I2 2-Ehin and Solvak Paper 4-11-2010

8/6/2019 I2 2-Ehin and Solvak Paper 4-11-2010

1/22

1

Party voters gone astray: explaining independent candidatesuccess in the 2009 European elections in Estonia

Paper prepared for the PIREDEU Final User Community Conference, Brussels,

18-19 November 2010

Piret Ehin, University of Tartu,[email protected]

Mihkel Solvak, University of Tartu,[email protected]

Institute of Government and Politics

University of Tartu

Tiigi 78-327, Tartu 50410, Estonia

+372 7375 583 (phone)

+372 7375 5832 (fax)

This paper is a draft. Comments and suggestions are most welcome.
mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]


2/22

2

Introduction

The 2009 European Parliament (EP) elections in Estonia produced something truly unique.

While government parties lost votes as predicted by the second-order national elections theory

(Reif and Schmitt 1980), the main beneficiaries were not opposition parties, large or small. Tothe surprise of observers, the six independent candidates included on the ballot took 30.1 per

cent of the nation-wide vote. A vast majority of these votes went to Indrek Tarand, a former

high-ranking civil servant who had run a campaign marked by strong anti-party rhetoric.

Tarand won a whopping 25.8 percent of the vote, a mere 1046 votes less than the best-

performing party. Having spent less than 2000 euros on his campaign (and borrowed money

from his mother to pay his registration deposit), Tarand secured for himself one of Estonias

six seats in the European Parliament.

This result is unprecedented on many levels. Being elected to the European Parliament as an

independent candidate is a rare accomplishment. Of the 736 MEPs elected to the EP for the

2009-2014 term, only three were elected as independents (one from Romania, one fromIreland, and one from Estonia). Between 1999 and 2007, five individuals entered the EP as

independent candidates (four from Ireland, one from Romania). While the success of

independents in Ireland can be attributed to the countrys candidate-centered electoral system

(Ireland and Malta are the only countries in the EU to use STV in EP elections), the seat

allocation procedure used in the 2009 EP elections in Estonia was closed-list proportional

representationthe type of electoral system considered to be least conducive to the electoral

strength of independent candidates (Brancati 2008). While Romania also uses closed-list PR,

the independents elected in 2007 and 2009 won less than 5% of the nation-wide vote an

outcome much more consistent with the expected effects of the electoral system. Both also

had close ties to certain political parties.1

Tarand, in contrast, was not backed by any political

party, significant interest group or civil society organization.

Tarands case is also unique in the context ofEstonias electoral history. The average share of

votes for independents over the last five parliamentary elections in Estonia has been 1.3

percent. In the general elections of 2007, independent candidates received only 0.1 percent of

the vote. In the countrys first-ever EP elections, held in May 2004, the combined vote share

of four independent candidates was 5.7. Thus, Tarands triumph in the 2009 contest

constitutes a rare example of strong independent performance in a party-centered system that

cannot be explained away by reference to electoral system effects, the peculiarities of the

national political traditions, overt or covert backing by political parties or other influential

organizations, or the candidates own material resources.

This article examines voting behaviour in the 2009 EP elections in Estonia in an attempt to

understand the logic that led over a quarter of the voters to cast a vote for an independent

candidate. The starting point for our analysis is the recognition that EP elections constitute

second-order national elections, where less is at stake for both parties and the voters, and

which the voters use not to express preferences about European integration but to send signals

to the national parties, and in particular, the political incumbents. Within this framework, we

develop two explanations for the voting behaviour observed in the Estonian EP election. The

first explanation builds on the strand of research that links the electoral performance of

independent candidates to the robustness of the voter-party linkage, attributing the triumph of

Tarand to widespread anti-party sentiment in the Estonian society. This explanation coincideswith the dominant domestic interpretation of the results of the 2009 vote in Estonia, and is in


3/22

3

line with the anti-party message that dominated Tarands campaign. A rival explanation,

which combines the predictions of the second-order national elections thesis with the logic of

spatial and cleavage voting, construes the success of Tarand as a variation on the familiar

theme of punishing the political incumbents. According to this explanation, Tarand

constituted a convenient instrumental choice for habitual government party supporters who

wished to punish the incumbents in a context where strong social cleavages and ideologicaldivides inhibited vote-switching to the opposition parties.

Both explanations yield a number of empirical predictions about patterns of vote-switching

and voter characteristics that we test with individual-level data from a postelection survey

conducted in the framework of the European Elections Studies. The main empirical question

addressed in this article is: what kind of voters voted for Indrek Tarand and how do they differ

from party voters and non-voters? The results of the empirical analysis enable us to answer

the question whether the vote for Tarand should be interpreted as a vote against political

parties or a vote against political incumbents.

While the immediate objective of this study is to understand the logic of the vote in theinherently intriguing Estonian case, the study also contributes to the study of voting behaviour

more generally. Although interest in independents appears to be on the rise, we still know

relatively little about why people vote for independent candidates and under what

circumstances they might do so en masse. The Estonian case is particularly well suited for

examining the question of how independent voters differ from party voters and non-voters:

due to the extraordinarily large number of votes for an independent candidate, the analysis

does not suffer from the problem of small n that has plagued studies on small party and

independent votes (van der Eijk et al 2006, 438). Second, an analysis of the Estonian case

helps cast light on the question of why people vote differently in EP elections and first-order

elections. While voting differently has been generally understood as voting for a different

party, the Estonian case suggests that the second-order setting has characteristics that may

lead voters (and candidates) to circumvent parties altogether.

The rest of this article is structured as follows. The first part explains how voting in second-

order elections differs from voting in first-order elections, and shows how certain structural

conditions associated with the SOE setting may strengthen the position of independent

candidates. The second part outlines two versions of protest-voting (anti-party vote or anti-

incumbent vote) and identifies a number of empirical propositions about voting behaviour

consistent with each account. The third section presents a brief overview of the 2009 EP

election in Estonia, assessing the degree to which the macro setting and aggregate results lend

plausibility to the competing explanations. The fourth section uses data from the EES toempirically test the propositions about voting behavior derived from the anti-party and anti-

incumbent accounts of mass vote for an independent candidate. After examining patterns of

vote-switching, it presents the results of descriptive discriminant analysis that tell us how

independent voters differed from party-voters and non-voters. The final section summarizes

the main results and elaborates on the implications of our findings.

Why voters vote differently in EP elections

People vote differently in EP elections than they do in general elections: fewer turn out, and

those who do are more inclined to support opposition parties than government parties. They

are also more prone to vote for small parties - including protest parties and ideologicallyextreme parties - than for the mainstream parties they would support in a general election. A


4/22

4

powerful explanation for these trends is offered by the second-order national elections (SOE)

thesis, a dominant approach to EP elections that has generated a wealth of theoretical and

empirical research (Reif and Schmitt 1980, Reif 1984, Marsh 1998, Koepke and Ringe 2006,

Hix and Marsh 2007, Schmitt 2009). According to this model, EP elections are national

because voters choose among national parties who run campaigns focusing on national issues.

They are second-order because in contrast to first-order elections, they have almost noimplications for the allocation and exercise of executive power. Because government

formation is not an issue, less is at stake in EP elections for both parties and their voters.

Against this backdrop, van der Eijk and Franklin (1996) have identified three modes of voting

in EP elections which they call voting with the head, voting with the heart, and voting

with the boot. Voting with the head implies that voters consider the strategic implications of

their vote just as they do in first-order elections. Voting with the heart, or sincere voting,

involves casting the ballot that the voters most prefer, without taking strategic considerations

into account. Finally, voting with the boot constitutes a form of expressive or protest voting in

which voters cast their ballot for a party they would not vote for in a real election in order to

send a message of distaste for the programs or candidates of the party they would normallyvote for (Franklin 2005: 5).

The prevalent understanding is that in EP elections, voters are much more likely to vote with

the heart or with the boot than in first-order elections. Because government formation is not at

stake, voters are freed from the need to consider the strategic implications of their vote. The

second-order setting thus encourages sincere voting i.e. voting based on ideological

proximity, group identities, or issue-positions. Voting with the boot, however, may also

involve insincere or instrumental voting (Marsh 1998). Voters who voted on the basis of

sincere preferences in first-order elections may, in second order contests, choose a party

different from the one they most closely identify with, in order to signal discontent with the

party that received their vote in general elections. The second-order setting allows the votersto safely engage in such punitive vote-switching because it does not alter the status quo in

terms of the exercise of executive power.

To the extent that strategic calculations distort true preferences in first -order elections by

making voters support larger parties, and to the extent voters are drawn to actors peripheral to

the political system to express protest in second-order elections, small and insignificant actors

are expected to do better in the SOE setting. This logic applies equally well to independent

candidates as to small and peripheral political parties. In fact, the SOE setting appears to offer

a number of advantages to independent candidates, relative to the first-order arena. The

propensity of voters to vote strategically by abandoning those considered unlikely to win seats

constitutes a major obstacle to the electoral strength of independent candidates in general

elections. Limited or non-existent government potential, however, does not constitute an

impediment in the second-order setting where the elections involve the selection of a

representative rather than a government (Marsh 1998: 593). Independent candidates have

better chances of success in a beauty contest where voters act based on sincere preferences

without considering the political weight of the parties and candidates appearing on the ballot.

Independents are also potential beneficiaries of the anti-incumbent trend, which makes a large

share of the voters to reconsider the choice they made in general elections, and increases their

susceptibility to the electoral appeal of alternative actors. Finally, the SOE context also

reduces the tremendous financial and organizational disadvantages that independent

candidates face when competing with political parties for political office. Because less is atstake at EP elections, parties spend significantly less on EP election campaigns than they do


5/22

5

in general elections. To the extent that party campaigns are low-cost and low-intensity,

compared to the first-order arena, independents have a better chance of gaining visibility and

getting their message across to the voters.

The above suggests that the second-order setting constitutes part of the explanation for the

300-fold difference in the vote share won by independent candidates in first-order and second-order elections in the Estonian case. While voting with the boot constitutes a plausible

explanation for mass vote for a candidate running on a protest platform, it is not immediately

clear who or what the Estonian voters protested against. Two possible explanations suggest

themselvesone derived from the literature on independent candidate performance, the other

building on the insights generated by the SOE model of EP elections.

Anti-party sentiment as a source of support for independents

The case-study rich literature on independent candidates has generated few general

explanations about the sources of the electoral appeal of independents. Aside from the body of

research focusing on election system effects (which will not be examined here because wehave a n of 1, and because the Estonian case conflicts with the basic propositions of this

approach), this literature suggests that the electoral strength of independents is linked to the

robustness of the voter-party linkage. The propositions associated with this approach are

straightforward. On the individual level, voters who do not feel close to any party are

considered to be more likely to vote for independents than voters with a stronger sense of

party attachment (Rosenstone et al 1984, 176-177; Donovan et al 2000, 60). On the aggregatelevel, independents can be expected to perform comparatively well in contexts characterized

by weak or undeveloped political parties, low levels of partisan identification or party

attachment, and high levels of electoral volatility (Gerring 2005). Thus, independents have

been shown to perform better in first democratic elections than in subsequent ones (Brancati

2008), and in new democracies compared to established ones.

A stronger version of this logic links independent success to the prevalence of anti-party

sentiment among the electorate. Discontent with political parties has been recognized as one

of the elements of the political malaise observed in advanced industrial democracies (Dalton

2004; Norris 1999), associated with growing political cynicism, low levels of political trust,

and declining levels of political participation. According to Poguntke (1996), anti-party

sentiment can be specific, reflecting voter dislike of major party alternatives, or generalized,

implying a rejection of party politics per se. Anti-partism has been associated with the

tendency to vote for independent candidates (e.g. Owen and Dennis, 1996) as well as minor

parties, especially if these parties adopt antiparty rhetoric or mobilize popular disenchantment

by offering new ways of doing politics (Belanger 2004). Specific events that increase

popular resentment towards party politicians (such as the MPs expenses scandal in the UK)

may increase support for independents. Also, a specific type of small parties - radical right

protest parties - have been shown to profit from popular dissatisfaction and anti-systemic

attitudes by appealing especially to a less educated and lower income voter segment (Lubbers

& Scheepers 2000). Aside from this possibility, existing studies suggest that independent

candidates and anti-party parties may perform important functions in a democratic political

system. They serve as vehicles channelingpolitical discontent in a pacific and democratic

way (Belanger 2004) and may even unlock voter apathy by providing disenchanted voters

more choice and alternative means of representation.


6/22

6

Punishing the incumbents: variations on a theme

The literature on European Parliament elections suggests a different interpretation of the

protest-vote observed in the Estonian case. One of the central claims of the SOE model is that

voters use EP elections to punish political incumbents. The fact that government parties suffer

losses in EP elections has been widely corroborated. This happens because EP elections takeplace in the middle of the national election cycle: as governments tend to disappoint voters,

support for incumbents is expected to decline as a function of time since last general

elections, with the anti-incumbent trend being most pronounced if EP elections are held in the

middle of the national election cycle (Marsh 1998, Reif 1984). Another reason why

government parties lose is that in second-order elections, voters can safely signal discontent

with the incumbents without having to consider the consequences of their protest votes for the

composition of the government.

The question whetherand under what conditions - the wish to punish incumbents could lead

voters to cast a ballot for an independent candidate leads us to consider the choices that

habitual government party voters face in EP elections. Basically, this choice can becharacterized in terms of the classic trilemma of loyalty, exit, or voice (Hirschman 1970).

Which option the voters choose will have major implications for both turnout and the

distribution of the vote. The choice between exit and voice will be strongly influenced by

the extent to which the electoral choice set includes convenient, low-cost options for signaling

dissatisfaction with the incumbents. While there is likely to be no shortage of actors

criticizing the government on the electoral menu, not all options will be equally acceptable or

attractive to the disenchanted government party voter. Indeed, instrumental vote-switching is

likely to entail potentially significant costs, as voters may have to negotiate ideological

distances and cross deeply-embedded social and political cleavages. The logic of spatial and

cleavage voting suggests that in choosing among potential receptacles of the anti-incumbent

vote, voters would prefer a party or a candidate whose ideological position and group identity

are similar to their ownas long as the choice does the job of sending a signal of discontent

with the government. This line of argument suggest that in a context where strategic

considerations matter less, or not at all, voters who wish to punish the incumbents will be

drawn to political actors that allow them to vote with the boot without betraying the heart.

Testable propositions

These two rival accounts of independent success yield a number of empirically testable

propositions. The key to decoding the signal the voters were sending in the Estonian EP

elections is finding out which voters abandoned their parties to vote for an independentcandidate. The explanation that combines the SOE model with the predictions of spatial and

cleavage voting yields the following empirical predictions about patterns of vote-switching,

relative to previous general elections, and the characteristics that distinguish independent

voters from party-voters and non-voters:

H1. Vote-switching from government parties to independent candidate was widespread.

H2. Vote-switching from government parties to opposition parties remained limited.

H3. The independent candidate mobilized few habitual abstainers.

H4. Voters who voted for an independent candidate did not differ from party-voters in terms

of the strength of partisan affiliations.

H5. Voters who voted for an independent candidate evaluated government performance more

negatively than voters who voted for government parties.


7/22


8/22

8

accession to the eurozone, energy security, EU policy towards Russia), almost no clearly

identifiable focal points for political debate emerged. Promises focusing on purely domestic

matters received more air time and media space than the parties visions of European

integration. Predictably, opposition parties used the elections as an opportunity to blame the

government for the economic crisis and social problems, zooming in on the hardships brought

by the budget cuts.

About a month before the election day, polls began to predict an EP seat to Indrek Tarand.

Tarand is a former high-ranking civil servant, a talk-show host, and the director of the

Estonian War Museum who had never held elected political office. He ran on an anti-party

platform, ridiculing the prevailing culture of party politics and criticizing the self-serving

ways of Estonias political elites. In particular, Tarand attacked the switchover from open to

closed party lists that the major parties had engineered specifically for the 2009 EP elections,

arguing that this change of electoral rules deprived voters of a real choice and made the

selection of Estonias MEPs a matter of party backroom politics . Although Tarand had very

limited campaign resources, bold and occasionally scandalous confrontations with the major

political parties helped him secure a strong media presence.

The seed sown by Tarand fell into a well-prepared soil. Popular frustration with the political

parties has been a major theme in the Estonian political discourse for well over a decade, andthe diagnosis of growing popular alienation from politics has been widely shared by the

media, politicians and opinion leaders, as well as social scientists (e.g. Arter 1996, Lagerspetz

and Vogt 2004). Surveys have recorded very low levels of popular trust in political parties

since the early 1990s (Ehin 2007); in June 2009, only 16% of the respondents in a

Eurobarometer survey said they trusted the parties (TNS Opinion & Social 2009). In terms of

electoral behavior, systemic dissatisfaction with the established parties has translated into

high levels of electoral volatility (the Pedersen index for the 1990-2003 period was 17.7%)

and a remarkable success of new parties promising new politics (Sikk 2006). For instance,

genuinely new parties gained 24.6% of the vote in the 2003 national election and 7.1% in

2007.

While Tarand sent few signals about his ideological leanings during the campaign, his track

record provides evidence of strong links to the social and political forces associated with the

parties in government. He is a prominent alumnus of an influential student society that, in the

early 1990s, had served as a breeding ground for the centre-right ethno-nationalist politicians

that had formed Estonias first post-communist government under the premiership of Mart

Laar. In 2009, Laar served as the incumbent chairman of IRL. Tarand is also the son of

Andres Tarand, one of the key figures in SDE, prime minister 1994-1995, and a member ofthe European Parliament 2004-2009. In addition to his centre-right leanings and embracement

of the ethnonationalist position associated with IRL, his standing vis--vis the still-relevant

question of assessing the countrys past was equally clear: in 2005, Tarand caused a scandal

by appearing in public wearing a T-shirt that called for burning the commies, followed by

the list of the names of active politicians who had been members of the Communist Party in

the Soviet era. Despite his close ties to IRL and the SDE, Tarand served a perfect choice for

punishing the Reform Party, due to a nasty, widely publicized personal conflict with Kristiina

Ojuland, former foreign minister and the top candidate on the Reform Partys electoral list.

Even though Estonia registered the greatest increase in voter turnout compared to the 2004 EP

elections among all EU countries (from 26.8 % in 2004 to 43.9 % in 2009), turnout in the EPcontest was much lower than in the previous national parliament elections (61.9 %) and


9/22

9

subsequent local elections (60.6%). The six mandates were allocated between four parties and

one independent candidate (see Table 1). The main opposition party, the Centre Party, gained

the largest share of the vote (26.1 %) and secured two seats in the EP. Tarand came in second,

having won only 1,046 votes less than the Centre Party. Indeed, Tarand gained the largest

share of the vote in all 17 regions (15 counties and two major cities) except Tallinn and Ida-

Viru county, which have a high percentage of Russian-speakers. The leading governmentforce, the Reform Party, was the main loser, as its vote share dropped from 27.8 % in the

2007 general elections to 15.3 % in EP elections. The two smaller parties represented in the

Riigikogu, the Greens and the Peoples Union, remained without a mandate, receiving 2.7 and

2.2 per cent of the vote, respectively.

Table 1. Results of 2009 European Parliament Elections in Estonia

EP party

group

affiliation (at

time of

election)

Number

of votes

% of votes (change compared

to 2007 general elections)

EP seats

Keskerakond

(Centre Party)

ALDE 103,506 26.1 (0) 2

Indrek Tarand

(independent candidate)

- 102,460 25.8 (NA) 1

Reformierakond (Reform Party) ALDE 60,877 15.3 (-12.5) 1

Isamaa ja Res Publica Liit

(ProPatria and Res Publica

Union)

EPP 48,492 12.2 (-5.7) 1

Sotsiaaldemokraatlik erakond

(Social Democratic Party)

PES 34,508 8.7 (-1.9) 1

Erakond Eestimaa Rohelised(Greens)

Greens-EFA 10,851 2.7 (-4.4) 0

Martin Helme (independentcandidate)

- 9,832 2.5 (NA) 0

Eestimaa Rahvaliit (Peoples

Union)

UEN 8,860 2.2 (-4.9) 0

Dimitri Klenski (independent

candidate)

- 7,137 1.8 (NA) 0

The table lists parties and independent candidates that received over 1% of the vote. Source: Estonian National

Electoral Committee, www.vvk

Who voted with the boot?

To test the hypotheses about vote-switching and differences among the groups of voters, weuse data from the voter survey of the European Elections Study (2009). A total of 1007

interviews (707 face-to-face and 300 phone interviews) were carried out in Estonia in June

2009, a few weeks after the elections. It is a nationally representative sample using non-

response weighting. Like most political surveys, the sample overrepresents voters at the

expense of non-voters but this does constitute a problem for our analysis because we are

concerned with comparing groups of voters, rather than estimating turnout or predictingvoting results.

2

For the purposes of this analysis, voters who voted for the Reform Party, IRL or SDE in the

2009 EP elections were grouped as coalition voters, while voters who voted for any other

party were classified as opposition voters. In the analysis that follows, we only include thoseindependent voters who voted for Tarand, while excluding the voters who voted for one of the


10/22

10

other five independent candidates on the ballot (who collectively received 4.3% of the vote).

Because the independent candidates in question differed greatly in terms of ideology, personal

profile, and campaign messages, we do not expect their voters to constitute a uniform group.

As this article is concerned with explaining the success of Tarand, grouping Tarand-voters

with the supporters of the other independent candidates would introduce unnecessary noise

into the analysis.

Patterns of vote-switching

A cross-tabulation of vote choice in EP elections by vote choice in 2007 general elections

(Table 2) shows that significant vote-switching took place and offers substantial evidence that

Tarand succeeded in attracting the votes of habitual government party voters. Over one half of

the voters (52.5 %) who voted for Tarand in 2009 had, in the 2007 general election, voted for

the three parties that formed the government (Reform Party, IRL, and SDE) following the

election. In contrast, only 17.3 % of the Tarand-voters had voted for any of the opposition

parties in 2007. Indeed, Tarand appears to have been the number one choice for voters

defecting the Reform Party, IRL or SDE. Vote-switching from the government to theopposition, in contrast, remained very limited. For instance, none of the IRL-voters in our

sample switched votes to the opposition parties, while 13.9 % of the Reform Party voters did

so. The mobilization of non-voters was limited and divided with opposition parties (mostly,

the Centre Party). Thus, these initial results offer strong support to all three expectations

about patterns of vote-switching: defection from the government parties was widespread, and

most of the defectors supported Tarand, while few voters who had supported the Reform

Party, IRL or SDE in 2007 switched to the opposition parties.

Table 2. Vote choice in 2009 EP elections by vote choice in 2007 general elections (n, %)

EP election 2009National election

2007

Reform

Party

IRL Centre

Party

SDE Peoples'

Union

Greens Tarand Didn't vote

Reform Party 49 (77.8) 13 (18.8) 16 (8.0) 12 (19.4) 1 (5.9) 0 (0.0) 54 (29.2) 45 (13.8)

IRL 3 (4.8) 39 (56.5) 0 (0.0) 4 (6.5) 0 (0.0) 0 (0.0) 22 (11.9) 18 (5.5)

Centre Party 1 (1.6) 3 (4.3) 126 (63.3) 7 (11.3) 1 (5.9) 1 (9.1) 16 (8.7) 46 (14.1)

SDE 2 (3.2) 6 (8.7) 3 (1.5) 22 (35.5) 1 (5.9) 1 (9.1) 21 (11.4) 16 (4.9)

Peoples Union 1 (1.6) 0 (0.0) 1 (0.5) 1 (1.6) 0 (0.0) 2 (18.2) 6 (3.2) 7 (2.1)

Greens 0 (0.0) 2 (2.9) 1 (0.5) 4 (6.5) 8 (47.1) 6 (54.6) 10 (5.4) 5 (1.5)

Didn't vote 3 (4.8) 0 (0.0) 32 (16.1) 2 (3.2) 4 (23.5) 0 (0.0) 23 (12.4) 131 (40.1)

DK 4 (6.3) 6 (8.7) 20 (10.1) 10 (16.1) 2 (11,8) 1 (9.1) 33 (17.8) 59 (18.0)

Total 63(100.0)

69(100.0)

199(100.0)

62(100.0)

17(100.0)

11(100.0)

185(100.0)

327(100.0)

Ideological constraints on vote-switching?

Why did voters disenchanted with the government not switch vote to one of the opposition

parties, such as the Centre Party that ran a campaign accusing the government of incompetent

and socially insensitive handling of the economic crisis? The mean ideological distances

between self and party, reported in Table 3, provide evidence of spatial voting in the Estonian

EP election (although it should be noted that the groups were not uniform in their evaluations,

as suggested by the large standard deviations). Overall, both coalition and opposition partyvoters appear to have voted for parties they considered to be closest to their own position on


11/22

11

the left-right scale. Although we lack data on how voters perceived the ideological position of

Tarand, the figures in Table 3 make clear that Tarand-voters perceived opposition parties

(with the exception of the Greens) to be ideologically much more distant from their own

position than the coalition parties. In particular, the perceived ideological remoteness of the

Centre Party suggests that for many habitual government party voters, switching vote to the

main opposition force to punish the incumbents would, on the terrain of Estonian partypolitics, amount to crossing the Rubicon. Overall, the figures in Table 3 lend strong support to

the conjecture that for voters belonging to the coalition party gene pool, switching vote to

the opposition parties (again, with the possible exception of the Greens) would have involved

betraying the heart.

Table 3 . Mean ideological distances between voters and parties

Coalition parties Opposition parties

VoterReform

Party IRL SDE

Centre

party Greens

Peoples

Union

Coalition -.42 (2.50) -.46 (2.41) 2.35 (2.60) 4.17 (3.88) 1.74 (2.85) 2.81 (2.94)

Opposition -2.75 (4.03) -2.30 (3.75) -.40 (3.24) -.07 (2.91) -.25 (3.02) .18 (3.48)

Tarand -.90 (3.74) -.68 (3.03) .52 (2.67) 2.75 (3.49) .39 (2.72) 1.56 (3.10)

Non-voter -.77 (3.80) -.50 (3.41) 1.02 (3.12) 1.38 (3.59) .34 (2.82) 1.18 (3.24)

The figures are mean distances between the placement of self and party on the left-right scale ranging from 0 to 10, with

standard deviations in parenthesis.

What distinguishes independent-voters from party-voters and non-voters?

Method and model specification

To explore whether, and in which way, independent voters differ from party-voters and non-

voters, we employ a statistical technique known as descriptive discriminant analysis or DDA

(Huberty 1994). The idea of DDA is to find out whether groups, as defined by the grouping

variable, differ statistically according to some function of the traits of the units (called

outcome variables in DDA). DDA is therefore well suited for an exploratory analysis on how

many latent dimensions and in what way contribute to group separation. This is achieved by

combining outcome variables mathematically into a synthetic variable that differentiates

maximally between the groups. In total n-1 such synthetic variables or discriminant functionsare calculated, were n is the number of groups or outcome variables, whichever is smaller.

The first function gives the best separation between the groups, while the second function,

which is orthogonal to the first, is calculated to separate best between the groups by using the

associations left out of the first function. This process is repeated until all discriminating

dimensions have been evaluated (Tabachnik and Fidell 2001: 459). The separating power of

each function can be evaluated to determine if some dimensions differentiate between all the

groups, some others between one group and all the rest and so forth. The technique gives

therefore more information on potential group differences than logistic regression which has

been used by previous studies examining vote choice in EP elections.3

Performing DDA requires that we define a set of outcome variables, or traits which webelieve distinguish the four groups from one another. In defining this set of variables, we rely


12/22

12

on existing studies of vote choice in EP elections, most notably van der Eijk & Franklin

(1996) and Oppenhuis (1995). These studies have constructed composite models of vote

choice, integrating insights from a range of voting theories. These models have included

individual-level measures of socio-political cleavages, ideology, attitudes towards European

integration, perceptions of government performance, partisanship, and socio-demographic

controls. We include essentially the same set of variables in our model, while making minoradjustments to adapt the model to the Estonian context as well as our data. Below, we offer a

brief summary of the variables included in the model, while a detailed description of survey

questions and coding choices are included in Appendix 2. We shall not repeat the theoretical

justification for the inclusion of the variables in the model because most of the items are

standard variables used in studies of vote choice, and because a more detailed theoretical

expositions can be found in the studies referred to above.

We employ three measures of social cleavages, including social class, ethnicity and the urban-

rural divide. While religion is widely used in Western European studies on vote choice, it

does not constitute an important cleavage in the non-religious Estonian society and is thus

excluded. Instead we use the ethnic cleavage. Although there are no significant ethnic partiesin Estonia, the Russian minority has been over time very uniform in its support for one of the

mainstream parties (the Centre Party). There are also significant and persistent differences

between the titular nationality and the Russian-speakers in terms of political values and

attitudes (Ehin 2007; Vihalemm & Kalmus 2009). The ethnic divide is hence politicised and,

to an extent, institutionalized. We use two measures to capture ideology: one is the traditional

left-right self-placement, while the other captures the new values dimension. The latter

measure in an index combining attitudes towards same sex marriages, right of abortion for

women, support of harsh punishment for criminals and teaching children authority in school.

To measure attitudes towards European integration, we employ two measures. The first

frequently used interval measure capturing attitudes towards further European integration (0

gone too far; 10 should be pushed further), while the other is assessment of countrys

membership in the EU (good, bad, neither). For government performance we use the standard

item on evaluation of the government record to date with a binary approve or disapprove

answer. For partisanship, we construct a measure combining the standard item on party

closeness with a three-category ordinal item on the strength of this bond. We also include

socio-demographic variables such as socio-economic status, education and age as these have

been demonstrated to influence electoral participation as well as vote choice (Blais et al 2009;

Aarts & Wessels 2005; Blais et al 2004; Caballero 2005; Franklin 2003; Matsusaka & Palda

1999). Those measures also act as proxies for personal efficacy, as a higher socio-economic

status together with education combine into a higher cognitive ability to understand politics.

Finally, the stability of communal ties is included as those form part of the wider stable socialbasis that has been demonstrated to stabilize voting decisions and influence political

participation (Giles & DAntico 1982; Zuckerman et al 1998). Thus, the socio-economic

variables included in the analysis are age, the level of education, a seven-category ordinal

item on self-reported family living standards and an item on how long the respondent has

been living in the community.

Results of the DDA

On the basis of variables inserted into the model, DDA produced three functions (analogous

to a factor in factor analysis) that differentiate between the groups (see Table 4).4

DDA needs

to be interpreted in three steps, first establishing the statistical and substantive significance ofthe results, then evaluating which variables define the bulk of the given functions and finally


13/22

13

exploring which groups score how on these functions. Wilkss lambda helps us to evaluate the

statistical significance of the function. It can range from 0 to 1 with a low value indicating

that group means differ and the value of 1 that they are the same. The substantive significance

of the functions can be evaluated by examining the canonical correlations (Rc). These show

the correlation between the grouping variable and the discriminant function created out of the

outcome variables. Squaring them (Rc2

) shows how much of the variance in this correlation isaccounted for by the particular function.

The first function has a large canonical correlation (.5) accounting for 25.0% of the variance

in the correlation between the grouping variable and the synthetic variable. The second

function has more modest explanatory power, accounting for 14.59% of this variance. Both

are statistically significant at p


14/22

14

Table 5. Relative contribution of the variables to group separation

Function 1 Function 2

Variables Coefficient rs rs2 Coefficient rs rs

2

Cleavage voting

Social class .207 .367 13.47 % .066 .175 3.06 %

Ethnicity -.382 -.597 35.64 % .317 .368 13.54 %

Rural-urban divide -.002 -.058 .34 % .389 .526 27.67 %

Ideology

Left-right .395 .609 37.09 % -.186 -.172 2.96 %

Social liberalism** .177 .139 1.93 % -.046 -.021 .04 %

European integration

EU integration -.020 .200 4.0 % .086 .033 0.11 %

Membership evaluation -.231 -.475 22.56 % -.191 -.101 1.02 %

Government performance -.440 -.648 41.99 % -.204 -.139 1.93%

Party attachment** .005 .144 2.07 % .410 .426 18.15 %

Controls

Age .183 .043 .18 % .176 .336 11.29 %

Education* .085 .203 4.12 % .388 .421 17.72 %

Living standards .036 .394 15.52 % -.048 -.150 2.25 %

Community roots* -.001 -.164 2.69 % .355 .504 25.40 %

* Square root transformed

** Reflected and square root transformed

The second function is defined primarily by the rural-urban divide, the length of living in the

community and party attachment. While the urban-rural divide contributes most to group

separation in this function, one cannot pin a neat overall label to the function because the

other variables are unrelated to this cleavage. Educational levels contributed also somewhat to

group separation in this function.

Group centroids are graphed in Chart 1. Centroids are the mean discriminant scores for the

four groups. In other words, they show which groups have more or less of the qualities

included in the function. Higher values on the first function stand for satisfaction with

government performance, subscribing to a right-of-the centre ideology, being ethnic Estonian,being satisfied with the countrys EU membership and higher self-ascribed class status. On

this function (Chart 1, x-axis), the group centroid is overwhelmingly highest for coalition

party voters, followed by Tarand-voters, non-voters, and opposition party voters.

The second function produces a different constellation of the four groups (Chart 1, y-axis).

Opposition and coalition party voters score closely and Tarand voters are clearly closer to

partisans than to non-voters. One could pit the non-voters against all the rest on this function.

Party voters, quite obviously, have a stronger sense of party attachment, are more urbanized

and have lived longer in their community. Tarand-voters are again closer to party voters than

to non-voters. This means that turnout was higher among the urbanized population and those

with a stronger sense of partisan ties.


15/22

15

Chart 1. Group centroids

Note: TVTarand voters (N=104); CVcoalition voters (88); OVopposition voters (164); NVnon-voters (160)

These results lead to the following conclusions about empirical support to the propositions we

derived from the SOE-model and the two rival versions of protest-voting. As predicted by the

SOE model and theories of spatial voting, vote-switching was rampant and occurred mostly

because voters switched from the coalition parties to Tarand. This largely confirms

expectations 1 and 2. The result that Tarand mobilized a relatively small share of habitual

non-voters (and was no more successful in this than the leading opposition party) lend

stronger support to expectation 3, associated with the anti-incumbent explanation, than toexpectation 10, derived from the anti-party sentiment explanation. Evidence for expectation 4

and 8 is somewhat mixed as the DDA showed Tarand voters to be closer to partisan voters on

the function that was influenced by partisanship, but the strength of partisan ties did not playany role in the first function that contributed much more strongly to group separation. Overall,

however, the results do not allow us to describe Tarand-voters as having strong anti-party

sentiments. Expectation 5 was also confirmed, as Tarand voters scored clearly lower on the

function that was defined, to a large extent, by evaluations of government performance. An

examination of ideological distances between groups of voters defined by vote choice and the

major parties produced strong evidence of spatial voting, thus lending initial support to

expectation 6. This result was corroborated by the multivariate analysis that showed Tarand-

voters to be closer to coalition party voters and different from opposition voters in terms ofideology, group identity, and social status. The results strongly reject the expectation (9) that

Tarand-voters share the profile of a typical protest-voter (less educated, lower socio-economic

status). Finally, evidence regarding expectation 7 was mixed: while EU membership

evaluation played a role in differentiating among the groups, its overall impact in the model

was however clearly smaller than that of cleavages or government performance. Overall, thus,

the results lend substantially stronger support to the claim that Tarand-voters were habitual

government party voters who had decided to punish the incumbents than to the alternative

explanation according to which the mass vote for Tarand signified widespread anti-party

sentiment.

TV (.329; -.051)

CV (1.062; .253)OV (-.568; .442)

NV (-.216; -.558)

-1

-0,5

0

0,5

1

-1,5 -1 -0,5 0 0,5 1 1,5

Function2

Function 1


16/22

16

Conclusions

This article set out to explain an anomaly: the spectacular electoral performance of an

independent candidate in a party-centered system. The explanation developed in this article

links the electoral triumph of an independent candidate to the structural conditions

characteristic of the second-order setting which free the voters from the need to consider thestrategic implications of their vote, while increasing the prevalence of sincere voting and

protest voting. Against this backdrop, the article developed two explanations for the mass vote

for Indrek Tarand in the Estonian 2009 EP elections, suggesting that the strong support for

Tarand could be interpreted either as anti-party vote or as anti-incumbent vote. It developed a

number of empirical propositions about voting behavior consistent with each account, and

assessed the empirical validity of these, using data from a postelection survey.

Our findings lend considerable support to the proposition that mass vote for an independent

candidate in the Estonian EP elections signified protest against the incumbent government in

mid-term elections that coincided with an economic crisis of unprecedented proportions. The

results of discriminant analysis show that the most important variable that distinguishedgovernment party voters, opposition voters, Tarand-voters and non-voters from one another

was the voters evaluation of government performance. In terms of socio-demographic

characteristics and ideological leanings, Tarand-voters were closer to coalition party voters

than to opposition party voters or non-voters. They were more likely to be ethnic Estonians

with higher self-ascribed class status who lived in the cities, subscribed to centre-right

ideological views and were happy with their countrys membership in the EU. The fact that

these features do not match the profile of a typical protest-voter (less educated and potentially

unemployed male with low socio-economic status) lends additional plausibility to the claim

that a vote for Tarand signified (temporary) dissatisfaction with the incumbent centre-right

government, rather than a rejection of the the system. Indeed, it is highly likely that many of

the voters who cast a ballot for an anti-party independent candidate in the relatively

inconsequential beauty contest that the EP elections constitute, will return to their parties at

the next general election to help prolong the emerging national tradition of centre-right

governments.

Overall, the findings are very much in line with the logic of the second-order model of EP

elections, while conflicting with the dominant domestic interpretation of Tarands success.

However, instead of just confirming a well-established theory, this study has also expanded

the explanatory scope of the SOE model. It has done so in three ways. First, by spelling out

the implications of the SOE setting for independents and empirically confirming the

predictions derived from the SOE approach in the Estonian case, this study has shown howthe dominant theoretical approach to EP elections can accommodate a highly unusual

electoral outcome. Second, this study suggests that the literature on EP elections would

benefit from a closer integration of the SOE model with general voting theories, such as

spatial voting or cleavage voting, as well as from a more careful analysis of electoral choice

sets. In this context, the question of how the structure of the choice sets influences the choice

between loyalty, exit, and voice that the government party voters face in EP elections, is a

particularly interesting and important one. Third, this study has helped refine the concept of

voting with the boot, suggesting that various forms of protest-voting may occur in EP

elections, and suggesting a way to assess the empirical validity of alternative explanations. On

a related note, the contribution of this study to the literature on anti-partism lies in a call for

caution when interpreting the intentions of voters who cast a ballot for parties or candidates


17/22

17

running on an anti-party platform. Beneath the rhetoric, other structural forces may be at

work.

The fact that we were not able to systematically examine the possible link between

independent success and electoral system change (from open to closed party lists) constitutes

a limitation of our study. Thus, our data does not allow us to answer the question whether theEstonian voters really have been socialized into voting for candidates, rather than parties, or

whether the voters purposefully punished the parties (or more specifically, government

parties) for tampering with the electoral rules. While we acknowledge the possibility that

amending the rules of the game may have helped to destabilize voting decisions, thus

contributing to the unprecedented success of an independent candidate, we hope that future

studies are able to cast more light on the effects of electoral rule change on voting in both

first-order and second-order elections.


18/22

18

Appendix 1. Variables included in the DDA

CleavagesIf you were asked to choose one of these five

names for your social class, which would you sayyou belong to?

1 - working class; 2 - lower middle class; 3 -

middle class; 4 - upper middle class; 5 - upperclass

Do you see yourself as Estonian or Russian? 1 - Estonian; 2Russian

Would you say you live in a... 1 - rural area or village; 2 - small or middle-sized

town; 3 - suburbs of large town or city; 4 - large

town or city

IdeologyIn political matters people talk of the left andthe right. What is your position? Pleaseindicate your views using any number on a scale

from 0 to 10, where 0 means left and 10 means

right.

0 - left; 10right

Index of the following items (For each of the

following statements, please tell me to what

degree you agree or disagree with each

statement):

- Same-sex marriages should be prohibited by

law

- Women should be free to decide on matters of

abortion

- People who break the law should be given

much harsher sentences than they are these days

- Schools must teach children to obey authority

1 - strongly agree; 2 - agree; 3 - neither agree nor

disagree; 4 - disagree; 5 - strongly disagree

European integrationSome say European unification should be pushed

further. Others say it already has gone too far.

What is your opinion?

0gone too far; 10should be pushed further

Generally speaking, do you think that Estoniasmembership of the European Union is a good

thing, a bad thing, or neither good nor bad?

1 - good thing; 2 - neither; 3 - bad thing

Government performanceDo you approve or disapprove the governmentsrecord to date?

1 - approve; 2 - disapprove

Party attachmentCombination of the following items:

-Do you consider yourself to be close to any

particular party?

-Do you feel yourself to be very close to this

party, fairly close, or merely a sympathiser?

1 - very close; 2 - fairly close; 3 - merely

sympathiser; 4 - not close to any party

ControlsAge

Level of education

What is your family living standard 1 - poor family; 7 - rich family

How long have you been living in your current

community [years]


19/22

19

References

Aarts, K. and B. Wessels. (2005). Electoral Turnout in Jacques Thomassen (Ed.), The European

Voter A Comparative Study Of Modern Democracies. Oxford University Press, pp. 64-83.

Arter, D. (1996). 'Estonia After the March 1995 Riigikogu Election: Still an Anti-Party System?', TheJournal of Communist Studies and Transition Politics, Vol 11 (3), 24971.

Arzheimer, K. (2005). Die Wahl extremistischer Parteien in Jrgen W. Falter; Harald Schoen(Hrsg.).Handbuch Wahlforschung . VS Verlag fr Sozialwissenschaften. pp. 389-421.

Belanger, E. (2004). Antipartyism and third-party vote choice: a comparison of Canada, Britain, andAustralia, Comparative Political Studies, Vol.37, Issue 9, pp 1054-1078.

Blais, A.; E. Gidengil; N. Nevitte and R. Nadeua. (2004)., Where does turnout decline come from?,European Journal of Political Research Vol.43, No. 1, pp. 221236,

Blais, A.; E. Gidengil; P. Fournier and N. Nevitte. (2009). Information, visibility and elections: Whyelectoral outcomes differ when voters are better informed European Journal of Political Research 48(2): 256280

Brancati, D. (2008). Winning Alone: The Electoral Fate of Independent Candidates Worldwide, TheJournal of Politics, Vol. 7, Issue 3, pp 648-662.

Caballero, C. (2005). Nichtwahl in Jrgen W. Falter; Harald Schoen (Hrsg.).HandbuchWahlforschung. VS Verlag fr Sozialwissenschaften. pp. 329-363.

Van der Eijk, C. and M. Franklin, eds. (1996). Choosing Europe? The European Electorate and

National politics in the Face of the Union. Ann Arbor: The University of Michigan Press

Van der Eijk, C; W. van der Brug; M Kroh and M Franklin. (2006). Rethinking the dependentvariable in voting behavior: On the measurement and analysis of electoral utilities. Electoral Studies,

Vol. 25, No.3, pp. 424-447.

Dalton, R. (2004).Democratic Challenges, Democratic Choices: The Erosion of Political Support in

Advanced Industrial Democracies. Oxford: Oxford University Press.

Donovan, T; S. Bowler and T. Terrio. (2000). Support for Third Parties in CaliforniaAmerican

Politics Quarterly, Vol.28, No. 2 pp 50-71.

Ehin, P. (2007). Political Support in the Baltic States 1993-2004.Journal of Baltic Studies, Vol.38(1): 1-20.

Eurobarometer 72, autumn 2009. available at

http://ec.europa.eu/public_opinion/archives/eb/eb72/eb72_ee_ee_nat.pdf

European Parliament Election Study 2009 [Voter Study], Advance Release, 16/04/2010,

(www.piredeu.eu). Marcel H. van Egmond, Eliyahu V. Sapir, Wouter van der Brug, Sara B. Hobolt,

Mark N. Franklin (2010) EES 2009 Voter Study Advance Release Notes. Amsterdam: University of

Amsterdam.
http://ec.europa.eu/public_opinion/archives/eb/eb72/eb72_ee_ee_nat.pdfhttp://ec.europa.eu/public_opinion/archives/eb/eb72/eb72_ee_ee_nat.pdfhttp://www.piredeu.eu/http://www.piredeu.eu/http://www.piredeu.eu/http://www.piredeu.eu/http://ec.europa.eu/public_opinion/archives/eb/eb72/eb72_ee_ee_nat.pdf


20/22

20

Franklin, M. (2003). The Dynamics of Electoral Participation in Lawrence LeDuc, Lawrence,

Richard G. Niemi; Pippa Norris (eds). Comparing Democracies 2: New Challenges in the Study of

Elections and Voting, London: Sage. pp. 148-168.

Franklin, M. (2005). The fading power of national politics to structure voting behavior in elections to

the European Parliament. Paper prepared for the Budapest Conference on the European ElectionStudy, May 2005, available athttp://www.ees-homepage.net/papers/budapest/franklin.pdf

Gerring, J. (2005). Minor Parties in Plurality Electoral Systems Party Politics, Vol. 11, No.1 pp 79-107.

Giles, M. W. and M. K. DAntico. (1982). Political participation and neighborhood social context

revisited.American Journal of Political Science Vol. 26, No. 1, pp. 144150.

Hirschman, A. (1970).Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and

States. Cambridge, MA: Harvard University Press.

Hix, S. and M. Marsh (2007). Punishment or Protest? Understanding European ParliamentElections, The Journal of Politics, Vol. 69, pp 495-510.

Huberty, C.. J. and J. M. Wisenbaker. (1986). Using Categorical Variables in DiscriminantAnalysis.,Multivariate Behavioral Research, Vol. 21, No. 4, pp. 479-496.

Huberty, C. J. (1994).Applied Discriminant Analysis. John Wiley & Sons, Inc.

Koepke, J. and N. Ringe (2006), The Second-order Election Model in an Enlarged Europe,European Union Politics, Vol. 7 (3): 321346.

Lagerspetz, M. and H. Vogt (2004), Estonia in S. Berglund, J. Ekman, and F. Aarebrot (eds.), The

Handbook of Political Change in Eastern Europe. Cheltenham: Edward Elgar, pp. 57-93.

Lubbers, M and P. Scheepers. (2000). Individual and contextual characteristics of the Germanextreme right-wing vote in the 1990s. A Test of Complementary Theories European Journal ofPolitical Research. Vol. 38, pp 63-94.

Marsh, M. (1998). Testing the second-order election model after four European Parliament elections.British Jounal for Political Science. Vol 28 (4): 591607.

Matsusaka, J.G. and F. Palda. (1999). Voter turnout: How much can we explain?Public Choice Vol.98, No. 3-4, pp. 431-446.

Norris, P. (ed.). (1999). Critical citizens: Global support for democratic governance. Oxford,

UK: Oxford University Press.

Oppenhuis, E. (1995). Voting Behavior in Europe. Amsterdam: Het Spinhuis.

Owen, D. and J. Dennis (1996). Anti-partyism in the USA and support for Ross Perot.EuropeanJournal of Political Research, 29, 383-400.

Poguntke, T. (1996). Anti-party sentimentConceptual thoughts and empirical evidence:Explorations into a minefield.European Journal of Political Research, 29, 319-344.

Reif, K. (1984). National electoral cycles and European elections.Electoral Studies, Vol. 3 (3):24455.
http://www.ees-homepage.net/papers/budapest/franklin.pdfhttp://www.ees-homepage.net/papers/budapest/franklin.pdfhttp://www.ees-homepage.net/papers/budapest/franklin.pdfhttp://www.ees-homepage.net/papers/budapest/franklin.pdf


21/22

21

Reif, K., and H. Schmitt (1980). Nine second order national elections: A conceptual framework for

the analysis of European election results.European Journal of Political Research, Vol. 8 (1): 344.

Rosenstone, S; R. Behr and Edward Lazarus. (1984). Third Parties in America: Citizen Response to

Major Party Failure. Princetone University Press.

Schmitt, H. (2009). European Parliament Elections after Eastern Enlargement, Journal of EuropeanIntegration, Vol. 31 (5).

Sikk, Allan. (2006). Highways to power: New party success in three young democracies. Tartu: Tartu

University Press.

Tabachnick, B. G. and L. S. Fidell. (2001). Using Multivariate Statistics. 4th ed. Allan & Bacon,

Boston.

Vihalemm, T. and V. Kalmus. (2009). Cultural Differentation of the Russian Minority.Journal ofBaltic Studies, 40(1): 95-119.

Wessels, B. and H-D. Klingemann. (2006). Parties and Voters Representative Consolidation inCentral and Eastern Europe? International Journal of Sociology, Vol 36, No. 2, pp 11-44.

Zuckerman, A. S.; L. A. Kotler-Berkowitz and L. A. Swaine. (1998). Anchoring politicalpreferences: The structural bases of stableelectoral decisions and political attitudes in Britain.European Journal of Political Research Vol.33, No. 3, pp. 285321.


22/22

22

Endnotes

1Lszl Tks, elected to the EP as an independent candidate in 2007, was backed by Fidesz, the main

opposition party in Hungary. Elena Bsescu, daughter of the incumbent president of Romania, had close ties

with the Democratic Liberal Party, which she rejoined immediately after the election.2 The resulting vote breakdown in the sample is as follows - 20.8% (N=194) coalition party voters, 24.3% (227)opposition party voters, 19.8% (185) independent voters and 35.0% (327) non-voters. The breakdown in reality

applying the same grouping rule was as follows - 15.8% coalition party voters, 13.5% opposition voters, 11.4%

independent voters, 56.2% non-voters (and 3.1% who voted for others).3 DDA is a normal-based analysis but it is to a certain degree robust to non-normality and ordinal or binary

variables can be used (Huberty & Wisenbaker 1986). Variables that required normalisation are marked in Table

4. Though multivariate normality was still not achieved DDA is robust to the violation of the assumption for

grouped data in case of similar group sizes and large overall sample size (Tabachnick & Fidell 2001, 462). As

sensitivity to outliers is a problem for the technique two multivariate outliers where removed with the help of

Mahalanobis distances. Boxs M test was used to evaluate equality of covariance matrices. As this test is overly

sensitive to any violations (ibid, 330) the scatterplots were examined for about equal scatter which allows to

proceed even in cases of highly significant Boxs M results (ibid., 463).4

The DDA used 516 cases in total because of listwise deletion of missing values resulting in the following

breakdown -104 Tarand voters, 88 coalition voters, 164 opposition voters and 160 non-voters.

I2 2-Ehin and Solvak Paper 4-11-2010

Documents

Transcript of I2 2-Ehin and Solvak Paper 4-11-2010