“Voting Correctly” in the 1999 European Parliament Election

Eric Hines University of Iowa [email protected]

Paper presented at the 2006 Annual Conference of the Southern Political Science Association, Atlanta, Georgia.

“Voting Correctly” in the 1999 European Parliament Election The general conclusion of two decades of research on EP elections is that they are "second-order" national elections where European issues are unimportant and which voters see as irrelevant because they have no impact on the makeup of national governments. The poor and declining turnout in EP elections compared to "first-order" national elections is often used as evidence of this latter point. Turnout is also used as evidence of a much more serious implication of research on EP elections, that they reflect a crisis of faith in the EU's democratic institutions. However, there is strong evidence that declining turnout in EP elections can be accounted for by structural factors and has nothing to do with voter attitudes (Franklin 2001). This does not mean, however, that democracy in the EU is particularly healthy; it just suggests that turnout is not an adequate measure of the legitimacy of EP elections. This begs the question, if not turnout, what can we use to evaluate the democratic credentials of EP elections? One possible alternative, introduced by Lau and Redlawsk (1997), is the idea of correct voting — the extent to which voters vote the way they would if they were fully informed about the issues and candidates running in the election. Their measure of voting correctly has been validated by both experimental data and survey data from US Presidential elections. Lau et al. (2005) argue that the voting correctly measure can be applied comparatively as a measure of democratic health because healthy democracies must not only encourage citizens to participate, but must enable citizens to make wise decisions in the voting booth. European elections are particularly fertile ground for using the concept of correct voting because their “second order” nature suggests voters should be unable to make informed, and therefore correct, voting decisions. In the following sections, this paper will outline in greater detail the potential problems in the health of EU level democracy, the concept of correct voting as a measure of democratic health, the reasons why EP elections offer a particularly fertile location for using the

Hines 2 concept, some theoretical expectations about the ability of voters to make correct voting decisions in EP elections, and finally offer a test of these expectations using the 1999 European Election Study (EES).

European Elections and the Democratic Deficit The accession of 10 new members states in May 2004 reinvigorated the debate over the growing perception of a disconnect between the institutions of the European Union (EU) and citizens of Europe, or the “democratic deficit.” A scholastic cottage industry has sprung around this perceived lack of democratic accountability. Even the European Commission acknowledges the need for change in the mechanisms of European governance. Their White Paper on European Governance (Commission, 2001) stems from a vague impression that the “permissive consensus” between the public and elites who have traditionally driven integration has broken down (Schutter, 2002). This has lead to an increase of criticism of the EU that focuses on the issues of political representation (input legitimacy) and policy outcomes (outcome legitimacy). There is general agreement that, at present, policy outcomes rather than representation are the primary mechanism for legitimating the EU, which gives credence to the argument that the EU is detached from the public. On the issue of representation, the primary criticism is that only one European institution is directly elected, the European Parliament (EP). The debate is not just over how much real influence the representatives of the peoples of Europe have in the policy process, although it is hard to ignore the EP’s “conditional agenda setting” ability (Tsebelis, 1994), but also whether the Parliament actually serves as a mechanism for representing the people of Europe in the first place. Of great concern is the continuing decline in voter turnout for European elections and the reality that candidates and parties continue to use the elections as forums for furthering national political agendas (Eijk and Franklin, 1996). In short, at first blush, European elections fall far

Hines 3 short of the normative ideal of the role of elections in a democratic polity. The problematic nature of EP elections continues to be of interest to political scientists because the “modern” liberal-democratic conception of democracy relies heavily on elections. Elections have gained this preeminent role because they provide political elites with incentives to develop rival policy ideas and propose rival candidates and voters with a mechanism for holding politicians accountable. For these reasons, elections are often viewed as a political version of the market where perfect electoral competition leads to optimal political outcomes. However, markets are always imperfect and elections by themselves do not guarantee perfect political competition or a vibrant democracy. If political elites have insufficient incentive to present alternative policies for voters to choose from, elections provide no real political competition and politics become oligarchic (Schattschneider 1960). Elections can also lead to a fragmentation of politics that leads to capture of the government by unrepresentative groups like in Sartori’s (1976) famous discussion of the “polarized pluralism” of the Italian party system with its permanent capture of government by centrist parties. Because of the close connection between elections and the health of a democracy, political science as a discipline has often taken on the task of monitoring them for danger signs, particularly levels of voter turnout. As a general trend, turnout has been declining in advanced democracies and this is often seen as an indication of citizen disengagement and lack of political efficacy among voters (Tiexera 1992, Dalton 2000, Wattenberg 2000; 2002). Others have used turnout as way to judge the relative health of democracies in a comparative context (Franklin 2002; Jackman and Miller 1995). This latter perspective has been part of a broader research program into the structural determinants of voter turnout that suggests things like the ease of registering to vote have a strong effect on turnout that is independent of voter attitudes (Jackman 1987). Unfortunately, the obsession media pundits and academics alike have with turnout rates

Hines 4 does not in the end always tell us much about the health of a democracy. For example, low turnout may be simply a function of generational change (Miller and Shanks 1996, Putnam 2000; Blais et al. 2001). The problem of using turnout as an indicator of democratic health is particularly relevant in the debate over the democratic deficit. As mentioned above, low and falling turnout in European parliament (EP) elections are the most frequently invoked sign of a democratic crisis in the European Union (EU). However, turnout may not be an appropriate measure of the EU's democratic health since Franklin (2001) shows that falling turnout in EP election can be explained without examining voter attitudes at all. Even more important, voters' specific attitudes towards Europe have no impact on turnout (Eijk and Franklin 1996). While low turnout in EP elections has nothing to do with support for the EU as an institution as is often argued, it does at least raise concerns that there is a potential problem with EP elections and their ability to help legitimate the EU. I propose using a slightly different standard to measure the effectiveness of democracy at the European level — correct voting.

Correct Voting Determining the correctness of someone’s vote might at first appear to be a difficult task because of many different values that could be used to judge it. However, the task is not as subjective as one might think. Lau and Redlawsk (1997) define a correct vote as one that is the same as one chosen under conditions of full information. This definition follows Dahl’s (1989) statement that “a person’s interest or good is whatever a person would choose with the fullest attainable understanding of the experiences resulting from that choice and its most relevant alternatives” (180-181). Unfortunately, human beings are cognitively limited and cannot possibly process all the information available to them during a campaign. Instead, voters must choose what information to collect and what to ignore. This means that what constitutes full

Hines 5 information for one voter may differ significantly from another. Ideally then, a measure of voting correctly examines vote choice under full information about those things the voter has determined are important for their voting decision. Lau and Redlawsk (1997) presented two measures of voting correctly. The first was based upon data collected in a controlled experiment. In this setting, complete information about mock candidates in a simulated election could be presented to participants after they had cast a vote. Participants who said they would not have changed their vote in light of this new information were said to have voted correctly. However, Lau and Redlawsk offered a second measure of voting correctly — a normative measure of naive vote preferences — that used the an objective, or normative, evaluation of the voter’s own, naive information gathering strategies to determine if they had cast a correct vote under the rule that the same criteria of judgment should be applied to all candidates or parties. While this second measure appeared valid in the experimental setting, Lau and Redlawsk also wanted to seek some measure of validity outside of the experiment. They applied the second procedure to data from the American National Election Study (ANES) from U.S presidential elections from 1972-1988. The authors offered three stylized hypotheses test their measure in real elections. They argued that given the limited cognitive resources, voters would be more likely to make correct voting decisions when there were fewer candidates, when candidates were more ideological distinct, and when campaign resources were reasonably balanced between candidates. The survey-based measure of voting correctly did appear to be valid, supporting each of these hypotheses in the American context. Since their initial article, very little work has been done using this measure of voting correctly. Only Lau et al. (2005) has attempted to apply this measure to other democratic systems and this work is very preliminary. Nevertheless, it suggests that the concept of voting

Hines 6 correctly can be applied in a comparative context. EP elections are no exception. In fact, EP elections are a particularly useful place to apply this measure because of two built in violations of normative democratic theory that could lead to low levels of correct voting. First, although the EU is a political system with a Parliament, it lacks the fully competitive party system taken for granted in a modern democracy. We expect parties to define and express a group’s needs and wants by building on shared ideas in order to bring together supporters. Parties contesting EP elections are criticized for failing to fulfill these functions at the European level. Second, survey after survey European voters know very little about European issues and politics. While individuals with low information can still vote correctly, low information can lead to poor or incorrect voting decisions.

General Expectations about Correct Voting in Europe While the idea of correct voting is actually quite atheoretical, we still need sort of framework upon which to base expectations about how the political environment will affect the ability of voters to get it right. For this purpose, I borrow Lau and Redlawsk’s (2005) processoriented model of voter decision-making, depicted in Figure 1, that allowed them to extend their initial analysis of correct voting at the systemic level to the individual level. This processoriented approach is designed to model the entire process behind vote choice, including how information is acquired and evaluated. It begins with four sets of factors that serve as the main independent variables in their analysis of vote choice and decision quality (correct voting). The model starts with a voter's demographic background characteristics, including partisan dispositions, which serve generally as controls. These in turn influence a voter's level of political sophistication or expertise and interact with "campaign factors" that are the contextual influences like the number of parties contesting an election that affect later stages of the decision process. In turn, these factors influence information processing and memory that Lau and Redlawsk

Hines 7 believe have important direct effects on the nature and quality of a vote decision. [Figure 1 about here] While the model itself provides a framework for more than just an analysis of decision quality, many of its implications can only be tested in an experimental setting. Therefore, my analysis of correct voting is restricted to what Lau and Redlawsk refer to as their “baseline model” (Lau and Redlawsk 2005, Chapter 10). This “standard political analysis” of what leads citizens to vote correctly examines the effects of demographic characteristics and general political dispositions, political sophistication and expertise, political knowledge (memory), and campaign factors. The baseline model, like the measure of correct voting itself, can be approximated using survey data, which Lau and Redlawsk subsequently did for American elections using the ANES. It is a simple task to replicate this analysis using the EES data. I now turn to defining the expected relationships between a number of indicators drawn from the four stages of the overall model that are included in the baseline version.

Demographic characteristics and general political dispositions Lau and Redlawsk (2006, forthcoming) identify several demographic variables that they expect to be correlated with correct voting. 1. First, a higher level of education is expected to be correlated with higher levels of correct voting. 2. Second, age is expected to be negatively correlated with correct voting because age is associated with a shallower search for information and less experience with the politics. 3. Third, the strength of political attachment should be correlated with correct voting. For a fourth characteristic, ideology, Lau and Redlawsk initially expected to find no

Hines 8 relationship, but instead found a slight negative correlation between correct voting and conservative ideology. In perhaps the largest departure from their original logic, I expect to find those expressing an extreme ideology (either left or right) are more likely to vote correctly because parties and individuals of extreme ideologies are often more concerned with issues of ideological purity and it should be easier for each to identify the other. 4. Voters with an extreme ideology (left or right) are more likely to vote correctly.

Political sophistication and knowledge 5. Those who express an interest in politics are likely to acquire the information necessary for voting correctly. 6. All else equal, the more political knowledge recalled by a voter the more likely he or she will be to have voted correctly.

Campaign factors Influence of political systems on levels of correct voting or “campaign factors” are the largest set of explanatory variables in this study. This set of variables includes several that influence correct voting at both the systemic level (overall level of correct voting) and an individual’s ability to cast a correct vote. Drawing upon Lau and Redlawsk (1997, 2005) and Lau et al. (2005), I offer some basic simple expectations of how campaign factors will influence the amount of correct voting in EP elections. Lau and Redlawsk’s (1997, 2005) initial hypotheses or expectations about the ability of voters to vote correctly in US presidential elections are a useful starting place.1 Their first expectation was simply that given voters’ limited cognitive resources, voters were more likely to 1

Another of Lau and Redlawsk’s (1997) findings was that equality of campaign spending impacts correct voting by making it harder for one candidate to “buy” incorrect votes. However, the lack of uniform reporting requirements on campaign spending in most countries for EP elections makes testing this expectation difficult.

Hines 9 make a correct decision when there are fewer candidates to choose from. In other words, the more alternatives in the choice set, the less correct voting. In the US cases, this appears to be the case. In the 1980, 1992, and 1996 elections, which of each featured a strong third party candidate, voters were significantly less likely to vote correctly. This should hold in EP elections as well and I expect to find that: 7. Voters are more likely to make correct decisions when there are fewer candidates from which to choose Lau and Redlawsk (1997, 2005) also expected to find that the position taking of the candidates would influence the ability of voters to vote correctly. They argued that the more ideological distinctive, or ideological space between the candidates, the easier it will be for voters to correctly identify the positions of the candidates and therefore the more accurate their voting will be. Since EU politics is seen as potentially being two-dimensional, this expectation can be extend to consider ideological distinctness not only on the traditional left-right dimension by also on the level of integration parties favor. 8. All else equal, the more distinct the left-right positions of the parties a voter has to choice from, the more likely a voter will vote correctly. 9. All else equal, the more distinct the positions of the parties a voter has to choice from on the EU, the more likely a voter will vote correctly. Moving to a cross-national analysis will introduce considerable more variation in the levels of correct voting then Lau and Redlawsk (1997) original expectations can explain. Lau et al. (2005), however, offer a long list of hypotheses that do. The operationalization of correct voting used in this study suggests that tactical voting violates the voters “best” interests. Therefore, voters who knowingly choose to vote for a party other then the one that best represents them will reduce the overall level of correct voting. This suggests several

Hines 10 expectations about correct voting and the proportionality of electoral systems, which heavily influences the amount of tactical voting in an election. I expect to find that: 10. All else equal, proportional representation (PR) should increase levels of voting correctly because it allows voters to vote sincerely. 11. Controlling for the number of parties contesting the elections, greater proportionality should increase the amount of correct voting. Finally, I expect to find that: 12. Compulsory voting laws should decrease correct voting because it leads to voting by uniformed and disinterested citizens. 13. Federalism should decrease voting correctly because it reduces the degree to which politics becomes nationalized. The persistence of territorial cleavages in federal systems will cloud voters’ abilities to gather information about the policy stands of national parties (e.g. the CSU in Bavaria is more conservative than the CDU/CSU at the federal level). 14. The greater the availability of political information in a country, all else equal, the higher the level of voting correctly because the easier it is to get information, the more likely voters will be able to do a full evaluation of all candidates or parties. 15. The more media accessed by a voter, the more likely they will vote correctly.

Factors unique to European Parliament Elections There are several factors that might influence correct voting in EP elections and reflect the existence of a democratic deficit. 16. Since European issues are more salient in new member states, voting correctly ought to decline the longer a voter’s nation has been in the EU.

Hines 11 17. Satisfaction with democracy in both the voter’s member state and the EU will influence how much information they acquire. A positive attitude should increase the likelihood of a correct vote. 18. Voters who believe the EU is a good thing for their country will be more likely to vote correctly than those who believe the EU is bad for their country. 19. Finally, since EP elections are generally treated as a referendum on national governments rather then as an election for the EP, voters who approve of the their current national government’s performance will be more likely to vote correctly as they have more incentive to vote sincerely.

Data and Methods To measuring correct voting in the 1999 EP election I utilize information available in the 1999 European Election Study.2 I choose 1999 because, while this survey is part of a quinquennial series of election surveys conducted in conjunction with European parliament elections, the 1999 study is the only version readily available that includes all the information necessary for estimating a measure of correct voting. I hope to extend this analysis to the 2004 parliamentary election once the embargo is removed from the latest EES. The goal for determining which party a respondent should have voted for is to include as many criteria for party evaluation as possible while still allowing the respondent’s own preferences to determine the outcome of the evaluation. The four categories of predictors available in the 1999 EES are party affiliation, broad agreement with a party’s policy stands (ideology), prospective evaluation of future support for a party, and incumbent performance evaluation. Each piece of information was weighted either 0 or 1 to capture the salience of a particular piece of information for that respondent. Each evaluation was scaled to range between 2

The dataset is available for download at http://www.europeanelectionstudies.net/EES %201999.htm

Hines 12 -1 and 1, except for party attachment and prospective support, which range from 0 to 1.3 The summary evaluation of each party was determined by the simple mean of all nonmissing evaluative items. A “correct” vote was for the party with the highest summary evaluation. Before turning to the results, I will detail the specifics behind each measure. Party attachment. Each respondent was asked if they felt an attachment to a particular political party. If the respondent answered yes, they were asked how strong this attachment was. The party attachment evaluation was weighted 0 for any respondents who did not provide an answer to one or both questions, thus adding nothing to that respondent’s summary evaluation. For those whose evaluation was weight one, the value of evaluation was based upon the strength of the attachment. Respondents reporting a very close attachment to a party were given +1 toward their summary evaluation of that party. Respondents reporting a close attachment got +.67 towards their evaluation and those reporting themselves to be “merely a sympathizer” received +.33 towards their evaluation of the party. Policy Agreement (Ideology). While the best measure of policy agreement between respondents and parties would look at specific issues, the EES does not include policy specific questions beyond asking individuals which issues are important at a given time. Therefore it is possible to know what issues the respondent thinks are important, but no way of measuring their policy preferences on those issues. This problem also exists in the Comparative Study of Electoral Systems (CSES) data used by Lau et al. (2005). Their fix was to use the respondent’s own ideological self-identification on an 11 point left-right scale as a proxy for general policy agreement with political parties. There really is no alternative to using this strategy with EES data. Luckily, the EES asks respondents to rate themselves on a ten-point scale for both left3

In Lau and Redlawsk (1997), each evaluation was scaled to range from -1 to 1. This was possible in part because of the two-party system in the US. In a multi-party system, a positive evaluation for a party does not necessarily translate into a negative evaluation of another. Thus, while a strong Democrat may have a deep dislike for a Republican candidate, a SPD supporter in Germany will not necessarily have a negative attitude towards the CDU. It may be possible to convert the prospective support for a party into this evaluation, but it is unclear to me at present how this could (or should) be done.

Hines 13 right ideology and for a general position on European integration. The survey also asked respondents to place each party on these same scales. Using these two pieces of information, and Rabinowitz and MacDonald’s (1989) “directional theory” of voting, it was possible to construct an ideological agreement evaluation for each respondent. Rabinowitz and MacDonald argue that being on the same “side” of an issue is more important than simple Euclidian distance between voter and party. For computation, they suggest subtracting the midpoint of the scale from both the voter’s position and the party’s position then multiplying them together. On the 10-point scale, 5.5 is the midpoint. This produces an initial product in the range of -20.25 to 20.25, but I scale it down to range between -1 and +1. The party’s positions were determined by using the mean position assigned by the respondents who correctly answered the single election knowledge question in the survey.4 These two evaluations were weighted zero if the respondent did not place themselves on the scales. Prospective Evaluation. Each respondent was asked on a scale of one to ten how probable it was they would ever vote for a party. To transform these responses into a prospective evaluation of each party, I scaled this variable to range between 0 and 1. A one indicates that a respondent sees himself or herself only voting for a single party, while a zero indicates the respondent does not planning on voting for a that party in the near future. One advantage of this measure is that it captures some of the dynamics of strategic voting since it indicates the willingness of the respondent to switch between certain parties. Incumbent’s Job Performance. Unlike the American system, where evaluation of incumbency is really only significant when a sitting president is up for re-election, in multi-party systems in Europe there is always an incumbent party or parties. Thus, satisfaction with the 4

Respondents were each asked which party in their country received the most votes in the actual election. Overall, about 66.27% of respondents answered this question correctly. Lau et al. (2005) used expert judgments for most of their measures of party position. In the future, I will compare the values gained using the “knowledgeable” respondents with expert judgments to see if they make any real difference in the calculation of the summary evaluations of correct voting.

Hines 14 current government’s job in office will always shape voting decisions. However, assigning a value for this evaluation is difficult for two reasons, one theoretical and one methodological. Theoretically, there is a strong justification for excluding incumbency as a form of evaluation for voting correctly in European elections since there is no direct connection between the Parliament and the governments of the member states. In fact, European elections are criticized for being more referenda on government performance than about Europe. However, governments are by their participation in the European Council and the Council of Ministers some of the most powerful actors in European politics. Therefore, a respondent’s satisfaction with the government’s overall performance will reflect not just performance in domestic politics, but the behavior of the government regarding their role in European politics as well. Including incumbency is normatively defensible, but constructing an evaluation of it offers a methodological issue. Most governments in Western Europe are coalition governments. If two or more parties share the blame (or the credit) for government performance, voters could choose to hold just the party of the prime minister responsible or all the coalition parties equally responsible. However, the standard for voting correctly says a voter is acting with full information. That standard plausibly implies that voters should be capable of assigning responsibility for government action in proportion to the role each coalition partner plays in decision-making. Therefore, not only is the incumbency evaluation weighted for salience, but is weighted in proportion to each coalition member’s relative size. The actual measure of incumbency in the survey is a simple dichotomous variable asking if the respondent approves or disapproves of the present government’s performance. If the respondent approves of the government’s performance, a value of +1, weighted again by the size of the party within a coalition, is added to each government party’s summary evaluation. If a respondent disapproves of the record, a -1 weighted by the party size is subtracted from each

Hines 15 government party’s evaluation.

Results Overall Levels of Correct Voting in 1999 EP Election The simplest question about correct voting is simply how many voters got it right? In the 1999 European Parliament elections, the answer across the EU15 appears to be about just short of 71 percent of the time. However, Table 1 makes it clear that the level of correct voting varies considerably between countries.5 In the 1999 EP elections, Portuguese voters appeared to do the best job making quality vote decisions with 92.82 percent of their votes representing their best fully informed preferences. On the other hand, barely a majority of Irish voters managed to cast a correct vote (53.45 percent). Interestingly, voters in Portugal and Ireland both had the smallest choice set of all the member states with just five parties demonstrating that the ability to make an informed vote is not just a factor of the number of choices available. Compared to the average level of correct voting in other elections — 70 percent in US presidential elections, 65 percent in British general elections, and 87 percent in Dutch general elections — an average of 71 percent in EP elections is surprisingly good.6 [Table 1 here]

Baseline Model of Correct Voting in 1999 EP Election Using the baseline model of correct voting described above, it is possible to test correct voting in the election study to see how well it conforms to expectations. Table 2 presents the 5

Belgium is absent from my analysis at present because of issues related to the unique division of their electoral system between the Dutch and French areas of the country. In effect, Belgium has two electoral systems and I have not yet determined how to take this into account. This issue should be resolved by the time the paper is formally presented. 6 The average correct vote for the US comes from Lau and Redlawsk (2006, forthcoming). Figures for Britain and the Netherlands come for separate analysis conducted by the author. The full analysis has not yet been conducted for these two cases, so these results are considered tentative.

Hines 16 results of a logistic regression of the baseline model weighted in order to more accurately capture the dynamics of the cross-national nature of this data.7 The model appears to be a relatively good fit with a pseudo-R2 of 0.20 and a Model χ2 of 995.81 that is significant with 24 degrees of freedom. The model predicts 62 percent of cases correctly and reduces errors in the prediction of a correct vote by 13.61 percent. [Table 2 about here] The first set of factors in the base model is the personal characteristics of the voter. The results from this set of variables are mixed. While education does not appear to affect the ability of voters to cast a correct vote, a voter’s age does have a small effect. Over a voter’s lifetime their ability to vote correctly, holding other factors at their sample median, declines by about fiver percent. These results are consistent with the findings from American presidential elections (Lau and Redlawsk 2006, forthcoming) and other countries (Lau et al. 2005). However, the influence of gender on correct voting in EP elections is inconsistent with findings from earlier research where gender never achieved statistical significance. In the 1999 EP election, holding all other factors at their sample median, a woman is 4 percent more likely to cast an incorrect vote then a man. The dummies for class are not significant, suggesting that upper class and working class voters are no more and no less likely to vote correctly than the middle class. Finally, religiosity (frequency of church attendance) also does not appear to affect the ability to vote correctly. In addition to these demographic characteristics, the baseline model also examines

7

The EES includes several different weighting variables. The weight used here is one designed for use in pooled analysis of systemic or contextual. In order for estimates of the effect of these variables to be unbiased each member states must be weighed equally in the analysis. The weight also takes in to consideration the actual voting behavior of voters in the member states and is designed to ensure that the sample distribution of voting behavior in each member state mirrors the actual outcome the different elections. This weight also removes the handful of responses from Northern Ireland.

Hines 17 political predispositions. The strength of a voter’s partisan attachment has both a statistical and substantive significance on the level of correct voting. Again holding other factors at their median values, a voter who expresses the highest level of partisan attachment is 15 percent more likely to vote correctly than a voter with no partisan attachment. The extremity of a voter’s selfidentified ideology is also significant, and holding all else at its sample median, a voter whose ideology is in the fist or last quartiles of the scale is 22 percent more likely to vote correctly than centrist voters. However, voters who express approval of their national government see no significant increase in their likelihood of voting correctly. The model also considered media access, but neither reading a newspaper nor watching television news had an effect on correct voting. The next set of factors the baseline model considers is political interest and knowledge. Only political interest proved to have a statistically significant impact on casting a correct vote. Moving the political interest index to the highest level from the median, holding everything else constant, increased the probability of a correct vote by 4 percent. Since the political knowledge measures in the survey are limited to a single question, the lack of significance may simply be a result of failing to really capture a voter’s political knowledge. The third set of factors the baseline considers deal with is “campaign factors” that alter the political context of the task of making a vote choice. These prove to be the most consistently significant in the model. First, the number of significant political parties among which voters have to choose has a significant impact on the ability of voters to vote correctly. Increasing the number of parties from the median number of parties (7) to the sample maximum (12) decreases that probability of voting correctly 12.1 percent. Further, the number of districts appears to be significant, albeit in the wrong direction as I expected. Increasing the number of constituencies from the sample median of a single national district to the maximum of 8 districts increases the

Hines 18 probability voting correctly by 11.7 percent. The overall size of the left-right dimension of national politics had no influence on the ability to vote correctly. However, voters in party systems where party’s positions on the EU are distinctive were more likely to vote correctly. Voters in the party system with the largest distance between parties were 10 percent more likely to vote correctly then their colleagues in party systems with the median amount of space between parties. Neither Federalism nor compulsory voting proved to be hindrances to voters. However, voters asked to cast a “Single Transferable Vote” (i.e. the Irish) were 24 percent less likely to vote correctly. One final political system or “campaign factor” that proved to be significant was the proxy measure of media availability measured by the per capita distribution of media sources (number of radio, tv, and newspapers per person). Voters with the highest access to media sources are 21 percent less likely to vote correctly then those with only the median amount of access. This suggests that exposure to media may “confuse” or mislead voters into voting against their own interests, which could be a serious problem. The final set of factors in the model dealt specifically with the European Union. Contrary to expectation, there appears to be little evidence of additional effects of a “democratic deficit” on correct voting levels in EP elections. There is no significant difference between voters in founding member states than in new member states. There is no significant difference between those who see EU membership as neutral and those who say it is either good or bad. There also appears to be no significant difference between those who are satisfied with democracy at home or in the European Union. However, it does appear that voters who disapprove of their national government’s job performance are 4 percent less likely to vote correctly.

Does it Make Any Difference? Although more then seventy percent of European voters voted their interests in the 1999 EP elections, the considerable variation between countries suggests that the number of seats held

Hines 19 by the party groups may not reflect the “correct vote.” Interpreting how incorrect voting affected the outcome of the elections is fairly. It simply requires reallocating the seats won by parties in each member state and then totaling those seats by party group in the EP. Table 3 presents the allocation of seats in the Parliament after the 1999 election using actual and correct vote totals. [Table 3 about here] What strikes me about the two seat allocations is how little using correct votes changed the contours of Parliament. Only 7.35% or 46 out 626 total seats in the Parliament changed parties. The change was more pronounced within countries with 17.4% of seats, or 107, changing parties at the national level. Again however, these changes were often eliminated at the supranational level. The largest party group, the European People’s Party (EPP), lost just ten seats. The Party of European Socialists (PES) gained 23 seats, but that represented only a 13 percent increase over their actual seat allocation. The two most substantial changes were with the parliaments third and fourth largest parties. The Liberals (ELDR) would have grown by nearly a third, while the Greens suffered the most substantive decline by losing 21 seats or 44 percent of their actual allocation. The left-wing EUL\NGL had the most stable allocation, losing just a single seat. The Union for Europe of the Nations (UEN) also saw an increase in their total seat allocation of about 19 percent. These shifts in allocations would have resulted in very few changes to the overall working of the Parliament. The EDLR, long the EU’s third party, would simply become even stronger and the EUL\NGL would overtake the Greens as the EP’s fourth largest party. The only really substantial change might be increased presence of the Euroskeptics in the UEN vis-à-vis Greens who become the second smallest in the Parliament. Overall, besides the sharp decline in the size of the Greens, a more informed citizenry would likely only lead to a slight reallocation of parliamentary resources and more balance between the EPP and the PES.

Hines 20

Discussion If we use the percentage of voters who cast a “correct vote” as a measure of the health of democracy in European Parliament elections, then things look pretty good. The percentage of European voters “getting it right” is the same as voters in American presidential elections and higher than the average in general elections in the United Kingdom. While less than the average in Dutch parliamentary elections, voters in several of the member states were close. In fact, there are really only three findings from this study that seem cause for concern. First, there is a small, but still troubling reduction in the ability of female voters to vote correctly in EP elections that does not appear in the US. Second, the Irish and the French seem to have a particularly hard time voting correctly. Besides the effects of Irelands unique STV method of voting and France’s high number of constituencies, there is nothing unique about these two cases to reduce correct voting so much.8 Third, media availability has a negative impact on the ability to vote correctly. This suggests that while actual use of media does not appear to influence the ability to vote correctly, something about a political system that has a high level of media saturation influences voters to vote against their best interests. This warrants further analysis. Evidence for a “democratic deficit” seems weak at best. None of the variables that might be expected to be significant if being a Euroskeptics or Euroenthusiasts affected the ability of voters to vote their interests. Besides the small affects of age and gender, demographic characteristics like education and class do not seem to be significant impediments to correct voting. Of course, levels of correct voting apply only to those who actually vote and if these factors influence certain voters to stay home, no amount of correct voting will overcome them. However, since the decline in turnout appears to be structural (Franklin 2001), and the impediments to correct voting are also structural, structural reform would appear to kill both 8

An entirely ad hoc explanation for the French may be their majoritarian electoral system for national elections causing some kind of confusion in EP elections.

Hines 21 birds with one stone. Ultimately, however, these results suggest to me that the claims of a democratic deficit are potentially overstated. Clearly voters could be more informed about their own interests and 30 percent of them could improve their vote choice, but that is a far cry from the total lack of democratic accountability the most pessimistic observers claim exists in the EU.

Hines 22

References Blondel, Jean, Richard Sinnott and Palle Svensson. 1998. People and Parliament in the European Union. Oxford: Oxford University Press. Commission (2001) White Paper on European Governance, 428 Final, 25 July, Luxembourg: Office of Official Publications of the European Union. Dalton, Russell. 2000. "Political support in advanced industrial democracies" pp. 57-77 in Russell Dalton and Martin Wattenberg, eds., Parties Without Partisans. New York: Oxford University Press. Franklin, Mark N. 2001. "How Structural Factors Cause Turnout Variations at European Parliament Elections." European Union Politics 2(3): 309-326. Jackman, Robert. 1987. "Political Institutions and Voter Turnout in the Industrial Democracies." American Political Science Review 81: 405-23. Jackman, Robert and Ross Miller. 1995. "Voter Turnout in the Industrial Democracies During the 1980s," Comparative Political Studies 27: 467-92. Lau, Richard R. and David P. Redlawsk. 1997. "Voting Correctly." American Political Science Review 91(September): 585-98. Lau, Richard R. and David P. Redlawsk. 2006, forthcoming. How Voters Decide: Information Processing in Election Campaigns. Cambridge: Cambridge University Press. Lau, Richard R., Parina Patel, Dalia F. Fahmy, and Robert F. Kaufman. 2005. "Correct Voting Across 30 Democracies (and 40 Elections)." Paper presented at the Annual Conference of the Midwest Political Science Association, Chicago, IL. Miller, Warren, and Merril Shanks. 1996. The New American Voter. Cambridge MA: Harvard University Press. Putnam, Robert. 2000. Bowling Alone. New York: Simon and Schuster. Sartori, Giovanni (1976) Parties and Party Systems: A Framework for Analysis, Volume I. Cambridge: Cambridge University Press. Schattschneider, Elmar E. (1960) The Semi Sovereign People: A Realist's View of Democracy in America, New York: Holt, Rinehart, and Winston. Schutter, Olivier de (2002) ‘Europe in Search of its Civil Society,’ European Law Journal 8/2: 158-217. Teixeira, Ruy. 1992. The Disappearing American Voter. Washington DC: Brookings.

Hines 23

Tsebelis, George (1994) “The Power of the European Parliament as a conditional agenda setter,” American Political Science Review, 88/1, 128-142. Van der Eijk, Cees, and Mark N. Franklin. 1996. Choosing Europe? The European Union and National Politics in the Face of Union. Ann Arbor: University of Michigan Press. Wattenberg, Martin. 2000. "The decline of party mobilisation" in Russell Dalton and Martin Wattenberg, eds., Parties Without Partisans. New York: Oxford University Press. Wattenberg, Martin. 2002. Where Have All the Voters Gone? Cambridge, MA: Harvard University Press.

Hines 24

Figure 1: Process-Oriented Model of Voter Decision-Making

Hines 25

Table 1: Percent of Voters voting correctly in the 1999 European Parliament Elections Country % Voting Correctly Austria 71.49 Belgium 70.08 Denmark 58.85 Finland 63.43 France 56.78 Germany 74.35 Greece 77.05 Ireland 53.45 Italy 71.09 Luxemburg 65.25 Netherlands 79.14 Portugal 92.86 Spain 84.77 Sweden 67.58 United Kingdom 66.02 EU15 70.57

Hines 26

Table 2: Logistic Regression of Baseline Model of Correct Voting in the 1999 European Parliament Elections Variables Education Age Female Working Class Upper class Religiosity Strength of Partisan Attachment Extreme Ideology Approve Government Record Watch TV New Read Newspapers Political Interest Index Political Knowledge Number of Political Parties Number of Constituencies Ideological Distinctiveness (LR) Ideological Distinctiveness (EU) Federalism Compulsory Voting STV Media Availability Years in European Union EU Good for Member State EU Bad for Member State Satisfied with Democracy (MS) Satisfied with Democracy (EU) Constant N Pseudo R2 Model χ2 (d.f.) Significance P.R.E. Percent Correctly Identified

Coefficients 0.002 -0.005* -0.258* 0.007 0.220 0.043 0.382*** 0.478* -0.075 0.027 0.029 0.181** -0.035 -0.081*** 0.058* -0.072 0.180*** 0.227 -0.108 -0.989*** -0.925*** -0.004 0.098 -0.079 -0.034 0.104 1.630*

Standard Errors 0.005 0.002 0.103 0.101 0.261 0.149 0.048 0.234 0.119 0.021 0.017 0.070 0.075 0.021 0.023 0.082 0.052 0.157 0.129 0.242 0.304 0.004 0.168 0.342 0.135 0.084 0.752

Increase P(CV) -0.046 -0.062

.152 .222

.04 -0.121 0.117 0.099

-0.24 -0.211

4963 0.08 256.77 (26) p < 0.001 12.82% 66%

Note: * represents significance at the .05 level, ** at the .01 level, and *** at the .001 level. The increase in the probability of a correct vote is calculated holding all other values equal to their sample median and increasing the variable of interest one unit. For dichotomous variables, the increase represents the change in probability between the absence and presence of the explanatory factor.

Hines 27

Table 3: Allocation of Seats in European Parliament after 1999 Election using Actual and Correct Votes

Party Group

Actual Vote Country EPP/ED PES ELDR Greens/EFA EUL/NGL UEN EDD N.I. Total

AT 7 7 0 2 0 0 0 5 21

BE 6 5 5 7 0 0 0 2 25

DE 53 33 0 7 6 0 0 0 99

DK 1 3 6 0 1 1 4 0 16

ES 28 24 3 4 4 0 0 1 64

FI 5 3 5 2 1 0 0 0 16

FR 21 22 0 9 11 13 6 5 87

GR 9 9 0 0 7 0 0 0 25

IE 5 1 1 1 1 6 0 0 15

IT 34 17 8 2 6 9 0 11 87

LU 2 2 1 1 0 0 0 0 6

NL 9 6 8 4 1 0 3 0 31

PT 9 12 0 0 2 2 0 0 25

SE 7 6 4 2 3 0 0 0 22

UK 37 30 10 6 0 0 3 1 87

Total 233 180 51 48 42 31 16 25 626

Correct Vote

Party Group

Country AT BE DE DK ES FI FR GR IE IT LU NL PT SE UK Total Change % 4 6 40 1 26 3 29 10 5 32 2 13 10 8 34 223 -10 -4 EPP/ED 6 2 49 2 24 3 37 7 3 23 1 6 12 4 24 203 23 13 PES 0 8 10 7 4 7 0 0 0 0 2 8 0 3 19 68 17 33 ELDR 7 6 0 0 4 0 0 0 2 0 1 0 0 1 6 27 -21 -44 Greens/EFA 0 0 0 1 5 2 10 8 1 4 0 3 1 6 0 41 -1 -2 EUL/NGL 0 0 0 1 0 0 11 0 4 19 0 0 2 0 0 37 6 19 UEN 0 0 0 4 0 0 0 0 0 0 0 1 0 0 3 8 -8 -50 EDD 4 3 0 0 1 1 0 0 0 9 0 0 0 0 1 19 -6 -24 N.I. Total 21 25 99 16 64 16 87 25 15 87 6 31 25 22 87 626 Change 5 4 26 1 2 4 23 2 3 16 1 6 1 4 9 46 % 24 16 26 6.3 3.1 25 26 8 20 18 17 19 4 18 10 7.35 Notes: In the second set of seat allocations, the last two rows represent the total number of seats that changed parties in the country and the percentage of all seats represented by those that changed. The last two columns represent the change in the number of seats held by each party groups and the percentage change that represents over their actual seat allocation.