IS COMPULSORY VOTING A REMEDY? EVIDENCE FROM THE 2001 POLISH PARLIAMENTARY ELECTIONS (draft version comments most welcome) Mikolaj Czesnik Institute of Political Studies, Polish Academy of Sciences, Warsaw and Institute of Political Science, Warsaw School of Social Psychology, Warsaw PAPER PREPARED FOR PRESENTATION AT THE ECPR JOINT SESSIONS OF WORKSHOPS, HELSINKI, MAY 7-12, 2007 PANEL 7 COMPULSORY VOTING: PRINCIPLES AND PRACTICE
Abstract Voter turnout has been decreasing in modern democracies. As high turnout is considered to be a desirable condition for democratic system, this decline can be perceived as a challenge to democracy. Many solutions are proposed and discussed. One of them is compulsory voting, which according to many scholars is effective, but at same time seems fairly controversial. Can it be a remedy for voter turnout decline? Would it work in new democracies? The main purpose of this paper is to study the effect of the (hypothetical) introduction of compulsory voting in Poland. More specifically, my aim is to answer the following questions: What would the introduction of compulsory voting change? Would it change election results? Would all the non-voters vote? If not, who would vote, and who would not? Polish National Election Study enables to investigate the influence of the (hypothetical) introduction of compulsory voting in Poland. The questionnaire of this study gives a unique opportunity to examine this issue in a more in-depth manner. Respondents (non-voters) in the survey were asked what they would do if voting were compulsory in Poland. Therefore the data enables to estimate: i) how would voter turnout change if voting were compulsory; ii) how would the results of elections change if voting were compulsory. The data also enables to examine two groups of non-voters: those, who are likely to transit from abstention to participation, and those, who are likely to remain non-voters. In the first section I refer to the question of significance of voter turnout whether it matters at all or not, which is, as I argue, crucial for studies on compulsory voting. In the second section an overview of voter turnout in CEE region is presented, while in the third section results of empirical analyses are discussed. The fourth section concludes the paper with discussion and final remarks. The hypotheses are tested on the Polish National Election Study data. 2
Introduction The main objective of this paper is to study compulsory voting in Poland. At first glance this venture might look quite risky and odd, since electoral participation in Poland is not (and never has been) mandatory. However, there are plausible reasons to discuss the issue: voter turnout in Poland is going down (and so it is in the whole CEE region) and various remedies, including compulsory voting, are discussed. Therefore, though Polish citizens are not (yet) forced to vote, the idea of introducing mandatory voting is more and more frequently discussed. As a consequence in this paper, instead of analyzing existing procedure, I rather tend to discuss a possible, but (yet) not existing phenomenon. More precisely, I aim to answer a following question: what would be the impact of introduction of compulsory voting in Poland? Does voter turnout at all matter? Each analysis of compulsory voting must be preceded by a more general discussion on importance and meaning of voter turnout in democracy. It is so due to the fact that our perspective on compulsory voting is strongly related to the fundamental question of whether (low) voter turnout at all matters or not. Hence it is inevitable to address this issue here. It is common to argue that low voter turnout is a threat to democracy. The very question whether (low) voter turnout matters or not is first of all theoretical. Indeed, the issue of citizens participation in activities of democratic community is central to theory of democracy. Two perspectives can be distinguished. Participationists claim that the most important element of the democratic rule is citizens participation in broadly defined governance. The ideal they refer to is Greek polis, where every member of demos could actually influence the fate of political community. According to this perspective, participation is the most crucial of democratic ideals and goals. Participationists claim that only through the citizens 3
participation the system may be truly democratic, and its citizens capable of self-fulfillment (Lively 1975: 134). In opposition to classical theory of democracy, which emphasizes the role of civic participation, emerges elitist theory of democracy, also referred to as the competitive theory of democracy (Sartori 1965: 126). Democracy is here reduced to a method: the democratic method is that institutional arrangement for arriving at political decisions in which individuals acquire the power to decide by means of a competitive struggle for the people s vote (Schumpeter 1957: 269). A constitutive element of the democratic system is not citizens participation, but competition among the elites: To simplify matters we have restricted the kind of competition for leadership which is to define democracy, to free competition for a free vote (Schumpeter 1957: 271). Consequently, two perspectives on the role of electoral participation in democracy, related to the two discussed concepts of democracy, are usually distinguished. One of them, related to the participatory theory of democracy, assumes that the democratic regime functions best when it is accompanied by a high level of civic involvement. A different approach indicates that low level of participation is an index of citizens satisfaction with the existing system: citizens do not vote, because democratic regime functions properly, it is not menaced, as neither are individual rights and liberties. Therefore there are no special incentives to participate in elections. I argue that the above distinction has an impact, often strongly underestimated or even neglected, on discussions about voter turnout and compulsory voting. Few enthusiasts of high voter turnout (among proponents of compulsory voting they constitute a gigantic majority) are brave enough to confess that their perspective on compulsory voting is somewhat blurred by the vision of democracy they have faith in. Unfortunately, these theoretical (or even ideological) foundations of our thinking are too often disregarded. 4
Meanwhile, our perspective on the importance of voter turnout, which is crucial for our evaluation of compulsory voting, should be also related to empirical evidence. It is true that wherever and whenever few people participate in the most basic democratic procedure, i.e. election, political scientists launch a debate on democratic deficit. It is also true that when electoral participation is in decline (which happens often in modern democracies) there is much ado about crisis of democracy. However, low voter turnout is not a threat to democracy per se. In fact, low voter turnout can matter in two instances: (i) firstly, when it overlaps with social and political inequalities; (ii) secondly, when it means de-legitimizing a regime (cf. McManus-Czubinska et al. 2004). Often these two situations take place simultaneously, which is not surprising since inequalities usually lead to de-legitimation of regime. But the only way to answer appropriately the question whether low voter turnout matters or not is to investigate it empirically. Indeed, since the very first studies on voting it is known that low voter turnout is associated with both unequal representation and de-legitimation of regime (c.f. Lijphart 1997). Several empirical studies show that non-voters are significantly different from voters with regard to many characteristics, including social backgrounds, ideological views, policy preferences, satisfaction with democracy etc. Even though most of these analyses pertain to Western democracies, it seems plausible to extrapolate their results to CEE countries, especially considering the fact that few analyses focusing on this region reveal fairly similar results. Consequently, in light of the arguments put forward above, it is easy to understand why many people in CEE countries are more and more concerned about lowering voter turnout in the region. Despite normative premises there are also convincing empirical arguments which imply thorough examination of the problem and justify increasing interest in finding its effective solution. Hence it seems that lowering voter turnout in CEE countries 5
really matters, and the whole process can be perceived as a serious threat to democracy. This is the very reason why reforms boosting voter turnout, including compulsory voting and arrangements alike, require both stimulating theoretical debate and in-depth empirical scrutiny. Voter turnout in Central Eastern Europe Before discussing necessity and prospects of introducing compulsory voting in CEE countries one must discuss voter turnout in the region. The very first thing is to present an overview of voter turnout in CEE countries. Figure 1 presents adequate data. Figure 1 about here Mean voter turnout in parliamentary 1 elections in post-communist period is presented. Quite striking is the fact that in this very small sample of countries there is a lot of variance. Mean voter turnout varies from 82% to 46%. On the one hand, quite high levels of voter turnout are observed in Czech Republic, Latvia, Slovakia and Slovenia. As far as voter turnout is concerned these countries are similar to Western democracies. On the other hand, Poland appears as a leader of electoral abstention in the region, with mean voter turnout that equals approximately 46% of eligible voters participating in parliamentary elections during democratic transition and afterwards. It is worth adding that in very few places around the world voter turnout is so low. Second issue worth discussing is the trend in voter turnout in CEE countries. Quite astonishingly for many, voter turnout in ten post-communist democracies has been going down since the very begging of democratic transition. Figure 2 presents this trend. Figure 2 about here 1 All countries under scrutiny are parliamentary democracies. Consequently, parliamentary elections, which are the main democratic procedures there, are analyzed. 6
Clear linear relationship can be observed. Simple bivariate linear regression model, used to estimate the effect of time on voter turnout, explains more than 33% of variance observed. According to this model voter turnout has been on average decreasing in CEE democracies by 1.7% per year. The relationship is statistical significant, despite the size of the sample. Figure 3 about here The trend is much stronger if Poland, an exceptional case where voter turnout has been very low and relatively stable during the period discussed (43.2% in the first parliamentary election after the transition and 40.6% in the last parliamentary election), is removed from the analysis. Then a simple bivariate linear regression model, used to estimate the effect of time on voter turnout in CEE countries, explains more than 52% of variance observed. According to this model voter turnout has been on average decreasing in CEE democracies (without Polish case) by almost 1.9% per year. The relationship is statistical significant. For many scholars this downward trend is difficult to explain. It is in fact hard to understand why people, who for many years were unable to influence political world around them, nowadays voluntarily refrain from the most fundamental democratic procedure, i.e. voting (which is, according to Dahl, the only procedure in modern democracy that allows citizens to influence the course of political affairs). However, in spite of rather low academic interest in studying voter turnout in the CEE region, some explanations of this phenomenon have been put forward. Some scholars suggest that this downward trend in voter turnout in CEE countries is not idiosyncratic; they argue it is a normal aspect of democratic transformation (Kostadinova 2003: 743). Others claim that it can be linked to a more general crisis of representative democracy (van Deth 2000), which leads to voter turnout decline everywhere, including post-communist democratic systems. Unfortunately, it is impossible to 7
say which of the proposed explanations is true, since most of them are not based on solid empirical evidence. But this downward trend has an impact on academic discourse. In the CEE countries there is much ado about voter turnout decline. Politicians, scholars, and journalists frequently talk about crisis of democracy, disengagement etc. As a result of the fact that voter turnout has been decreasing, many reforms have been proposed, and some have even been introduced. Consequently, voting has been simplified: in almost every country something has been done in order to boost voter turnout. Voting facilities, such as proxy voting, mail voting, two-day voting or e-voting, has been more and more popular. In some countries even more exotic reforms were put forward for instance in Bulgaria during the 2005 parliamentary election an electoral lottery was organized. But all these actions have not stopped the downward trend in voter turnout. Thus in public debate a fundamental reform, i.e. introduction of compulsory voting, has been appearing more and more frequently. This growing popularity 2 brings about the very question of what would be the impact of implementation of compulsory voting in CEE countries. The main problem is that we know very little about the hypothetical introduction of compulsory voting in CEE countries. Although the issue is more and more frequently discussed, the debate is for the most part of normative character. Solid knowledge, based on empirical studies, is very limited. In light of these facts it seems crucial to go beyond normative arguments and attempt to estimate empirically the possible results of such a fundamental reform in the region. This is the focal aim of this study. 2 I am not claiming here that the majority of CEE citizens supports introduction of compulsory voting. Quite contrary, when I am discussing voting under communist rule, I argue that CEE citizenries are nowadays rather against implementation of such law. But there is no doubt that the idea of compulsory voting has been in recent times more and more often put forward and discussed (probably due to the fact that other reforms implemented has not given satisfactory results). For this reason I dare to talk about growing popularity of compulsory voting. 8
Data In empirical analyses below Polish National Election Study (PNES) data is used. It is a research project affiliated with the Institute of Political Studies, Polish Academy of Sciences, which covers most important national-level electoral events in Poland. PNES collaborates with the Comparative Study of Electoral Systems (CSES). The data is available in the public domain. Empirical analyses In empirical analyses below three issues are to be addressed. Firstly, the effect of hypothetical introduction of compulsory voting in Poland is estimated. The very question of how many non-voters would vote if voting were compulsory is answered. Secondly, differences between groups of non-voters, which can differently react to the introduction of compulsory voting, are discussed; their socio-demographic backgrounds, as well as their values, ideological positions and political preferences are investigated. The third purpose of this section is to estimate the results of parliamentary elections if everyone had voted (if voting had been compulsory in the 2001 parliamentary election). Hypothetical behavior of non-voters if voting were compulsory There are many different ways of estimating the hypothetical behavior of non-voters. The main challenge is to estimate non-voters electoral preferences. In other words the central issue is whom they would vote for if they at all participated in elections. One of the modes of estimating non-voters electoral preferences is to ask them directly about it. This is the strategy applied in PNES. In the 2001 survey non-voters were explicitly asked whom they would vote for if they participated in elections. Frequencies of this question are presented in table 1. Four categories of respondents are distinguished: i) non-voters who know whom they 9
would vote for if they participated in elections; ii) non-voters who would not vote despite the fine sanction; iii) non-voters who would cast an invalid ballot; iv) non-voters who do not know what they would do (if voting were compulsory). Table 1 about here The biggest (49.8%) group consists of those citizens who would participate in elections if voting were compulsory. There are relatively few hard non-voters; only 7% would not vote even if abstention resulted in financial sanction. More than 11% would cast an invalid ballot. The group of indecisive respondents is quite numerous; approximately one third of non-voters does not know what they would do if voting were compulsory. These results mean that introduction of compulsory voting (with fine sanction) could substantially increase voter turnout in Poland. When studying such reform one must bear in mind that change in voter turnout, and its impact on democracy, is closely related to the overall level of electoral participation in a given country. If voter turnout is high, say 80% or 90%, then transition of 50% of non-voters from abstention to participation does not change the whole situation dramatically. But if non-voters are more numerous than voters, which is the case in Poland, then such change would mean a lot. Voter turnout would increase from the low level of 40%-50% (this is the level of voter turnout in Poland nowadays) to the level of 60%-70%. This would indicate achieving a normal European level of electoral participation. Thus on the whole it is plausible to argue that introduction of compulsory voting, notwithstanding its probability, would, at least to a certain extent, solve the so-called voter turnout puzzle in Poland 3. Very interesting is the fact that so many non-voters are unable to decide how they would behave if voting were compulsory. Almost one third of non-voters answer in this manner. The size of this group impedes accurate estimation of the overall voter turnout if 3 I of course acknowledge the fact that such increase would be purely technical ; problems and deficiencies of Polish democracy, which might contribute to low voter turnout, would almost certainly remain. 10
voting were compulsory in Poland: it is impossible to anticipate their behavior. Several explanations of this situation can be put forward here. But the most plausible is the most obvious one these citizens really do not know how they would behave if voting were compulsory. I hypothesize that their answers are strongly related to the character of the question they were asked, which is very general and lacks some important details 4 (e.g. the level of fine sanction, inevitability of fine etc.). Moreover, the situation described in the question (introduction of compulsory voting) is rather unlikely to become true. As a result many citizens cannot answer to this question in any other way. I assume that these indecisive citizens lack clear, strong attitudes towards compulsory voting. Maybe some of them do not even understand completely the question they were asked. Surely they are neither strong supporters of such procedure nor strong opponents of it. Thus it is plausible to argue that their behavior would be determined by very specific arrangements and factors. For instance, these non-voters should be particularly sensitive about the amount of fine sanction. If it was low, or easy to avoid, then they would abstain. If, on the other hand, it was high and hard to avoid, then they probably would participate in election. Finally, it is also interesting to discuss the two smaller groups distinguished: hard non-voters, who would not participate in elections despite the fine sanction, and those who would cast an invalid ballot. It is optimistic that citizens who are clearly against the most fundamental democratic procedure, i.e. voting, constitute only an inferior minority in Polish society. But it is intriguing why opponents of elections and voting would react differently to the introduction of compulsory voting. Why some non-voters, who clearly do not want to vote, decide not to change their habits? Why others, who neither are enthusiasts of voting, decide to participate in election and cast an invalid ballot? The most plausible, ad-hoc hypothesis links their behavior to economic factors and suggests that the difference observed 4 It is obvious on the other hand that such question, for methodological reasons, cannot be too detailed and too long. 11
is due to dissimilarities in economic status. Casting an invalid ballot can be a mode of avoiding the fine sanction. Richer citizens, who are not threatened by the imposition of the fine sanction, are more likely to be hard non-voters. Poorer citizens, in order to avoid the fine sanction, are more likely to cast an invalid ballot. A comparison of the two groups, which is presented below, should provide an empirical test for this hypothesis. Differences between groups of non-voters A second task I deal with in this paragraph is related to the question of who are the respondents in four groups distinguished in the question about hypothetical behavior if voting were compulsory in Poland. Do they come from similar social strata, or are they anyhow different with regard to specific socio-demographic characteristics? Are their ideological views, political preferences or opinions about democracy similar or different? How do they explain their abstention? Socio-demographic differences. Firstly, social backgrounds of the four groups distinguished must be investigated. Simple bivariate analyses suggest that there are significant, though not very strong differences between the groups with regard to specific socio-demographic variables. Table 2 reports measures of association (Cramer s V) and tests of statistical significance (Pearson s Chi-square) of the relationships under scrutiny. Table 2 about here The relationships between dependent variable and six socio-demographic independent variables are explored: sex, age, education, occupational status, place of residence and income 5 (a dummy variable discriminating between highest quartile of household income and 5 Independent variables are coded in the following way: gender (0-female; 1-male), age (6 categories: 18-25, 26-35, 36-45, 46-55, 56-65, 66+), education level (4 categories: primary, vocational, secondary, higher), occupational status (a dummy variable reporting whether individual works or not), place of residence (6 categories: village, less than 20000 inhabitants, 20000 to 50000 inhabitants, 50000 to 100000 inhabitants, 100000 to 500000 inhabitants, more than 500000 inhabitants), and income (a dummy variable discriminating between highest quartile of household income and the rest). 12
the rest). Some findings are worth mentioning at this point 6. Non-voters who would vote (and who know how they would vote) are overrepresented: among respondents with secondary education and among respondents holding university degree; among respondents with high household income; among urban citizens; among respondents who work. Hard non-voters are relatively many among respondents who do not work. Non-voters who would cast an invalid ballot are overrepresented among males, respondents with vocational education and respondents who work. Indecisive non-voters are unproportionally many in the following groups: females, elderly respondents (66+), respondents with primary education, respondents living in rural areas and respondents who do not work. More complex, multivariate analysis sheds some new light on the relationships between dependent variable and six socio-demographic independent variables. Multinomial logistic regression enables to estimate the pure impact of each of independent variables, controlling for all the other variables. Table 3 shows results of this analysis. Table 3 about here The model explains almost 10% of variance observed. Overall, 53.1% of the cases are classified correctly. The likelihood ratio tests check the contribution of each effect to the model. The table shows that the effects of each independent variable change little after controlling for all the others, apart from age 7 and income 8, which become insignificant. However, as the dependent variable has more than two categories, the effect of each independent variable is not so evident. One must bear in mind that each of them can differently influence dependent variable: at the same time it can increase probability of choosing one category (with respect to the reference category) and decrease probability of 6 It is important to bear in mind that distributions of these variables among non-voters are different from distributions in the whole sample (society). All six variables under scrutiny here are found to be significant predictors of voter turnout. 7 In this model a continuous variable is used. 8 Likelihood ratio test for this variable is statistically insignificant; however, its effect on the depend variable is statistically significant in one case (see below). 13
choosing another one (with respect to the reference category). Multinomial logistic regression model compares two alternatives and estimates the impact of each independent variable in this very specific situation 9. Parameters with significant negative coefficients decrease the probability of that response category with respect to the reference category. Parameters with positive coefficients increase the probability of that response category with respect to the reference category. The parameter estimates tables summarize the effect of each variable; in each table reference category is different. Tables 4 to 6 about here If the reference category is the DK answer, then place of residence increases probability of voting for party, education increases probability of voting for party and being hard non-voter 10, being a female decreases probability of being hard non-voter, not having a job increases probability of being hard non-voter and having lower household income increases probability of casting an invalid ballot. If the reference category is the voter answer, then not having a job increases probability of being hard non-voter and having lower household income increases probability of casting an invalid ballot. Finally, if the reference category is the hard non-voter answer, then not having a job decreases probability of casting an invalid ballot. On the whole, the analyses presented above show that answers to the question about the hypothetical introduction of compulsory voting are, at least to some extent, determined by socio-demographic traits of respondents. Not surprisingly, citizens with higher socioeconomic status are more likely to vote, while lower socio-economic status increases probability of being indecisive. Hard non-voting is first of all related to occupational status 9 For example, gender can increase probability of voting for party A, when compared to voting for party B, but at the same time it can decrease probability of voting for party A, when compared to voting for party C. 10 This example illustrates very well peculiarities of multinomial logistic regression model. Though the finding may look odd, it in fact means only that both respondents who would vote and respondents who would not vote are on average, all other things being equal, significantly better educated than respondents who are uncertain about their behavior if voting were compulsory. 14
people who do not work are more likely, ceteris paribus, to remain non-voters, even if fine sanction is at risk. It probably means that for them non-voting is a way of protest they do not vote in order to de-legitimize the system. For this reason they claim they would not vote even if voting were made compulsory. Ideological and political differences. My subsequent aim is to examine ideological and political differences between the groups distinguished. If there are any significant differences between the groups as far as socio-demographic backgrounds are concerned, then it is plausible to expect that respondents from these groups will have different ideological positions, divergent political likes and dislikes, diverse policy preferences (given that usually social background is strongly related to ideology and policy preferences). Thus beneath these (contingent) differences are discussed. Firstly, I examine ideological positions. Table 7 about here In order to infer respondents ideological positions a left/right self-placement scale is used. Each respondent is asked to locate himself on a 0-10 scale, representing this fundamental ideological cleavage. The four groups distinguished differ quite significantly to the extent that ideological positions are concerned. Hard non-voters are more rightist than the other groups. In fact, post hoc test (Bonferroni) shows that this is the most deviant group; other groups are not different or, to be more precise, the differences are not statistically significant. It is worth adding that there are no differences with regard to ideology between voters and non-voters (the average ideological position in two groups is exactly the same, equaling 4.66). Table 8 about here Secondly, I aim to discuss (contingent) differences in policy positions. PNES provides a very convenient tool for analyzing policy preferences. Each respondent in this study is asked to locate himself on a number of 0-10 scales, representing main political issues. Table 8 15
presents differences between the groups under scrutiny. Surprisingly enough, respondents from the four groups distinguished do not differ significantly as far as policy preferences are concerned. In two cases only (crime and taxes) there are statistically significant, though rather small, differences between the groups. Opinions and views on other policy issues are not significantly different. Table 9 about here Thirdly, I examine differences in political likes and dislikes between the groups. Table 9 presents the adequate data. Eight key Polish political parties, which were competing in the 2001 parliamentary election, are evaluated: each respondent is asked to evaluate each party on a 0-10 like/dislike scale. As it can be noticed in case of three parties there are statistically significant differences between the groups. However, the relationships are not strong. Attitudes towards the remaining five parties are not significantly differentiated. Satisfaction with democracy. Another factor that can differentiate non-voters is an attitude towards political system. Hence it is interesting to examine differences in satisfaction with democracy, which may exist between the four groups distinguished. It must be noticed however in the very beginning that non-voters in general are not satisfied with the way democracy works in Poland the difference between voters and non-voters is substantial and statistically significant (c.f. Czesnik 2006). Table 10 presents the results of this comparison. Table 10 about here It turns out that non-voters are rather homogeneous in this regard. The differences are small and almost statistically insignificant. The Levene s test equals 2.418 (df=3) and lacks statistical significance (p=0.065), which means that the null hypothesis (that the group variances are equal) cannot be rejected. Therefore it appears that satisfaction with the way democracy works in Poland is quite equally distributed in different groups of non-voters and 16
as such does not have any significant impact on the decision whether to vote or to abstain if voting were compulsory. Circumstantial vs. voluntary abstention. When examining differences between the four groups distinguished in the question about hypothetical behavior if voting were compulsory in Poland it is worth to look at the motives of non-voting, because propensity to vote (if voting were compulsory) might be related to reasons of non-voting. Thus firstly the distinction of circumstantial vs. voluntary abstention must be discussed. This distinction is based on the assumption that electoral behavior is not solely the function of socialized habit. It is of course true that in general citizens vote or abstain because they want and decide to do so. But sometimes other determinants of voting behavior can be distinguished. It is crucial to remember that sometimes citizens abstain due to circumstantial reasons. Hence in this paper a key distinction between circumstantial and voluntary abstention, proposed in earlier studies on voter turnout (cf. Blondel et. al. 1998), must be employed. According to this proposition, abstention can be caused by circumstantial features: illness, work, trip etc. Citizens, who abstain due to circumstantial reasons, should be different from voluntary abstainers: they should be more likely to vote if voting were compulsory, because their abstention is caused by independent factors. Table 11 about here Table 11 reports results of this analysis. In fact individuals who did not vote due to circumstantial reasons are more likely to participate in elections than voluntary abstainers. Among those ready to participate (who are even able to name the party they would vote for) if voting were compulsory circumstantial non-voters constitute more than 37%. On the contrary they are much less likely to appear among hard non-voters and those who intend to cast an invalid ballot. The relationships are statistically significant. Among undecided non-voters 17
(those who choose a do not know answer) proportion of circumstantial abstainers is the same as in the whole subsample of non-voters. Generally speaking, the above analyses show that introduction of compulsory voting could have a positive impact on Polish democracy. As the differences between different groups of non-voters are rather weak or at all insignificant, it is plausible to assume that the reform would substantially improve the situation of Polish democratic system. As far as the two criteria of significance of voter turnout are concerned, i.e. equality and legitimacy, introduction of compulsory voting would both: i) reduce the socio-demographic and ideological bias, existing between voters and non-voters, and ii) lessen the number of unsatisfied (or less satisfied) citizens who stay outside the political system 11. What if everyone voted? Finally, I estimate the results of the 2001 parliamentary election if voting had been compulsory. It is achievable, since we know party preferences of all the voters who would vote if voting were compulsory (they were asked not only whether they would vote or no, but also whom they would vote for see appendix). Table 11 presents adequate data. Table 12 about here The distribution of the vote among non-voters is astonishingly similar to the distribution among respondents who voted (who claim they voted 12 ). There are only minor differences in support for particular parties. Two parties (SRP and PiS) get slightly higher support among non-voters, while two other parties (PO and LPR) do relatively worse in this group. Overall distribution of party preferences, calculated for all the possible voters (voters + 11 One can argue that inclusion of non-voters would substantially change their evaluation of the system. On average they are less satisfied than voters with the way democracy works. It is partly due to the fact that they are excluded from political process. Once they participated in election, which would result in the re-inclusion `into political system (aggregation of their preferences, increased feeling of political efficacy etc.), their negative attitudes towards the system could change significantly. 12 We do not know exactly how many respondents, who claim they voted, actually participated in the election. So-called overreporting bias is difficult to estimate, since in Poland it is impossible to validate information about voter turnout. 18
non-voters who would participate in election), is almost equal to the official results. In case of only one party (the winner SLD) the deviation does exceed three points; in other cases deviations are about two points or less 13. It all means that compulsory voting would have had only a limited impact on the results of the 2001 parliamentary election. If voting had been compulsory then, voter turnout would have increased substantially, but the consequences of election would not have changed. Although much more citizens would have voted, the winner of election would have been the same. Moreover, the distribution of seats would have remained unchanged, since additional votes would not have influenced the overall distribution of party preferences. An argument quite often cited in such situations (if distributions of party preferences among voters and non-voters are similar) is that non-voters must refer to the situation on political scene in a given country. Even if their ideological and political preferences are significantly different (which is not the case in Poland), they can take into consideration, once asked whom they would vote if voting were compulsory, only these parties, which exist and operate in given system. Therefore the supply side of political market inevitably structures non-voters attitudes and preferences. Nevertheless, this line of reasoning does not explain why distributions of party preferences in both groups discussed are so similar. There is little doubt that non-voters must refer to the existing party system; but this notion does not imply equal distributions of the variable under scrutiny. Actually, there are more reasons to believe that these distributions should be substantially different, given the fact that there are significant dissimilarities between the two groups. In light of this argument the above finding appears as even more attention-grabbing and valuable. 13 These deviations are relatively easy to explain. In general post-electoral survey results reveal higher support for winners and lower support for losers. This is the case with the 2001 PNES survey: winners (SLD, PO, SRP, PiS) gain slightly higher support, while losers (incumbents AWSP and UW) get less than on the polling day. 19
If there are so many similarities between voters and non-voters with regard to party preferences, it is interesting to see whether determinants of the vote (voting for particular party) are also similar in the two groups analyzed. Consequently, models explaining the vote in two groups are compared. Tables 13 and 14 shows results of this analysis. Table 13 and 14 about here Multinomial logistic regression models are used. Final versions of the model, including socio-demographic, political and ideological variables, are presented. Vote patterns investigated in two groups are surprisingly similar. Almost the same sets of variables have a statistically significant effect on the decision whom to vote for, both among voters and nonvoters. The models explain respectively 63% (voters) and 60% (non-voters) of the variance observed. These results show that there are much more similarities than expected: not only distributions of party preferences are alike, but even their correlates and determinants are comparable. Conclusions The paper makes three general conclusions. Firstly, empirical analyses show that introduction of compulsory voting would be a very effective tool for boosting voter turnout in Poland. If Polish citizens had been forced to vote in the 2001 parliamentary elections, then vast majority of non-voters would have participated in the elections. As a consequence voter turnout would have increased substantially. Second conclusion is that various social groups would differently react to introduction of compulsory voting. Reactions to this reform are first of all related to socio-demographic backgrounds. Gender, education, place of residence and occupational status are found to be statistically significant predictors of the dependent variable. Also ideological self-placement is significantly associated with the dependent variable. Policy preferences and satisfaction with 20
democracy do not discriminate between the groups investigated. It is worth noticing however that differences observed are not so big. This is important, because it means that the group, which will remain outside electoral politics after introduction of compulsory voting, will not be so different from the rest of society. Such situation (lack of significant socio-demographic or ideological bias) should contribute positively to the quality of democracy and its legitimacy. My third conclusion is that introduction of compulsory voting would not have an impact (or would have a limited impact) on the election results. This finding is the most surprising one. It seems that introduction of compulsory voting, which would change quite a lot with regard to voter turnout, system legitimacy etc., would probably change only a little the situation on the Polish political scene. This argument is very important for politicians: if compulsory voting would not change substantially the results of election, then there are few incentives to introduce it. This notion is based on observation that political parties and their leaders are rational: if no one is able to attract majority of the new voters (which is the most likely scenario given their party preferences), then it is cheaper not to introduce compulsory voting (given the fact that the more voters on electoral market, the more expensive each electoral campaign). My major task in this paper is to answer a fundamental question of whether compulsory voting is a remedy for low voter turnout countries like Poland. On the whole, the paper gives a positive answer compulsory voting would substantially increase voter turnout. If low voter turnout is defined as a sign of democratic rule crisis in Poland (most people define the situation in this way), then the implementation of compulsory voting appears as a good solution of many problems Polish democracy has been lately suffering. Obviously, many problems would remain such mechanical increase of voter turnout according to some scholars would strengthen crisis of democracy, instead of overcoming it. 21
It is open to discussion whether and to what extent these results can be extrapolated to other countries. Poland, as far as voter turnout is concerned, is quite unique. Therefore relationships observed and conclusions reached in this paper are of limited significance for other post-communist countries. However, the findings, despite their uniqueness and problems with extrapolating them to other countries, can be useful for future empirical research on compulsory voting, which is needed given the downward trend in voter turnout observed in Europe. 22
References Blondel, Jean, Richard Sinnott and Palle Svensson. 1998. People and Parliament in the European Union: Participation, Democracy and Legitimacy. Oxford: Clarendon Press. Czesnik, Mikolaj. 2006. Voter Turnout and Democratic Legitimacy in Central Eastern Europe. Polish Sociological Review 156: 449-470. Deth, Jan W. van. 2000. Interesting but Irrelevant: Social Capital and the Saliency of Politics in Western Europe. European Journal of Political Research 37: 115-147. Kostadinova, Tatiana, 2003. Voter turnout dynamics in post-communist Europe. European Journal of Political Research 42: 741-759. Lijphart, Arend. 1997. Unequal Participation: Democracy s Unresolved Dillema. American Political Science Review 91: 1-14. Lively, Jack. 1975. Democracy. Oxford: Basil Blackwell. McManus-Czubinska, Clare, William L. Miller, Radoslaw Markowski and Jacek Wasilewski. 2004. When Does Turnout Matter? The Case of Poland. Europe-Asia Studies 56: 401-420. Sartori, Giovanni. 1965. Democratic Theory. New York: Frederic A. Praeger Publishers. Schumpeter, Joseph A. 1957. Capitalism, Socialism, Democracy. New York: Harper and Raw. 23
Appendix Question used to investigate (estimate) electoral behavior of non-voters in the 2001 Polish National Election Study. In some democratic countries, e.g. Belgium or Australia, citizens are obliged to vote under fine sanction. If this were the case in Poland (i.e. compulsory voting), for which party would you choose to vote? (INTERVIEWER: Please show CARD 2)... 92) I would not vote despite the sanction 93) I would cast an invalid ballot 97) Hard to say 24
Tables and figures Figure 1. Mean voter turnout in CEE countries. 90,00 80,00 70,00 60,00 68,07 78,02 63,11 68,64 76,21 58,09 69,03 81,87 72,65 50,00 45,99 40,00 30,00 BLG CzR EE HU LA LITH POL ROM SLV SLO Source: Own calculations on basis of IDEA data. Figure 2. Downward trend in voter turnout in CEE countries. 100,00 90,00 80,00 turnout 70,00 60,00 50,00 R kwadrat dla Liniowej = 0,336 40,00 1990 1991 1992 1993 1994 1995 1996 1997 1998 year 1999 2000 2001 2002 2003 2004 2005 Source: Own calculations on basis of IDEA data. 25
Figure 3. Downward trend in voter turnout in CEE countries (without Poland). 100,00 90,00 80,00 turnout 70,00 60,00 50,00 R kwadrat dla Liniowej = 0,527 40,00 1990 1991 1992 1993 1994 1995 1996 1997 1998 year 1999 2000 2001 2002 2003 2004 2005 Source: Own calculations on basis of IDEA data. Table 1. Electoral behavior if voting were compulsory. Frequency Percent Valid Percent Valid Party 366 20.4 49.8 NV 51 2.9 7.0 Invalid 81 4.5 11.1 DK 236 13.1 32.1 Total 734 40.9 100.0 Missing Voted 1060 59.1 Total 1794 100.0 Source: PNES 2001. 26
Table 2. Social backgrounds of electoral behavior if voting were compulsory. Variables Cramer s V Pearson s Chi-square Gender 0.12 10.36* (df=3) Age 0.13 39.37** (df=15) Education 0.12 34.04** (df=9) Income 0.11 8.67* (df=3) Place of residence 0.12 31.57** (df=15) Occupational status 0.15 16.17** (df=3) Source: PNES 2001. ** p 0.01, * p 0.05 Table 3. The likelihood ratio test. Effect -2 Log Likelihood of Reduced Model Chi-Square df Significance Intercept 1531.064(a) 0.000 0 - Place of residence 1541.029 9.965 3 0.019 Age 1532.378 1.315 3 0.726 Education 1542.388 11.324 3 0.010 Gender 1540.517 9.453 3 0.024 Income 1538.277 7.213 3 0.065 Occupational status 1542.071 11.007 3 0.012 Nagelkerke R-Square 9.9% Cases classified correctly 53.1% Source: PNES 2001. 27
Table 4. Determinants of non-voters hypothetical behavior parameter estimates table. vote choice if voting compulsory(a) B Standard error Wald df Sig. Exp(B) 1.00 party Intercept -.201.478.177 1.674 resid350.149.050 9.036 1.003 1.161 Respage -.004.005.610 1.435.996 Edurec.330.112 8.715 1.003 1.391 [ngender=.00] -.327.177 3.390 1.066.721 [ngender=1.00] 0(b).. 0.. [nincome=.00] -.105.248.180 1.671.900 [nincome=1.00] 0(b).. 0.. [njob=.00] -.044.196.049 1.824.957 [njob=1.00] 0(b).. 0.. 2.00 nv Intercept -3.091.915 11.403 1.001 Resid350.019.093.044 1.834 1.020 Respage.001.010.004 1.950 1.001 Edurec.488.202 5.863 1.015 1.630 [ngender=.00] -.746.320 5.428 1.020.474 [ngender=1.00] 0(b).. 0.. [nincome=.00].356.511.486 1.486 1.428 [nincome=1.00] 0(b).. 0.. [njob=.00].832.376 4.903 1.027 2.298 [njob=1.00] 0(b).. 0.. 3.00 invalid Intercept -1.786.760 5.527 1.019 Resid350.101.075 1.832 1.176 1.106 Respage -.009.009.927 1.336.992 Edurec.304.168 3.261 1.071 1.355 [ngender=.00] -.646.270 5.751 1.016.524 [ngender=1.00] 0(b).. 0.. [nincome=.00].871.438 3.960 1.047 2.390 [nincome=1.00] 0(b).. 0.. [njob=.00] -.513.289 3.149 1.076.599 [njob=1.00] 0(b).. 0.. a The reference category is: 4.00 dk. b This parameter is set to zero because it is redundant. Source: PNES 2001. 28
Table 5. Determinants of non-voters hypothetical behavior parameter estimates table. vote choice if voting compulsory(a) B Standard error Wald df Sig. Exp(B) 2.00 nv Intercept -2.890.882 10.742 1.001 Resid350 -.130.089 2.143 1.143.878 Respage.005.009.278 1.598 1.005 Edurec.158.193.671 1.413 1.171 [ngender=.00] -.419.308 1.845 1.174.658 [ngender=1.00] 0(b).. 0.. [nincome=.00].462.488.894 1.344 1.587 [nincome=1.00] 0(b).. 0.. [njob=.00].876.361 5.901 1.015 2.401 [njob=1.00] 0(b).. 0.. 3.00 invalid Intercept -1.585.716 4.895 1.027 Resid350 -.048.069.483 1.487.953 Respage -.004.008.249 1.618.996 Edurec -.026.157.028 1.867.974 [ngender=.00] -.320.254 1.585 1.208.726 [ngender=1.00] 0(b).. 0.. [nincome=.00].977.409 5.689 1.017 2.655 [nincome=1.00] 0(b).. 0.. [njob=.00] -.470.270 3.022 1.082.625 [njob=1.00] 0(b).. 0.. a The reference category is: 1.00 party. b This parameter is set to zero because it is redundant. Source: PNES 2001. Table 6. Determinants of non-voters hypothetical behavior parameter estimates table. vote choice if voting compulsory(a) B Standard error Wald df Sig. Exp(B) 3.00 invalid Intercept 1.306 1.061 1.514 1.219 resid350.082.105.608 1.436 1.085 Respage -.009.012.622 1.430.991 Edurec -.184.230.641 1.423.832 [ngender=.00].099.370.072 1.789 1.104 [ngender=1.00] 0(b).. 0.. [nincome=.00].515.608.718 1.397 1.673 [nincome=1.00] 0(b).. 0.. [njob=.00] -1.345.419 10.333 1.001.260 [njob=1.00] 0(b).. 0.. a The reference category is: 2.00 nv. b This parameter is set to zero because it is redundant. Source: PNES 2001. 29
Table 7. Means of ideological position in four groups. N Mean Standard deviation Party 308 4.44 2.442 NV 28 5.96 2.385 Invalid 47 4.79 2.023 DK 99 4.90 1.894 Total 481 4.66 2.322 Source: PNES 2001. Eta-square=0.026, p 0.01 Table 8. Means of policy positions in four groups. Eta Eta square Significance Crime 0.118 0.014 0.018 Privatization 0.065 0.004 0.427 Role of Church 0.067 0.005 0.368 Nomenklatura 0.007 0.000 0.998 Unemployment 0.067 0.004 0.358 Taxes 0.109 0.012 0.046 EU 0.105 0.011 0.074 Subsidies for farmers 0.102 0.010 0.067 Welfare state 0.099 0.010 0.072 Foreign capital 0.047 0.002 0.703 Source: PNES 2001. 30