscores the lack of a systematic effect of economic conditions on turnout.
13 We exclude parliamentary runoffs in mixed or SMD systems, since voting occurs only in certain electoral districts in these cases. We also exclude referenda.
14 The main source for turnout statistics is the IDEA website http://www.idea.int/vt/index.cfm supplemented where necessary by national election statistics. Details on the coding of all variables can be found in Appendix Table A5.
15 All four are economic variables: human development indicators (7 missing); unemployment lagged one year (18 missing); inflation lagged one year (18 missing); and per-capita TVs (21 missing).
16 For dangers of list-wise deletion, see King et al. 2001. Despite these concerns, we reran the analysis using list-wise deletion and found almost no difference in our results; see Model 5 of Appendix Table A2.
17 We do not impute missing data because most of our missing data is from elections that occur very early in the transition period, and we fear that a data generating process based primarily on data from later in the transition period might not be appropriate for the early transition period.
18 The models presented were run using the xtpcse command in Stata 9.2.
19 Readers should also note that all models include a constant term, but given the fact that we are omitting numerous variables from the presentation of the regression tables, we also omit the constant term. Without the full set of coefficients, the constant is intuitively meaningless. We thank Larry Bartels for highlighting this point.
20 We code presidential, parliamentary, and semi-presidential systems on the basis of an updated version of the Frye, Hellman, and Tucker (2000) executive power index. If the president scores above a 10 in executive powers, the country is coded as a presidential system. If the president scores above a five but less than 11 and is popularly elected, the country is coded as a semi-presidential system. If the president is not popularly elected or scores below a six, the country is coded as a parliamentary system. We also code all Soviet republics as parliamentary systems through the August 1991 coup attempt, at which point they are all assigned their starting value based on the index. Mongolia, which is not included in the index, is coded as a semi-presidential system. Readers should note that these objective rules lead us to very similar coding decisions as found in Armingeon (2005).
21 We have included simultaneous presidential and parliamentary election in semi-presidential systems as high importance elections on the grounds that if you are electing both your parliament and your president, that is just as important as only a president or parliament in a presidential or parliamentary system.
22 The omitted category is presidential election in a presidential system. Readers should note that for the purpose of properly specifying these interaction effects, the parliamentary election variables excludes parliamentary elections held simultaneously with presidential elections in presidential systems, which in practice means that one parliamentary election (the 1995 Georgian election) is not coded as a parliamentary election by this variable.
23 Predicted values from a statistical model that includes the relevant interaction terms between election sequence and election importance reveal an almost identical pattern.
24 More specifically, from 1990 to 1996, the Visegrad 4 (Poland, Hungary, the Czech Republic, and Slovakia) are coded as first-tier countries, the former Soviet Republics, Mongolia, Albania, and the former Yugoslavia (with the exception of Slovenia) are coded as third-tier countries, and the remaining countries are placed in the second tier. From 1997 to 2002, we include in the first tier any country that had opened negotiations with the EU for membership. Starting from 1997, that includes Estonia, Hungary, Poland, Slovenia, and the Czech Republic; in 2000 it expands to include Bulgaria, Latvia, Lithuania, Romania, and Slovakia. Former Soviet republics, Mongolia, and members of the former Yugoslavia (excluding Slovenia) are placed in the third tier for this entire period, and the remaining countries are placed in the second tier. From 2002 on, we place those countries that were going to be included in the first wave of expansion in the first tier (Czech Republic, Hungary, Estonia, Latvia, Lithuania, Poland, Slovakia, and Slovenia), former Soviet republics and Mongolia are in the third tier, and the remaining countries—now including the states of the former Yugoslavia (aside from Slovenia)—are in the second tier.
25 Predicted values from a more complex version of Model 3 in Table 4 that includes the relevant interaction effects reveal a very similar pattern.
26 This is measured by the change in Freedom House political and civil rights scores since the preceding year; see Appendix A5 for details. Given concerns with the quality of Freedom House data, we reran this test using Polity data (see Table A3, Model 6). Using Polity data introduced a host of methodological concerns as well, the most serious being that Polity scores often change around the date of an election; we also lost a few cases for which there were no Polity scores. As a result, we tried a variety of different ways of coding change in democracy at the time of an election using Polity scores. The bottom line was that no matter how we coded the variable, the coefficient on Change in Democratization was always in the correct direction, and always hovered around a p<=.10 (two tailed) or p<=.05 (one-tailed) significance level. The results we have included in Table A3 are indicative. As a result, we can at the very least conclude that the Polity data do not falsify our findings using the Freedom House data.
27 The measure, published by the UNDP in its annual World Development Report, measures human development achievements in terms of life expectancy, educational attainment, and adjusted real income.
28 In our models, we include economic conditions lagged by one year. Using current year economic conditions produces the same general pattern, although in most cases the standard errors are quite a bit larger.
29 For example, in 2004 a one-standard deviation (21%) increase in GDP as a percentage of 1989 was associated with a substantively large 5.9 % reduction in predicted turnout (significant at .01).
30 We come to the same conclusion using Polity scores; see Appendix Table A3, Model 6.
31 Similarly, as was mentioned in an earlier note, we did not find any relationship between levels of corruption and turnout in a subset of the data.
References
Aguilar, Edwin Eloy and Alexander Pacek. 2000. “Macroeconomic Conditions, Voter Turnout, and the Working Class/Economically Disadvantaged Party Vote in Developing Countries.” Comparative Political Studies 33 (8): 995–1017. Aldrich, John H. 1993. “Rational Choice and Turnout.” American Journal of Political Science 37 (1): 246–78. Arcelus, F, and A. Meltzer. 1975. “The Effect of Aggregate Economic Variables on Congressional Elections.” American Political Science Review 69 (4): 1232–65. Armingeon, Klaus. 2005. “Comparative Political Data Set II: 28 Post-Communist Countries.” Bern, Switzerland: Institut fur Politikwissenschaft-Universitat Bern. Bahry, Donna, and Christine Lipsmeyer. 2001. “Economic Adversity and Public Mobilization in Russia.” Electoral Studies 20 (3): 371–98. Bell, Janice. 2001. The Political Economy of Reform in Post-Communist Poland. Northampton, MA: E. Elgar. Bernhagen, Patrick, and Michael Marsh. 2007. “Voting and Protesting: Explaining Citizen Participation in Old and New European Democracies.” Democratisation 14 (1): 44–72. Blais, Andre, and Agnieszka Dobrzynska. 1998. “Turnout in Electoral Democracies.” European Journal of Political Research 33: 239–61. Brambor, Thomas, William Roberts Clark, and Matt Golder. 2006. “Understanding Interaction Models: Improving Empirical Analyses.” Political Analysis 14 (1): 63–82. Caldeira, Gregory, Samuel Patterson, and Gregory Markko. 1985. “Getting Out the Vote.” Journal of Politics 47 (3): 490–509. Cokgezen, Murat. 2004. "Corruption in Kyrgyzstan: The Facts, Causes, and Consequences." Central Asian Survey 23 (1): 79–94. Colton, Timothy. 2000. Transitional Citizens: Voters and What Influences Them in the New Russia. Cambridge, MA: Harvard University Press. Duch, Raymond. 2001. “A Developmental Model of Heterogeneous Economic Voting in New Democracies.” American Political Science Review 98 (4): 895–910. Fearon, James. 1999. “Electoral Accountability and the Control of Politicians: Selecting Good Types vs. Sanctioning Poor Performance.” In Democracy, Accountability, and Representation, eds. Bernard Manin, Adam Przeworski, and Susan Stokes. Cambridge: Cambridge University Press, 55–97. Fiorina, Morris P. 1978. “Economic Retrospective Voting in American National Elections: A Micro-Analysis.” American Journal of Political Science. 22 (2):426–43. Fish, M. Steven. 2000. “The Executive Deception: Superpresidentialism and the Degradation of Russian Politics.” In Building the Russian State: Institutional Crisis and the Quest for Democratic Governance, ed. V. Sperling. Boulder, CO: Westview Press, 177–92. Fornos, C., Timothy Power, and James Garand. 2004. “Explaining Voter Turnout in Latin America, 1980–2000.” Comparative Political Studies 37: 909–40. Fowler, B. 2004. “Hungary: Unpicking the Permissive Consensus.” West European Politics 27 (4): 624–51. Franklin, Mark. 2004. Voter Turnout and the Dynamics of Electoral Competition in Established Democracies since 1945. New York: Cambridge University Press. Friedman, Jeffrey. 1996. The Rational Choice Controversy: Economic Models of Politics Reconsidered. New Haven, CT: Yale University Press. Frye, Timothy, Joel Hellman, and Joshua Tucker. 2000. “Data Base on Political Institutions in the Post-Communist World.” The Ohio State University. Unpublished Data Set. Green, Donald P., and Ian Shapiro. 1994. Pathologies of Rational Choice Theory: A Critique of Applications in Political Science. New Haven, CT: Yale University Press Grofman, Bernard. 1993. “Is Turnout the Paradox that Ate Rational Choice Theory?” In Information, Participation, and Choice: An Economic Theory of Democracy in Perspective, ed. Bernard Grofman. Ann Arbor: University of Michigan Press, 93–105. Hellman, Joel S., Geraint Jones, and Daniel Kaufmann. 2000. “Seize the State, Seize the Day”: State Capture, Corruption, and Influence in Transition. The World Bank, World Bank Institute/Governance, Regulation and Finance Division and Europe and Central Asia Region/ Public Sector Group and European Bank of Reconstruction and Development/ Office of Chief Economist. Hirshman, Albert. 1970. Exit, Voice, and Loyalty: Response to Decline in Firms, Organizations, and States. Cambridge, MA: Harvard University Press. Hutcheson, Derek. 2004. “Protest and Disengagement in the Russian Federal Elections of 2003–04.” Perspectives on European Politics and Society 5 (2): 304–30. Jackman, Robert. 1987. “Political Institutions and Voter Turnout in the Industrial Democracies.” American Political Science Review 81 (2): 405–24. Jackman, Robert, and Ross Miller. 1995. “Voter Turnout in the Industrial Democracies during the 1980s.” Comparative Political Studies 27 (4): 467–92. Jesuit, David. 2003. “The Regional Dynamics of European Electoral Politics: Participation in National and European Contests in the 1990s.” European Union Politics 4 (2): 139–64. King, Gary, James Honaker, Anne Joseph, and Kenneth Scheve. 2001. “Analyzing Incomplete Political Science Data: An Alternative Algorithm for Multiple Imputation.” American Political Science Review 95 (1): 45–69. Kobach, Kris. 2001. “Lessons Learned in the Participation Game.” In Direct Democracy: The Eastern and Central European Experience, ed. Andreas Auer and Michael Butzer. Burlington, VT: Ashgate Publishing, 292–309. Kostadinova, Tatiana. 2003. “Voter Turnout Dynamics in Post-Communist Europe.” European Journal of Political Research 42: 741–59. Krastev, Ivan. 2002. “The Balkans: Democracy without Choices.” Journal of Democracy 13 (3): 39–53. Mason, David S. 2003/04. “Fairness Matters: Equity and the Transition to Democracy.” World Policy Journal 20 (4): 48–56. Morton, Rebecca B. 1991. “Groups in Rational Turnout Models.” American Journal of Political Science 35 (3): 758–76. Morton, Rebecca B. 2006. Analyzing Elections. 1st ed. New York: Norton. Nannestad, P. and M. Paldam. 1997. “The Grievance Asymmetry Revisited: A Micro Study of Economic Voting in Denmark.” European Journal of Political Economy 13 (1): 81–99. Pacek, Alexander. 1994. “Macroeconomic Conditions and Electoral Politics in East-Central Europe.” American Journal of Political Science 38 (3): 723–44. Pammett, Jon. 1999. “Elections and Democracy in Russia.” Communist and Post-Communist Studies 32: 45–60. Perea, E. 2002. “Individual Characteristics, Institutional Incentives, and Electoral Abstention in Western Europe.” European Journal of Political Research 41 (5): 643–73. Perez-Linan, A. 2001. “Neoinstitutional Accounts of Voter Turnout: Moving Beyond Industrial Democracies.” Electoral Studies 20 (2): 281–97. Pop-Eleches, Grigore. 2009. From Economic Crisis to Reform: IMF Programs in Latin America and Eastern Europe. Princeton, NJ: Princeton University Press. Powell, G. Bingham. 1986. “American Voter Turnout in Comparative Perspective.” American Political Science Review 80 (1): 17–43. Radcliff, Benjamin. 1992. “The Welfare State, Turnout, and the Economy: A Comparative Analysis.” American Political Science Review 86 (2): 444–56. Roller, Edeltraud, Dieter Fuchs, Hans-Dieter Klingemann, Bernhard Wessels, and Janos Simon. 2005. Consolidation of Democracy in Central and Eastern Europe. Cologne, Germany: Zentralarchiv ZA4054. Rosenstone, Steven. 1982. “Economic Adversity and Voter Turnout.” American Journal of Political Science 26 (1): 25–46. Rosenstone, Steven, and John Mark Hansen. 1993. Mobilization, Participation, and Democracy in America. New York: MacMillan Schlozman, Kay, and Sydney Verba. 1979. Injury to Insult. Cambridge, MA: Harvard University Press. Sniderman, Paul, and Richard Brody. 1977. “Coping: the Ethic of Self-Reliance.” American Journal of Political Science 21 (3): 501–23. Southwell, Priscilla. 1988. “The Mobilization Hypothesis and Voter Turnout in Congressional Elections.” Western Political Science Quarterly 41: 173–88. Stone, Randall W. 2002. Lending Credibility: The International Monetary Fund and the Post-Communist Transition. Princeton, NJ: Princeton University Press. Tucker, Joshua A. 2006. Regional Economic Voting: Russia, Poland, Hungary, Slovakia, and the Czech Republic, 1990–99. New York: Cambridge University Press. Tucker, Joshua A. 2007. “Enough! Electoral Fraud, Collective Action Problems, and Post-Communist Colored Revolutions.” Perspectives on Politics 5 (3): 537–53. Vachudova, Milada Anna. 2005. Europe Undivided: Democracy, Leverage, and Integration after Communism. New York: Oxford University Press. White, Stephen, and Ian McAllister. 2004. “Dimensions of Disengagement in Post-Communist Russia.” Journal of Communist Studies and Transition Politics 20 (1): 81–97. Wolfinger, Raymond, and Steven Rosenstone. 1980. Who Votes. New Haven, CT: Yale University Press. Wyman, Matthew, and Stephan White. 1995. “Public Opinion, Parties, and Voters in the December 1993 Russian Elections.” Europe-Asia Studies 47 (4): 195–228. Zaller, John. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press.
